Geographic Information

Systems:
Concepts, Methodologies, Tools,
and Applications
Information Resources Management Association
USA

Volume I
Managing Director: Lindsay Johnston
Editorial Director: Joel Gamon
Assistant Acquisitions Editor: Kayla Wolfe
Book Production Manager: Jennifer Romanchak
Publishing Systems Analyst: Adrienne Freeland
Development Editor: Chris Wozniak
Assistant Production Editor: Deanna Jo Zombro
Cover Design: Nick Newcomer

Published in the United States of America by

Information Science Reference (an imprint of IGI Global)
701 E. Chocolate Avenue
Hershey PA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: cust@igi-global.com
Web site: http://www.igi-global.com

Copyright © 2013 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or
companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.

Library of Congress Cataloging-in-Publication Data

Geographic information systems (2013)
Geographic information systems : Concepts, methodologies, tools, and applications / Information Resources Management
Association, Editor.
pages cm
Includes bibliographical references and index.
Summary: “This reference is a collection of knowledge on the latest advancements and research of geographic information
systems, offering background for those involved in geographical data”--Provided by publisher.
ISBN 978-1-4666-2038-4 (hardcover) -- ISBN (invalid) 978-1-4666-2039-1 (ebook) -- ISBN (invalid) 978-1-4666-2040-7
(print & perpetual access) 1. Geographic information systems. 2. Geographic information systems--Research. 3. Geospa-
tial data. I. Information Resources Management Association. II. Title.
G70.212.G4257 2013
910.235--dc23
2012029283

British Cataloguing in Publication Data

A Cataloguing in Publication record for this book is available from the British Library.

The views expressed in this book are those of the authors, but not necessarily of the publisher.
Editor-in-Chief
Mehdi Khosrow-Pour, DBA
Editor-in-Chief
Contemporary Research in Information Science and Technology, Book Series

Associate Editors
Steve Clarke, University of Hull, UK
Murray E. Jennex, San Diego State University, USA
Annie Becker, Florida Institute of Technology, USA
Ari-Veikko Anttiroiko, University of Tampere, Finland

Editorial Advisory Board

Sherif Kamel, American University in Cairo, Egypt
In Lee, Western Illinois University, USA
Jerzy Kisielnicki, Warsaw University, Poland
Keng Siau, University of Nebraska-Lincoln, USA
Amar Gupta, Arizona University, USA
Craig van Slyke, University of Central Florida, USA
John Wang, Montclair State University, USA
Vishanth Weerakkody, Brunel University, UK
List of Contributors

Ababsa, Fakhreddine / University of Evry Val d’Essonne, France................................................. 2164

Abadía, Mercedes Farjas / Universidad Politécnica de Madrid, Spain............................................ 970
Abdelouahad, Abdelkaher Ait / University of Mohammed V-Agdal, Morocco................................ 616
Abdullah, Alias / International Islamic University Malaysia, Malaysia............................................ 991
Abdullah, Muhammad Faris / International Islamic University Malaysia, Malaysia..................... 991
Abid, Mohamed Amine / University of Manouba, Tunisia.............................................................. 1345
Aboutajdine, Driss / University of Mohammed V-Agdal, Morocco................................................... 616
Adamchuk, Viacheslav / McGill University, Canada...................................................................... 1368
Aguilar, José Alberto Hernandez / Autonomous University of Morelos, Mexico........................... 1809
Aguilera, Fatima Sayuri Quezada / Autonomous University of Aguacalientes, Mexico................ 1809
Akinyede, J. O. / African Regional Centre for Space and Technology Education, Nigeria............. 1515
Akinyemi, Felicia O. / Kigali Institute of Science and Technology, Rwanda................................... 2040
Alabi, Tunrayo / International Institute of Tropical Agriculture, Nigeria......................................... 912
Alam, Bhuiyan Monwar / The University of Toledo, USA.............................................................. 1385
Algeo, Katie / Western Kentucky University, USA............................................................................ 1309
Ali, Hesham A. / Mansoura University, Egypt................................................................................. 1448
Ali, Mohammed H. / Mansoura University, Egypt.......................................................................... 1448
Allen, Jeffery / Clemson University, USA......................................................................................... 1297
Al-Marri, Salem / Leeds Metropolitan University LEEDS, UK....................................................... 1604
Annegarn, Harold J. / University of Johannesburg, South Africa..................................................... 880
Argyreas, Nikolaos / National Center of Scientific Research “Demokritos,” Greece....................... 726
Arnold, Lesley / Landgate, Australia................................................................................................. 638
Ayansina, Ayanlade / Obafemi Awolowo University, Nigeria............................................................ 938
Ayats, Victor Pascual / Spatial Data Infrastructure of Catalonia (IDEC), Spain............................. 465
Balasubramaniam, S. / JSS University, India.................................................................................. 1988
Barker, Bradley / University of Nebraska-Lincoln, USA................................................................. 1368
Barth, Andreas / Beak Consultants GmbH, Germany........................................................................ 765
Batsakis, Sotirios / Technical University of Crete, Greece............................................................... 1751
Bekkers, Victor / Erasmus University Rotterdam, The Netherlands................................................ 1974
Belghith, Abdelfettah / University of Manouba, Tunisia................................................................. 1345
Beniston, Martin / University of Geneva, Switzerland.................................................................... 1260
Bennewitz, Evelyn / Beak Consultants GmbH, Germany.................................................................. 765
Berglund, Ulla / Swedish University of Agricultural Sciences, Sweden............................................. 799
Bermudez, Luis / Southeastern Universities Research Association, USA......................................... 890
Bernabé-Poveda, Miguel-Ángel / Technical University of Madrid, Spain............................................ 1
Bernadó, Oriol / Science and Technology Park for Tourism and Leisure, Spain............................. 1100
Bigorra, Albert / Science and Technology Park for Tourism and Leisure, Spain............................ 1100
Bimonte, Sandro / Laboratoire d’Informatique en Images et Systèmes d’information, France &
Cemagref, UR TSCF, France..................................................................................................... 91, 377
Bizani, Erasmia / Aristotle University of Thessaloniki, Greece....................................................... 1559
Blackwell, P. R. / Stephen F. Austin State University, USA................................................................ 926
Blank, Daniel / University of Bamberg, Germany.............................................................................. 666
Bol, Peter K. / Harvard University, USA.......................................................................................... 1764
Borràs, Joan / Science and Technology Park for Tourism and Leisure, Spain................................. 2208
Brimicombe, Allan J. / University of East London, UK..................................................................... 151
Brisaboa, Nieves R. / University of A Coruña, Spain......................................................................... 358
Brister, Evelyn / Rochester Institute of Technology, USA................................................................ 2063
Brown, Stacey R. / Oklahoma State University, USA...................................................................... 1864
Brunner, Dominik / Fraunhofer Institute of Optronics, Germany................................................... 1587
Bruno, Fernanda / Universidade Federal do Rio de Janeiro, Brazil.................................................. 66
Brunt, Matthew / Western Kentucky University, USA..................................................................... 1309
Bunch, Rick / University of North Carolina at Greensboro, USA.................................................... 1612
Burns, Meghan D. / Montana Natural Heritage Program, USA..................................................... 1170
Butterworth, Edwin / CACI International, USA............................................................................. 1665
Buurma, J.S. / Wageningen UR, The Netherlands........................................................................... 1883
Calvo, Pedro Ángel Merino / Universidad Politécnica de Madrid, Spain........................................ 970
Cameron, Nairne / Algoma University, Canada.............................................................................. 1665
Cangel, Ayça / International Competitiveness Research Institute, Turkey....................................... 1392
Carroll, Jon W. / Michigan State University, USA........................................................................... 1328
Cerney, Dawna L. / Youngstown State University, USA................................................................... 1665
Chandramouli, Magesh / Purdue University, USA............................................................................ 216
Chen, Yenming J. / National Kaohsiung First University of Science and Technology, Taiwan.......... 81
Cherifi, Hocine / University of Burgundy, France.............................................................................. 616
Christophoridis, Christophoros / Aristotle University of Thessaloniki, Greece............................. 1559
Clave, Salvador Anton / University Rovira i Virgili, Spain................................................... 1100, 2208
Coetzee, Serena / University of Pretoria, South Africa ................................................................... 1794
Comer, Jonathan C. / Oklahoma State University, USA.................................................................. 1864
Connolly, Matthew H. / Texas State University-San Marcos, USA................................................. 1953
Conover, Helen / The University of Alabama in Huntsville, USA...................................................... 890
Cooper, Antony K / CSIR, South Africa & University of Pretoria, South Africa............................. 1794
Corcho, Oscar / Universidad Politécnica de Madrid, Spain.............................................................. 434
Crumlish, John / University of Ulster, UK......................................................................................... 540
Curran, Kevin / University of Ulster, UK.......................................................................................... 540
Curtin, Kevin M. / George Mason University, USA............................................................................ 39
D’Ulizia, Arianna / Consiglio Nazionale delle Ricerche, IRPPS, Italy.............................................. 848
Dao, Hy / University of Geneva and UNEP, Switzerland.................................................................. 1260
Dartevelle, Sébastien / Los Alamos National Laboratory, USA....................................................... 1697
Day, Patrice / University of Wisconsin-Milwaukee, USA................................................................. 1525
De Bono, Andrea / University of Geneva and UNEP, Switzerland.................................................. 1260
de Kool, Dennis / Center for Public Innovation, The Netherlands.................................................. 1974
de la Flor, Joan / Science and Technology Park for Tourism and Leisure, Spain............................ 2208
Dean, Denis J. / University of Texas at Dallas, USA........................................................................ 1828
DeGroote, John / University of Northern Iowa, USA......................................................................... 480
DeMers, Michael N. / New Mexico State University, USA............................................................... 1541
Díaz, Laura / Universitat Jaume I de Castellón, Spain & Institute of New Imaging Technologies,
Universitat Jaume I, Spain....................................................................................................... 305, 952
Didier, Jean-Yves / University of Evry Val d’Essonne, France........................................................ 2164
Dolney, Timothy J. / Penn State University – Altoona, USA........................................................... 2128
Duckham, Matt / University of Melbourne, Australia......................................................................... 24
Dugas, Daniel P. / New Mexico State University, USA..................................................................... 1541
Duke-Williams, Oliver / University of Leeds, UK........................................................................... 1675
Eckert, Jeanette / The University of Toledo, USA............................................................................ 1385
El Asmar, Toufic / University of Florence, Italy................................................................................. 228
El Hassouni, Mohammed / University of Mohammed V-Agdal, Morocco........................................ 616
Epelbaum, Samuel / Pace University, USA..................................................................................... 1724
Epperson, Ann / Western Kentucky University, USA....................................................................... 1309
Everding, Thomas / University of Muenster, Germany.................................................................... 2020
Fernández-Wyttenbach, Alberto / Technical University of Madrid, Spain.......................................... 1
Ferri, Fernando / Consiglio Nazionale delle Ricerche, IRPPS, Italy................................................ 848
Fisher, Gavin / University of Ulster, UK............................................................................................. 540
Fitzner, Daniel / Fraunhofer IGD, Germany.................................................................................... 1619
Fonollosa, Arturo Beltran / Universitat Jaume I de Castellón, Spain............................................ 2176
Fraser, Kym / University of South Australia, Australia................................................................... 2150
Fridman, Alexander / Russian Academy of Science, Russia........................................................... 1724
Fugazza, Cristiano / Institute of Environment and Sustainability, Italy............................................ 305
Fytianos, Konstantinos / Aristotle University of Thessaloniki, Greece........................................... 1559
Gallegos, Julio Cesar Ponce / Autonomous University of Aguacalientes, Mexico.......................... 1809
Gallo, Giovanni / University of Catania, Italy................................................................................. 1132
Gangopadhyay, Aryya / University of Maryland - Baltimore County, USA........................................ 47
Gardarin, Georges / PRiSM Laboratory, France.............................................................................. 550
Gaur, S. / Ecole Nationale Supérieure des Mines, France................................................................. 736
Gelernter, Judith / Carnegie Mellon University, USA..................................................................... 1941
Gensel, Jerome / Laboratoire d’Informatique de Grenoble, France.................................................. 377
Ghose, Rina / University of Wisconsin-Milwaukee, USA................................................................. 1525
Gil, José / Institute of New Imaging Technologies, Universitat Jaume I, Spain................................. 952
Gillavry, Edward Mac / Webmapper, The Netherlands................................................................... 1231
Giuliani, Gregory / University of Geneva and UNEP, Switzerland................................................. 1260
Goh, Mark / University of South Australia, Australia & National University of Singapore,
Singapore....................................................................................................................................... 2150
Gorod, Alex / Stevens Institute of Technology, USA......................................................................... 1724
Gould, Michael / Universitat Jaume I, Spain..................................................................................... 493
Graham, Amy K. / Oklahoma State University, USA...................................................................... 1864
Graillot, D. / Ecole Nationale Supérieure des Mines, France............................................................ 736
Grandgenett, Neal / University of Nebraska at Omaha, USA.......................................................... 1368
Granell, Carlos / Universitat Jaume I de Castellón, Spain & Institute for Environment and
Sustainability, Italy.................................................................................................................. 305, 952
Graves, Sara / The University of Alabama in Huntsville, USA.......................................................... 890
Greenlee, Janet C. / New Mexico State University, USA................................................................. 1541
Gribb, William J. / University of Wyoming, USA............................................................................ 1665
Grifoni, Patrizia / Consiglio Nazionale delle Ricerche, IRPPS, Italy................................................ 848
Guan, Weihe (Wendy) / Harvard University, USA.......................................................................... 1764
Guarnera, Giuseppe Claudio / University of Catania, Italy........................................................... 1132
Guo, Zhen-Sheng / Hokkaido University, Japan.............................................................................. 1216
Gupta, Phalguni / Indian Institute of Technology Kanpur, India....................................................... 703
Hammitzsch, Martin / Helmholtz Centre Potsdam, GFZ German Research Centre for Geosciences,
Germany........................................................................................................................................... 449
Hane, Elizabeth / Rochester Institute of Technology, USA.............................................................. 2063
Hanke, Henrik / University of Duisburg-Essen, Germany................................................................... 17
Hansen, Stefan / Spatial/Information Systems Ltd./LISAsoft, Australia............................................ 250
Hawthorne, Timothy L. / Columbus State University, USA............................................................ 1848
Haynes, Kingsley E. / George Mason University, USA.................................................................... 1665
Hegedüs, Péter / Budapest University of Technology and Economics, Hungary............................... 609
Henrich, Andreas / University of Bamberg, Germany....................................................................... 666
Herold, Sam / University of Ottawa, Canada..................................................................................... 175
Ho, Chao-Ching / National Yunlin University of Science and Technology, Taiwan........................... 783
Hodge, Bill / City of Midland, USA................................................................................................... 1665
Honea, Robert B. / KU Transportation Research Institute, USA..................................................... 1665
Hosszú, Gábor / Budapest University of Technology and Economics, Hungary............................... 609
Huang, Bo / Chinese University of Hong Kong, China...................................................................... 216
Huerta, Joaquín / Institute of New Imaging Technologies, Universitat Jaume I, Spain.................... 952
Hyyppä, Juha / Finnish Geodetic Institute, Finland........................................................................ 1011
Ioja, Cristian / University of Bucharest, Romania........................................................................... 1276
Iqbal, Muhammad Usman / University of New South Wales, Australia........................................... 399
Irkhin, Alexey / Beak Consultants GmbH, Germany......................................................................... 765
Isern, David / Universitat Rovira i Virgili, Spain............................................................................. 2208
Jaakkola, Anttoni / Finnish Geodetic Institute, Finland................................................................. 1011
Jegede, O. O. / Obafemi Awolowo University, Nigeria....................................................................... 938
Jenihhin, Maksim / Tallinn University of Technology, Estonia......................................................... 407
Jha, Manoj K. / Morgan State University, USA................................................................................. 857
Joshi, P. K. / TERI University, India................................................................................................. 1913
Jutman, Artur / Tallinn University of Technology, Estonia............................................................... 407
Kaartinen, Harri / Finnish Geodetic Institute, Finland................................................................... 1011
Kakaletris, George / University of Athens, Greece............................................................................ 258
Kathlene, Lyn / Colorado State University, USA............................................................................... 685
Katsianis, Dimitris / University of Athens, Greece............................................................................ 258
Kaushik, Vandana Dixit / Harcourt Butler Technological Institute, India........................................ 703
Keiser, Ken / The University of Alabama in Huntsville, USA............................................................. 890
Khant, Aung Phey / Asian Institute of Technology, Thailand............................................................ 815
Kholladi, Mohamed-Khireddine / University Mentouri of Constantine, Algeria............................ 750
Klimaszewski-Patterson, Anna / New Mexico State University, USA............................................ 1541
Klippel, Alexander / University of Melbourne, Australia.................................................................. 250
Korfmacher, Karl / Rochester Institute of Technology, USA........................................................... 2063
Kourie, Derrick G / University of Pretoria, South Africa . ............................................................. 1794
Koussouris, Sotirios / National Technical University of Athens, Greece......................................... 1243
Kovács, Ferenc / Budapest University of Technology and Economics, Hungary............................... 609
Ku, Wei-Shinn / Auburn University, USA.......................................................................................... 631
Lancaster, Anne / Independent Researcher, USA............................................................................... 529
Larsen, Lillian I. / University of Redlands, USA................................................................................ 832
Lehmann, Anthony / University of Geneva and UNEP, Switzerland.............................................. 1260
Li, Yang / University of East London, UK.......................................................................................... 151
Lim, Samsung / University of New South Wales, Australia............................................................... 399
Lindquist, Peter S. / The University of Toledo, USA........................................................................ 1385
Liu, Zhengjun / Chinese Academy of Surveying and Mapping, China............................................ 1011
Llaves, Alejandro / University of Muenster, Germany..................................................................... 2020
Loi, Nguyen Kim / Nong Lam University, Vietnam.......................................................................... 2080
Lourenço, Júlia M. / University of Minho, Portugal....................................................................... 1476
Lu, Kang Shou / Towson University, USA........................................................................................ 1297
Luaces, Miguel R. / University of A Coruña, Spain........................................................................... 358
Lutz, Collin / Virginia Polytechnic Institute and State University, USA.......................................... 1368
Manso-Callejo, Miguel-Angel / Universidad Politécnica de Madrid, Spain.................................. 2176
Mansouri, Mo / Stevens Institute of Technology, USA..................................................................... 1724
Martins, Hugo / Universidade Nova de Lisboa, Portugal & Universidade Técnica de Lisboa,
Portugal......................................................................................................................................... 1773
Maskey, Manil / The University of Alabama in Huntsville, USA....................................................... 890
Mathews, Leah Greden / University of North Carolina at Asheville, USA....................................... 529
Matyas, Corene J. / University of Florida, USA.............................................................................. 1069
Mazimpaka, Jean Damascène / National University of Rwanda, Rwanda....................................... 295
Mazzetti, Paolo / National Research Council, Italy........................................................................... 305
McDonald, Darrel / Stephen F. Austin State University, USA............................................................ 926
McGuire, Steven / Aberystwyth University, UK................................................................................. 108
McKnight, Kevin P. / Michigan Department of Transportation & Michigan State University,
USA................................................................................................................................................ 1170
Meenar, Mahbubur R. / Temple University, USA.............................................................................. 122
Men, Hao / Stevens Institute of Technology, USA............................................................................... 502
Messina, Joseph P. / Michigan State University, USA..................................................................... 1170
Messineo, Antonio / Kore University of Enna, Italy......................................................................... 1411
Mimoun, D. / Ecole Nationale Supérieure des Mines, France........................................................... 736
Misra, Santosh K. / Cleveland State University, USA........................................................................ 336
Moody, Rebecca / Erasmus University Rotterdam, The Netherlands.............................................. 1974
Moore, Alyssa K. / University of Redlands, USA............................................................................... 832
Moreno, Antonio / Universitat Rovira i Virgili, Spain..................................................................... 2208
Morgan, John / Towson University, USA......................................................................................... 1297
Morgan, Kathy / University of Nebraska-Lincoln, USA.................................................................. 1368
Nakato, Teddy / Ministry of Agriculture, Uganda.............................................................................. 938
Nativi, Stefano / National Research Council, Italy............................................................................ 305
Neumann, Alf / University of Cologne, Germany................................................................................ 17
Niţă, Mihai / University of Bucharest, Romania............................................................................... 1276
Noack, Silke / Beak Consultants GmbH, Germany............................................................................ 765
Norder, John W. / Michigan State University, USA......................................................................... 1328
Nordin, Kerstin / Swedish University of Agricultural Sciences, Sweden........................................... 799
Nugent, Gwen / University of Nebraska-Lincoln, USA.................................................................... 1368
Oduwole, Olusoji / Cocoa Research Institute of Nigeria, Nigeria..................................................... 912
Okafor, Christopher / International Institute of Tropical Agriculture, Nigeria................................ 912
Olaleye, V. F. / Obafemi Awolowo University, Nigeria....................................................................... 938
Olufemi, Bolarin / University of Ilorin Nigeria................................................................................. 938
Onose, Diana / University of Bucharest, Romania........................................................................... 1276
Opeyemi, Zubair A. / National Space Research and Development Agency, Nigeria...................... 1515
Orosz, Mihály / Budapest University of Technology and Economics, Hungary................................ 609
Oryspayev, Dossay / University of Northern Iowa, USA.................................................................. 480
Panno, Domenico / University of Palermo, Italy.............................................................................. 1411
Pappathanasi, Kym / Salem State University, USA......................................................................... 1090
Paradzayi, Charles / University of Johannesburg, South Africa & Midlands State University,
Zimbabwe......................................................................................................................................... 880
Patent-Nygren, Megan / University of Nebraska-Lincoln, USA..................................................... 1368
Pathak, Prasad A. / University of North Carolina at Greensboro, USA.................................. 130, 1612
Patroescu, Maria / University of Bucharest, Romania..................................................................... 1276
Pazos, Andrés / Universitat Jaume I, Spain........................................................................................ 493
Peduzzi, Pascal / UNEP, Switzerland............................................................................................... 1260
Pérez, Yolanda / University Rovira i Virgili, Spain................................................................ 1100, 2208
Peters, Rob / University of Amsterdam, The Netherlands................................................................ 1243
Petrakis, Euripides G.M. / Technical University of Crete, Greece................................................. 1751
Pigozzi, Bruce W. / Michigan State University, USA....................................................................... 1170
Pochiraju, Kishore / Stevens Institute of Technology, USA............................................................... 502
Poveda, José / University of Texas, USA............................................................................................. 493
Prasetyo, Lilik B. / Bogor Agriculture University, Indonesia.......................................................... 1901
Priyanka, Neena / TERI University, India........................................................................................ 1913
Pultar, Edward / University of California – Santa Barbara, USA................................................... 2006
Quddus, Mohammed A. / Loughborough University, UK................................................................. 658
Quintana, Manuel Sillero / Universidad Politécnica de Madrid, Spain........................................... 970
Rahaman, Khan R. / University of Minho, Portugal....................................................................... 1476
Raik, Jaan / Tallinn University of Technology, Estonia...................................................................... 407
Ramachandran, Muthu / Leeds Metropolitan University, LEEDS UK........................................... 1604
Ray, Nicolas / University of Geneva and UNEP, Switzerland.......................................................... 1260
Reid, Neil / The University of Toledo, USA....................................................................................... 1053
Rex, Art / Appalachian State University, USA.................................................................................... 529
Riccioli, Francesco / University of Florence, Italy............................................................................. 228
Richter, Kai-Florian / Universität Bremen, Germany....................................................................... 250
Rocha, Jorge G. / Universidade do Minho, Portugal....................................................................... 1773
Rozylowicz, Laurenţiu / University of Bucharest, Romania............................................................ 1276
Russo, Antonio P. / University Rovira i Virgili, Spain............................................................ 1100, 2208
Saleh, Ahmed I. / Mansoura University, Egypt................................................................................ 1448
Samanta, Sutapa / Maryland State Highway Administration, USA................................................... 857
Santoro, Mattia / National Research Council, Italy........................................................................... 305
Saquicela, Victor / Universidad Politécnica de Madrid, Spain......................................................... 434
Sauser, Brian / Stevens Institute of Technology, USA....................................................................... 1724
Savary, Lionel / PRiSM Laboratory, France...................................................................................... 550
Sawada, Michael C. / University of Ottawa, Canada........................................................................ 175
Schmidt, Frank / Beak Consultants GmbH, Germany....................................................................... 765
Schwarzer, Stefan / UNEP, Switzerland........................................................................................... 1260
Seco, Diego / University of A Coruña, Spain...................................................................................... 358
Setiawan, Yudi / Bogor Agriculture University, Indonesia.............................................................. 1901
Shampur, Nikhil V. / Cleveland State University, USA........................................................................ 47
Shortridge, Ashton M. / Michigan State University, USA............................................................... 1170
Siabato, Willington / Technical University of Madrid, Spain................................................................ 1
Sikder, Iftikhar U. / Cleveland State University, USA............................................. 47, 336, 1183, 1421
Sinton, Diana Stuart / University of Redlands, USA......................................................................... 832
Smith, Bruce W. / Bowling Green State University, USA................................................................ 1053
Sonder, Kai / International Maize and Wheat Improvement Center, Mexico..................................... 912
Sorrentino, John A. / Temple University, USA................................................................................... 122
Sphicopoulos, Thomas / University of Athens, Greece...................................................................... 258
Sportolaro, Carlo / Agronomist, Perugia, Italy................................................................................ 1429
Sridhar, R. / Sri Ramakrishna Mission Vidyalaya College of Arts and Science, India.................... 1988
Srivastava, Rajeev / Institute of Technology, Banaras Hindu University (ITBHU), India................ 569
Stanco, Filippo / University of Catania, Italy................................................................................... 1132
Stefansson, Gunnar / Chalmers University of Technology, Sweden & University of Iceland,
Iceland............................................................................................................................................ 2051
Stillwell, John / University of Leeds, UK......................................................................................... 1675
Sugumaran, Ramanathan / University of Northern Iowa, USA....................................................... 480
Tanaka, Yuzuru / Hokkaido University, Japan................................................................................ 1216
Tanasi, Davide / Arcadia University, USA........................................................................................ 1132
Tapp, Anna / University of North Carolina at Greensboro, USA..................................................... 1612
Thomopoulos, Stelios C. A. / National Center of Scientific Research “Demokritos,” Greece......... 726
Todaro, Rudj Maria / Insula S.p.A., Italy........................................................................................ 1392
Torquati, Biancamaria / University of Perugia, Italy...................................................................... 1429
Towers, George / Concord University, USA..................................................................................... 1151
Toxopeus, Albertus G. / University of Twente, The Netherlands..................................................... 2102
Tripathi, Nitin Kumar / Asian Institute of Technology, Thailand..................................................... 815
Trisurat, Yongyut / Kasetsart University, Thailand......................................................................... 2102
Tsitsanis, Anastasios / National Technical University of Athens, Greece........................................ 1243
Ubar, Raimund / Tallinn University of Technology, Estonia.............................................................. 407
Valcik, Nicolas A. / The University of Texas at Dallas, USA............................................................ 1502
Valls, Aida / Universitat Rovira i Virgili, Spain................................................................................ 2208
van Westen, Cees / University of Twente, The Netherlands............................................................. 1120
Van Woerden, Jaap / UNEP, Switzerland........................................................................................ 1260
Varoutas, Dimitris / University of Athens, Greece............................................................................. 258
Veenendaal, Bert / Curtin University, Australia................................................................................ 638
Vilches-Blázquez, Luis. M. / Universidad Politécnica de Madrid, Spain......................................... 434
Villalón, Christian José Correa / Institute of Education of Aguascalientes, Mexico...................... 1809
Villanova-Oliver, Marlène / Laboratoire d’Informatique de Grenoble, France............................... 377
Vizzari, Marco / University of Perugia, Italy................................................................................... 1429
Vogt, Brandon J. / University of Colorado at Colorado Springs, USA................................. 1200, 1665
Volpe, Roberto / University of Palermo, Italy.................................................................................. 1411
Wallgrün, Jan Oliver / Universität Bremen, Germany.................................................................... 1639
Wang, Haojun / University of Southern California, USA.................................................................. 631
West, Geoff / Curtin University, Australia.......................................................................................... 638
Whalen, Stephen / University of North Carolina at Chapel Hill, USA.............................................. 130
Whitesides, Clayton J. / Texas State University-San Marcos, USA................................................. 1953
Whitford, Walter G. / New Mexico State University, USA.............................................................. 1541
Wijaya, Chandra Irawadi / Bogor Agriculture University, Indonesia............................................ 1901
Witt, Ron / UNEP, Switzerland......................................................................................................... 1260
Woldai, Tsehaie / University of Twente, The Netherlands................................................................ 1120
Wolter, Diedrich / Universität Bremen, Germany............................................................................ 1639
Woodside, Joseph M. / Cleveland State University, USA................................................................ 1183
Wu, Xiaoying / Curtin University, Australia...................................................................................... 638
Xia, Jianhong (Cecilia) / Curtin University, Australia...................................................................... 638
Yesmin, Sharmin / Temple University, USA....................................................................................... 122
Young, Stephen / Salem State University, USA................................................................................ 1090
Zadorozhny, Vladimir I. / University of Pittsburgh, USA................................................................... 30
Zahari, Rustam Khairi / International Islamic University Malaysia, Malaysia............................... 991
Zeitouni, Karine / PRiSM Laboratory, France.................................................................................. 550
Zendjebil, Iman Maissa / University of Evry Val d’Essonne, France............................................. 2164
Zeug, Gunter / European Environment Agency, Denmark............................................................... 1587
Zhu, Lingli / Finnish Geodetic Institute, Finland............................................................................. 1011
Zimmermann, Roger / National University of Singapore, Singapore............................................... 631
Table of Contents

Volume I

Section 1
Fundamental Concepts and Theories
This section serves as a foundation for this exhaustive reference tool by addressing underlying principles
essential to the understanding of Geographic Information Systems. Chapters found within these pages
provide an excellent framework in which to position Geographic Information Systems within the field of
information science and technology. Insight regarding the critical incorporation of global measures into
Geographic Information Systems is addressed, while crucial stumbling blocks of this field are explored.
With 15 chapters comprising this foundational section, the reader can learn and chose from a compen-
dium of expert research on the elemental theories and technological building blocks underscoring the
Geographic Information Systems discipline.

Chapter 1
Showing the Past: Integrating Cartographic Collections through Virtual Map Rooms........................... 1
Willington Siabato, Technical University of Madrid, Spain
Alberto Fernández-Wyttenbach, Technical University of Madrid, Spain
Miguel-Ángel Bernabé-Poveda, Technical University of Madrid, Spain

Chapter 2
Location-Based Services: A Taxonomy on Theory and Practice........................................................... 17
Henrik Hanke, University of Duisburg-Essen, Germany
Alf Neumann, University of Cologne, Germany

Chapter 3
Location Privacy.................................................................................................................................... 24
Matt Duckham, University of Melbourne, Australia

Chapter 4
Location-Based Performance Tuning in Mobile Sensor Networks....................................................... 30
Vladimir I. Zadorozhny, University of Pittsburgh, USA
Chapter 5
Routing ................................................................................................................................................. 39
Kevin M. Curtin, George Mason University, USA

Chapter 6
Web-Based Geospatial Services: Implementing Interoperability Specifications................................... 47
Iftikhar U. Sikder, Cleveland State University, USA
Aryya Gangopadhyay, University of Maryland - Baltimore County, USA
Nikhil V. Shampur, Cleveland State University, USA

Chapter 7
A Brief Cartography of Smart Cameras: Proactive Surveillance and Control....................................... 66
Fernanda Bruno, Universidade Federal do Rio de Janeiro, Brazil

Chapter 8
GIS, Grid Computing and RFID in Healthcare Information Supply Chain: A Case for Infectious
Disaster Management............................................................................................................................. 81
Yenming J. Chen, National Kaohsiung First University of Science and Technology, Taiwan

Chapter 9
On Modeling and Analysis of Multidimensional Geographic Databases . ........................................... 91
Sandro Bimonte, Laboratoire d’Informatique en Images et Systèmes d’information, France

Chapter 10
Technology and the Multipolar Global Economy: Implications for European Competitiveness........ 108
Steven McGuire, Aberystwyth University, UK

Chapter 11
Coupling GPS and GIS........................................................................................................................ 122
Mahbubur R. Meenar, Temple University, USA
John A. Sorrentino, Temple University, USA
Sharmin Yesmin, Temple University, USA

Chapter 12
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition of
Arctic Lakes, Toolik Lake Region, Alaska.......................................................................................... 130
Prasad A. Pathak, University of North Carolina at Greensboro, USA
Stephen Whalen, University of North Carolina at Chapel Hill, USA

Chapter 13
Mobile Geographic Information Systems . ......................................................................................... 151
Yang Li, University of East London, UK
Allan J. Brimicombe, University of East London, UK
Chapter 14
A Review of Geospatial Information Technology for Natural Disaster Management in Developing
Countries ............................................................................................................................................. 175
Sam Herold, University of Ottawa, Canada
Michael C. Sawada, University of Ottawa, Canada

Chapter 15
Virtual Environments for Geospatial Applications.............................................................................. 216
Magesh Chandramouli, Purdue University, USA
Bo Huang, Chinese University of Hong Kong, China

Section 2
Development and Design Methodologies
This section provides in-depth coverage of conceptual architecture frameworks to provide the reader
with a comprehensive understanding of the emerging developments within the field of Geographic In-
formation Systems. Research fundamentals imperative to the understanding of developmental processes
within Geographic Information Systems are offered. From broad examinations to specific discussions on
methodology, the research found within this section spans the discipline while offering detailed, specific
discussions. From basic designs to abstract development, these chapters serve to expand the reaches
of development and design technologies within the Geographic Information Systems community. This
section includes 15 contributions from researchers throughout the world on the topic of Geographic
Information Systems.

Chapter 16
GIS Technique for Territorial Analysis: Spatial MultiCriteria Decision Analysis............................... 228
Francesco Riccioli, University of Florence, Italy
Toufic El Asmar, University of Florence, Italy

Chapter 17
Cognitively Ergonomic Route Directions............................................................................................ 250
Alexander Klippel, University of Melbourne, Australia
Kai-Florian Richter, Universität Bremen, Germany
Stefan Hansen, Spatial/Information Systems Ltd./LISAsoft, Australia

Chapter 18
Design and Implementation Approaches for Location-Based, Tourism-Related Services.................. 258
George Kakaletris, University of Athens, Greece
Dimitris Varoutas, University of Athens, Greece
Dimitris Katsianis, University of Athens, Greece
Thomas Sphicopoulos, University of Athens, Greece

Chapter 19
Methodical Spatial Database Design with Topological Polygon Structures ...................................... 295
Jean Damascène Mazimpaka, National University of Rwanda, Rwanda
Chapter 20
Methodologies for Augmented Discovery of Geospatial Resources .................................................. 305
Mattia Santoro, National Research Council, Italy
Paolo Mazzetti, National Research Council, Italy
Stefano Nativi, National Research Council, Italy
Cristiano Fugazza, Institute of Environment and Sustainability, Italy
Carlos Granell, Universitat Jaume I de Castellón, Spain
Laura Díaz, Universitat Jaume I de Castellón, Spain

Chapter 21
Semantic Interoperability of Geospatial Services................................................................................ 336
Iftikhar U. Sikder, Cleveland State University, USA
Santosh K. Misra, Cleveland State University, USA

Chapter 22
New Discovery Methodologies in GIS: Improving the Information Retrieval Process...................... 358
Nieves R. Brisaboa, University of A Coruña, Spain
Miguel R. Luaces, University of A Coruña, Spain
Diego Seco, University of A Coruña, Spain

Chapter 23
A Multidimensional Model for Correct Aggregation of Geographic Measures ................................. 377
Sandro Bimonte, Cemagref, UR TSCF, France
Marlène Villanova-Oliver, Laboratoire d’Informatique de Grenoble, France
Jerome Gensel, Laboratoire d’Informatique de Grenoble, France

Chapter 24
Location Privacy in Automotive Telematics . ..................................................................................... 399
Muhammad Usman Iqbal, University of New South Wales, Australia
Samsung Lim, University of New South Wales, Australia

Chapter 25
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams . .............................. 407
Raimund Ubar, Tallinn University of Technology, Estonia
Jaan Raik, Tallinn University of Technology, Estonia
Artur Jutman, Tallinn University of Technology, Estonia
Maksim Jenihhin, Tallinn University of Technology, Estonia

Chapter 26
Semantic Annotation of Geospatial RESTful Services Using External Resources............................. 434
Victor Saquicela, Universidad Politécnica de Madrid, Spain
Luis. M. Vilches-Blázquez, Universidad Politécnica de Madrid, Spain
Oscar Corcho, Universidad Politécnica de Madrid, Spain
Chapter 27
Framework for Graphical User Interfaces of Geospatial Early Warning Systems ............................. 449
Martin Hammitzsch, Helmholtz Centre Potsdam, GFZ German Research Centre for
Geosciences, Germany

Chapter 28
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery ........................................... 465
Victor Pascual Ayats, Spatial Data Infrastructure of Catalonia (IDEC), Spain

Chapter 29
Development and Application of a Spreadsheet-Based Spatial Decision Support
System (SDSS).................................................................................................................................... 480
Dossay Oryspayev, University of Northern Iowa, USA
Ramanathan Sugumaran, University of Northern Iowa, USA
John DeGroote, University of Northern Iowa, USA

Chapter 30
A Package-Based Architecture for Customized GIS .......................................................................... 493
Andrés Pazos, Universitat Jaume I, Spain
José Poveda, University of Texas, USA
Michael Gould, Universitat Jaume I, Spain

Section 3
Tools and Technologies
This section presents extensive coverage of various tools and technologies available in the field of
Geographic Information Systems that practitioners and academics alike can utilize to develop different
techniques. These chapters enlighten readers about fundamental research on the many tools facilitat-
ing the burgeoning field of Geographic Information Systems. It is through these rigorously researched
chapters that the reader is provided with countless examples of the up-and-coming tools and technologies
emerging from the field of Geographic Information Systems. With 18 chapters, this section offers a broad
treatment of some of the many tools and technologies within the Geographic Information Systems field.

Chapter 31
Algorithms for 3D Map Segment Registration.................................................................................... 502
Hao Men, Stevens Institute of Technology, USA
Kishore Pochiraju, Stevens Institute of Technology, USA

Chapter 32
TVAL-Farm: A Qualitative Enhancement of the LESA Model........................................................... 529
Art Rex, Appalachian State University, USA
Leah Greden Mathews, University of North Carolina at Asheville, USA
Anne Lancaster, Independent Researcher, USA
Chapter 33
OpenStreetMap.................................................................................................................................... 540
Kevin Curran, University of Ulster, UK
John Crumlish, University of Ulster, UK
Gavin Fisher, University of Ulster, UK

Chapter 34
GeoCache: A Cache for GML Geographical Data............................................................................... 550
Lionel Savary, PRiSM Laboratory, France
Georges Gardarin, PRiSM Laboratory, France
Karine Zeitouni, PRiSM Laboratory, France

Volume II
Chapter 35
PDE-Based Image Processing: Image Restoration.............................................................................. 569
Rajeev Srivastava, Institute of Technology, Banaras Hindu University (ITBHU), India

Chapter 36
Multicast over Location-Based Services ............................................................................................ 609
Péter Hegedüs, Budapest University of Technology and Economics, Hungary
Mihály Orosz, Budapest University of Technology and Economics, Hungary
Gábor Hosszú, Budapest University of Technology and Economics, Hungary
Ferenc Kovács, Budapest University of Technology and Economics, Hungary

Chapter 37
A New Image Distortion Measure Based on Natural Scene Statistics Modeling................................ 616
Abdelkaher Ait Abdelouahad, University of Mohammed V-Agdal, Morocco
Mohammed El Hassouni, University of Mohammed V-Agdal, Morocco
Hocine Cherifi, University of Burgundy, France
Driss Aboutajdine, University of Mohammed V-Agdal, Morocco

Chapter 38
Modern Navigation Systems and Related Spatial Query . .................................................................. 631
Wei-Shinn Ku, Auburn University, USA
Haojun Wang, University of Southern California, USA
Roger Zimmermann, National University of Singapore, Singapore

Chapter 39
Managing Schema Evolution in a Federated Spatial Database System............................................... 638
Xiaoying Wu, Curtin University, Australia
Jianhong (Cecilia) Xia, Curtin University, Australia
Geoff West, Curtin University, Australia
Lesley Arnold, Landgate, Australia
Bert Veenendaal, Curtin University, Australia
Chapter 40
Map Matching Algorithms for Intelligent Transport Systems............................................................. 658
Mohammed A. Quddus, Loughborough University, UK

Chapter 41
Describing and Selecting Collections of Georeferenced Media Items in Peer-to-Peer Information
Retrieval Systems................................................................................................................................. 666
Daniel Blank, University of Bamberg, Germany
Andreas Henrich, University of Bamberg, Germany

Chapter 42
Cognitive Mapping and GIS for Community-Based Resource Identification .................................... 685
Lyn Kathlene, Colorado State University, USA

Chapter 43
Complementary Part Detection and Reassembly of 3D Fragments ................................................... 703
Vandana Dixit Kaushik, Harcourt Butler Technological Institute, India
Phalguni Gupta, Indian Institute of Technology Kanpur, India

Chapter 44
Biometric Authentication in Broadband Networks for Location-Based Services............................... 726
Stelios C. A. Thomopoulos, National Center of Scientific Research “Demokritos,” Greece
Nikolaos Argyreas, National Center of Scientific Research “Demokritos,” Greece

Chapter 45
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping
Wells.................................................................................................................................................... 736
D. Mimoun, Ecole Nationale Supérieure des Mines, France
S. Gaur, Ecole Nationale Supérieure des Mines, France
D. Graillot, Ecole Nationale Supérieure des Mines, France

Chapter 46
Last Online Deposits Spatial Data in the Web..................................................................................... 750
Mohamed-Khireddine Kholladi, University Mentouri of Constantine, Algeria

Chapter 47
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks
and GIS ............................................................................................................................................... 765
Silke Noack, Beak Consultants GmbH, Germany
Andreas Barth, Beak Consultants GmbH, Germany
Alexey Irkhin, Beak Consultants GmbH, Germany
Evelyn Bennewitz, Beak Consultants GmbH, Germany
Frank Schmidt, Beak Consultants GmbH, Germany
Chapter 48
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings ....................................... 783
Chao-Ching Ho, National Yunlin University of Science and Technology, Taiwan

Section 4
Utilization and Application
This section discusses a variety of applications and opportunities available that can be considered by
practitioners in developing viable and effective Geographic Information Systems programs and pro-
cesses. This section includes 16 chapters that review topics from case studies in Africa to best practices
in Asia and ongoing research in the United States. Further chapters discuss Geographic Information
Systems in a variety of settings (environment, city planning, disease/disaster management, etc.). Con-
tributions included in this section provide excellent coverage of today’s IT community and how research
into Geographic Information Systems is impacting the social fabric of our present-day global village.

Chapter 49
Children’s Maps in GIS: A Tool for Communicating Outdoor Experiences in Urban Planning......... 799
Kerstin Nordin, Swedish University of Agricultural Sciences, Sweden
Ulla Berglund, Swedish University of Agricultural Sciences, Sweden

Chapter 50
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS ........................................ 815
Nitin Kumar Tripathi, Asian Institute of Technology, Thailand
Aung Phey Khant, Asian Institute of Technology, Thailand

Chapter 51
The Spatially Interactive Literature Analysis System Study Tool: A GIS-Based Approach to
Interpreting History in the Classroom.................................................................................................. 832
Alyssa K. Moore, University of Redlands, USA
Lillian I. Larsen, University of Redlands, USA
Diana Stuart Sinton, University of Redlands, USA

Chapter 52
Geographic Visual Query Languages and Ambiguities Treatment...................................................... 848
Arianna D’Ulizia, Consiglio Nazionale delle Ricerche, IRPPS, Italy
Fernando Ferri, Consiglio Nazionale delle Ricerche, IRPPS, Italy
Patrizia Grifoni, Consiglio Nazionale delle Ricerche, IRPPS, Italy

Chapter 53
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window: Theory, Solution and
Application........................................................................................................................................... 857
Sutapa Samanta, Maryland State Highway Administration, USA
Manoj K. Jha, Morgan State University, USA
Chapter 54
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture
Radar (SAR)......................................................................................................................................... 880
Charles Paradzayi, University of Johannesburg, South Africa & Midlands State University,
Zimbabwe
Harold J. Annegarn, University of Johannesburg, South Africa

Chapter 55
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web...................................... 890
Manil Maskey, The University of Alabama in Huntsville, USA
Helen Conover, The University of Alabama in Huntsville, USA
Ken Keiser, The University of Alabama in Huntsville, USA
Luis Bermudez, Southeastern Universities Research Association, USA
Sara Graves, The University of Alabama in Huntsville, USA

Chapter 56
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa:
A Case Study of Nigeria...................................................................................................................... 912
Tunrayo Alabi, International Institute of Tropical Agriculture, Nigeria
Kai Sonder, International Maize and Wheat Improvement Center, Mexico
Olusoji Oduwole, Cocoa Research Institute of Nigeria, Nigeria
Christopher Okafor, International Institute of Tropical Agriculture, Nigeria

Chapter 57
The Columbia Regional Geospatial Service Center System: A Proven Model for the
United States........................................................................................................................................ 926
P. R. Blackwell, Stephen F. Austin State University, USA
Darrel McDonald, Stephen F. Austin State University, USA

Chapter 58
Mapping the Distribution of Tsetse Flies in Eastern Uganda: A Geoinformatics Approach............... 938
Teddy Nakato, Ministry of Agriculture, Uganda
O. O. Jegede, Obafemi Awolowo University, Nigeria
Ayanlade Ayansina, Obafemi Awolowo University, Nigeria
V. F. Olaleye, Obafemi Awolowo University, Nigeria
Bolarin Olufemi, University of Ilorin Nigeria

Chapter 59
Open Source Based Deployment of Environmental Data into Geospatial Information
Infrastructures...................................................................................................................................... 952
José Gil, Institute of New Imaging Technologies, Universitat Jaume I, Spain
Laura Díaz, Institute of New Imaging Technologies, Universitat Jaume I, Spain
Carlos Granell, Institute for Environment and Sustainability, Italy
Joaquín Huerta, Institute of New Imaging Technologies, Universitat Jaume I, Spain
Chapter 60
Application of Topographical Capture Techniques for Modelling Virtual Reality: From the Static
Object to the Human Figure................................................................................................................. 970
Mercedes Farjas Abadía, Universidad Politécnica de Madrid, Spain
Manuel Sillero Quintana, Universidad Politécnica de Madrid, Spain
Pedro Ángel Merino Calvo, Universidad Politécnica de Madrid, Spain

Chapter 61
GIS Implementation in Malaysian Statutory Development Plan System . ......................................... 991
Muhammad Faris Abdullah, International Islamic University Malaysia, Malaysia
Alias Abdullah, International Islamic University Malaysia, Malaysia
Rustam Khairi Zahari, International Islamic University Malaysia, Malaysia

Chapter 62
3D City Modeling and Visualization for Smart Phone Applications ................................................ 1011
Juha Hyyppä, Finnish Geodetic Institute, Finland
Lingli Zhu, Finnish Geodetic Institute, Finland
Zhengjun Liu, Chinese Academy of Surveying and Mapping, China
Harri Kaartinen, Finnish Geodetic Institute, Finland
Anttoni Jaakkola, Finnish Geodetic Institute, Finland

Chapter 63
Assessing the Success of an Industrial Cluster ................................................................................. 1053
Neil Reid, The University of Toledo, USA
Bruce W. Smith, Bowling Green State University, USA

Chapter 64
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after
Landfall ............................................................................................................................................. 1069
Corene J. Matyas, University of Florida, USA

Section 5
Organizational and Social Implications
This section includes a wide range of research pertaining to the social and behavioral impact of Geo-
graphic Information Systems around the world. Chapters introducing this section critically analyze and
discuss trends in Geographic Information Systems, such as historical GIS, volunteered GIS, and health-
care applications of GIS. Additional chapters included in this section look at collaborative mapping,
agriculture, and conservation. Also investigating a concern within the field of Geographic Information
Systems is research which discusses the effect of trust and security on Geographic Information Systems.
With 19 chapters, the discussions presented in this section offer research into the integration of global
Geographic Information Systems as well as implementation of ethical and workflow considerations for
all organizations.
Chapter 65
The Digital Geography Lab at Salem State University: The Evolution of One of the Oldest
Educational Digital Geospatial Labs................................................................................................. 1090
Kym Pappathanasi, Salem State University, USA
Stephen Young, Salem State University, USA

Chapter 66
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS........................................ 1100
Oriol Bernadó, Science and Technology Park for Tourism and Leisure, Spain
Albert Bigorra, Science and Technology Park for Tourism and Leisure, Spain
Yolanda Pérez, University Rovira i Virgili, Spain
Antonio P. Russo, University Rovira i Virgili, Spain
Salvador Anton Clave, University Rovira i Virgili, Spain

Chapter 67
The RiskCity Training Package on Multi-Hazard Risk Assessment................................................. 1120
Cees van Westen, University of Twente, The Netherlands
Tsehaie Woldai, University of Twente, The Netherlands

Volume III
Chapter 68
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style.............. 1132
Filippo Stanco, University of Catania, Italy
Davide Tanasi, Arcadia University, USA
Giuseppe Claudio Guarnera, University of Catania, Italy
Giovanni Gallo, University of Catania, Italy

Chapter 69
Cultural Dasymetric Population Mapping with Historical GIS: A Case Study from the Southern
Appalachians...................................................................................................................................... 1151
George Towers, Concord University, USA

Chapter 70
Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile
Virus in Detroit, Michigan................................................................................................................. 1170
Kevin P. McKnight, Michigan Department of Transportation & Michigan State University, USA
Joseph P. Messina, Michigan State University, USA
Ashton M. Shortridge, Michigan State University, USA
Meghan D. Burns, Montana Natural Heritage Program, USA
Bruce W. Pigozzi, Michigan State University, USA
Chapter 71
GIS Application of Healthcare Data for Advancing Epidemiological Studies ................................. 1183
Joseph M. Woodside, Cleveland State University, USA
Iftikhar U. Sikder, Cleveland State University, USA

Chapter 72
Colorado 14ers, Pixel by Pixel........................................................................................................... 1200
Brandon J. Vogt, University of Colorado at Colorado Springs, USA

Chapter 73
A Component-Based 3D Geographic Simulation Framework and its Integration with a
Legacy GIS........................................................................................................................................ 1216
Zhen-Sheng Guo, Hokkaido University,, Japan
Yuzuru Tanaka, Hokkaido University, Japan

Chapter 74
Collaborative Mapping and GIS: An Alternative Geographic Information Framework................... 1231
Edward Mac Gillavry, Webmapper, The Netherlands

Chapter 75
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal
Systems ............................................................................................................................................. 1243
Anastasios Tsitsanis, National Technical University of Athens, Greece
Sotirios Koussouris, National Technical University of Athens, Greece
Rob Peters, University of Amsterdam, The Netherlands

Chapter 76
Sharing Environmental Data through GEOSS................................................................................... 1260
Gregory Giuliani, University of Geneva and UNEP, Switzerland
Nicolas Ray, University of Geneva and UNEP, Switzerland
Stefan Schwarzer, UNEP, Switzerland
Andrea De Bono, University of Geneva and UNEP, Switzerland
Pascal Peduzzi, UNEP, Switzerland
Hy Dao, University of Geneva and UNEP, Switzerland
Jaap Van Woerden, UNEP, Switzerland
Ron Witt, UNEP, Switzerland
Martin Beniston, University of Geneva, Switzerland
Anthony Lehmann, University of Geneva and UNEP, Switzerland

Chapter 77
Agriculture and Conservation in the Natura 2000 Network: A Sustainable Development
Approach of the European Union...................................................................................................... 1276
Cristian Ioja, University of Bucharest, Romania
Laurenţiu Rozylowicz, University of Bucharest, Romania
Maria Patroescu, University of Bucharest, Romania
Mihai Niţă, University of Bucharest, Romania
Diana Onose, University of Bucharest, Romania
Chapter 78
A Neural Network for Modeling Multicategorical Parcel Use Change . .......................................... 1297
Kang Shou Lu, Towson University, USA
John Morgan, Towson University, USA
Jeffery Allen, Clemson University, USA

Chapter 79
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park . ...................... 1309
Katie Algeo, Western Kentucky University, USA
Ann Epperson, Western Kentucky University, USA
Matthew Brunt, Western Kentucky University, USA

Chapter 80
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region of Ontario,
Canada................................................................................................................................................ 1328
John W. Norder, Michigan State University, USA
Jon W. Carroll, Michigan State University, USA

Chapter 81
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs .............. 1345
Mohamed Amine Abid, University of Manouba, Tunisia
Abdelfettah Belghith, University of Manouba, Tunisia

Chapter 82
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience................ 1368
Viacheslav Adamchuk, McGill University, Canada
Bradley Barker, University of Nebraska-Lincoln, USA
Gwen Nugent, University of Nebraska-Lincoln, USA
Neal Grandgenett, University of Nebraska at Omaha, USA
Megan Patent-Nygren, University of Nebraska-Lincoln, USA
Collin Lutz, Virginia Polytechnic Institute and State University, USA
Kathy Morgan, University of Nebraska-Lincoln, USA

Chapter 83
A Reflection on the Ph.D. Program in Spatially Integrated Social Science at the University of
Toledo ............................................................................................................................................... 1385
Bhuiyan Monwar Alam, The University of Toledo, USA
Jeanette Eckert, The University of Toledo, USA
Peter S. Lindquist, The University of Toledo, USA
 Section 6
Managerial Impact
This section presents contemporary coverage of the social implications of Geographic Information
Systems, more specifically related to the corporate and managerial utilization of information sharing
technologies and applications, and how these technologies can be extrapolated to be used in Geographic
Information Systems. Core ideas such as urban geospatial management, environmental quality monitor-
ing, city competitiveness, and other determinants that affect the intention to adopt technological innova-
tions in Geographic Information Systems are discussed. Equally as crucial, chapters within this section
discuss how leaders can utilize Geographic Information Systems applications to get the best outcomes
from their governors and their citizens.

Chapter 84
Urban Geospatial Management System............................................................................................. 1392
Rudj Maria Todaro, Insula S.p.A., Italy
Ayça Cangel, International Competitiveness Research Institute, Turkey

Chapter 85
Technical and Economical Feasibility of Biomass Use for Power Generation in Sicily . ................ 1411
Antonio Messineo, Kore University of Enna, Italy
Domenico Panno, University of Palermo, Italy
Roberto Volpe, University of Palermo, Italy

Chapter 86
Managing Uncertainty in Geospatial Predictive Models .................................................................. 1421
Iftikhar U. Sikder, Cleveland State University, USA

Chapter 87
Participatory GIS for Integrating Local and Expert Knowledge in Landscape Planning . ............... 1429
Biancamaria Torquati, University of Perugia, Italy
Marco Vizzari, University of Perugia, Italy
Carlo Sportolaro, Agronomist, Perugia, Italy

Chapter 88
Location Management in PCS Networks Using Base Areas (BAs) and 2 Level Paging (2LP)
Schemes............................................................................................................................................. 1448
Hesham A. Ali, Mansoura University, Egypt
Ahmed I. Saleh, Mansoura University, Egypt
Mohammed H. Ali, Mansoura University, Egypt

Chapter 89
The Role of GIS in City Competitiveness......................................................................................... 1476
Khan R. Rahaman, University of Minho, Portugal
Júlia M. Lourenço, University of Minho, Portugal
Chapter 90
Using Geospatial Information Systems for Strategic Planning and Institutional Research............... 1502
Nicolas A. Valcik, The University of Texas at Dallas, USA

Chapter 91
Harnessing Nigeria’s Investment in Satellite Technology for Sustainable Agriculture and Food
Security ............................................................................................................................................. 1515
Zubair A. Opeyemi, National Space Research and Development Agency, Nigeria
J. O. Akinyede, African Regional Centre for Space and Technology Education, Nigeria

Chapter 92
E-Planning through the Wisconsin Land Information Program: The Contexts of Power, Politics
and Scale............................................................................................................................................ 1525
Patrice Day, University of Wisconsin-Milwaukee, USA
Rina Ghose, University of Wisconsin-Milwaukee, USA

Chapter 93
Rapid Evaluation of Arid Lands (REAL): A Methodology............................................................... 1541
Daniel P. Dugas, New Mexico State University, USA
Michael N. DeMers, New Mexico State University, USA
Janet C. Greenlee, New Mexico State University, USA
Walter G. Whitford, New Mexico State University, USA
Anna Klimaszewski-Patterson, New Mexico State University, USA

Chapter 94
Environmental Quality Monitoring, Using GIS as a Tool of Visualization, Management and Decision-
Making: Applications Emerging from the EU Water Framework Directive EU 2000/60................. 1559
Christophoros Christophoridis, Aristotle University of Thessaloniki, Greece
Erasmia Bizani, Aristotle University of Thessaloniki, Greece
Konstantinos Fytianos, Aristotle University of Thessaloniki, Greece

Chapter 95
Disaster Management and Virtual Globes: A High Potential for Developing Countries................... 1587
Gunter Zeug, European Environment Agency, Denmark
Dominik Brunner, Fraunhofer Institute of Optronics, Germany

Chapter 96
Global Emergency-Response System Using GIS.............................................................................. 1604
Salem Al-Marri, Leeds Metropolitan University LEEDS, UK
Muthu Ramachandran, Leeds Metropolitan University, LEEDS UK

Chapter 97
Leveraging the Science of Geographic Information Systems............................................................ 1612
Rick Bunch, University of North Carolina at Greensboro, USA
Anna Tapp, University of North Carolina at Greensboro, USA
Prasad Pathak, University of North Carolina at Greensboro, USA
Chapter 98
Formalizing Cross-Parameter Conditions for Geoprocessing Service Chain Validation ................. 1619
Daniel Fitzner, Fraunhofer IGD, Germany

Section 7
Critical Issues
This section contains 16 chapters, providing a wide variety of perspectives on Geographic Information
Systems and its implications. Such perspectives include readings in ethics, theory, practice, and a great
deal more. The section also discusses new educational possibilities for uses of GIS. Within the chapters,
the reader is presented with an in-depth analysis of the most current and relevant issues within this
growing field of study. Crucial questions are addressed and alternatives offered, such as what makes
up communities, and how geographic information systems and other spatial technologies help define
such groups.

Chapter 99
Qualitative Spatial Reasoning for Applications: New Challenges and the SparQ Toolbox.............. 1639
Diedrich Wolter, Universität Bremen, Germany
Jan Oliver Wallgrün, Universität Bremen, Germany

Chapter 100
Applied Geography Education in Focus: Strategic Panel Session..................................................... 1665
Nairne Cameron, Algoma University, Canada
Edwin Butterworth, CACI International, USA
Dawna L. Cerney, Youngstown State University, USA
William J. Gribb, University of Wyoming, USA
Kingsley E. Haynes, George Mason University, USA
Bill Hodge, City of Midland, USA
Robert B. Honea, KU Transportation Research Institute, USA
Brandon J. Vogt, University of Colorado Colorado Springs, USA

Chapter 101
Temporal and Spatial Consistency . .................................................................................................. 1675
Oliver Duke-Williams, University of Leeds, UK
John Stillwell, University of Leeds, UK

Volume IV
Chapter 102
Validation Approaches to Volcanic Explosive Phenomenology ....................................................... 1697
Sébastien Dartevelle, Los Alamos National Laboratory, USA
Chapter 103
Target Evaluation and Correlation Method (TECM) as an Assessment Approach to Global Earth
Observation System of Systems (GEOSS)........................................................................................ 1724
Samuel Epelbaum, Pace University, USA
Mo Mansouri, Stevens Institute of Technology, USA
Alex Gorod, Stevens Institute of Technology, USA
Brian Sauser, Stevens Institute of Technology, USA
Alexander Fridman, Russian Academy of Science, Russia

Chapter 104
Integrated Ontologies for Spatial Scene Descriptions ...................................................................... 1751
Sotirios Batsakis, Technical University of Crete, Greece
Euripides G.M. Petrakis, Technical University of Crete, Greece

Chapter 105
Embracing Geographic Analysis beyond Geography: Harvard’s Center for Geographic Analysis
Enters 5th Year................................................................................................................................... 1764
Weihe (Wendy) Guan, Harvard University, USA
Peter K. Bol, Harvard University, USA

Chapter 106
Distributed Geospatial Data Management for Entomological and Epidemiological Studies . ......... 1773
Hugo Martins, Universidade Nova de Lisboa, Portugal & Universidade Técnica de Lisboa, Portugal
Jorge G. Rocha, Universidade do Minho, Portugal

Chapter 107
An Assessment of Several Taxonomies of Volunteered Geographic Information............................. 1794
Antony K Cooper, CSIR, South Africa & University of Pretoria, South Africa
Serena Coetzee, University of Pretoria, South Africa
Derrick G Kourie, University of Pretoria, South Africa

Chapter 108
Logistics for the Garbage Collection through the use of Ant Colony Algorithms ........................... 1809
Julio Cesar Ponce Gallegos, Autonomous University of Aguacalientes, Mexico
Fatima Sayuri Quezada Aguilera, Autonomous University of Aguacalientes, Mexico
José Alberto Hernandez Aguilar, Autonomous University of Morelos, Mexico
Christian José Correa Villalón, Institute of Education of Aguascalientes, Mexico

Chapter 109
The Efficacy of Aerial Search during the Battle of Midway............................................................. 1828
Denis J. Dean, University of Texas at Dallas, USA

Chapter 110
Communities, Cartography and GIS: Enhancing Undergraduate Geographic Education with Service
Learning............................................................................................................................................. 1848
Timothy L. Hawthorne, Columbus State University, USA
Chapter 111
Relating Transportation Quality Indicators to Economic Conditions in the South-Central U.S....... 1864
Jonathan C. Comer, Oklahoma State University, USA
Amy K. Graham, Oklahoma State University, USA
Stacey R. Brown, Oklahoma State University, USA

Chapter 112
Monitoring, Analyzing and Understanding the Dynamics of Complex Processes: The Case of the
Public Debate on Pesticides in The Netherlands............................................................................... 1883
J.S. Buurma, Wageningen UR, The Netherlands

Chapter 113
Spatial Model Approach for Deforestation: Case Study in Java Island, Indonesia........................... 1901
Lilik B. Prasetyo, Bogor Agriculture University, Indonesia
Chandra Irawadi Wijaya, Bogor Agriculture University, Indonesia
Yudi Setiawan, Bogor Agriculture University, Indonesia

Chapter 114
Geo-Informatics for Land Use and Biodiversity Studies . ................................................................ 1913
P. K. Joshi, TERI University, India
Neena Priyanka, TERI University, India

Section 8
Emerging Trends
This section highlights research potential within the field of Geographic Information Systems while
exploring uncharted areas of study for the advancement of the discipline. Introducing this section are
chapters that set the stage for future research directions and topical suggestions for continued debate,
centering on the new venues and forums for discussion. A pair of chapters on data mining and urban
planning makes up the middle of the section of the final 16 chapters, and the book concludes with a
look ahead into the future of the Geographic Information Systems field, with “Toward a Roadmap to
E-Government for a Better Governance.” In all, this text will serve as a vital resource to practitioners
and academics interested in the best practices and applications of the burgeoning field of Geographic
Information Systems.

Chapter 115
Neogeography ................................................................................................................................... 1941
Judith Gelernter, Carnegie Mellon University, USA

Chapter 116
Estimating Fractional Snow Cover in Mountain Environments with Fuzzy Classification ............. 1953
Clayton J. Whitesides, Texas State University-San Marcos, USA
Matthew H. Connolly, Texas State University-San Marcos, USA
Chapter 117
Virtual Neighborhoods and E-Government: A Case Study Comparison........................................... 1974
Rebecca Moody, Erasmus University Rotterdam, The Netherlands
Dennis de Kool, Center for Public Innovation, The Netherlands
Victor Bekkers, Erasmus University Rotterdam, The Netherlands

Chapter 118
Fractal Estimation Using Extended Triangularisation and Box Counting Algorithm for any Geo-
Referenced Point Data in GIS............................................................................................................ 1988
R. Sridhar, Sri Ramakrishna Mission Vidyalaya College of Arts and Science, India
S. Balasubramaniam, JSS University, India

Chapter 119
Data Mining Location-Based Social Networks for Geospatial Discovery........................................ 2006
Edward Pultar, University of California – Santa Barbara, USA

Chapter 120
Discovering Geosensor Data By Means of an Event Abstraction Layer........................................... 2020
Alejandro Llaves, University of Muenster, Germany
Thomas Everding, University of Muenster, Germany

Chapter 121
Towards a Rwandan NSDI: An Update............................................................................................. 2040
Felicia O. Akinyemi, Kigali Institute of Science and Technology, Rwanda

Chapter 122
Intelligent Freight Transportation System: The Effects on Transportation Operations
Performance....................................................................................................................................... 2051
Gunnar Stefansson, Chalmers University of Technology, Sweden & University of Iceland, Iceland

Chapter 123
Visualizing Plant Community Change Using Historical Records .................................................... 2063
Evelyn Brister, Rochester Institute of Technology, USA
Elizabeth Hane, Rochester Institute of Technology, USA
Karl Korfmacher, Rochester Institute of Technology, USA

Chapter 124
Sustainable Land Use and Watershed Management in Response to Climate Change Impacts:
Overview and Proposed Research Techniques.................................................................................. 2080
Nguyen Kim Loi, Nong Lam University, Vietnam

Chapter 125
Modeling Species Distribution.......................................................................................................... 2102
Yongyut Trisurat, Kasetsart University, Thailand
Albertus G. Toxopeus, University of Twente, The Netherlands
Chapter 126
A GIS Methodology for Assessing the Safety Hazards of Abandoned Mine Lands (AMLs):
Application to the State of Pennsylvania........................................................................................... 2128
Timothy J. Dolney, Penn State University – Altoona, USA

Chapter 127
Innovative ICT Applications in Transport and Logistics: Some Evidence from Asia....................... 2150
Mark Goh, University of South Australia, Australia & National University of Singapore, Singapore
Kym Fraser, University of South Australia, Australia

Chapter 128
3D Camera Tracking for Mixed Reality using Multi-Sensors Technology....................................... 2164
Fakhreddine Ababsa, University of Evry Val d’Essonne, France
Iman Maissa Zendjebil, University of Evry Val d’Essonne, France
Jean-Yves Didier, University of Evry Val d’Essonne, France

Chapter 129
Automatic Metadata Generation for Geospatial Resource Discovery............................................... 2176
Miguel-Angel Manso-Callejo, Universidad Politécnica de Madrid, Spain
Arturo Beltran Fonollosa, Universitat Jaume I de Castellón, Spain

Chapter 130
Semantic Enrichment for Geospatial Information in a Tourism Recommender System................... 2208
Joan de la Flor, Science and Technology Park for Tourism and Leisure, Spain
Joan Borràs, Science and Technology Park for Tourism and Leisure, Spain
David Isern, Universitat Rovira i Virgili, Spain
Aida Valls, Universitat Rovira i Virgili, Spain
Antonio Moreno, Universitat Rovira i Virgili, Spain
Antonio Russo, Universitat Rovira i Virgili, Spain
Yolanda Pérez, Universitat Rovira i Virgili, Spain
Salvador Anton-Clavé, Universitat Rovira i Virgili, Spain
xxxii

Preface

The constantly changing landscape of Geographic Information Systems makes it challenging for experts
and practitioners to stay informed of the field’s most up-to-date research. That is why Information Sci-
ence Reference is pleased to offer this three-volume reference collection that will empower students,
researchers, and academicians with a strong understanding of critical issues within Geographic Informa-
tion Systems by providing both broad and detailed perspectives on cutting-edge theories and develop-
ments. This reference is designed to act as a single reference source on conceptual, methodological,
technical, and managerial issues, as well as provide insight into emerging trends and future opportunities
within the discipline.
Geographic Information Systems: Concepts, Methodologies, Tools, and Applications is organized
into eight distinct sections that provide comprehensive coverage of important topics. The sections are:
(1) Fundamental Concepts and Theories, (2) Development and Design Methodologies, (3) Tools and
Technologies, (4) Utilization and Application, (5) Organizational and Social Implications, (6) Manage-
rial Impact, (7) Critical Issues, and (8) Emerging Trends. The following paragraphs provide a summary
of what to expect from this invaluable reference tool.
Section 1, Fundamental Concepts and Theories, serves as a foundation for this extensive reference
tool by addressing crucial theories essential to the understanding of Geographic Information Systems.
Introducing the book is “Showing the Past” by Willington Siabato, Alberto Fernández-Wyttenbach, and
Miguel-Ángel Bernabé-Poveda, a great foundation laying the groundwork for the basic concepts and
theories that will be discussed throughout the rest of the book. Another chapter of note in Section 1 is
titled “GIS, Grid Computing and RFID in Healthcare Information Supply Chain” by Yenming J. Chen,
which discusses the growing use of GIS and distributed computing in the health informatics and clinical
technologies industry. Section 1 concludes, and leads into the following portion of the book with a nice
segue chapter, “Virtual Environments for Geospatial Applications,” by Magesh Chandramouli and Bo
Huang. Where Section 1 leaves off with fundamental concepts, Section 2 discusses architectures and
frameworks in place for Geographic Information Systems.
Section 2, Development and Design Methodologies, presents in-depth coverage of the conceptual
design and architecture of Geographic Information Systems, focusing on aspects including territorial
analysis, spatial database design, augmented discovery, semantic interoperability, location privacy, semantic
annotation, and many more topics. Opening the section is “GIS Technique for Territorial Analysis” by
Francesco Riccioli and Toufic El Asmar. This section is vital for developers and practitioners who want to
measure and track the progress of Geographic Information Systems on a local, national, or international
level. Through case studies, this section lays excellent groundwork for later sections that will get into
present and future applications for Geographic Information Systems, including, of note: “Methodolo-
gies for Augmented Discovery of Geospatial Resources” by Mattia Santoro, Paolo Mazzetti, Stefano
 xxxiii

Nativi, Cristiano Fugazza, Carlos Granell, and Laura Díaz, and “User-Friendly Geoportal Interfaces
for Geospatial Resource Discovery” by Victor Pascual Ayats. The section concludes with an excellent
work by Hazel Owen, titled “A Package-Based Architecture for Customized GIS.”
Section 3, Tools and Technologies, presents extensive coverage of the various tools and technologies
used in the implementation of Geographic Information Systems. Section 3 begins where Section 2 left
off, though this section describes more concrete tools at place in the modeling, planning, and applications
of Geographic Information Systems. The first chapter, “Algorithms for 3D Map Segment Registration,”
by Hao Men and Kishore Pochiraju, contains a great example of the types of works that can be found in
this section, a perfect resource for practitioners looking into new technologies within the field. Section
3 is full of excellent chapters like this one, including such titles as “A New Image Distortion Measure
Based on Natural Scene Statistics Modeling,” “Map Matching Algorithms for Intelligent Transport
Systems,” and “Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater
Pumping Wells” to name a few. Where Section 3 described specific tools and technologies at the disposal
of practitioners, Section 4 describes successes, failures, best practices, and different applications of the
tools and frameworks discussed in previous sections.
Section 4, Utilization and Application, describes how the broad range of Geographic Information
Systems efforts has been utilized and offers insight on and important lessons for their applications and
impact. Section 4 includes the widest range of topics because it describes case studies, research, method-
ologies, frameworks, architectures, theory, analysis, and guides for implementation. Topics range from
landscape biodiversity characterization, geographic visual query languages, and estimating biomass, to
open source deployment, topographical capture, and geospatial analysis. The first chapter in the section
is titled “Children’s Maps in GIS,” which was written by Kerstin Nordin and Ulla Berglund. The breadth
of topics covered in the chapter is also reflected in the diversity of its authors, from countries all over
the globe, including Malaysia, Thailand, Sweden, Italy, Nigeria, Uganda, Zimbabwe, South Africa,
the United States and more. Section 4 concludes with an excellent view of a case study in technology
implementation and use, “A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones
after Landfall” by Corene J. Matyas.
Section 5, Organizational and Social Implications, includes chapters discussing the organizational
and social impact of Geographic Information Systems. The section opens with “The Digital Geography
Lab at Salem State University” by Stephen S. Young and Kym Pappathanasi. Where Section 4 focused
on the broad, many applications of Geographic Information Systems technology, Section 5 focuses ex-
clusively on how these technologies affect human lives, either through the way they interact with each
other, or through how they affect behavioral/workplace situations. Other interesting chapters of note
in Section 5 include “Using Volunteered Geographic Information to Assess the Spatial Distribution of
West Nile Virus in Detroit, Michigan” by Kevin P. McKnight, Joseph P. Messina, Ashton M. Shortridge,
Meghan D. Burns, and Bruce W. Pigozzi, and “Cartography and Stability to Enhance and Self Regulate
Proactive Routing in MANETs” by Mohamed Amine Abid and Abdelfettah Belghith. Section 5 concludes
with a fascinating study of a new development in Geographic Information Systems, in “A Reflection
on the Ph.D. Program in Spatially Integrated Social Science at the University of Toledo,” written by
Bhuiyan Monwar Alam, Jeanette Eckert, and Peter S. Lindquist.
Section 6, Managerial Impact, presents focused coverage of Geographic Information Systems as it
relates to effective uses of urban geospatial management, location management, city competitiveness,
e-planning, environmental quality monitoring, disaster management, and much more. This section serves
as a vital resource for developers who want to utilize the latest research to bolster the capabilities and
xxxiv

functionalities of their processes. The section begins with “Urban Geospatial Management System,” a
great look into how city planners and civil engineers can use the latest geospatial technologies to more
effectively practice their work. The 15 chapters in this section offer unmistakable value to managers
looking to implement new strategies that work at larger bureaucratic levels. The section concludes with
“Formalizing Cross-Parameter Conditions for Geoprocessing Service Chain Validation” by Daniel
Fitzner. Where Section 6 leaves off, section seven picks up with a focus on some of the more content-
theoretical material of this compendium.
Section 7, Critical Issues, presents coverage of academic and research perspectives on Geographic
Information Systems tools and applications. The section begins with “Qualitative Spatial Reasoning for
Applications,” by Diedrich Wolter and Jan Oliver Wallgrün. Other issues covered in detail in Section
7 include geography education, taxonomies of volunteered information, community and GIS, complex
processes in GIS, and much more. The section concludes with “Geo-Informatics for Land Use and Biodi-
versity Studies” by P. K. Joshi and Neena Priyanka, a great transitional chapter between Sections 7 and 8
because it examines an important question going into the future of the field. The last chapter manages to
show a theoretical look into future and potential technologies, a topic covered in more detail in Section 8.
Section 8, Emerging Trends, highlights areas for future research within the field of Geographic
Information Systems, opening with “Neogeography” by Judith Gelernter. Section 8 contains chapters
that look at what might happen in the coming years that can extend the already staggering amount of
applications for Geographic Information Systems. Other chapters of note include “Data Mining Location-
Based Social Networks for Geospatial Discovery” and “3D Camera Tracking for Mixed Reality using
Multi-Sensors Technology.” The final chapter of the book looks at an emerging field within Geographic
Information Systems, in the excellent contribution, “Semantic Enrichment for Geospatial Information
in a Tourism Recommender System” by David Isern, Antonio Moreno, Joan de la Flor, Joan Borràs, Aida
Valls, Antonio Russo, Yolanda Pérez, and Salvador Anton-Clavé.
Although the primary organization of the contents in this multi-volume work is based on its eight
sections, offering a progression of coverage of the important concepts, methodologies, technologies,
applications, social issues, and emerging trends, the reader can also identify specific contents by utilizing
the extensive indexing system listed at the end of each volume. Furthermore to ensure that the scholar,
researcher, and educator have access to the entire contents of this multi volume set as well as additional
coverage that could not be included in the print version of this publication, the publisher will provide
unlimited multi-user electronic access to the online aggregated database of this collection for the life
of the edition, free of charge when a library purchases a print copy. This aggregated database provides
far more contents than what can be included in the print version, in addition to continual updates. This
unlimited access, coupled with the continuous updates to the database ensures that the most current
research is accessible to knowledge seekers.
As a comprehensive collection of research on the latest findings related to using technology to
providing various services, Geographic Information Systems: Concepts, Methodologies, Tools, and
Applications, provides researchers, administrators and all audiences with a complete understanding
of the development of applications and concepts in Geographic Information Systems. Given the vast
number of issues concerning usage, failure, success, policies, strategies, and applications of Geographic
Information Systems in countries around the world, Geographic Information Systems: Concepts, Meth-
odologies, Tools, and Applications addresses the demand for a resource that encompasses the most
pertinent research in technologies being employed to globally bolster the knowledge and applications
of Geographic Information Systems.
 Section 1
Fundamental Concepts and
Theories

This section serves as a foundation for this exhaustive reference tool by addressing underlying principles essential
to the understanding of Geographic Information Systems. Chapters found within these pages provide an excellent
framework in which to position Geographic Information Systems within the field of information science and
technology. Insight regarding the critical incorporation of global measures into Geographic Information Systems
is addressed, while crucial stumbling blocks of this field are explored. With 15 chapters comprising this founda-
tional section, the reader can learn and chose from a compendium of expert research on the elemental theories and
technological building blocks underscoring the Geographic Information Systems discipline.
 1

Chapter 1
Showing the Past:
Integrating Cartographic Collections
through Virtual Map Rooms

Willington Siabato
Technical University of Madrid, Spain

Alberto Fernández-Wyttenbach
Technical University of Madrid, Spain

Miguel-Ángel Bernabé-Poveda
Technical University of Madrid, Spain

ABSTRACT
Spatial Data Infrastructures have become a methodological and technological benchmark enabling
distributed access to historical-cartographic archives. However, it is essential to offer enhanced virtual
tools that imitate the current processes and methodologies that are carried out by librarians, historians
and academics in the existing map libraries around the world. These virtual processes must be supported
by a generic framework for managing, querying, and accessing distributed georeferenced resources and
other content types such as scientific data or information. The authors have designed and developed
support tools to provide enriched browsing, measurement and geometrical analysis capabilities, and
dynamical querying methods, based on SDI foundations. The DIGMAP engine and the IBERCARTO
collection enable access to georeferenced historical-cartographical archives. Based on lessons learned
from the CartoVIRTUAL and DynCoopNet projects, a generic service architecture scheme is proposed.
This way, it is possible to achieve the integration of virtual map rooms and SDI technologies bringing
support to researchers within the historical and social domains.

INTRODUCTION accessing and visualizing historical cartographic

At the present time, almost all map libraries on
the Internet are image collections generated by the
digitization of early maps. These types of graphic
files provide researchers with the possibility of
DOI: 10.4018/978-1-4666-2038-4.ch001
information keeping in mind that this information
has a degree of quality that depends upon elements
such as the accuracy of the digitization process
and proprietary constraints (e.g. visualization,
resolution downloading options, copyright, use
constraints). In most cases, access to these map
libraries is useful only as a first approach and it

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

is not possible to use those maps for scientific
work due to the sparse tools available to measure,
match, analyze and/or combine those resources
with different kinds of cartography (Fernández-
Wyttenbach, Siabato, Bernabé, & Wachowicz,
2010). It is necessary to design a set of tools that
offer the advantages of real map libraries into a
virtual environment, offering advanced online
measurement and georeferencing tools, providing
the capability of comparative analysis between
two or more maps and specific tools in order to
facilitate querying and visualizing process.
Two such online sources are The American
Geographical Society Library Digital Map Collec-
tion (2009) and the LUNA Browser (Rumsey, 2010)
of the David Rumsey Map Collection (Rumsey,
2011). The American Geographical Society col-
lection contains over 500 maps ranging from early
maps of Asia to historical maps of Wisconsin and
Milwaukee. The David Rumsey Map Collection,
by contrast, has over 25,000 maps and images
online, this collection focuses mainly on rare
18th and 19th century maps of North and South
America as well as other cartographic materials.
The user interface allows accessing and querying
certain characteristics but they do not provide tools
to analyze the maps, much less integrate them in
a geographical context. Another relevant online
historical content provider is The World Digital
Library (U.S. Library of Congress & UNESCO,
2010). The WDL makes it possible to discover,
study, and enjoy cultural treasures from around
the world on a single website. Its cultural treasures
include but are not limited to maps. Nevertheless,
WDL is once again an interesting data repository
in which the opportunity to provide geographic
tools was not taken.
Applications that are more recent provide the
geographical context but there are not enough
tools for getting the most from the maps. A first
approach was the Google Maps Rumsey Historical
Maps portal of the David Rumsey Historical Maps
collection (Rumsey, 2008). These new interfaces
(Google Maps and Google Earth) allow visualizing
2
Showing the Past
the early maps on Google´s globe. This way, it is
possible to visualize them georeferenced and to
compare them with contemporaneous cartography.
Over one hundred and twenty historical maps
have been selected by David Rumsey from his
collection of more than 150,000 historical maps
to be shown on the Google viewers. This small
sample is a good example of how geographic
environments can be used for sharing early maps.
One step ahead is The Alexandria Digital
Library (2009) geographical browser. This in-
teresting geographic approach provides tools for
discovering and accessing maps. It offers a rich
interface for querying the resources providing up
to four different panels of search parameters. The
main difference between Alexandria’s interface
and the Rumsey approach is that the first one
comprises a single interface for accessing any
map from the collections and it is possible to see
more than one resource at the same time. This
service is also based on Google Maps in order to
provide the geographical context. A comprehen-
sive description of the evolution of Digital Map
Libraries and the most relevant online services,
stressing the challenges in the design of the next
generation of Virtual Map Rooms (VMR), is
available in Evolution of Digital Map Libraries
towards Virtual Map Rooms: new challenges for
the historical research (Fernández-Wyttenbach,
Siabato, Bernabé, & Wachowicz, 2010).
But while the Web mapping/visualization tools
developed by Google and Microsoft provide very
fast, easy-to-access views of images and maps,
they are not suited for complex work and analy-
ses. Nonetheless, the pervasive use of these new
environments offers an opportunity for sharing
data and specifically early maps. Therefore, it is
essential provide users with environments like
these but offering enough tools and elements for
using early maps properly, not just for seeing them.
Thus, setting up robust VMR as a tool for sup-
porting historical projects in which a large num-
ber of national and international researches and
institutions are involved, represents an advance
Showing the Past
in reducing the lack of applications that could
provide historians and documentary experts with
remote access to the existing information from
any map library around the world and also with
a comprehensive set of tools for exploiting and
analyzing these documents. Web-based mapping
has been very important and now we are moving
from simple mapping and geospatial visualization
to full online geoservices with which, as Dan-
germond (2008) asserted, ‘the GeoWeb allows
extensive sharing of maps and data and opens up
access to geographic applications to everyone.
This, together with the growing availability of
georeferenced content and the ability to easily
search, discover, and mash up these service, is
enabling a whole new pattern and architecture for
geographical applications.’ This pattern empha-
sizes open and interoperable services that can be
used to support a broad array of geographically
related applications such as the proposed VMR.
Because of that and following the evolution of
the geographic services on the Web, the VMR
and tools will be available through one single and
usable portal on the Internet.
The proposed VMR facilitates access to
digitized resources (restricted and unrestricted)
and working with maps located in different map
libraries all over the world. It also provides a set of
tools implemented for the geographical treatment
and manipulation of these resources. In addition,
it simplifies the task of the institutions that are in
charge of preserving those archives: optimising
human resources, guarding against misappropria-
tions and avoiding direct manipulation of those
archives being in a fragile conservation state.
This job is at the same time an instance of
how the Geographic Information Technologies
may be useful for the professional and research
work of historians and social scientists in general,
remarkably contributing to the dissemination of the
History of Cartography and Cartographic Heritage
in the most specialized academic environments.
The paper is structured as follows; part two
presents the research context of the project and how
a Virtual Map Room could be useful for historians
and researchers in general. Part three describes the
implementation of the VMR, including concepts
about the data repository, the architecture of the
service, the data server and the implementation of
a Web client. Part four presents an enriched graphic
user interface that contains new tools for interact-
ing with the data repository and maps, including a
querying-timeline component for accessing maps
easily and the MapAnalyst component software
for geometrical analysis. Finally, conclusions are
presented, and further steps to strengthen the plat-
form implementing other services are discussed.
RESEARCH CONTEXT
It is possible to state that Spatial Data Infrastruc-
tures (SDI) are by now a methodological and
technological benchmark for publication of the
cartographic heritage (Fernández-Wyttenbach,
Álvarez, Bernabé, & Borbinha, 2007). The use
of the standards and recommendations defined by
the Open Geospatial Consortium (OGC) enables
access to historical-cartographic information with
a degree of detail only restricted by the quality of
the digitalized documents and associated informa-
tion issued by each supplier institution.
Thus we underline the high potential of map
servers consistent with the OpenGIS standards and
specifications, especially OGC Web Map Service
Interface Standard (WMS), (Open Geospatial
Consortium, 2006a) and its capability to set up a
platform enabling distributed access to different
historical archives. Despite this, it is necessary to
find the way for creating and providing the most
appropriate virtual tools in order to undertake the
usual studies that are being carried out on the old
maps by librarians or researchers, either intrinsi-
cally on the document or in relation to the real
world. This way, metric tools will be available
to find out real distances and surfaces among
other characteristics and also to make possible
analysis that researchers can formulate on the
3

maps (georeferenced documents). In addition to
the conventional GIS tools, new ones have been
developed to facilitate access to data repositories
and manipulation of the results of the queries.
From the beginning, it was observed that the
appropriate scenario for development and integra-
tion of a virtual map room would be within the
context of historical social research projects and
better if these projects would have international
and local repercussion. Such is the case of the
DynCoopNet and CartoVIRTUAL projects in
which the need to rely on distributed access of
the diverse cartographic archives was recognised.
On the one hand, the aim of DynCoopNet
is to provide new theoretical insights about the
dynamics of complex systems (the evolution of
cooperation tying together the self-organizing
commercial networks of the first global age)
through the implementation of GIS as data inte-
gration engine, fed by distributed databases which
store large data sets and information collections,
and by using tools for management and discovery
of information and tools for temporal analysis
and visualization.
On the other hand, the CartoVIRTUAL proj-
ect, from the R&D programme of the Ministry
of Science and Innovation (Spain), intends to
design a methodology for the creation and
implementation of distributed virtual historical
map libraries based on geographic information
and geomatics standards (ISO 19100 standards
series), OGC specifications, and SDI services.
We set up a prototype (Mercator, 2011) having the
advantages of real map libraries with advanced
online measurement and geo-referencing tools
(Jenny, Weber, & Hurni, 2007) integrated into
a virtual environment. In this project, we also
take advantage of digital rights considerations
(GeoDRM Reference Model (Open Geospatial
Consortium, 2006b)) and the MARC-ISO19115
metadata crosswalk developed in order to facilitate
the publication of cartographic heritage through
SDI (Crespo et al., 2010).
4
Showing the Past
Both projects have kicked off with the aim
of creating specialized virtual map libraries. The
results of the DIGMAP (2009) project have been
taken advantage of, providing solutions for general
users and map libraries with the purpose of pro-
moting the cartographic heritage and supporting
research as well as library and documentary in-
formation systems. Beyond the merely distributed
access to all resources and simple visualization,
the projects intend to provide new virtual tools
for analysis and query of documents, favoring
the safekeeping of documents and restricting
physical access. A set of methods will be sought
for digitization, geo-referencing, cataloguing,
publishing, querying and visualization of histori-
cal cartographic documents within the historical
map libraries on the Internet.
From our viewpoint, it is advisable to develop
and set up a VMR based on open standards and
taking into account the mentioned above tools. This
VMR will enable historians and researchers from
CartoVIRTUAL, DynCoopNet, and other projects
to access the huge amount of information which
encompasses the First Global Age, and/or any kind
of data collections that are available to be shared.
Figure 1 shows a snapshot of the proposed Web
client for the VMR.
IMPLEMENTATION OF THE VIRTUAL
MAP ROOM
The VMR has been implemented as a network of
cooperating physical servers providing data and
services. This enables savings in development
time, operating costs, and allows for uniformity
of data supplier and resources. The organizational
structure, operating environment, technical ar-
guments and technological components of this
proposal, including the relationships between its
parts, and the principles and guidelines, entirely
belong to a typical Spatial Data Infrastructure
architecture. These principles are robust enough
and mature for sharing any kind of geographic data.
Showing the Past
Figure 1. Some searching results in the Virtual Map Room (http://www.cartovirtual.es/navegar)
In order to achieve the integration of map li-
braries and their resources and archives, the team
project has set up a virtual space based on Web
servers, map servers, CGI and PHP/MapScript
support, AJAX support, and other technical ele-
ments which provide enough tools and services
to offer a rich, robust and reliable VMR. These
components have been selected as part of the SDI
architecture. The main element is the map server,
through which the publishing of geographic and
spatial data is feasible. This map server has been
recompiled for fulfilling the requirements of re-
searchers, fine-tuning the service, and improving
the performance of OGC specifications.
On the other hand, a specific Web thin client
has also been designed for enabling access to
the DIGMAP search engine and to the historical
cartography repository. This Web client, which is
an online interface, synchronises the search engine
with a geographic browser that allows carrying
out geographic browsing, querying operations, and
analytical processes once the searching process
has been done.
The software platform of the project is based
on Open Source projects in both the server and
client side; most of them using the GNU General
Public License which allows others to use, change,
and share the software and secondary applications
without restrictions and makes sure it remains free
software for all its users. This means the proposed
framework is also under the GNU GPL.
Historical Data Repository
The information used in the test and implementa-
tion processes has been provided by DIGMAP, an
online repository of historical-cartographic docu-
ments. This portal is the outcome of a European
project co-funded by the Community programme
eContentplus and published on November 2008
(DIGMAP, 2008). The aim of DIGMAP was
to turn its portal into an international reference
gateway to old maps and related bibliography. The
project proposed the development of solutions for
digital libraries especially focused on cartography
that promotes the cultural and scientific heritage
5

(Borbinha, Pedrosa, Gil, Martins, Freire, Dobreva,
& Fernández-Wyttenbach, 2007).
This project provides easy access to the
thousands of early maps of national libraries and
collections around the world through the Internet.
This potential is undoubtedly useful for the cre-
ation of a virtual map room that aims to provide
information for historical and social research in-
terests, since it allows access to a huge amount of
cartographic information and related documents,
no matter where it is stored. The main service is a
specialised digital library that recovers metadata
from European national libraries and offers online
map searching and access to the contents. Meta-
data from other sources are also reused as well
as descriptions and references to other relevant
external resources (Pedrosa, Luzio, Manguinhas,
Martins, & Borbinha, 2008).
It is important to stress that DIGMAP is a
virtual digital library, in the sense that it holds
only the metadata that describes the resources, but
not the resources themselves, which remain in the
local libraries or Web sites (this means that the
VMR access the original resources through these
metadata). The resources also can be (1) digital-
born, (2) digitized, or even (3) physical resources
existing only in the shelves of the libraries. When
the resources are digitized maps, it is possible to
index them by their geographic boundaries. This
characteristic allows the geographic browser
to show the map in the right position when the
resource is queryable, that is to say, when the
resource is georeferenced; otherwise, it will not
be possible to load and display the map on the
geographic Web client.
Independent of its technical and technological
contributions, the real added value of the DIGMAP
Portal as data engine for the development of virtual
map rooms is the large number of digital maps that
has previously been indexed and georeferenced. It
is noteworthy that the main resources come from
the IBERCARTO Working Group, a specific as-
sociation in which Spanish and Portuguese map
libraries are working together. It is a consortium
6
Showing the Past
made up of over 40 map libraries from national
libraries, archives, universities, and army and
civil map services.
Figure 2 shows the georeferencing process
in the DIGMAP Geographic Index module. It
should be noted that the geographic referencing
process in DIGMAP is not quite accurate, it just
supplies an approximate location of the maps
based on the maximum and minimum coordinates
of a bounding box. This reduces the accuracy of
measurements and operations on the map. It is,
nonetheless, valid enough as a first approximation.
There are methods that deal with the geometric
and projective characteristics and properties of
early cartographic representations (Livieratos,
2006; Boutoura & Livieratos, 2006) that could
be taken into account in the future in order to
improve this weakness for online resources.
However, if the user needs further accuracy and
to avoid distortion, the VRM have integrated the
MapAnalyst tool (Jenny, Weber, & Hurni, 2007)
to offer transforming options and to analyse
geometrical aspects such as the distortion grid,
vectors of displacement, and scale isolines of the
maps. This tool is quite useful for manipulating
and transforming old maps if necessary.
Although DIGMAP data engine is a great
source of information as described so far, its
graphical user interface is not very efficient. The
geographic browsing tools associated with the
search engine are excessively simple and do not
follow appropriate usability criteria (Fernández-
Wyttenbach, Moya, Álvarez, & Bernabé, 2009).
For this reason, it is necessary to develop a new
user interface that allows researchers to access
this huge collection more easily and to use the
maps in a better way.
So that this interface could be really useful and
assure system interoperability, it has been chosen
to incorporate with the virtual map room the basic
concepts of the SDI reference framework. Thus
the results (maps and documents) are presented
on a web-mapping framework in accordance with
the usual OGC services, providing cartographic
Showing the Past
Figure 2. Geographical indexing and referencing of historical contents in DIGMAP
information using independent layers, assuring
process and platform compatibility, and therefore,
making easier the documentary research for librar-
ians, historians and other researchers. The next
sections describe the server and client processes.
Geographic Data Server
After having identified the characteristics of the
data repository, and knowing the benefits and
disadvantages of geographic servers available
currently on the market (Jansen, 2009; Aime,
2009; Aime & McKenna, 2009), the geographic
data server MapServer has been chosen as the
most appropriate option for management and
publishing of cartographic and historical informa-
tion. This server offers enough quality, reliability
and functionality for sharing data on the Internet.
It supports display and querying of hundreds of
cartographic data sources and formats, render-
ing and showing them using different reference
systems and projections through the on-the-fly
projection capabilities. In addition, it assures high
quality rendering. These characteristics are very
useful for the integration of early maps, WMS
data layers and other cartographic resources. For
instance, transparency and anti-aliasing properties
are quite efficient for showing early cartography
on a reference WMS layer as shown in Figure 3.
MapServer is a general platform for publishing
spatial data and other geographic applications to
the Web. Based on the language ANSI C/C++, it
may be executed, configured and customised in
a large number of operating systems and environ-
ments. This means that the framework developed
on this proposal can run almost on any platform
configuration of any map library. Also, MapServ-
er works with numerous optional external librar-
ies, allowing its extensibility, and supports many
raster, vector and database formats. This charac-
teristic proves essential when accessing, extract-
ing and serving data from different historical data
repositories, such as the ones described above in
the DIGMAP Project.
The flexibility of use that the server offers
through the exchange of parameters using con-
ventional HTTP request methods, GET and POST,
allows it to dynamically configure the appropriate
variables to meet the specific needs of a particular
user at a given instant. It thus solves the drawback
of generating maps using a static configuration file
which does not allow modifications by external
7

Figure 3. Transparency and anti-aliasing examples
users. Due to the development characteristics of
the DIGMAP components in which the message
interchange is handled through this technique,
MapServer is a candidate that covers interaction
with the historical contents of the map libraries
and other future collections that someone wants
to integrate.
Thus from a technical point of view, it is
possible extract the information and visualize
it on the client by means of parameterized URL
query strings. On one hand, the geoinformation
is extracted from the historical documentation
repositories to generate a raster file as a result,
and on the other hand, a text file with additional
information is also obtained for subsequent geo-
referencing process. The generated information
is accessible from a temporary URL that points
to a public directory of the Web server in which
MapServer is executed as a Common Gateway
Interface application (CGI). Thus, the server has
available the image and the information required
for georeferencing.
8
Showing the Past
Once the access to the repository data has been
solved and the platform has been defined, next step
is to visualize the information and to offer users
(librarians, historians, researchers) the necessary
tools (Saracevic, 2000; Marchionini, Plaisant, &
Komlodi, 2003) to be able to interact with the
identified resources.
Service Architecture Scheme
The typical application flow for data services
based on MapServer is diagrammed in Figure 4.
The basic architecture of MapServer applications
contains a Mapfile, geographic data, HTML pages,
MapServer CGI application and the HTTP Server.
The Mapfile is a structured text configuration
file that defines where the data is and where to
output the generated images. It also defines map
layers, including their data source, projections,
and symbology. Through these elements we can
define sources for early cartography and load them.
Showing the Past
Figure 4. Typical application flow of MapServer
These sources define the second element of
the basic architecture: geographic data. The geo-
graphic server supports several data input formats
by default, but in our case we compiled it with
the open source libraries GDAL and OGR in
order to enable more formats and offer higher
quality images. For this service, we are support-
ing, among others, WMS layers, shapefile spatial
data format, and alpha channel values (through
the AGG driver); the last one is very important
because it lets the system overlay maps and im-
prove query tasks. This way, we ensure the ap-
plication can read and load any file from any library
or data source.
HTML pages are the interface between the user
and MapServer. These are fully provided by the
p.mapper framework (described more fully in the
next section), thus it is possible to focus only on
the data and functionality. The MapServer CGI
application and the HTTP Server are the core of
the systems. The CGI is an executable file that
receives requests and returns images, data, and
other elements that are used by the HTTP Server
for serving up the HTML pages when the user
(researchers) clicks on the browser. In this case,
the Apache 2 HTTP Server supports the VMR.
Based on Figure 4, it is quite easy to describe
data flow of the VMR. In brief, MapServer gets
the resources through the DIGMAP collection
by URL requests, and displays them on the Web
client (p.mapper). Figure 5 illustrates the process
graphically.
The Early Cartography Web Client
To interact with the queried resources it is neces-
sary to rely on a stable and functional interface
which suits the purpose of a virtual map room.
Taking into account that currently on the market
there is a large number of robust and high-per-
formance thin-client web-mapping frameworks, it
was decided to build our proposal based on these
clients and not to develop a new one from scratch
since most of them comply with the minimum
historians’ requirements (Crespo-Sanz, 2009)
to interact with early cartography. Among the
desirable selection criteria should be emphasized:
software license, programming language, model
9

Figure 5. VMR flow chart
and structure, ease of customisation, language used
by the API or application interface, OGC services
supported, dependence on the map server, inclu-
sion of metadata components, interoperability and
usability of its tools, and perhaps most relevant, an
Open Source framework for its easy adaptability
to specific needs.
After having analyzed the available options,
it was decided that the visualisation framework
best fitting the specific needs of map libraries is
p.mapper, a thin Web client that offers wide range
of functionality and multiple options for facilitat-
ing the configuration and installation of applica-
tions based on MapServer and PHP/MapScript.
It works following the multi-layer model and the
client-server architecture which suits perfectly for
geospatial data visualisation and its manipulation
through basic browsing and querying tools. As we
said above, it is characterized by using the GNU
General Public Licence which ensures being
able to modify the application and redistribute it
without restrictions. On the other hand, it relies on
a modular programming structure favouring the
development and incorporation of new functional-
ities or modification of the existing ones, according
to requirements. In addition to the characteristics
already mentioned, p.mapper is programmed un-
der the object-oriented programming paradigm,
10
Showing the Past
so useful for increasing functionality without
modifying the original structure.
As a thin client, p.mapper takes advantage of
the latest advances in Web technologies such as
AJAX, enabling transfer of data from client to
server and return of results asynchronously, updat-
ing information for the client without having to
reload the website. This functionality is exploited
to achieve a more natural and intuitive interaction
with early maps. In addition, p.mapper supports
several relational DBMS and contains JavaScript
libraries that facilitate the programmer’s task when
it is necessary to update functionalities or create
new ones. It handles the light format for data
interchange JSON that minimises the bandwidth
expended in client-server communications, with
the consequent increment in response time and
speed.
This way, we have used MapServer and
p.mapper for providing access to historical-
cartographic archives through a graphical Web
interface. The server platform is composed as
follows: Linux Debian Server as Operating Sys-
tem, Apache 2 as HTTP Server, PHP5 support as
scripting language; on the client side, usual Web
mechanism such as JavaScript (language script),
Cascading Style Sheets (layout presentation) and
XML (configuration) were used. As we said above,
communication between server and client has
Showing the Past
been established through AJAX to provide natural
interaction with maps. These tools allow any user
access to data, regardless their choice of computer,
software, browser or other specifications.
The results and performance obtained so far
are reliable and the tests carried out indicate that
this framework has been a good choice, espe-
cially for developing new specific tools for early
cartography and integrating with the DIGMAP
data repository.
Enriched Graphic User Interface
In order to offer historians and researchers an in-
tuitive interface for accessing and querying maps,
we have improved the client-framework develop-
ing new tools and functionalities. The framework
has a flexible layout structure for including new
elements into it through frame windows. Some
frames that remain empty can be used for almost
any task if the right techniques are applied. We
have used these frames for including the DIGMAP
engine and other external applications that provide
users with a single graphic interface for searching,
querying, analyzing, and browsing data.
Due to the fact that the DIGMAP metadata
repository had already been set up, it was neces-
Figure 6. VMR customised functions for early cartography searching and querying
sary to synchronize both the data engine and the
framework. This job was done using session vari-
ables. Through this method, we have achieved our
goal of incorporating metadata and data naturally
so that users can manipulate them easily.
Once the synchronization process has been
done, a new Table of Contents (TOC) is loaded in
the browser. This new TOC shows the footprints
of the available georeferenced sources, it must
be kept in mind that not all the metadata records
contain indexed geospatial sources. Figure 6 shows
a customized TOC.
Once the metadata have been loaded it is pos-
sible to describe the queried map. For this task,
there is a specific tool that loads and shows the
information by clicking on the map. After this, a
table describing maps is showed. This table con-
tains a link to the original resource and a further
description. Transparency properties are also
handled for this table in order not to obscure the
geographic information. Finally, when user needs
to change parameters or make new queries with-
out taking into account the previous ones, it is
possible to reset the client and delete the meta-
data and data previously loaded. A single X
posted in the TOC will delete the current map.
11

Another useful tool is the incorporation of
timeline browsing synchronized with text searches
on the catalogue and the geographic browser. This
tool is based on the SIMILE© Project Timeline
widget (http://www.simile-widgets.org/timeline).
In the same way as we used an empty frame for
including the DIGMAP engine, we have used a
bottom frame available in the p.mapper’s layout.
This tool is useful for knowing the main description
of the map and, even better, to show an overall
view of the maps in a single timeline. Through
this tool, researchers will be able to query just
the maps they really need independent of search-
ing criteria, locating them by date. It works as a
complementary filter for the searching that the
user has previously done.
Finally, as explained in the preceding para-
graphs, the MapAnalyst component has been
included in the VRM. In addition to the above,
it should be mentioned that MapAnalyst is soft-
ware developed by Institute of Cartography of
Swiss Federal Institute of Technology Zurich
for the accurate analysis of early maps. This
high-quality tool provides historians, librarians,
and others users with the possibility of analyzing
and browsing the old maps in a modern context,
understanding distortions and checking historical
Figure 7. The MapAnalyst software component for geometrical analysis
12
Showing the Past
evolution of cities and places. Its main purpose
is to compute distortion grids and other types of
visualizations that illustrate the geometrical ac-
curacy and distortion of old maps. The software
uses pairs of control points on an early map and
on a modern reference map. The control points
are used to construct geometrical analysis (Jenny
& Weber, 2011). Figure 7 shows and example of
a distortion grid analysis.
CONCLUSION AND FUTURE
ADVANCES
Virtual Map Rooms are tools of high interest for
historical and social research projects. They pro-
vide specialised access to distributed cartographic
collections and especially early maps collections.
The success and sustainability of these tools will
be influenced by their design and usability pat-
terns; this shall be carried out according to the
criteria of publishing of geographic information
issued by SDI specifications. They are at the
same time an instance of how the Geographic
Information Technologies may be useful for the
professional and research work of historians and
social scientists in general, contributing to the
Showing the Past
dissemination of the History of Cartography and
Cartographic Heritage in the most specialized
academic environments.
This paper has described two projects in which
a Web-mapping platform has been developed
in order to supply valuable and comprehensive
instruments for both representing and promot-
ing the knowledge of cartographic heritage to a
large number of people. As far as the platform
has been developed, the result is a web-mapping
portal accessible directly from the Internet that
makes historical-cartographic resources available
for everybody. In this sense, some tasks for the
internationalization of the platform have been
done but not enough for providing universal ac-
cess. Some improvements that cover this aspect
will be included in the near future.
As a Web mapping platform, the VMR brings
together the SDI Architecture into a single, easy-
to-use Web Portal that provides access to histori-
cal data repositories. Any data repository can be
included in this portal; any library which provides
access to its digital contents through the DIGMAP
Architecture will be part of the proposed VMR
automatically. This arrangement is the fastest way
to provide access to the cartographic heritage on
the web, leveraging the power of some of the best
open source geospatial platform and the biggest
repository of historic-cartographic data on the
Internet.
Despite the existing proposals, the Virtual
Map Room proposed in this paper is the larg-
est georeferenced historical-cartographic map
collection on the Internet. This is due to the fact
that the application uses the DIGMAP metadata
repository, which provides access to over 2,500
georeferenced historical maps, unlike other col-
lections that provide access to 150 (The David
Rumsey Map Collection) or 300 (The World Digital
Library) georeferenced digital maps. Thus the
VMR provides access up to eight times more than
the other prestigious collections. However, the
quality of the maps must be highlighted. Although
these collections offer fewer maps, they maintain
a high degree of quality. Actually, the georefer-
encing process in these collections is better than
the one done in the DIGMAP engine, especially
on the Rumsey collection. Nonetheless, the VMR
includes a geometrical analysis and correction
tool to improve maps if necessary.
If one compares digital access to georeferenced
early maps, using web-mapping interfaces, one
finds that the tools available on the VMR are quite
useful, intuitive and usable. The server has been
set up for completeness, reliability and availability.
The VMR relies on the SDI architecture founda-
tions. This work has supplied a supported web
mapping platform for larger configurations related
with early cartography. Finally, we have added key
components that simplify and enhance the user
experience and support the work of libraries in
charge of preserving cartographic archives. Some
usability aspects have been taken into account to
develop an intuitive GUI.
This project will incorporate some significant
improvements into the Virtual Map Room in the
near future by introducing new technical and
conceptual contributions that will enhance its us-
ability according to users’ requirements. To that
end, it is necessary to address the following issues:
1. Incorporation of semantic and spatio-
temporal components. These components
play an important role to complement
the research work carried out within the
proposed platform. Their presentation and
implementation in a coordinated fashion
will mean a considerable advance in initial
management of information and will lead
to a substantial change in the search habits
that historians and researchers now use.
2. Quality of the contents. The Virtual Map
Room cannot be responsible for the quality
of the information accessed. It is not uncom-
mon to find out that there are still collections
without georeferenced archives or with very
poor descriptions (metadata). However,
the application is capable of automatically
13

incorporating all the improvements that
may be made to the contents in the future,
as well as new catalogues or collections. At
the same time, in spite of incorporating the
MapAnalyst tool to allow the user to check
quality and reliability of the resources, it
would also be useful to improve the geo-
graphical indexing and referencing system
provided by DIGMAP.
Once viability is ascertained, the CartoVI-
RTUAL project will move forward towards the
creation of an institutional prototype of a National
Virtual Map Library: a geoportal providing access
to historical map libraries, museums and archives.
The success and sustainability of the project will
also depend on adequate training programs for
installation and maintenance of map libraries.
We are currently designing an academic program
focus on Geographic Information Technologies
applied to map libraries. This course will soon be
accessible from the Virtual Map Library, supported
on an eLearning platform.
Finally, we are also working to access and
compare most of the historical and cartographic
information that is scattered and inaccessible in
the Latin American map libraries through a single
Web site. To this end, the Ibero-American Virtual
Map Libraries Network (CartoRED) project has
started. This network has initial support from
eleven geographical institutions representing
seven countries, which are responsible for safe-
guarding the historical and cartographic funds in
Latin America. We expect to complete a 100%
knowledge and technology transfer process in
order to offer the whole possibilities of integra-
tion to the Latin American libraries and related
institutions. The use of open source tools will be a
guarantee for the success of the network, in which
raw seeking low-cost technological solutions and
collaborative development networks.
14
Showing the Past
ACKNOWLEDGMENT
This research work has been partially funded by
CartoVIRTUAL initiative (Ref. CSO2008-03248/
GEOG) from the R&D Programme 2008-2011 of
the Ministry of Science and Innovation of Spain.
REFERENCES
Aime, A. (2009). GeoServer in production. Paper
presented at the Free and Open Source Software
for Geospatial Workshop, Sydney, Australia.
Aime, A., & McKenna, J. (2009). Web mapping
performance shoot-out. Paper presented at the
Free and Open Source Software for Geospatial
Workshop, Sydney, Australia.
Alexandria Digital Library. (2009). Globetrotter
- NGDA interface. Retrieved from http://clients.
alexandria.ucsb.edu/globetrotter/
American Geographical Society Library. (2009).
Digital map collection. Retrieved from http://
www.uwm.edu/Library/digilib/maps/index.html
Arrowsmith, J. R., Glenn, N., Crosby, C. J., &
Cowgill, E. (2008). Current capabilities and com-
munity needs for software tools and educational
resources for use with LiDAR high resolution
topography data. Retrieved from http://opentopo.
sdsc.edu/docs/LIDAR_software_tools_needs_
Aug_2008.pdf
Borbinha, J., Pedrosa, G., Gil, J., Martins, B.,
Freire, N., Dobreva, M., et al. (2007). Digital
libraries and digitised maps: An early overview
of the DIGMAP project. In D. Goh, T. Cao, I.
Sølvberg, & E. Rasmussen (Eds.), Proceedings of
the 10th International Conference on Asian Digital
Libraries: Looking Back 10 Years and Forging
New Frontiers (LNCS 4822, pp. 383-386).
Showing the Past
Boutoura, C., & Livieratos, E. (2006). Some fun-
damentals for the study of the geometry of early
maps by comparative methods. e-Perimetron.
International Web Journal on Sciences and Tech-
nologies Affined to History of Cartography and
Maps, 1(1), 60–70.
Crespo, M., Criado, M., Capdevila, J., Sánchez, A.,
Soteres, C., & Juanatey, M. (2010). El patrimonio
cartográfico en las infraestructuras de datos espa-
ciales. Revista Catalana de Geografia, 15(41), 4.
Crespo Sanz, A. (2009). The atlas of El Escorial,
a forgotten map. Boletín de la Real Sociedad
Geográfica, 2626, 117–142.
Dangermond, J. (2008). GIS and the GeoWeb.
ArcNews, 30, 1–21.
DIGMAP. (2009). Discovering our past world
with digitised maps. Retrieved from http://www.
digmap.eu
Fernández-Wyttenbach, A., Álvarez, M., Bernabé,
M. Á., & Borbinha, J. (2007). Digital map libraries
services in the spatial data infrastructure (SDI)
framework: The DIGMAP project. Paper presented
at the 23rd International Cartographic Conference,
Moscow, Russia.
Fernández-Wyttenbach, A., Moya, J., Álvarez, M.,
& Bernabé, M. Á. (2008). First approaches to the
usability of digital map libraries. e-Perimetron:
International Web journal on Sciences and Tech-
nologies Affined to History of Cartography and
Maps, 3(2), 63-76.
Fernández-Wyttenbach, A., Siabato, W., Bernabé,
M. Á., & Wachowicz, M. (2010). Evolution of
digital map libraries towards virtual map rooms:
New challenges for the historical research. In Kriz,
K., Cartwright, W., & Hurni, L. (Eds.), Mapping
different geographies (Vol. 12, pp. 129–140).
Heidelberg, Germany: Springer-Verlag.
Jansen, M. (2009). Comparing apples and or-
anges: Uncovering the mystery of component
selection in WebGIS projects. Paper presented at
the Free and Open Source Software for Geospatial
Workshop, Sydney, Australia.
Jenny, B., & Weber, A. (2011). MapAnalyst - The
map historian’s tool for the analysis of old maps.
Retrieved from http://mapanalyst.cartography.ch/
Jenny, B., Weber, A., & Hurni, L. (2007). Visual-
izing the planimetric accuracy of historical maps
with MapAnalyst. Cartographica: The Interna-
tional Journal for Geographic Information and
Geovisualization, 42(1), 89–94. doi:10.3138/
carto-v42-1-089
Livieratos, E. (2006). On the study of the geometric
properties of historical cartographic representa-
tions. Cartographica: The International Journal
for Geographic Information and Geovisualization,
41(2), 165–176. doi:10.3138/RM86-3872-8942-
61P4
Marchionini, G., Plaisant, C., & Komlodi, A.
(Eds.). (2003). The people in digital libraries:
Multifaceted approaches to assessing needs and
impact. Cambridge, MA: MIT Press.
Martins, B., Manguinhas, H., Borbinha, J., &
Siabato, W. (2009). A geo-temporal information
extraction service for processing descriptive
metadata in digital libraries. e-Perimetron. Inter-
national Web Journal on Sciences and Technolo-
gies Affined to History of Cartography and Maps,
4(1), 25–37.
Mercator Research Group. (2011). CartoVIR-
TUAL - Prototipo de cartoteca histórica nacional
distribuida. Retrieved from http://www.cartovir-
tual.es/navegar
Open Geospatial Consortium. (2006a). OpenGIS®
web map server implementation specification 1.3.0
(Tech. Rep. No. OGC 06-042). Bergen, Norway:
Open Geospatial Consortium Inc.
15

Open Geospatial Consortium. (2006b). Geospatial
digital rights management reference model - Geo-
DRM RM (Tech. Rep. No. OGC 06-004r3). Bergen,
Norway: Open Geospatial Consortium Inc.
Pedrosa, G., Luzio, J., Manguinhas, H., Martins,
B., & Borbinha, J. (2008). DIGMAP: A digital
library reusing metadata of old maps and enriching
it with geographic information. In B. Christensen-
Dalsgaard, D. Castelli, B. Ammitzbøll Jurik, &
J. Lippincott (Eds.), Proceedings of the 12th Eu-
ropean Conference on Research and Advanced
Technology for Digital Libraries (LNCS 5173,
pp. 434-435).
Rumsey, D. (2008). David Rumsey historical
map collection. Retrieved from http://rumsey.
geogarage.com/
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 4,
edited by Donald Patrick Albert, pp. 93-108, copyright 2011 by IGI Publishing (an imprint of IGI Global).
16
Showing the Past
Rumsey, D. (2010). David Rumsey map collec-
tion - Luna browser. Retrieved from http://www.
davidrumsey.com/view/luna
Rumsey, D. (2011). David Rumsey map collection.
Retrieved from http://www.davidrumsey.com
Saracevic, T. (2000). Digital library evaluation:
Toward evolution of concepts. Library Trends,
49(2), 350–369.
U.S. Library of Congress & UNESCO. (2010).
The world digital library (WDL). Retrieved from
http://www.wdl.org/en/
Wachowicz, M., & Owens, J. B. (2009). Space-
time representations of complex networks: What
is next? GeoFocus, 9(1), 1–8.
 17

Chapter 2
Location-Based Services:
A Taxonomy on Theory and Practice

Henrik Hanke
University of Duisburg-Essen, Germany

Alf Neumann
University of Cologne, Germany

ABSTRACT
The provisioning of Location-Based Services (LBS) follows the chain of determination of a position,
mapping this information onto a natural language-based description of this position and performing the
service itself. The evolution of technologies regarding applications and infrastructure, standards and
contents has brought up various streams that have influenced the development of this chain over the past
years (Zeimpekis et al., 2003). On the one hand, emerging theoretical concepts have been showing the
way for many commercial and non-commercial services. On the other hand, the conceptual evolution
has been accompanied by significant investments of mobile technology companies and service providers
to the further development of practical solutions (Gessler and Jesse, 2001).

INTRODUCTION and consumed by customers. It sheds light on the

A wide field for technological innovation, the
conceptual discussion of LBS has widely remained
a technology issue dominated by the development
of positioning techniques, infrastructure and data
transmission concepts. This chapter re-emphasizes
the term service, including information and func-
tionality, which is offered by LBS applications
DOI: 10.4018/978-1-4666-2038-4.ch002
ubiquitous information management approach as
important foundation for advanced mobile data
services (Acharya et al., 2004).
Furthermore, the chapter provides an overview
of the essential service concepts and relevant
implications, challenges, and opportunities that
can be derived from the application context of
LBS. Finally, a taxonomy on theory and practice
is presented that draws the link from the technol-
ogy to the service.

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

UBIQUITOUS INFORMATION
MANAGEMENT
Along with the sophistication and increasing
performance of communication devices, such as
Personal Digital Assistants (PDAs), mobile phones
as well as wireless communication networks, the
environment and the world increasingly adopts a
mobile character. In this respect, a very important
driver is constituted by a ubiquitous information
management concept, which is free from temporal
and, in general, also from spatial constraints. In
such mobile computing, ubiquitous computing or
pervasive computing environments, mobile data-
bases and the data dissemination infrastructure are
two integral components especially in the context
of LBS. Data dissemination can follow push-based
or pull-based information flows depending on
where location and application data is processed.
This can be done either on the server side or on
the device, i.e., client side (Acharya et al., 2004).
In push-based systems, data is pushed into a
wireless channel and a user tunes in and down-
loads the required data. This approach can process
read-only transactions and may include popular
data like the stock quotes, news, weather infor-
mation, traffic information. On the other side,
in pull-based wireless services, a user induces
the server to process specific transactions and
send the result to the user through a back chan-
nel. These transactions can be simple queries or
update transactions.
The two-tier concept of information manage-
ment to disseminate, process and further store
data can be collectively termed information layer.
LOCATION-ADAPTED SERVICES
LBS provide users of mobile devices personal-
ized services tailored to their current location.
These central information services fall into three
broad categories that also emphasize the added
value for consumers: positioning and location
18
Location-Based Services
identification, contextual and environmental loca-
tion information as well as navigation between
different locations.
There exists a vast body of literature on po-
sitioning technologies reaching from the early
Active Badge indoor locating solutions to the
Global Positioning System (GPS) and the more
recent Wireless Local Area Network (WLAN) and
Bluetooth concepts (King et al., 2006).
The diversity of the underlying technological
basis as well as the opportunities and limitations
among these approaches in design and character-
istic means of data networks correspond to the
increasing need for adapted LBS infrastructures.
These are tailored to the specific requirements
of different types of locations, different modes
of mobility and distance (Zeimpekis et al., 2003).
The precision of location information and the
distance of a mobile device to a Location Service
Provider (LSP) are two factors that play a key role
in this context. They present the overall frame-
work in which the service layer of key features
of applications and infrastructure technology is
embedded.
Physical and Symbolic Location
A LSP can provide two kinds of location informa-
tion: physical and symbolic. Physical positions are
determined by longitude, latitude and altitude. In
contrast, symbolic location encompasses abstract
ideas of where something is: in a certain room, in
a certain town or next to a certain object (Gessler
and Jesse, 2001).
A system providing a physical position can
usually be augmented to provide the correspond-
ing symbolic location information with additional
information, infrastructure, or both. For example,
a PDA equipped with a GPS receiver may access
a separate database that contains the positions
and geometric service regions of other objects to
provide applications with symbolic information
(Brumitt et al., 2000). Linking real-time train
positions to the reservation and ticketing database
Location-Based Services

can help locate a passenger on a train. Applica-
tions can use the physical position to determine an
entire range of symbolic information (Hightower
& Borriello 2001).
The distinction between physical position
and symbolic location is more pronounced with
some technologies than others. GPS is clearly a
physical positioning technology. Bluetooth Indoor
Positioning Service (BIPS) systems are symbolic
location technologies mostly based on proximity
to known objects. However, some systems such
as Cricket can be used in either mode, depending
on their specific configuration (King et al., 2006).
Location Distance and Mobility
for the appropriate context. Remote mobility is
related to movements between different distant
locations.
Following the interdependencies between dif-
ferent types of mobility and the design of LBS, a
compatible concept of distance may be developed
from short to long range with similar implications
on infrastructure requirements, techniques and
precision of positioning systems, context and navi-
gation services. Table 1 provides a taxonomy of
LBS functions within the service and information
layer framework and summarizes the implications
on the key features of LBS data dissemination,
infrastructure and applications.

Tilson et al. (2004) identify three types of mobility: LOCATION-ADAPTED TECHNOLOGY

micro, local and remote. Micro mobility is related
to moveable artifacts in very confined spaces.
Relative location information is more useful than
absolute position. The need for very short-range
positioning favors particular technologies, for
example, Ultra Wideband (UWB). Local mobility,
for example indoors, is related to the mobility of
people or objects at a particular location and the
need for local awareness. This normally implies
stringent positioning requirements. Applications
in this context require location information to be
determined down to a few meters and translated
As mentioned above, LBS exploit knowledge of
the location of a mobile device to provide services
to it and utilize the location information in some
way or the other. Hence, tracking a mobile device
and the delivery of location-adapted services
prerequisites the process of gaining concrete
position information, that can be carried out in
different ways, with different location context-
related technological implications, opportunities
and challenges.

Table 1. LBS functions – a service and information taxonomy

Service layer
Positioning Context Navigation
data available
global positioning, traffic
Push mobile advertisement from read-only
indoor positioning alerts Key features of
Informa- channel
data dissemina-
tion layer emergency response, mobile yellow pages, data processed and
routing, direction tion
Pull proximity informa- location-aware updated
finding
tion service billing on-demand
support of micro,
precision for local location description,
local and remote
and micro mobility contextual alerts
mobility
Key features of applications and
infrastructure technology

19

LBS management can be classified into two
dimensions: collection of position-specific data in
order to realize the service and the implementation
and operation of an application in terms of the
service itself (Weinberg, 2004). Each dimension
itself can be subdivided in relation to the orga-
nization of the infrastructure. On the one hand,
a dimension might be network-based. It would
consist of centralized and aggregated information,
which are dependent on core network features
and resources, such as mobile phone antennas.
On the other hand, distributed network-centric
features can be classified as organizational edge
of the infrastructure.
In order for LBS to add value at all, minimum
precision requirements concerning tracking ex-
ist. While network coverage and hence tracking
and positioning precision of a device is high in
densely populated areas, it is rather low outside
of cities, as the distances between reference cells
grow. While within cities a coverage using tradi-
tional technology up to 100m is possible, it can
reduce down to 10 km in rural areas (Steinfield,
2004). Hence, coverage is negatively correlated
to population density.
Location Tracking and
Location Awareness
Traditionally, a mobile device is traceable by
using the service infrastructure it communicates
with or is tied to. As a handset involved in the
process of offering LBS has no built-in position-
finder or location-determination technology,
the process is referred to as location tracking
(Seydim et al., 2001). The infrastructure is made
up of several antennas, called towers or cells.
The tower transmitting the signal to devices is
referred to as Cell Of Origin (COO). On a basis
of the knowledge of the COO, the Time Of Ar-
rival (TOA) of signals between the handset and
the tower is measured in order to determine the
device’s position. Although this method provides
distinct location information, it lacks precision.
20
Location-Based Services
In order to get more precise location information,
the structure of the cell needs to be revived. A cell
usually consists of three or more antennas, which
raises the opportunity to aim at the difference at
which devices signals reach theses three or more
towers. This process is called Time Difference Of
Arrival (TDOA). As these need no further hand-
set, client-side, technology other than a simple
antenna, supplementary technology is needed in
order to generate high-level tracking precision.
By measuring the time at which signals from the
towers reach the handset, precision levels up to
30 m are possible. Hence, the TDOA functionality
is being reversed. The Enhanced Observed Time
Difference (EOTD) thus needs the handset to be
much more involved in position determination.
Encountering the core mechanism, location
awareness is referred to as the situation in which
the handset does not need any further server-side
technologies beneath the retrieval of location-
initiated data, such as geographic position, in
order to provide LBS (Steinfield, 2004). Hence,
the position-originating and position-finding
technologies are integrated within the handset.
In this respect, the handset gains ability to gener-
ate knowledge of position by itself. Requiring a
highly equipped handset, accuracy of the locating-
process can achieve a maximum level, enabling a
location-range of less than 10 m (Agarwal et al.,
2002). Therefore, if satellite communication can
be ensured, handsets might be traced anywhere,
anytime.
Enhancing GPS capabilities, it is possible to
make a combined use of network technologies and
equipment. By implementing such hybrid systems,
a major weakness of GPS is addressed. Hence, the
provision of tracking-capabilities in aggravating
circumstances (Zeimpekis et al., 2003), such as
erratic satellite communication, can be ensured in
use of assisted GPS (A-GPS). Therefore, A-GPS
can provide and speed up tracking with further
robustness against up-link and down-link data
transmission failures and uncertainty.
Location-Based Services
Location Data Interpretation
Following the collection of adequate and crucial
position information, these data need to be inter-
preted and prepared in the next step according to
the LBS task. This process can again be structured
into core and edge technology based systems
(Weinberg, 2004).
Using network resources for the provisioning
of LBS in terms of prepared data, the handset does
not process any intelligent software by itself. It is
rather relying on server-side application outputs.
Active uplink data transmission provides the server
with the prerequisiting location information. The
server-side application analysis, prepares and
outputs the data in order for the handset to draw
the requested information. This might be accom-
plished by using the simplest handset software,
such as the Wireless Application Protocol (WAP)
browser to which the data is downloaded and
displayed in according to the LBS. In this sense,
many sources of uncertainty and instability can
Figure 1. Table showing LBS technology: A taxonomy on edge and core properties
arise and conflict with the LBS primary goal of ro-
bustness, for example, data communication issues,
server-side processing errors or even downtimes.
Providing faster services in terms of appli-
cation processing, flexibility and mobility can
be ensured by the translocation of the services
operation. Hence, the computing task regarding
the processing of the application itself might be
executed on the client-side. Provided the handset
has appropriate software implemented, it can there-
fore perform the LBS self-sufficiently: once the
location is determined, the application runs locally
with the user, hence being independent from core
network resources. The built-in software processes
the information gained through communication
by itself, for example, placing the user on a map
stored in the memory of the handset.
21

CONCLUSION
While it is technically possible to precisely locate
a device within a few meters or less, the device
needs to meet specific requirements. These might
reduce the proposed usefulness of LBS, as they
possibly conflict with other non-LBS provision-
ing goals, such as small and handy terminals.
Higher precision in localization demands higher
costs in infrastructure as well as in handset setup
(Patterson et al., 2003). Implementing GPS
competency into a handset raises the need for
further built-in hardware-components, as well as
higher conductible battery and processing power,
which is resulting in higher expenses, as well as
distorted devices.
The interaction between the above mentioned
edge-core and collection-operation relation
delimit not only the implementation of a LBS
concerning the underlying infrastructural oppor-
tunities in terms of the alignment of a service, but
also LBS goals such us precision or processing
power (Weinberg, 2004). For example, while it
is possible to use core technologies within a data
collection process and provide a precise location
information, the further generation of a value-
added services might not be: routing information
within a navigation application requires continu-
ous communication in order to compute the correct
route. Continuous communication itself induces
higher network traffic and hence network affliction
and requires stronger battery-power. On the other
hand, in order to use this service with a GPS-based
technology, stronger processing power and other
handset hardware is required; yet, network afflic-
tion would be reduced.
The challenges and opportunities of LBS in
terms of services adapted to the specificities of
different locations and context-related infor-
mation needs of LBS users fall into two broad
categories: mobility and distance requirements,
accuracy and precision. Therefore, it is primarily
22
Location-Based Services
the diversity that stands out regarding the currently
state-of-the-art of LBS devices, applications and
the technology infrastructure at the micro, local
and remote level. The different location context-
related prerequisites make it hard to integrate
the existing variety of systems into one scalable,
capable and cost efficient solution (Hightower
and Borriello, 2001). Figure 1 provides a final
overview on the location- and positioning-related
technology issues within the dichotomy of edge
and core properties and makes clear reference to
the precision challenge.
REFERENCES
Acharya, D., Kumar, V., & Dunham, M. H. (2004).
Infospace: Hybrid and adaptive public data dis-
semination system for ubiquitous computing.
Journal for Wireless Communications and Mobile
Computing.
Agarwal, N., Basch, J., Beckmann, P., Bharti,
P., Bloebaum, S., Casadei, C., Chou, A., Enge,
P., Fong, W., & Hathi, W., Mann, w., Stone, J.,
Tsitsiklis, J., & Van Roy, B. (2002). Algorithms
for GPS operation indoors and downtown. GPS
Solutions, 6, 149–160. doi:10.1007/s10291-002-
0028-0
Brumitt, B., Krumm, J., Meyers, B., & Shafer,
S. (2000). Ubiquitous computing and the role of
geometry. IEEE Personal Communications (Spe-
cial Issue on Smart Spaces and Environments),
7-5, 41-43.
Gessler, S., & Jesse, K. (2001). Advanced location
modeling to enable sophisticated LBS provision-
ing in 3G networks. In M. Beigl, P. Gray, & D.
Salber (Eds.) Proceedings of the Workshop on
Location Modeling for Ubiquitous Computing.
Atlanta, USA.
Location-Based Services
Hightower, J., & Borriello, G. (2001). A survey
and taxonomy of location systems for ubiquitous
computing. In Computer Science and Engineering
(Ed.) Technical Report (document number UW-
CSE 01-08-03). Washington, DC: University of
Washington.
King, T., Haenselmann, T., Kopf, S., & Effelsberg,
W. (2006). Positionierung mit wireless-LAN und
bluetooth. Praxis der Informationsverarbeitung
und Kommunikation, 29(1), 9–17. doi:10.1515/
PIKO.2006.9
Patterson, C. A., Muntz, R. R., & Pancake, C. M.
(2003). Challenges in location-aware computing.
In IEEE CS & IEEE ComSoc.
Seydim, A. Y., Dunham, M. H., & Kumar, V.
(2001). Location dependent query processing.
Proceedings of the 2nd ACM international work-
shop on data engineering for wireless and mobile
access, pp.47-53, Santa Barbara,CA, USA.
Steinfield, C. (2004). The development of loca-
tion based services in mobile commerce. In: B.
Priessl, H. Bouwman, & C. Steinfield (Eds.),
E-Life after the dot.com bust. Berlin, Germany:
Springer Verlag.
Weinberg, G. (2004). Location based services
case study. MIT Center for E-Business. Boston.
Zeimpekis, V., Giaglis, G. M., & Lekakos, G.
(2003). A taxonomy of indoor and outdoor
positioning techniques for mobile location
services. SIGecom Exchange, 3(4), 19–27.
doi:10.1145/844351.844355
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 269-276,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
KEY TERMS AND DEFINITIONS
Bluetooth Indoor Positioning Service
(BIPS): A symbolic location technology based
on proximity to known objects.
Cell of Origin (COO): In a traditional com-
munication environment a COO is referred to as
the cell or tower, a handset device is tied to.
Enhanced Observed Time Difference
(EODT): Reversing the TDOA-principle, EOTD
measures the transferring-time of signals from the
towers to the device.
Location Service Provider (LSP): A provider
of a value-adding service, which is dependent on
location information.
Time of Arrival (TOA): The position of
a device can be determined by measuring the
transferring-time of a signal between the device
and the COO.
Time Difference of Arrival (TDOA): Deter-
mining a more precise position information of a
device by taking advantage of a cells infrastructure
and measuring the transferring time of a device
to three or more antennas.
Ubiquitous Information Management
(UIM): A communication concept, which is
free from temporal and, in general, from spatial
constraints.
Ultra Wideband (UWB): A technology which
enables very short-range positioning information.
23
24

Chapter 3
Location Privacy
Matt Duckham
University of Melbourne, Australia

ABSTRACT
In this chapter, the author raises a number of issues surrounding the ever-growing capabilities of geo-
informatics. Location privacy can be defined as the claim of individuals to determine for themselves
when, how, and to what extent information about their geographic location is communicated to others.
Location privacy has become an especially important issue in geoinformatics because of the emergence
of location-aware computing. This implication of the interaction between these technology and personal
rights is explored in this chapter.

INTRODUCTION • Inferring other personal information:

A person’s geographic location is personal and
sensitive information that needs to be protected.
Unprotected information about where a person
is located can be misused in a number of ways,
including:
• Causing personal harm: Information
about a person’s location may be used to
cause that person harm. For example, a va-
riety of criminal activities, such as stalking
or even physical assault, require that indi-
vidual’s location is known to the criminal.
DOI: 10.4018/978-1-4666-2038-4.ch003
Geographic location fundamentally lim-
its many of the activities a person can be
engaged in. For this reason, location can
be used to infer other personal informa-
tion about an individual. For example, the
information that Anna is regularly located
in a hospital oncology unit might be used
to infer that Anna is suffering from cancer.
An insurance company might then unfairly
use this inference to deny health insurance
to Anna.
Location privacy can be defined as the claim
of individuals to determine for themselves when,
how, and to what extent information about their

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Location Privacy
geographic location is communicated to others.
Location privacy is a special type of information
privacy. Information privacy is usually defined as
the “claim of individuals, groups, or institutions to
determine for themselves when, how, and to what
extent information about them is communicated to
others” (Westin, 1967, p. 7). As Figure 1 shows,
information privacy is just one of several types
of privacy. Privacy in general is regarded as a
fundamental human right, recognized in Article
12 of the UN Universal Declaration of Human
Rights (1948).
PRIVACY IN LOCATION-AWARE
ENVIRONMENTS
Location privacy has become an especially
important issue in geoinformatics because of
the emergence of location-aware computing.
Location-aware computing environments com-
bine high-power mobile computer platforms,
like personal digital assistants (PDAs) or cellular
phones; location-sensing technology, like global
positioning system (GPS); and wireless commu-
nication, like WiFi or Bluetooth. Using location-
aware computing environments it is possible to
track an individual’s location in real-time, and
Figure 1. Classification of types of privacy, after Rotenberg and Laurant (2004)
communicate that information to interested par-
ties anywhere in the work (e.g., via the Internet).
Users of location-aware technology can po-
tentially take advantage of many new and highly
beneficial location-based services. Examples of
increasingly common location-based services
include navigation and emergency response sys-
tems. People using location-based services must
reveal information about their location to a service
provider. One of the first location-based services,
the AT&T “Find Friends” service, enabled friends
and family to share information about their location
to each other via their mobile phones (Strassman
and Collier, 2004).
In some situations individuals may not even
be explicitly aware that their location is being
tracked. For example, not all cell phone users
are aware that cellular phone systems must track
the approximate location of service subscribers
in order to route calls to and from subscribers’
mobile handsets. With respect to location privacy,
three types of location-sensing technique can be
distinguished (Kaasinen, 2003):
1. Client-based positioning systems: use
a person’s mobile device to compute that
person’s location. Client-based position sys-
tems may be used without revealing location
25

information to a third party. However, most
location-based services still require such
information be disclosed. GPS is an example
of a client-based positioning system.
2. Network-based positioning systems: com-
pute a person’s location using the network
infrastructure. Network-based positioning
systems require that a person’s location must
be revealed to the network administrators.
For example, CGI (cell global identity) is a
common network-based positioning tech-
nique used in cell-phone networks
3. Network-assisted positioning systems: use
a combination of client-based and network-
based technology to compute a person’s
location. Network assisted positioning sys-
tems require that some information about
a person’s location must be revealed to the
network administrators, but not all. For ex-
ample, A-GPS (assisted GPS) combines CGI
positioning with GPS to provide increased
positioning speed and efficiency.
As location-aware technology becomes more
commonplace, increasing volumes of information
about individuals’ locations is being generated,
communicated, and stored. Correspondingly, con-
cern for protecting the location privacy of these
individuals is increasing. An example of one area
of concern is the so-called “location-based spam,”
where an unscrupulous business sends a person
unsolicited marketing material for products or
services related to that person’s location (Schilit
et al., 2003).
PROTECTING LOCATION PRIVACY
There are four different approaches to protecting
an individual’s location privacy:
• Regulatory strategies: aim to establish
and enforce rules that govern the fair use
of personal location information. For ex-
26
Location Privacy
ample, most countries have fair informa-
tion practice or data protection laws that
govern the use of all personal informa-
tion, including location information (e.g.,
OECD, 1980).
• Privacy policies: are machine-readable
specifications of acceptable uses of loca-
tion information used to monitor or en-
force location privacy automatically. For
example, PAWS (privacy awareness sys-
tem) provides a technological framework
that enables individuals to automatically
keep track of the storage and usage of per-
sonal location information by third parties
(Langheinrich, 2001).
• Anonymity strategies: dissociate infor-
mation about an individual’s location from
information about that individual’s iden-
tity. For example, the “geonotes” system
allows users to read and write digital notes
attached to particular geographic locations.
Users of the geonotes system can protect
their privacy by using an alias (pseud-
onym) instead of their real name (Espinoza
et al., 2001).
• Obfuscation strategies: deliberately de-
grade the quality of information about an
individual’s location. For example, instead
of specifying a precise coordinate location,
a person might obfuscate his or her location
by only specifying in which street, suburb,
city, or even region he or she is currently
located (Duckham and Kulik, 2005).
Unfortunately, no single privacy protection
strategy can offer a complete solution to location
privacy issues. Regulatory strategies and privacy
policies provide mechanisms to proscribe certain
uses of personal information, but such strategies
cannot prevent inadvertent or malicious disclosure
of this information. Anonymity and obfuscation
provide mechanisms to hide personal informa-
tion. In the case of anonymity, a user’s identity
is hidden. In the case of obfuscation, a user’s
Location Privacy
Figure 2. Summary of location privacy protection strategies
exact location is hidden. However, anonymity and
obfuscation are not suitable for all situations and
may be vulnerable to attacks aimed at invading
an individual’s location privacy. For example,
because our precise location can uniquely identify
us, analysis of an anonymous user’s location over
time may be used to infer that user’s identity.
Consequently, combinations of these different
privacy protection strategies, summarized in
Figure 2, need to be used.
FUTURE OF LOCATION PRIVACY
Society’s attitudes to privacy change over time.
As an example, in 1753 an attempt to establish a
national census was defeated in the British parlia-
ment as “subversive” to British citizens’ privacy
and liberty. Today, censuses are a basic function
of modern governments that few people object to
(Rule, 1973). Changing attitudes have led some
authors to argue for greater transparency instead
of greater privacy (Brin, 1999). Other authors
point to practical difficulties of maintaining
high-levels of privacy in the face of changing
technologies, like location-aware computing.
For example, in 1999 Scott McNealy (CEO of
Sun Microsystems) controversially exclaimed
in response to privacy concerns: “You have zero
privacy anyway, get over it!” (Sprenger, 1999). It
is possible that in the future we may, as a society,
discard the notion of a right to location privacy.
However, such a dramatic change in attitudes
seems unlikely to happen soon. Already, location
privacy concerns have led to a number of high
profile legal cases (e.g., Chicago Tribune, 1999;
Lee, 2004). Whichever way societal attitudes
change in the longer term, it seems certain that
in the near future concerns about location privacy
will continue to increase. Therefore, an essential
component of the successful development of new
applications of location-aware technology is the
continuing development of ways to enable access
to location-based information and services at the
same time as protecting the individual’s right to
location privacy.
REFERENCES
Brin, D. (1999). The Transparent Society. Perseus.
27

Chicago Tribune. (2001). Rental firm uses GPS in
speeding fine. Chicago, IL: Associated Press (p. 9).
Duckham, M., & Kulik, L. (2005). A formal
model of obfuscation and negotiation for loca-
tion privacy. In H. W. Gellersen, R. Want, & A.
Schmidt, (Eds.), Pervasive 2005. Springer: Berlin.
(pp. 152-170).
Espinoza, F., Persson, P., Sandin, A., Nyström, H.,
Cacciatore, E., & Bylund, M. (2001). GeoNotes:
Social and navigational aspects of location-based
information systems. In G. D. Abowd, B. Brumitt,
& S. Shafer (Eds.), Ubicomp 2001: Ubiquitous
Computing. Berlin: Springer. (pp. 2-17).
Kaasinen, E. (2003). User needs for location-
aware mobile services. Personal and Ubiquitous
Computing, 7(1), 70–79. doi:10.1007/s00779-
002-0214-7
Langheinrich, M. (2001). Privacy by design -
principles of privacy-aware ubiquitous systems.
In G. D. Abowd, B. Brumitt, & S. Shafer (Eds.),
Ubicomp 2001: Ubiquitous Computing, Berlin:
Springer. (pp. 273-291).
Lee, J. (2004). Location-Tracking Sparks Privacy
Concerns. Korea Times. http://times.hankooki.
com, Accessed 26 July 2005.
Organization for Economic Co-operation and
Development (OECD). (1980). Guidelines on the
Protection of Privacy and Transborder Flows of
Personal Data. http://www.oecd.org, Accessed
25 July 2005.
Rotenberg, M., & Laurant, C. (2004). Privacy
and human rights 2004: An international survey
of privacy laws and developments. Privacy In-
ternational. http://www.privacyinternational.org/
survey, Accessed 26 July 2005.
Rule, J. B. (1973). Private Lives and Public Sur-
veillance. London: Allen Lane.
28
Location Privacy
Schilit, B. N., Hong, J., & Gruteser, M. (2003).
Wireless location privacy protection. IEEE Com-
puter, 36(12), 135–137.
Sprenger, P. (1999). Sun on privacy: Get over
it. Wired (San Francisco, Calif.), (January): 26.
Strassman, M., & Collier, C. (2004). Case study:
Development of the Find Friend application. In
J. Schiller & A. Voisard (Eds.), Location-based
services, Morgan Kaufmann. (pp. 27-39).
United Nations (1948). Universal Declaration
of Human Rights. United Nations Resolution
217 A (III).
Westin, A. F. (1967). Privacy and Freedom. New
York: Atheneum.
KEY TERMS AND DEFINITIONS
Cell Global Identify (CGI): Unique iden-
tification number assigned to each cell within a
cellular communications network. Also the term
for positioning systems that use CGI.
Global Positioning System (GPS): Position-
ing system based on triangulation of radio signals
from orbiting GPS satellites.
Location-Aware Computing: Computing
environment that combines high-power mobile
computer platform, location-sensing technology,
and wireless communication.
Location-Based Services: Application of
location-aware computing to provide mobile user
with information relevant to his or her current
location.
Location-Based Spam: Unsolicited market-
ing material (spam) related to an individual’s
current or recent location.
Location Privacy: The claim of individuals
to determine for themselves when, how, and to
what extent information about their geographic
location is communicated to others.
Location Privacy

Personal Digital Assistant (PDA): Mobile Wireless Fidelity (WiFi): Wireless digital
handheld computing device for common business network communication system using any of the
and organizational applications. IEEE 802.11 family of standards.

This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 254-259,
copyright 2009 by Information Science Reference (an imprint of IGI Global).

29
30

Chapter 4
Location-Based
Performance Tuning in
Mobile Sensor Networks
Vladimir I. Zadorozhny
University of Pittsburgh, USA

ABSTRACT
The author of this chapter considers the location-based approach for performance tuning that significantly
facilitates the challenge of utilizing Mobile Sensor Networks. The authors introduce mobile nodes that
can be deployed in conjunction with stationary sensor nodes to perform mission critical surveillance
and monitoring tasks. It details the past advances in this field and discusses other approaches to this
challenge.

MOBILE SENSOR NETWORKS Sensor nodes can be attached to mobile devices

Recent advances in wireless communications and
microelectronics have enabled wide deployment
of smart sensor networks. Such networks naturally
apply to a broad range of applications that involve
system monitoring and information tracking (e.g.,
airport security infrastructure, monitoring of
children in metropolitan areas, product transition
in warehouse networks, fine-grained weather/
environmental measurements, etc.). Wireless
sensors are small resource-constrained devices
with wireless communication capability, process-
ing power, and environment sensing equipment.
DOI: 10.4018/978-1-4666-2038-4.ch004
such as mobile robots forming a Mobile Sensor
Network (MSN).
There has been considerable research on
designing mobile platforms to transport wireless
sensors. The Millibot project at Carnegie Mel-
lon University (Bererton et al., 2000) focused on
constructing heterogeneous distributed robots that
combine mobile platforms with different sensor
devices such as cameras, temperature sensors,
movement sensors, and so forth. Robomote (Sibley
et al., 2002), designed in the Robotic Embebbed
Systems Laboratory at the University of Southern
California is a robot platform that functions as a
single mobile node in a mobile sensor network.
Figure 1 shows actual pictures of the millibot and

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Location-Based Performance Tuning in Mobile Sensor Networks
Figure 1. Millibot (left) and robomote (right)
robomote platforms. Another example includes
efforts on using commercial-off-the-shelf com-
ponents to build inexpensive and modular robots
(Bergbreiter and Pister, 2003).
Mobile Sensor Network consists of mobile
platform (e.g., mobile robots) carrying wireless
sensor devices that can be deployed in conjunction
with stationary sensor nodes to acquire and process
data for surveillance and tracking, environmental
monitoring for highly sensitive areas, or execute
search and rescue operations. Resource constraints
of MSNs make it difficult to utilize them for
advanced environmental monitoring that requires
data intensive collaboration between the robots
(e.g., exchange of multimedia data streams)
(Scerri et al., 2003; Scerri, Xu et al., 2004). To
meet the application requirements, the data ex-
change must be performed over a wireless link.
Meanwhile, even high rate wireless networks
(e.g., 802.11 networks) use a best-effort service
that has limitations of data intensive multimedia
applications since it can lead to packet loss, delay
and jitter (Kurose and Ross, 2005). The problem
aggravates in low rate wireless sensor networks,
(e.g., 802.15.4 networks) (Zheng and Lee, 2004).
In this chapter, we consider location-based
approach for performance tuning in MSNs. This
approach assumes that each node in MSN is aware
of its geographic location. Note, that using Global
Positioning System (GPS) is not always possible
in such systems because of severe energy and
location precision constraints. Commonly MSNs
utilize ad-hoc localization methods based on
nodes calculating their coordinates using special
beacon nodes whose positions are known. Further
consideration of this subject is beyond the scope
of this chapter.
PERFORMANCE TUNING IN
MOBILE SENSOR NETWORKS
Applications using MSNs have stringent require-
ments for efficient mechanisms of querying sensor
data and delivering the query result. Minimizing
sensor query response time becomes crucial in mo-
bile sensor networks. At the same time, minimizing
energy consumption per query is equally crucial
for these battery-powered devices. In general, the
time/energy trade-offs involve energy and time
gain/loss associated with specific layouts of the
nodes. Proper positioning (relocation) of mobile
sensors combined with changing the transmission
ranges of sensors have a considerable impact on
the time/energy trade-off. Specifically, both factors
impact the following characteristics of a Mobile
Sensor Network:
• Collision-free concurrency: Packet col-
lisions is one of the major sources of en-
ergy and time waste in MSN. Once any
two or more nodes in the same Collision
Domain (CD) transmit packets at the same
time, a collision occurs, and packets are
31
 Location-Based Performance Tuning in Mobile Sensor Networks
corrupted and discarded. Packet collisions
can be avoided by minimizing the number
of intersecting CDs and by synchronizing
data transmissions among nodes within the
same CD. Both relocation and changing the
transmission range of sensor nodes could
result in changing the number of potential
collision-free concurrent transmissions.
• Filtering factor: Relocation and chang-
ing the transmission range of sensors can
result in changing the number of hops and
intermediate transmission nodes involved
in query execution. This, however, brings
both benefits and penalties. If the filtering
factor of the intermediate node is low (i.e.,
it just retransmits the data) then it can intro-
duce some time and energy loss due to ex-
tra hop. From the other side, the intermedi-
ate node does reduce the data transmission
ranges, which results in energy conserva-
tion. If the intermediate node does a lot of
filtering, the benefits include spending less
energy in order to transmit less data.
A mobile sensor query is characterized by large
data streams among participating nodes with pos-
sible in-node data filtering/aggregation, which can
be described as a tree-like data delivery pattern
(query routing tree). Mobile sensors are moved
into target positions according to the selected query
routing tree and mobile sensor deployment plans
taking into consideration the current topology of
stationary sensor nodes, the applications’ cover-
age requirements, and the collision domains of
the sensor nodes.
Figure 2 elaborates on the concept of colli-
sion domains in a typical wireless network such
as IEEE 802.15.4 (Zheng and Lee, 2004) and
illustrates how collisions are handled in such
a network. Consider two nodes n1 and n2 that
wish to communicate. Assuming that all sensor
nodes use the same frequency band for trans-
mission, two transmissions that overlap will get
corrupted (collide) if the sensor nodes involved
32
in transmission or reception are in the same col-
lision domain CD(n1,n2) defined as the union of
the transmission ranges of n1 and n2. In Figure
2, nodes n1, n2, n3, and n4, n5 and n6 are in the
same collision domain. This implies that when n1
and n2 are communicating, n3, n4, n5 and n6 can-
not participate in any communications. A typical
wireless network handles collisions using carrier
sense multiple-access with collision avoidance
(CSCMA-CA) (Zheng and Lee, 2004). In general,
before starting a transmission, nodes must sense
the channel for a predetermined amount of time
(waiting time). If the channel is busy, the nodes
wait for the predetermined amount of time after the
channel becomes free. In addition, nodes backoff
for a random time to avoid the possibility that two
or more nodes transmit at the same time after the
waiting period. For this entire period, the node
must sense the channel and this consumes energy.
Each packet also needs to be acknowledged by the
receiver since wireless channels are unreliable.
In general, mobile sensors should position
themselves and adjust their transmission power
so as to minimize overlap of CDs in the query
tree. In some cases, however, this general strat-
egy may result in time and/or energy loss. In
order to capture the associated tradeoff, we intro-
duce a concept of collision-free concurrency (cfc)
of a query tree. We say that query tree T1 has
higher cfc than an equivalent query tree T2, if T1
Figure 2. Collision domain of two communicat-
ing sensors
Location-Based Performance Tuning in Mobile Sensor Networks
enables more concurrent transmissions without
risk of collisions than T2. A query tree has cfc=1,
if it allows for all potentially concurrent transmis-
sion pairs to occur. Two elementary transmissions
et1 and et2 are potentially concurrent in a query
tree T if they do not share the same destination
and there is no strict order between et1 and et2.
For example, consider the query tree in Figure 3a
which was generated for some query Q. The
query tree in Figure 3a is associated with cfc=1/3,
since it allows for one transmission pair (n4~n2,
n5~n3) to occur without risk of collisions. Note
that the query tree includes three potentially con-
current transmission pairs: (n4~n2, n5~n3),
(n4~n2, n3~n1), and (n5~n3, n2~n1). Here ni~nj
denotes an elementary one-hop transmission from
a sensor node ni to node nj. On the other hand,
Figure 3b illustrates a query tree with cfc=0 since
none of the three potentially concurrent transmis-
sion pairs can be performed without risk of col-
lisions.
An optimization technique should be applied
to teamwork planning in the MSN to take into
account the time/energy trade-offs. In general, the
trade-offs involve energy and time gain/loss as-
sociated with specific layouts of the wireless
sensors. The mobile robots should position them-
selves and adjust their transmission power so as
to minimize overlap of CDs in the query tree. The
optimizer is responsible for the choice of the best
query routing tree. The mobile robots should move
into target positions according to the selected
Figure 3. Explanation of collision-free concurrency
query tree. Some of the mobile robots can facili-
tate data delivery acting as intermediate nodes
rather than data acquisition nodes. Such mobile
facilitators can introduce extra hops in order to
reduce transmission ranges of the data acquisition
nodes. In addition, the facilitators can also act as
filters aggregating/compressing the input data and
so decreases the amount of data transmitted from
the remote nodes to the root node.
Figure 4 shows a query tree topology with four
previously positioned nodes s0, s1, s2, s3 and three
different positions of a mobile facilitator m. The
facilitators consume extra energy and introduce
some extra processing delay. However, by reduc-
ing the transmission range and data stream sizes,
they are also capable of reducing the overall query
time and energy consumption. To sum up, given a
query, the coverage requirements, and the initial
position of the mobile robots, the query optimizer
shifts through possible mobile robots positions
in order to generate the candidate trees with ac-
ceptable response time and energy consumption.
ALGEBRAIC OPTIMIZATION
In this section we consider an algebraic query
optimization technique based on a Data Transmis-
sion Algebra (DTA) (Zadorozhny et al., 2005) that
allows a query optimizer to generate query routing
trees to maximize collision-free concurrent data
33
 Location-Based Performance Tuning in Mobile Sensor Networks
Figure 4. Re-positioning of mobile facilitator
transmissions taking into account intermediate
hops and filtering factors of facilitators.
The DTA consists of a set of operations that
take transmissions between wireless sensor nodes
as input and produce a schedule of transmissions
as their result. A one-hop elementary transmis-
sion from sensor node ni to node nj is denoted
as ni~nj. Each transmission ni~nj is associated
with a collision domain CD(ni, nj) as defined in
the previous section. A transmission schedule is
either an elementary transmission, or a composi-
tion of elementary transmissions using one of the
operations of the DTA. The basic DTA includes
three operations that combine two transmission
schedules A and B:
1. o(A,B): This is a strict order operation, that
is, A must be executed before B.
2. a(A,B): This is an overlap operation, that is,
A and B can be executed concurrently.
3. c(A,B): This is a non-strict order operation,
that is, either A executes before B, or vice
versa.
For an example of the DTA operations consider
the query tree in Figure 5. It shows some DTA
specifications that reflect basic constraints of
the query tree. For instance, operation o(n4~n2,
n2~n1) specifies that transmission n2~n1 occurs
after n4~n2 is completed. This constraint reflects a
part of the query tree topology. Operation c(n2~n1,
n3~n1) specifies that there is an order between
34
transmissions n2~n1 and n3~n1 since they share
the same destination. However this order is not
strict. Operation a(n4~n2, n5~n3) specifies that
n4~n2 can be executed concurrently with n5~n3,
since neither n3 nor n5 belongs to CD(n4,n2), and
neither n4 nor n2 are in CD(n5,n3).
Each operation of the DTA specification de-
fines a simple transmission schedule that consists
of two elementary transmissions. The DTA intro-
duces a set of transformation rules (Zadorozhny
et al., 2004) that can be used to generate more
complex schedules. Figure 5 shows an example
of a complete schedule that includes all elemen-
tary transmissions of the query tree. Figure 5 also
shows the initial DTA specification reflecting
basic constraints of the query tree. The initial
specification consists of a set of elementary trans-
missions reflecting the tree topology imposed by
the query semantics, as well as order and overlap
operations over the elementary transmissions.
Figure 6 shows query routing trees from Figure
4 with four previously positioned nodes s0, s1,
s2, s3 and three different positions of a mobile
facilitator m. Note how the re-positioning of the
facilitator is reflected in the initial DTA specifica-
tions is1, is2 and is3.
Out of the many possible query routing trees
and transmission schedules the optimizer should
select an option with an acceptable query response
time and overall energy consumption. This is a
multi-objective optimization (MOP) problem
(Miettinen, 1999). In general, the MOP aims at
Location-Based Performance Tuning in Mobile Sensor Networks
Figure 5. Example of DTA specifications
Figure 6. Impact of mobility on DTA specification
minimizing values of several objective functions
f1,…fn under a given set of constraints. To choose
between different vectors of the optimization
objectives, the optimizer utilizes the concept of
Pareto optimality (Miettinen, 1999). Informally,
an objective vector is said to be Pareto optimal
(also called Pareto front) if all other feasible vec-
tors in the objective space have a higher value for
at least one of the objective functions, or else have
the same value for all objectives. Among all Pa-
reto optimal solutions, the optimizer should chose
one using an application-dependent utility func-
tion. The optimizer should evaluate time and
energy gains/losses and make a preference con-
sidering the relative importance of time and en-
ergy in the context of a specific query.
Figure 6 reports on Pareto fronts for a simple
two-hop query tree of 8 nodes with some data
aggregation/filtering at intermediate nodes. For
example, filtering factor 0.2 means that 20% of
data delivered to an intermediate node will be
forwarded to the base station. A major observation
here is an increase of variability in both time and
energy consumption with a decrease of the facili-
tator filtering factor. This means that in general,
the optimizer can benefit from higher filtering
factors (the lower filtering factor reduces more
input data). However, there is a considerable risk
for the optimizer to behave as badly as in the case
of a high filtering factor. The optimizer explores
related time/energy tradeoffs maximizing benefits
and avoiding risks of selecting bad schedules.
The assumption is that the energy source used for
35
 Location-Based Performance Tuning in Mobile Sensor Networks
Figure 7. Actual Pareto fronts explored by optimizer
moving the platform is separate from the sensor
batteries. Considering both energy and time to
move is an interesting issue that should be explored
separately and more work in this area is required.
Note that the algebraic optimization may be
expensive due to its combinatorial nature. The
number of alternative query trees and schedules
grow at least exponentially with the number
of sensor nodes and elementary transmissions
participating in the query. In order to handle the
optimization complexity, the optimizer should use
scalable techniques, such as utilize heuristic-based
pruning that eliminates suboptimal alternatives
and randomized algorithms (Ioannidis and Kang,
1990).
CONCLUSION
Mobile Sensor Networks include a large collec-
tion of collaborating mobile sensor devices. The
mobile nodes can be deployed in conjunction
with stationary sensor nodes to perform mission
critical surveillance and monitoring tasks. Such
applications require efficient collaboration be-
tween mobile sensors, which imply intensive data
exchange over wireless links. Meanwhile, resource
constraints of MSNs make it difficult to utilize
them for advanced data-intensive monitoring
tasks. In this article we considered location-based
approach for performance tuning that significantly
facilitates this challenge.
36
Other approaches can also be adopted to
optimize data exchange over a wireless link in
mobile environments. For example, Particle sys-
tems utilize nondestructive bit wise arbitration
for channel access that considerably reduces the
collisions rates (Decker et al., 2005). Several
distributed time slot scheduling algorithms for
collision-free communications were implemented
as an extension of basic Time Division Multiple
Access (TDMA) protocols. Ammar and Stevens
(1991) proposed a distributed TDMA protocol with
a goal to permit mobile sensors to move and then
reallocate themselves a time slot without involv-
ing the entire network. Ali et al. (2002) proposed
a distributed and adaptive TDMA algorithm for
multi-hop mobile networks. One concern with this
design is that dynamic topology changes may lead
to frequent exchanges of control packets that could
consume bandwidth and energy resources. In gen-
eral, the distributed TDMA scheduling schemes
have considerable control message overhead for
building data delivery schedules.
REFERENCES
Ali, F., Appani, P., Hammond, J., Mehta, V., Non-
eaker, D., & Russell, H. (2002). Distributed and
Adaptive TDMA Algorithms for Multiple-Hop
Mobile Networks, Proceedings of MILCOM,
546-551.
Location-Based Performance Tuning in Mobile Sensor Networks
Ammar, M., & Stevens, D. (1991) A Distributed
TDMA Rescheduling Procedures for Mobile
Packet Radio Networks, Proceedings of IEEE
International Conference on Communications
(ICC), 1609-1613.
Bererton, C., Navarro-Serment, L., Grabowski,
R., Paredis, C., & Khosla, P. (2000). Millibots:
Small distributed robots for surveillance and map-
ping. Proceedings of Government Microcircuit
Applications Conference, 162-167.
Bergbreiter, S., & Pister, K. S. (2003) CotsBots: an
Off-the-Shelf Platform for Distributed Robotics.
Proceedings of in the 2004 IEEE/RSJ International
Conference on Intelligent Robots and Systems,
1632-1637.
Decker, C., Krohn, A., Beigl, M., & Zimmer,
T. (2005). The Particle Computer System. Pro-
ceedings of the ACM/IEEE Fourth International
Conference on Information Processing in Sensor
Networks (IPSN05), 443-448.
Ioannidis, Y., & Kang, Y. (1990). Randomized
algorithms for optimizing large join queries.
Proceedings of ACM SIGMOD, 312-321.
Kurose, J. F., & Ross, K. W. (2005). Computer
Networking: A Top-down Approach. Addison-
Wesley. 712 p.
Miettinen, K. (1999). Nonlinear Multiobjective
Optimization Kluwer Academic Publisher, 298
Scerri, P., Pynadath, D., Johnson, L., Rosenbloom,
P., Si, M., Schurr, N., & Tambe, M. (2003). A
Prototype Infrastructure for Distributed Robot
Agent Person Teams. Proceedings of 2nd Int. Joint
Conference on Autonomous Agents and Multiagent
Systems, 433-440.
Scerri, P., Xu, Y., Liao, E., Lai, J., & Sycara, K.
(2004). Scaling Teamwork to Very Large Teams.
Proceedings of AAMAS, 888-895.
Sibley, G., Rahimi, M., & Sukhatme, G. (2002).
Robomote: A Tiny Mobile Robot Platform for
Large-Scale Ad Hoc Sensor Networks. Proceed-
ings of the Intl. Conference on Robotics and
Automation, 1143-1148.
Zadorozhny, V., Chrysanthis, P., & Krishnamurthy,
P. (2004). A Framework for Extending the Synergy
between MAC Layer and Query Optimization in
Sensor Networks. Proceedings of DMSN VLDB
Workshop, 68-77.
Zadorozhny, V., Sharma, D., Krishnamurthy, P., &
Labrinidis, A. (2005). Tuning query performance
in sensor databases. Proceedings of MDM. 247-
251.
Zheng, J., & Lee, M. (2004). Will IEEE 802.15.4
make ubiquitous networking a reality? a discussion
on a potential low power, low bit rate standard.
IEEE Communications Magazine, 42(6), 140–146.
doi:10.1109/MCOM.2004.1304251
KEY TERMS AND DEFINITIONS
Collision Domain: The union of the trans-
mission ranges of two communicating wireless
sensors.
Collision-Free Concurrency: A quality
measure of a query routing tree in a mobile sen-
sor network reflecting the number of concurrent
transmission pairs that can be performed within
the query routing tree without risking packet
collisions.
Data Transmission Algebra (DTA): An alge-
braic formalism consisting of a set of operations
that take transmissions between wireless sensors
as input and produce a schedule of transmissions
as their result. MSN query optimizer utilizes DTA
to select a query routing tree and transmission
schedules with an acceptable query response time
and overall energy consumption.
Mobile Facilitator: A mobile sensor device
that facilitates data delivery in a mobile sensor
37
 Location-Based Performance Tuning in Mobile Sensor Networks

network acting as an intermediate node rather
than a data acquisition wireless sensor.
Mobile Sensor Device: A mobile platform
that functions as a single mobile node carrying a
wireless sensor in a mobile sensor network.
Mobile Sensor Network (MSN): A collection
of collaborating mobile sensor devices deployed
in conjunction with stationary sensor nodes to
perform surveillance and monitoring tasks.
Packet Collision: A phenomena that occurs
once any two or more nodes in the same collision
domain transmit packets at the same time. When
collision occurs packets are typically corrupted
and discarded.
Query Routing Tree: A tree-like data delivery
pattern generated by a query in mobile sensor
networks.
Wireless Sensor: A small resource-con-
strained device with wireless communication
capability, processing power and environment
sensing equipment.

This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 260-268,
copyright 2009 by Information Science Reference (an imprint of IGI Global).

38
 39

Chapter 5
Routing
Kevin M. Curtin
George Mason University, USA

ABSTRACT
Routing is the act of selecting a course of travel. Routing problems are one of the most prominent and
persistent problems in geoinformatics. This large research area has a strong theoretical foundation
with ties to operations research and management science. There are a wide variety of routing models
to fit many different application areas, including shortest path problems, vehicle routing problems, and
the traveling salesman problem, among many others. There are also a range of optimal and heuristic
solution procedures for solving instances of those problems. Research is ongoing to expand the types of
routing problems that can be solved, and the environments within which they can be applied.

INTRODUCTION tion networks. Routing is the most fundamental

Routing is the act of selecting a course of travel.
This process is undertaken by nearly every active
person every day. The route from home to school
or work is chosen by commuters. The selection
of stops one will make for shopping and other
commercial activities and the paths between those
stops is a routing activity. Package delivery ser-
vices plan routes for their trucks in such a way
that packages are delivered within specified time
windows. School buses are assigned routes that
will pick up and deliver children in an efficient
manner. Less tangible objects such as telephone
calls or data packets are routed across informa-
DOI: 10.4018/978-1-4666-2038-4.ch005
logistical operation for virtually all transportation
and communications applications.
As in the examples above, routing is frequently
seen as a practical effort to accomplish some goal.
Its importance to geoinformatics, however, lies in
the nature of routing as a general problem. Trans-
portation, communications, or utility systems can
all be modeled as networks—connected sets of
edges and vertices—and the properties of networks
can be examined in the context of the mathematical
discipline of graph theory. Routing procedures can
be performed on any network dataset, regardless of
the intended application. This chapter will discuss
the formulation of routing problems including
shortest path problems, and will review in detail
general vehicle routing problems and the travel-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

ing salesman problem. Solution procedures for
routing problems are discussed and future trends
in routing research are outlined.
BACKGROUND
Generally, a routing procedure is based on an
objective—or goal—for the route, and a set of
constraints regarding the route’s properties. By
far the most common objective for routing prob-
lems is to minimize cost. Cost can be measured
in many different ways, but is frequently defined
as some function of distance, time, or difficulty
in traversing the network. Thus the problem of
locating the least cost or shortest path between
two points across a network is the most common
routing problem. It is also a problem for which
there are several extremely efficient algorithms
that can determine the optimal solution. The most
widely cited algorithm that solves the least cost
path problem on directed graphs with non-negative
weights was developed by Edsgar Dijkstra (1959),
and an even more efficient version of this algo-
rithm—the two-tree algorithm—exists (Dantzig,
1960). Alternative algorithms have been presented
that will solve this problem where negative weights
may exist (Bellman, 1958), where all the short-
est paths from each node to every other node are
determined (Dantzig, 1966; Floyd, 1962), and
where not only the shortest path but also the 2nd,
3rd, 4th, or kth shortest path must be found (Evans
& Minieka, 1992).
NETWORK DESIGN PROBLEMS
The shortest path problem is just one of a class
of related routing problems that can be described
as network design problems. Network design
problems require that some combination of the
elements of a network (edges and vertices) be cho-
sen in order to provide a route (or routes) through
the network. This group includes the minimal
40
Routing
spanning tree problem, the Steiner tree problem,
the Traveling Salesman Problem, and the vehicle
routing problem, among many others (Magnanti
& Wong, 1984). The modeling of these problems
frequently takes the form of integer programming
models. Such models define an objective and a set
of constraints. Solution procedures are applied that
require decisions to be made that generate a route
that optimizes the objective while respecting the
constraints. Given the limited space in this forum,
the following sections will focus on the modeling
of two significant routing problems in an effort
to demonstrate the characteristics of the general
class. Vehicle Routing Problems are presented in
order to discuss the range of possible objectives
for routing problems, and the Traveling Salesman
Problem is presented to demonstrate the formula-
tion of the objectives and constraints.
Vehicle Routing Problems
Vehicle Routing Problems (VRPs) are those that
seek to find a route or routes across a network
for the delivery of goods or for the provision of
transport services. From their earliest incarnations
VRPs have been formulated as distance or cost
minimization problems (Clarke & Wright, 1964;
Dantzig & Ramser, 1959). This overwhelming
bias persists to this day. Nine out of ten research
articles regarding route design in the context of
transit routing written between 1967 and 1998 and
reviewed by Chien and Yang (2000) employed
a total cost minimization objective. When the
route is intended as a physical transport route,
the cost objective is nearly always formulated
as a generalized measure of operator costs (List,
1990), user costs (Dubois et al., 1979; Silman et
al., 1974), or both operator and user costs (Ceder,
2001; Chien et al., 2001; Lampkin & Saalmans,
1967; Newell, 1979; Wang & Po, 2001).
The few exceptions include a model that maxi-
mizes consumer surplus (Hasselström, 1981), a
model that seeks to maximize the number of public
transport passengers (van Nes et al., 1988), a model
Routing
that seeks equity among users (Bowerman et al.,
1995), a model that seeks to minimize transfers
while encouraging route directness and demand
coverage (Zhao & Gan, 2003), and a model that
seeks to maximize the service provided to the
population with access to the route (Curtin &
Biba, 2006). VRPs for transport services can
be designed to either determine single optimal
routes, or a system of routes (Ceder & Wilson,
1986; Chakroborty & Dwivedi, 2002; List, 1990;
Silman et al., 1974)
A substantial subset of the literature posits that
routing problems are not captured well by any
single optimization objective, but rather multiple
objectives should be considered (Current & Marsh,
1993). Among the proposed multi-objective
models are those that tradeoff maximal covering
of demand against minimizing cost (Current &
Pirkul, 1994; Current et al., 1984, 1985; Current
& Schilling, 1989), those that seek to both mini-
mize cost and maximize accessibility in terms of
distance traveled (Current et al., 1987; Current &
Schilling, 1994), and those that tradeoff access
with service efficiency (Murray, 2003; Murray
& Wu, 2003).
Regardless of the objective that is deemed
appropriate for a routing application, the prob-
lem will frequently be posited in the form of a
structured mathematical model. In the next section
the Traveling Salesman Problem is presented to
demonstrate how such models are formulated.
The Traveling Salesman Problem
The Traveling Salesman Problem (TSP) is argu-
ably the most prominent problem in combinatorial
optimization. The simple way in which the prob-
lem is defined in combination with its notorious
difficulty has stimulated many efforts to find an
efficient solution procedure. The TSP is a classic
routing problem in which a hypothetical salesman
must find the most efficient sequence of destina-
tions in his territory, stopping only once at each,
and ending up at the initial starting location. The
TSP has its origins in the Knight’s Tour problem
first formally identified by L. Euler and A. T.
Vandermonde in the mid-1700s. In the 1800s, the
problem was identified as an element of graph
theory and was studied by the Irish mathemati-
cian, Sir William Rowan Hamilton. The problem
was named the Hamiltonian cycle problem in his
honor (Hoffman A. J. & Wolfe P., 1985).
The first known mention of the TSP under that
name appeared in a German manual published
in 1832, and this was followed by four applied
appearances of the problem in the late 1800s and
early 20th century (Cook, 2001). The mathemati-
cian and economist Karl Menger publicized the
TSP in the 1920s in Vienna (Applegate D., 1998),
then introduced it in the United States at Harvard
University as a visiting lecturer, where the prob-
lem was discussed with Hassler Whitney who at
that time was doing his Ph.D. research in graph
theory. In 1932, the problem was introduced at
Princeton University by Whitney, where A. W.
Tucker and Merrill Flood discussed the problem
in the context of Flood’s school-bus routing study
in New Jersey (Schrijver, 2004). Flood went on to
popularize the TSP at the RAND Corporation in
Santa Monica, California in late 1940s. In 1956
Flood mentioned a number of connections of the
TSP with the Hamiltonian paths and cycles in
graphs (Flood, 1956). Since that time the TSP
has been considered one of the classic models in
combinatorial optimization, and is used as a test
case for virtually all advancements in solution
procedures.
There are many mathematical formulations for
the TSP, with a variety of constraints that enforce
the requirements of the problem. Since this is not
the appropriate forum for reviewing all of the po-
tential formulations, the formulation attributed to
Vajda (Vajda, 1961) has been chosen in order to
demonstrate how such a formulation is specified.
The following notation is used:
41
 Routing

n = the number of cities to be visited;
i and j = indices of cities that can take integer
values from 1 to n
t = the time period, or step in the route between
the cities
xijt = 1 if the edge of the network from i to j is
used in step t of the route, and 0 otherwise
dij = the distance or cost from city i to city j
The objective function is to minimize the sum
of all costs (distances) of all of the selected ele-
ments of the tour:
n n n
In addition to the above constraints the deci-
sion variables are constrained to be integer values
in the range of 0 to 1: 0 ≤ x ijt ≤ 1.
Like any routing problem structured as an inte-
ger program, in order to solve the TSP a procedure
must be employed that allows decisions to be made
regarding the values of the decision variables. The
choice of a solution procedure depends in part on
the difficulty of the routing problem and the size of
the problem instance being solved. The following
section describes the combinatorial complexity
of routing problems and the solution procedures
that can be used to solve them.

∑ ∑ ∑d x ij ijt

i =1 j =i t =1
SOLUTION PROCEDURES
The tour is subject to the following constraints: FOR ROUTING PROBLEMS
For all values of t, some arc must be traversed,
hence The TSP and most VRPs are considered to be in a
class of problems that are highly combinatorially
complex. There are, for example, (n – 1)! possible
∑∑x ijt
= 1 for all t.
tours for the TSP. Therefore, as the number of
i j
cities to visit, n, grows, the number of possible
tours grows very rapidly. So rapidly, in fact, that
For all cities, there is just one other city which
even small instances of these problems cannot
is being reached from it, at some time, hence
be solved by enumeration (the inspection of all
possible combinations).
∑∑x ijt
= 1 for all i. If this is the case these integer programming
j t
problems may be solved optimally using a ver-
sion of the simplex method to generate fractional
For all cities, there is some other city from optimal solutions from the linear programming
which it is being reached, at some time, hence relaxation of the integer program, followed by
a branch and bound search procedure to identify
∑∑x ijt
= 1 for all j. integer optimal solutions. A variety of reformula-
i t
tion techniques, preprocessing routines, and con-
straint elimination methods can assist in solving
When a city is reached at time t, it must be these problems.
left at time t + 1, in order to exclude disconnected However, these problems grow so quickly
subtours that would otherwise meet all of the above that it may be impractical or impossible to obtain
constraints. These subtour elimination constraints optimal solutions for large problem instances
are formulated as: even with specialized solution procedures. When
this is the case, alternate—though not guaranteed
∑x ijt
= ∑x jkt +1
for all j and t. optimal—solution procedures can be employed.
i k

42
Routing
These include formal heuristic (or approximate)
methods to quickly find good routes (Bowerman et
al., 1995; Chien & Yang, 2000; Fan & Machemehl,
2004; van Nes et al., 1988; Zhao & Gan, 2003),
heuristics based on genetic algorithms (Chien et
al., 2001; Tom & Mohan, 2003) or other proce-
dures with a stochastic element, heuristics that
incorporate expert user input in the process (Baaj
& Mahmassani, 1995; Ceder & Wilson, 1986;
Lampkin & Saalmans, 1967; Shih et al., 1998),
and heuristics that are entirely based on expert user
experience. The last of these is the most widely
used and perhaps the most important technique and
is sometimes referred to as manual route planning
(Dubois et al., 1979; Moorthy, 1997). The manual
method can result in routes that are extremely
near optimal, and the ability of such experts to
quickly react to customer demands and complaints
is a valuable asset (Newell, 1979). However, the
solutions determined heuristically in this way are
likely to be suboptimal and without an optimal
solution process there is no way to determine
how far from optimal the manual solutions are.
Lastly, there has recently been increased interest
in the use of simulation methods, such as agent
based modeling and cellular automaton models
to generate optimal routes (Jiang et al., 2003).
FUTURE TRENDS IN
ROUTING RESEARCH
Since routing is such a fundamental activity, and
the problems can be so difficult to solve, there is
an ongoing need to improve the modeling of the
problems themselves and the procedures used to
solve them. In addition to these long term goals,
routing researchers are finding ways to take advan-
tage of the flood of data regarding networks and
locations of objects and activities in order to use
these data to identify optimal routes. Of particular
interest is the modeling of “best” routes for trip
planning that do not necessarily minimize cost, but
rather maximize some function of benefit for the
traveler. More frequently, routing procedures are
being conducted while conditions are constantly
subject to change (such as traffic conditions). The
issue of routing in these dynamic environments
is a growing research area.
CONCLUSION
Routing problems are one of the most promi-
nent and persistent problems in geoinformatics.
This large research area has a strong theoretical
foundation with ties to operations research and
management science. There are a wide variety of
routing models to fit many different application
areas, and there are a range of optimal and heuristic
solution procedures for solving instances of those
problems. Research is ongoing to expand the types
of routing problems that can be solved, and the
environments within which they can be applied.
REFERENCES
Applegate D., B. R., Chvtal V., & Cook W.. (1998).
On the Solution of Traveling Salesman Problems.
Documenta Mathematica, 645-656.
Baaj, M. H., & Mahmassani, H. S. (1995). Hybrid
route generation heuristic algorithm for the design
of transit networks. Transportation Research
Part C, Emerging Technologies, 3(1), 31–50.
doi:10.1016/0968-090X(94)00011-S
Bellman, R. (1958). On a routing problem. Quar-
terly of Applied Mathematics, 16(1), 87–90.
Bowerman, R., Hall, B., & Calamai, P. (1995). A
multi-objective optimization approach to urban
school bus routing: formulation and solution
method. Transportation Research Part A, Policy
and Practice, 29(2), 107–123. doi:10.1016/0965-
8564(94)E0006-U
43
 Routing

Ceder, A. (2001). Operational objective functions Current, J. R., ReVelle, C. S., & Cohon, J. L.
in designing public transport routes. Journal of (1987). The median shortest path problem: a
Advanced Transportation, 35(2), 125–144. multiobjective approach to analyze cost vs.
accessibility in the design of transportation net-
Ceder, A., & Wilson, N. (1986). Bus network
works. Transportation Science, 21(3), 188–197.
design. Transportation Research Part B: Meth-
doi:10.1287/trsc.21.3.188
odological, 20B(4), 331–344. doi:10.1016/0191-
2615(86)90047-0 Current, J. R., & Marsh, M. (1993). Multiobjec-
tive transportation network design and routing
Chakroborty, P., & Dwivedi, T. (2002). Optimal
problems: Taxonomy and annotation. Euro-
route network design for transit systems using
pean Journal of Operational Research, 65, 4–19.
genetic algorithms. Engineering Optimization,
doi:10.1016/0377-2217(93)90140-I
34(1), 83–100. doi:10.1080/03052150210909
Current, J. R., & Pirkul, H. (1994). Efficient
Chien, S., & Yang, Z. W. (2000). Optimal feeder
algorithms for solving the shortest covering path
bus routes on irregular street networks. Journal
problem. Transportation Science, 28(4), 317–327.
of Advanced Transportation, 34(2), 213–248.
doi:10.1287/trsc.28.4.317
Chien, S., Yang, Z. W., & Hou, E. (2001). Genetic
Current, J. R., & Schilling, D. A. (1989). The
algorithm approach for transit route planning and
covering salesman problem. Transportation Sci-
design. Journal of Transportation Engineering,
ence, 23(3), 208–213. doi:10.1287/trsc.23.3.208
127(3), 200–207. doi:10.1061/(ASCE)0733-
947X(2001)127:3(200) Current, J. R., & Schilling, D. A. (1994). The
median tour and maximal covering tour prob-
Clarke, G., & Wright, J. W. (1964). Scheduling
lems: Formulations and heuristics. European
of vehicles from a central depot to a number of
Journal of Operational Research, 73, 114–126.
delivery points. Operations Research, 12(4),
doi:10.1016/0377-2217(94)90149-X
568–581. doi:10.1287/opre.12.4.568
Curtin, K. M., & Biba, S. (2006). The Transit Route
Cook, W. J. (2001). The Traveling Salesman
Arc-Node Service Maximization Problem. Euro-
Problem Paper presented at the Conference Name|.
pean Journal of Operations Research, In Review.
Retrieved Access Date| from URL|.
Dantzig, G. B. (1960). On the shortest route
Current, J. R., ReVelle, C. S., & Cohon, J. L.
through a network. Management Science, 6,
(1984). The application of location models to the
187–190. doi:10.1287/mnsc.6.2.187
multiobjective design of transportation networks.
Regional Science Review, 14, 1–23. Dantzig, G. B. (1966). All Shortest Routes in
a Graph. Paper presented at the International
Current, J. R., ReVelle, C. S., & Cohon, J. L.
Symposium, Rome.
(1985). The maximum covering / shortest path
problem: A multiobjective network design and Dantzig, G. B., & Ramser, J. H. (1959). The truck
routing formulation. European Journal of Opera- dispatching problem. Management Science, 6(1),
tional Research, 21, 189–199. doi:10.1016/0377- 80–91. doi:10.1287/mnsc.6.1.80
2217(85)90030-X
Dijkstra, E. W. (1959). A note on two problems in
connexion with graphs. Numerische Mathematik,
1, 269–271. doi:10.1007/BF01386390

44
Routing
Dubois, D., Bel, G., & Llibre, M. (1979). A set
of methods in transportation network synthesis
and analysis. The Journal of the Operational
Research Society, 30(9), 797–808.
Evans, J. R., & Minieka, E. (1992). Optimization
Algorithms for Networks and Graphs (2nd ed.).
New York: Marcel Dekker, Inc.
Fan, W., & Machemehl, R. B. (2004). Optimal
Transit Route Network Design Problem: Algo-
rithms, Implementations, and Numerical Results
(No. SWUTC/04/167244-1). Austin, TX: Center
for Transportation Research.
Flood, M. M. (1956). The traveling-salesman
problem. The Journal of the Operational Research
Society, 4, 61–75.
Floyd, R. W. (1962). Algorithm 97, shortest
path. Communications of the ACM, 5(6), 345.
doi:10.1145/367766.368168
Hasselström, D. (1981). Public Transportation
Planning: Mathematical Programming Approach.
University of Gothenburg, Gothenburg, Sweden.
Hoffman, A. J., & Wolfe, P. (1985). History. In
The Traveling Salesman Problem - A Guided
Tour of combinatorial Optimization (pp. 1-17).
Chichester: John Wiley & Sons.
Jiang, R., Hu, M. B., Jia, B., & Wu, Q. S. (2003).
Realistic bus route model considering the capacity
of the bus. The European Physical Journal B,
34(3), 367–372. doi:10.1140/epjb/e2003-00233-1
Lampkin, W., & Saalmans, P. D. (1967). The de-
sign of routes, service frequencies, and schedules
for a municipal bus undertaking: a case study.
Operations Research, 18(4), 375–397.
List, G. F. (1990). Toward optimal sketch-level
transit service plans. Transportation Research
Part B: Methodological, 24(5), 325–344.
doi:10.1016/0191-2615(90)90007-L
Magnanti, T. L., & Wong, R. T. (1984). Network
design and transportation planning: models and
algorithms. Transportation Science, 18(1), 1–55.
doi:10.1287/trsc.18.1.1
Moorthy, N. V. R. (1997). Planning of integrated
transit network for bus and LRT. Journal of Ad-
vanced Transportation, 31(3), 283–309.
Murray, A. T. (2003). A coverage model for im-
proving public transit system accessibility and
expanding access. Annals of Operations Research,
123, 143–156. doi:10.1023/A:1026123329433
Murray, A. T., & Wu, X. (2003). Accessibility
tradefoffs in public transit planning. Journal of
Geographical Systems, 5, 93–107. doi:10.1007/
s101090300105
Newell, G. F. (1979). Some issues relating to the
optimal design of bus routes. Transportation
Science, 13(1), 20–35. doi:10.1287/trsc.13.1.20
Schrijver, A. (2004). On the history of combina-
torial optimization. from http://homepages.cwi.
nl/~lex/files/histco.pdf
Shih, M., Mahmassani, H. S., & Baaj, M. H.
(1998). Planning and design model for transit
route networks with coordinated operations.
Transportation Research Record, 1623, 16–23.
doi:10.3141/1623-03
Silman, L. A., Barzily, Z., & Passy, U. (1974).
Planning the route system for urban buses.
Computers & Operations Research, 1, 201–211.
doi:10.1016/0305-0548(74)90046-X
Tom, V. M., & Mohan, S. (2003). Transit route
network design using frequency coded genetic
algorithm. Journal of Transportation Engineer-
ing, 129(2), 186–195. doi:10.1061/(ASCE)0733-
947X(2003)129:2(186)
Vajda, S. (1961). Mathematical Programming.
Reading: Addison-Wesley Publishing Company.
45

van Nes, R., Hamerslag, R., & Immers, B. H.
(1988). Design of public transport networks.
Transportation Research Record, 1202, 74–83.
Wang, J. J., & Po, K. (2001). Bus routing strate-
gies in a transit market: A case study of Hong
Kong. Journal of Advanced Transportation,
35(3), 259–288.
Zhao, F., & Gan, A. (2003). Optimization of Transit
Network to Minimize Transfers (No. BD015-
02). Miami: Lehman Center for Transportation
Research.
KEY TERMS AND DEFINITIONS
Graph Theory: The mathematical discipline
related to the properties of networks.
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 246-253,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
46
Routing
Heuristics: Procedures for quickly finding
good alternate—though not guaranteed optimal—
solutions to routing problems.
Network: A connected set of edges and ver-
tices.
Network Design Problems: A set of combi-
natorially complex network analysis problems
where routes across or flows through the network
must be determined.
Routing: The act of selecting a course of travel.
Shortest Path Problem: The routing prob-
lem of finding the shortest—or least cost—path
through a network.
Traveling Salesman Problem: The most
prominent problem in combinatorial optimization,
defined as the routing problem where a hypotheti-
cal salesman must find the most efficient sequence
of destinations in his territory, stopping only once
at each, and end up at the initial starting location.
 47

Chapter 6
Web-Based Geospatial Services:
Implementing Interoperability
Specifications

Iftikhar U. Sikder
Cleveland State University, USA

Aryya Gangopadhyay
University of Maryland - Baltimore County, USA

Nikhil V. Shampur
Cleveland State University, USA

ABSTRACT
This chapter characterizes the requirements of Geographic Information Systems (GIS) middleware and
its components for dynamic registering and discovering of spatial services specifically for collaborative
modeling in environmental planning. The chapter explores the role of Web services with respect to imple-
mentation standard and protocols and identifies implementation features for exposing distributed GIS
business logic and components via Web services. In particular, the chapter illustrates applications of the
interoperability specifications of Open GIS Consortium’s (OGC) Web Mapping Service and (WMS), Web
Processing Standards (WPS) with respect to implementation feature. The chapter demonstrates a proto-
type implementation of collaborative environmental decision support systems (GEO-ELCA- Exploratory
Land Use Change Assessment) where Web service-enabled middleware adds core functionality to a Web
mapping service. The application demonstrates how individual workspace-based namespaces can be used
to perform Web mapping functionality (such as spatial analysis in visualization) through the integration
of environmental simulation models to explore collective planning scenario. Built on OGC compliant
connector and supports WMS and WPS, the system includes interactive supports for geospatial data
query, mapping services and visualization tools for multi-user transactions.

DOI: 10.4018/978-1-4666-2038-4.ch006

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

INTRODUCTION
Both public and private enterprises have recently
produced a surge of interest in Web applications
for Geographic Information Systems (GIS). In
recent years service oriented middleware has
emerged as an essential ingredient in distributed
systems (Alonso, 2004; Chang & Park, 2006;
Chatterjee & Webber, 2004). This has triggered
a new wave of enthusiasm in composition of
complex services in a meaningful way involv-
ing not only traditional alphanumeric data but
complex geographic data and services (Jones &
Taylor, 2004; Sikder & Gangopadhyay, 2003,
2004). In particular, collaborative and groupware
researches are being directed towards develop-
ing reusable generic model and procedures,
which can be made to communicate each other
in any distributed system in a heterogeneous
environment. Integrating Web service with
collaborative Geographic Information Systems
(GIS) has triggered new wave of researchers
who are composing dynamic services involving
complex geospatial objects and models (Balram
& Dragic´evic, 2006). The growing need for a
service oriented middleware for GIS is especially
realized in three main contexts: (1) to access GIS
data from anywhere (2) to disseminate spatial
information of analysis and exploration of spa-
tial patterns and relationships between disparate
GIS datasets and (3) to allow GIS modeling/
processing tools and services to be downloaded
or uploaded over the internet by remote users to
work interactively by using existing Web brows-
ers rather than installing proprietary GIS software
locally on their machines (Peng & Tsou, 2003).
The recent trend of geospatial computing is a
gradual shift of traditional desktop GIS towards
distributed GIS (also referred to as GIServices).
With technology moving at such a fast pace,
the services expected by a GIS user (including
mobile ones) are quite demanding. The growing
demand for users’ need to view relationships
48
Web-Based Geospatial Services
between several geographically separate datas-
ets and perform varying degree of analysis and
geo-processing would inevitably require service
oriented architecture (SOA) of core GIS which
would define the use of services to support the
requirements of software users. In this context,
GIServices may be defined as a self-contained,
stateless spatial processing function which ac-
cepts one or more requests and returns one or
more responses through a well-defined, standard
interface. By having such services distributed
all over the Internet and accessible in a uniform
standard manner, it is possible to envision the
integration of several spatial services (chaining of
services) to provide higher levels of functionality
to existing services (Peng & Tsou, 2003). For
example, a typical GIS query “Find the nearest
Japanese restaurant along the highway” could
possibly be answered by chaining Web services
such as geocoding points of interest, integrating
transport networks, dynamic segmentation of
network, providing routing network, cartographic
map rendering, and text-to-voice conversion.
This paper explores the role of Web services
and their implementation standards and protocols
and characterizes features for distributed GIS for
modeling collaborative spatial processes with
respect to the current interoperability standards
and specifications. In particular the paper the
identifies the interoperability requirements of
OGC’s (Open GIS Consortium) Web Mapping
Service and (WMS), Web Processing Standards
(WPS) with respect to implementation features.
The rest of the paper is organized as follows:
section 2 discusses the framework of service
integration in distributed GIS; section 3 discusses
implementation standard for interoperability-
based distributed GIS; section 4 characterizes the
essential features of distributed GIS components;
and finally section 5 illustrates a prototype imple-
mentation of a system for geospatial resource
integration in environmental planning.
Web-Based Geospatial Services
Progress in Distributed Spatial
Services: Related Works
Collaborative GIS have been used in many
planning problems for solving semi-structured
or loosely structured decision problems in en-
vironmental planning (Angelides & Angelides,
2000; Balram & Dragic´evic, 2006; Balram et
al.,2003; Kingston et al., Carver, 2000) The Web
GIS implementation area mainly includes envi-
ronmental planning (ISikder & Gangopadhyay,
2002; Tuchyna, 2006), data dissemination (Hu,
1999; Schuurman & Leszczynski, 2006), commu-
nity planning (Al-Kodmany, 2000; MacEachren
et al., 2006; Rao et al., 2007). While researchers
continue to argue for an integration and structuring
of collaborative mapping and visualization tech-
nologies into spatial decision making (Armstrong,
1994; Balram & Dragic´evic, 2006; Jankowski &
Nyerges, 2001; MacEachren, 2001; Nyerges &
Jankowski, 2001), a Web-based GIS framework
designed to integrate stakeholders into the plan-
ning process has yet to be realized. One of the
major impediments to developing GIServices for
collaborative modeling is the lack of interoper-
able component technologies. Heterogeneity of
geo-spatial systems has plagued GIS since its
inception (Goodchild et al., 1992; Stoimenov &
Djordjevic-Kajan, 2005; Worboys & Deen, 1991).
Different agencies had built many different geo-
graphic data models and systems, following their
native organizational interest and problem domain
(Egenhofer & Herring, 1991). The benefit of col-
lective learning has not yet been fully realized,
due to a lack of mechanism for reusable service
and models in participatory systems.
As a precursor to Web services, CORBA and
Microsoft’s DCOM or Java-based RMI were often
used for distributed access and query of spatial
data, sometimes integrated with DSS (Eldrandaly,
2006). However, being “tightly coupled” with na-
tive data structure, the broker-based services are
unable to make sure that whenever an organization
makes changes in their native data structure or
services the corresponding change is automati-
cally reflected in all other organizations sharing
the same resource. Moreover, middleware-based
access through a broker relies on a standard defini-
tion of “interfaces.” Geo-processing services can
become very cumbersome in the absence of such
interfaces. From the decision support point of
view, having data access at the client’s end with-
out robust geo-processing capabilities amounts to
little help. In a broker-based solution, the client
has to pull massive amounts of data at his/her end
and manage it locally. Such approaches assume
the client’s explicit ability to manipulate server
connections and invoke remote objects. Thus, a
frequent spatial process, such as a spatial join
between data from two different servers, needs
to be coordinated at the client’s end. Such object
level manipulations of spatial processes often fail
to provide a high-level view to the application
developer. Paradoxically, in a spatial decision
support system the user or decision maker’s view
on spatial features or geometry needs to be real-
ized at a higher level of abstraction while at the
same time maintaining the transparency of system
processes (Anderson & Moreno-Sanchez, 2003;
Rao et al., 2007; Tait, 2005).
Unlike current component technologies,
however, Web Services do not use object model-
specific protocols such as DCOM, RMI, or IIOP
that require specific, homogeneous infrastructures
on both the client and service machines. While
implementations tightly coupled to specific com-
ponent technologies are perfectly acceptable in a
controlled environment, they become impractical
on the Web. As the set of participants in an inte-
grated business process changes and as technology
changes over time, it becomes very difficult to
guarantee a single, unified infrastructure among
all participants. Web Services take a different
approach; they communicate using ubiquitous
Web protocols and data formats, such as HTTP
and XML. Any system supporting these Web
standards will be able to support Web Services
(Peng & Tsou, 2003).
49

In static web mapping the web client is a simple
web browser with the capability of handling HTML
and Web Forms. A Web Form is a HTML Page
with data entry fields for user input. The user
inputs are collected and sent by the browser to
the server in a HTTP message. The web server
receives the message but cannot respond to it.
Since it does not understand any requests other
than for HTML or other MIME-type documents,
it passes the request to a back-end program. The
back end program is a traditional server side ap-
plication that does the actual processing. In the
case of web mapping, these back-end programs
include map servers and DBMS servers. The web
server then returns the results to the client. Here,
the web server becomes middleware, connecting
the web client and the back end applications(Peng
& Tsou, 2003).
Web mapping of GIS applications has evolved
from the concept of simple static web publishing
of maps on a web page as static map images in
appropriate graphic formats (GIF, JPEG, PNG,
PDF etc). The client (web browser) makes re-
quests using standard HTTP constructs such as
GET and POST to a web server, which returns
the requested map image file in appropriate file
format. However, an effective Web mapping
requires more than just making static making
maps; it requires processing queries and doing
some limited spatial analysis on the server before
presenting the output in a specialized cartographic
format on the standard web browser. Historically,
the Xerox PARC Map Viewer developed in 1994
was the earliest attempt to distribute GIS using
HTTP server and CGI program(Putz, 1994). This
was followed by the development of an on-line
GIService of GRASSLinks prototype that mim-
icked traditional GISystem functions, such as map
browsing and buffering overlay (Huse, 1995).
The Alexandria Digital Library offered a “digi-
tal library” metaphor for publishing multimedia
content including georeferenced data. Among the
commercially available Web service-enabled ap-
plications, ESRI’s ArcWeb Services offer a way
50
Web-Based Geospatial Services
to include GIS content and capabilities, Intranet
applications, or ArcGIS—on demand. ESRI’s
ArcExplorer viewer is designed to support Web
services provided through its proprietary ArcIMS
server software. Microsoft’s MapPoint offers a
street-mapping program that includes some de-
mographic overlays. Users can buy the package
or access MapPoint Web Services, which can be
utilized to integrate other applications: location-
based services, GPS car navigation, and decision
support systems. MapInfo supports both OpenGIS
and the World Wide Web Consortium’s Web
services standards, including SOAP, WSDL and
UDDI, in products such as its miAware software
(MapInfo, 2007) (mobile location services plat-
form for MapInfo) for developing location-based
services. MapServer, an open environment, sup-
ports the creation of maps at the client’s side by
serving as a map engine to provide database and
spatial content through the client’s browser. It also
supports many OGC’s specifications like WMS,
non-transactional WFS, WCS, WMC, SLD, and
GML (MapServer, 2007).
The concept of dynamic distributed computing
paradigm calls for a metaphorical counterpart of
the so called LEGO blocks that can be interlocked
and stacked to create complex geospatial services
(Tsou & Buttenfield, 2002). The GIService and
modules like LEGOs can be rearranged and
restacked in a different configuration do a dif-
ferent task (Zipf & Jost, 2006). In Web services,
a series of protocols such as eXtensible Markup
Language (XML); Simple Object Access Proto-
col (SOAP); Web Service Description Language
(WSDL); and Universal Description, Discovery,
and Integration (UDDI) - provides standards for
communication and collaborative processing ca-
pacity among Web-service compliant architecture.
A key advantage is that various GIS layers can
be dynamically queried and integrated while still
maintaining independence in a distributed envi-
ronment. One of the promises of Web Services
in the GIS world is up-to-date information since
the data is stored with the service provider and
Web-Based Geospatial Services
expectedly kept up-to-date when accessed by the
GIS web service consumer. From organizational
point of view, this may be very appealing, for
local governments (such as counties and other
organizations) can still independently collect and
manage data locally and integrate information and
services using Web services. A client, for example,
a transportation company, could directly access a
local government’s base map without maintaining
its own. At the same time the client can update
local government’s data from its own record. As
far as data interoperability is concerned, extended
collaboration and partnerships using Web Services
could provide opportunity to open interfaces
and communication mechanisms for distributed
computing. Web services extend the use of GIS
by making the integration with other applications
easier. Despite the promise of GIS Web services,
interoperability among GIS programs and data
sources is far from seamless. Vendor adoption
of GIS Web services is still a work in progress.
Interactive Web mapping demands more interac-
tions between the user and the client interface and
more client-size processing and functionalities
than the static web mapping applications. Also,
CGI extensions such as Servlets, ASP, etc. are used
to mitigate shortcomings of CGI. Most current
interactive Web CGI programs are based on this
model (dynamic viewer with CGI extensions).
Interactive viewers include 1) DHTML viewer –
makes static HTML pages dynamic using client
side scripting (VBscript, javascript), DOM and
CSS; 2) Java Applets – Executable java code
downloaded from the server and executed on
the client at runtime. Applets can be integrated
inside the web browser; and 3) ActiveX Con-
trols – Modular pieces of software that perform
tasks and communicate with other programs and
modules over the Internet via OLE.
Implementing Interoperability
Standards in Distributed GIS
A suite of emerging Web services standards
from the Open GIS Consortium Inc. (OGC) is
facilitating the transition of historically standalone
GIServices by enabling interoperability. Web
services can put relevant GIS applications on the
end user’s desktop by embedding them within
familiar applications. The most intensively used
standards include Web Map Service, Web Feature
Service and the XML-based Geography Markup
Language (GML)(OGC, 2007; Peng & Zhang,
2004; Siyuan, Griffiths, & Paton, 2007); they al-
low applications to access distributed spatial data
across the Internet to any OGC-enabled repository.
Interoperability Requirements of
OGC’s WMS, WPS Standards
Although interactive mapping programs are
popular, they suffer from common problems
such as poor performance and limited functions.
However the biggest problem is that they are
often proprietary and not necessarily interoper-
able. Different web mapping programs were
developed in different database frameworks and
use different technologies. Integrating and sharing
information among web mapping programs are
arduous tasks. Furthermore, migrating technology
from one platform to another is very difficult.
Open GIS Consortium (OGC) has been making
efforts to develop a set of standards to guide the
development of web mapping programs so that
they can be interoperable. OGC developed Web
Map Server (WMS) implementation information
specifications that are the first effort towards stan-
dardizing the implementations of web mapping
programs (Peng & Tsou, 2003).
51

Web Map Server (WMS)
Implementation Specifications
A Web Map Service (WMS) produces maps of
spatially referenced data dynamically from geo-
graphic information. A “map” is defined to be a
portrayal of geographic information as a digital
image file suitable for display on a computer
screen. The WMS specification stipulates that
a mapping service should be able to at least 1)
produce a map (as a picture, series of graphical
elements, feature data); 2) answer basic queries;
and 3) inform other programs about its capabili-
ties (what maps it can produce and which can be
queried further)(OGC). OGC initially came up
with thin, medium and thick clients in its specifica-
tion, but abandoned them for they were inherently
ambiguous in classifying clients this way. Instead
they use the kind of information presented at to
web client to categorize web mapping services
into three cases:
1. Picture case: Here a client’s request is an-
swered by a picture of the map
2. Graphic element case: Here the web cli-
ent receives a set of graphic elements (e.g.,
Scalable Vector Graphics (SVG)). The
picture element case may be considered as
a subset of graphic element case.
3. Data or feature case: Here geographic
feature data is sent from the server to the
client.
Currently the picture case is the most popular
framework adopted by the GIS industry. It provides
only for limited map display functions and user
interactions. Along with progress of web mapping
and information technologies, the data case and
the graphic element case have become slightly
more popular. WMS-produced maps are generally
rendered in a pictorial format such as PNG, GIF or
JPEG, or occasionally as vector-based graphical
elements in Scalable Vector Graphics (SVG) or
52
Web-Based Geospatial Services
Web Computer Graphics Metafile (WebCGM)
formats.
Web Map Service operations can be invoked
using a standard web browser by submitting re-
quests in the form of Uniform Resource Locators
(URLs). The content of such URLs depends on
which operation is requested. In particular, when
a client requests a map, the URL indicates what
information is to be shown on the map, what
portion of the Earth is to be mapped, the desired
coordinate reference system, and the output image
width and height. When two or more maps are
produced with the same geographic parameters and
output size, the results can be accurately overlaid
to produce a composite map. The use of image
formats that support transparent backgrounds
(e.g., GIF or PNG) allows underlying maps to
be visible. Furthermore, individual maps can be
requested from different servers. The Web Map
Service thus enables the creation of a network of
distributed map servers from which clients can
build customized maps. A basic WMS classifies
its geographic information holdings into “Layers”
and offers a finite number of predefined “Styles”
in which to display those layers.
Requirements for a WMS over HTTP
The International Standards defines the implemen-
tation of the Web Map Service on a distributed
computing platform (DCP) comprising Internet
hosts that support the Hypertext Transfer Pro-
tocol (HTTP). Thus the online resource of each
operation supported by a server is an HTTP URL
(Uniform Resource Locator). The URL may be
different for each operation, or the same, at the
discretion of the service provider. Each URL is
implementation-dependent, with only the query
portion comprising the service request itself as
defined by the International Standard. The HTTP
supports two request methods: GET and POST.
One or both of these methods may be offered by
a server, and the use of the Online Resource URL
differs in each case. Support for the GET method
Web-Based Geospatial Services
is mandatory; support for the POST method is
optional. A Web Map Service shall support the
“GET” method of the HTTP protocol while the
Web Map Service may support the “POST” method
of the HTTP protocol.
Operations on Web Map Services
The three operations defined for a Web Map
Service are as follows:
1. GetCapabilities: The purpose of the manda-
tory GetCapabilities operation is to obtain
service metadata, which is a machine read-
able (and human-readable) description of the
server’s information content and acceptable
request parameter values.
2. GetMap: The GetMap operation returns a
map. Upon receiving a GetMap request, a
WMS shall either satisfy the request or issue
a service exception.
3. GetFeatureInfo: GetFeatureInfo is an
optional operation. The GetFeatureInfo
operation is designed to provide clients of a
WMS with more information about features
in the pictures of maps that were returned by
previous Map requests. The canonical use
case for GetFeatureInfo is that a user sees
the response of a Map request and chooses
a coordinate (i, j) on that map from which
to obtain more information. The basic op-
eration provides the ability for a client to
specify which pixel is being asked about,
which layer(s) should be investigated, and
what format the information should be
returned in. Since WMS protocol is state-
less, the GetFeatureInfo request indicates
to the WMS what map the user is viewing
by including most of the original GetMap
request parameters. The actual semantics of
how a WMS decides what to return more
information about or what exactly to return
are left up to the WMS provider.
Web Processing Service
Web processing service exposes pre-programmed
calculations for geospatial data to the Internet. The
Web Processing Service (WPS) interface specifies
WPS operations that can be requested by a client
and performed by a WPS server. Those operations
are as follows:
• GetCapabilities: This operation allows a
client to request and receive back service
metadata (or Capabilities) documents that
describe the abilities of the specific server
implementation.
• DescribeProcess: This operation allows a
client to acquire more specific information
about an Execute operation provided by
the WPS, including the input parameters
and formats, and similarly the outputs.
• Execute: This operation allows a client to
run a specified process of the WPS with
qualified input parameters and values.
Implementing GetCapabilites and
DescribeProcess through WSDL
These operations have many similarities to other
OGC Web Services such as Web Mapping Service
(WMS), Web Feature Service (WFS), etc. The
GetCapabilites and DescribeProcess are imple-
mented using web services through WSDL which
is an XML format for describing Web Services.
WSDL describes the public interface to the web
service using XML-based service description.
It specifies how to communicate using the web
service, namely the protocol bindings and message
formats required to interact with the web services
listed in its directory. The supported operations
and messages are described abstractly, and then
bound to a concrete network protocol and mes-
sage format. WSDL is often used in combination
with SOAP and XML Schema to provide web
services over the Internet (Chang & Park, 2006).
A client (program) connecting to a web service
53

can read the WSDL to determine what functions
are available on the server. Any special data types
used are embedded in the WSDL file in the form
of XML Schema. The client can then use SOAP
to actually call one of the functions listed in the
WSDL. This amounts to achieving the same ob-
jectives as OGC WPS operations, GetCapabilities
and DescribeProcess.
Characterizations of Components
in a Distributed GIService
What follows is the characterization of the com-
ponent of distributed mapping services of GIS
with respect to dynamically distributed GIS.
We have seen that that although web server and
map server perform important tasks, they cannot
directly communicate since each talk different
languages. Web servers use HTTP and HTML
while map servers have different query structures
and formats, so they rely on in-between programs
to translate between them. Therefore, for a robust
distributed GIservice-oriented middleware should
include the following things:
• Establishing connections between web
server and map server
• Translating user requests from the web
server and passing them to the map server
for processing
• Translating output from map server to out-
put expected by the web server (such as
HTML) which will be forwarded to the
web browser
In addition, GIS middleware performs value
added service such as
• Managing concurrent requests, load
balancing
• Managing state, transaction management
• Security
54
Web-Based Geospatial Services
There already exists a wealth of business
logic that has been implemented for providing
various web mapping services. Exposing existing
GIS business functionality (geospatial analysis,
processing, etc.) via web services is an essential
aspect. Web services provide standard interfaces
via WSDL needed for a distributed GIS architec-
ture. Server side components can be implemented
using any distributed computing framework such
as J2EE or.NET. However to provide the neces-
sary standard interface, the functionality must be
exposed via web services.
Figure 1 shows the various components
typically found in the architecture of a distributed
GIS provider. It consists of a web server as the
front end and interfaces with the client. The web
server communicates with the application server
and other related middleware, all of which are
responsible for the business login that fulfills the
user’s request. The map server is responsible for
generating the requested maps. The data server is
the data store where the geographic data is housed.
To provide for higher scalability, we use multiple
map servers and multiple web servers. Since
many map servers and data servers are involved,
we should also include catalog services, data
repositories and load balance services. A catalog
service is used to keep track of what functions
each map server can provide. Data repository is
a registration service that keeps track of the data
types and location of data sets; it also manages
multiple data servers by filtering the right data
set from the appropriate data server.
GEO-ELCA: COLLABORATIVE
GISERVICE ARCHITECTURE
In this section we illustrate a prototype imple-
mentation of a Web-based spatial decision support
system for collaborative planning of urban land use
change evaluation. GEO-ELCA allows the various
features of GIS services on the Web. The system
allows dynamic selection of a feature type (i.e.,
Web-Based Geospatial Services
Figure 1. Exposing component functionality via Web service
polygon – from land use theme) interactively, so
that a user can change attribute items and identify
a feature property. A user can initiate a change in
land use type by graphically selecting a polygon.
The server side application processes the request
and makes necessary updates in the database
to reflect the corresponding the changes of the
pollutant coefficients. Every request to change
in land use category results in a recalculation of
the mass export of pollutants and corresponding
statistics. The processed result is sent back to
the Web server and then to the client side. The
user decision profile is then input to a simulation
model to estimate the yearly pollution load. The
system integrates a simulation model commonly
used in urban hydrology--the so-called “Simple
Method”(Schueler, 1999) for estimating exports
of various pollutants runoff from different land
uses. The output is then visualized as pollutant
distribution in terms of different classification
schemes (e.g., standard deviation, plain break,
quantile, etc.) with a modified map legend.
Multi-User Context in Distributed GIS
In view of a decision support framework, col-
laborative modeling presupposes multiple parties
with different perspectives working together in
a complex emergent environment. These parties
(henceforth termed as “agent, “decision maker”,
or “stakeholder” interchangeably) must have an
integrated data access from heterogeneous sources
to integrate with transparent high performance
computing resources to compose decision models
dynamically. However, in real life situations, it is
often difficult to achieve the stakeholders’ views
or effective patterns of social interactions in the
planning process. For example, decisions on how
current land use should be changed depend on
legal, environmental, regulatory constraints as
well as biases and preferences of different group
or institutions. In particular, understanding urban
land use dynamics involves considerations of the
complex behavior of individual decision makers
and the interaction of local and regional institutions
in multiple scales. Moreover, such decisions are
inherently spatial in nature because the change
in a particular parcel may have direct or indirect
consequences to the neighboring parcels.
For example, EPA’s Brownfields development
program involves the expansion and redevelop-
ment of urban areas that may be complicated by
the potential presence of a hazardous substance,
pollutant, or contaminant (EPA, 2007). While
cleaning up and reinvesting in these properties
takes development pressures off of undevel-
55

oped, open land and improves the environment,
evaluating a candidate property to determine if
it meets the criteria for redevelopment is inher-
ently a collective decision process. For instance,
an individual landowner may act from his or her
individual interest; however, in the long run the
overall land use scenario may be undesirable to
everyone. Moreover, changes in land use may
create concern for environmental impact in the
surrounding region. While land-use changes are
often identified as a major driving force of ecologi-
cal changes, with the conversion of land use from
one category to another there is an overall change
of hydrological characteristics resulting from the
changes of impervious areas. Consequently, there
is an increase of volume and peak flow of pos-
sible increase in the concentration of pollutants,
which potentially could deteriorate the environ-
ment. Hence, a centralized planning process is
essentially inadequate to reflect group dynamics.
Such group-individual dilemmas make it ideally
suited for collaborative planning in a distributed
environment, particularly in the seamless integra-
tion of Web-services in accessing geospatial data
and models for environmental planning.
The resulting pollution map can be visualized
with multiple theme overlay (Figure 2). A user can
Figure 2. User specific visualization to change polygon attributes in GEO-ELCA
56
Web-Based Geospatial Services
review the planning scenario of other users. The
system logs individual users’ preferences which
can be used as input into mediating algorithm
(e.g., genetic algorithm) to resolve conflicts among
user preferences of land use choice and thereby
optimize pollution scenarios.
Distributed Mapping Services
GEO-ELCA offers cartographic representation
with a specialized visualization feature aided with
dynamic legend rendering and symbol manipu-
lation. The GEO-ELCA’s cartographic features
include the following:
• A visualization of pollution potential
map based on different themes (e.g., total
Nitrogen, total BOD etc.)
• Multiple representations and visualizations
of the same data based on different statisti-
cal classification schemes (Equal Interval,
Standard Deviation, etc.)
• Visualization of a user-specified color
ramp of customized legend
• Enhanced navigational options for spatial
overlay of multiple geographic layers
Web-Based Geospatial Services
Figure 2 shows a typical example where users
graphically select a polygon and choose a land
use category from the pop up list (upon right
click) to simulate the “what if” kind of pollution
scenario. The simulation model estimates the
pollution potential as a result of land use change.
For example, users can visualize the pollution
characteristics of total Nitrogen or total BOD,
etc., of the selected polygon. The changes made
in the database by the user (user-1) are completely
invisible to other users (e.g., user-2). User-2 has
an option to concurrently perform the similar
operation independently without being effected
by User-1’s. The result of the simulation will
be different if different options are chosen. The
visualization and cartographic representations
can be locally customized by users. In each case,
the corresponding legends and color rendering
services are implemented accordingly.
Service Integration in GEO-ELCA
GEO-ELCA’s main feature provides users with
a meaningful composition of services to perform
an exploratory analysis and assessment of the
consequence of users’ decisions in environmen-
tal planning and compares the results with the
collective decision. The system is built on an
OGC-compliant connector that supports access
to WMS and WPS.
The client sends WMS Request (GetMap,
GetCapabilities, GetFeatures) and WPS Execute
Figure 3. Overview of service integration mechanism of GEO-ELCA
Operation (changeLandUsageType) messages
which are XML encoded and enclosed in SOAP
1.1 messages (see Figure 3). The construction of
SOAP messages is done using client side script-
ing such as JavaScript, VBScript and can also
makes use of advanced features such as AJAX
(Sayar, Pierce, & Fox, 2006). Built on top of the
ESRI’s ArcIMS, the system uses ArcXML (native
XML encoding of spatial object) to communicate
between the custom middleware and the Web
mapping server. The services offered by the Web
Mapping Server are similar to the requirements
of OGC’s implementation specification for a web
map service (WMS). The only difference is that
the communication language is ArcXML whereas
the OGC international standard defines a more
generic XML format. The communication with
client and Web server involves ArcXML in both
directions, whereas OGC specifies that only re-
sponse parameters need to be an XML document
while request parameters may be in formats such
as Key Value Protocol (KVP) encoding-- com-
monly used in HTTP GET and POST. It should be
noted that a bidirectional mapping of request and
response parameters from OGC format to ArcXML
format is quite simple and can be achieved by a
middleware solutions. Table 1. shows ArcXML
equivalent of OGC WMS operation.
The following example illustrates how the
services rendered by the customized middleware
uses ArcXML to map directly to the operations
needed to conform to the OGC standard for a
57
 Web-Based Geospatial Services

Table 1. ArcXML equivalent of OGC Web mapping

AYERS=polygon&STYLES=None&SRS=None&BB
service operations
OX=418755.237974071,4569670.72466017,
OGC WMS Operation ArcXML Request Operation 468691.884383251,4608998.26066327&WID
GetCapabilities GET_SERVICE_INFO
TH=786&HEIGHT=400&FORMAT=image/gif
GetMap GET_IMAGE
GetFeatures GET_FEATURES
This is mapped in GEO-ELCA according to the
corresponding request to the web mapping server
which is in ArcXML. GET_IMAGE Request sent
to the web map server (in ArcXML) looks like
WMS (and a WPS). The operation offered by the
what follows:
WPS which is implemented by our middleware
are as follows:
<?xml version=”1.0” encoding=”UTF-8”
?>
void changeLandUsageType(string: use-
<ARCXML version=”1.1”>
rID, long: minX, long: minY, long:
<REQUEST>
maxX, long: maxY, int: newLandType)
<GET_IMAGE>
<PROPERTIES>
The operation allows users to change the land
<ENVE-
type graphically underlying GIS data. The input
LOPE minx=”418755.237974071”
agreements are as follows:
miny=”4569670.72466017”
maxx=”468691.884383251
userID - Identifier of the user (to
”
perform data manipulation on user
maxy=”4608998.26066327” />
specific data)
<IMAGESIZE
minX, minY, maxX, maxY – bounding co-
height=”400” width=”786” />
ordinates of land whose land type is
to change
newLandType – Identifier of class of
<LAYERS>
Land Type to which the land must be
<LAYER
changed.
name=”polygon” id=”0” />
<LAYERS>
For example, a simple request for change of
<LEG-
land use is encoded using KVP encoding
END title=”Legend” font=”Arial”
width=”170” height=”300”
http://localhost/service?user=tester
autoextend=”true” backgroundcol-
&minX=23.45&minY=24.34&maxX=45.67&ma
or=”255,255,255” />
xY=67.8
</PROPERTIES>
OGC Complaint GET MAP Request (Encod-
<LAYER type=”acetate”
ed using KVP Key-Value Pair and send
name=”theNorthArrow”>
using HTTP-GET)
<OBJECT
http://localhost:8080/poly_Project/sw
units=”pixel”>
itchUser?VERSION=1.1&REQUEST=GetMap&L
<NORTHAR-
ROW type=”4” size=”15” coords=”20
35” shadow=”32,32,32” angle=”0”

58
Web-Based Geospatial Services

antialiasing=”True” /> and have the specified spatial reference system,

</OBJECT> bounding box, size, and format. Every GET_IM-
</LAYER> AGE request to the web map server returns the
<LAYER type=”acetate” URL of the generated file which has a unique
name=”theScaleBar”> filename. The URL of the image returned by the
<OBJECT web map server complies with the expected OGC
units=”pixel”> Complaint GET_MAP Response.
<SCALE-
BAR screenlength=”157” coords=”471 ‘Virtual’ Private User Workspace
3” mode=”geodesic” fontcolor=”0,0,0” for Model Preference
fontstyle=”Regular” width=”5” />
</OBJECT> In this section we examine the database design
</LAYER> which allows the WPS operation to be performed
</GET_IMAGE> and also allows for collaborative modeling us-
</REQUEST> ing a single underlying GIS data set. In order
</ARCXML> to generate multi-use profiles for collaborative
modeling, a relational model was developed
GET_IMAGE Response received from Map providing each user a ”virtual private” (note: can
Server: you say “virtually private”?) workspace. Each
user is associated with a native profile database,
<?xml version=”1.0” while in fact the actual visualization is rendered
encoding=”UTF-8”?> from a single dataset. The GIS layer includes
<ARCXML version=”1.1”> shapefile--a non-topological data structure com-
<RESPONSE> monly used to represent GIS data in polygon
<IMAGE> features. Every polygon has an associated area,
<ENVE- perimeter, and land usage class (represented by
LOPE minx=”405084.257055615” class ID and described in human-readable form as
miny=”4569670.72466017” a Class Label field). We demonstrate two use case
maxx=”482362.865301707” scenarios of the collaborative modeling. In case
maxy=”4608998.26066327” /> I, the user performs default services that involve
<OUTPUT url=”http:// normal backup, replacement, map generation and
maestro/output/poly_project_MAE- restoration of data (Figure 4). In case II the user
STRO47223362.jpg” /> specifies a land change operation, so after the
<LEGEND url=”http:// backup, the user’s table is updated to reflect the
maestro/output/poly_project_MAE- land change request and then the same replace,
STRO47223363.jpg” /> map generation and restore operations follow,
</IMAGE> similar to case I. The sequences are as follows:
</RESPONSE>
</ARCXML> 1. The operations begin with a request from
the client sent to the web server (not shown
This is mapped to an OGC complaint GET_ in diagram).
MAP response. The response to a valid GetMap 2. The middleware receives the request, backs
request must be a map of the georeferenced up the master data (backing up only the re-
information layer requested, in the desired style,

59

Figure 4. Collaborative feature of GEO-ELCA in the distributed GIS systems
quired amount of data) into a backup data
store.
3. The middleware then updates the spatial
layer with user specific data.
4. Then the middleware forwards the request for
map to the web map server after converting
the request to a format understandable by the
web map server. In GEO-ELCA ArcXML
is the language of communication with
ArcIMS.
5. The web map server queries the data store
to retrieve the data it needs to generate the
map
6. The records returned consist of current user’s
data.
7. A map image is generated using the updated
(current user’s) data and the map image is
returned to the middleware solution.
8. The master data is restored to original values
by updating records from backup data store
9. The middleware solution returns the map
image (or URL) of the newly generated user
specific Map.
A change of land type request issued involves
a number steps when compared to GetFeatures
or GetMap requests. The additional step that is
performed by the middleware are steps 3 and 4
updates the user’s table to reflect the change in
60
Web-Based Geospatial Services
polygon values prior to switching master data
contents with user data contents. The changes
made by the users are reflected on their respec-
tive data table and hence changes can be tracked.
We note that the response to a changeLandType
request is a Map Image response showing the
visualization Map reflecting changed land type
on map (see Figure 6).
Figure 5 illustrates the sequence of operations
when the user performs a land use change. The
sequences are as follows:
1. The operations begin with a request from
the client sent to the web server (not shown
in diagram).
2. The middleware receives the request, and
then it updates the specific user’s data.
3. It then backs up the master data (backing
up only the required amount of data) into a
backup data store.
4. The middleware then updates the spatial
layer with user specific data.
5. Then the middleware forwards the request for
a map to the web map server after converting
the request to a format understandable by the
web map server. In our case ArcXML is the
language of communication with ArcIMS.
Web-Based Geospatial Services

Figure 6. Sequence diagram for changing land use category by collaborative editing of polygon attributes

6. The web map Server queries the data store
to retrieve the data it needs to generate the
map
7. The records returned consist of current user’s
data.
8. A map image is generated using the updated
(current user’s) data and the map image is
returned to the middleware solution.
9. The master data is restored to original values
by updating records from backup data store.
10. The middleware solution returns the map
image (or URL) of the newly generated user
specific Map.

Figure 5. Using Case Scenario II – User requests to edit a map to change polygon land type

61

CONCLUSION
This article presents a conceptual and prototypical
implementation of a Web service-based geospatial
resource integration scheme. Since Web services
avoid problems of tightly coupled distributed
object techniques and exposes an application pro-
gramming interface over the Web, it holds promise
for distributed resource sharing and collabora-
tive environmental planning. By wrapping and
dynamically integrating remote geo-processing
services from multiple sources, one can develop
an emergent collaborative system using interop-
erable standards of XML and SOAP (Chang &
Park, 2006). We introduced architecture for Web
service-based collaborative modeling applications
for environmental planning. The prototype system
is based on the conversion of the OGC compat-
ible GIS services to Web Services for flexible
collaboration. The prototype architecture offers
functionalities to integrate disparate spatial data
processing services into meaningful compositions
to perform complex exploratory analysis such as
the assessment of the environmental consequences
of individual decision makers’ choices and the
collective behavior that results.The significant
advantage of the architecture is that it provides a
means for self-describing services in which the
end user (any WMS/WPS client) can invoke a
set of services (e.g., catalog services of spatial
map). Specifically the system makes use of OGC’s
GetCapabilities, GetMap and GetFeaturesInfo
to create a user-specific “on-the-fly” map layer
from existing spatial data layers. The composition
GIS components allow development of collab-
orative applications (e.g., dynamic user specific
plan profiling and customization of cartographic
visualization and rendering). In terms of collab-
orative decision-making, the added advantage
is that community based geo-spatial vocabulary
and the corresponding modeling semantics can be
communicated effectively (e.g., the consequence
of user decision can be simulated graphically to
estimate the yearly pollution load as a result of
62
Web-Based Geospatial Services
individual decisions). Each user’s planning profile
can finally be compared for further modification.
The collaborative model espoused here is geared
towards archiving different opinions regarding
visualization and negotiation. We have also noted
that before GIservices can be wrapped with SOAP
API, the generic geo-processing model formalism
is needed to link models to the domain specific
application. A formal description of the spatial data
component and process model will allow better
interoperation among heterogeneous systems.
Future extension of GEO-ELCA will include an
ontology-driven semantic layer for high level
communication with geo-processing services.
REFERENCES
Al-Kodmany, K. (2000). Using Web-based tech-
nologies and geographic information systems in
community planning. Journal of Urban Technol-
ogy, 7, 1–30. doi:10.1080/713684108
Alonso, G. (2004). Web services: concepts, ar-
chitectures and applications. Berlin: Springer.
Anderson, G., & Moreno-Sanchez, R. (2003).
Building Web-based spatial information solu-
tions around open specifications and open source
software. Transactions in GIS, 7(4), 447–466.
doi:10.1111/1467-9671.00158
Angelides, M., & Angelides, M. C. (2000). Using
multimedia database information systems over
the internet for enhancing the planning process
for dealing with the built heritage. International
Journal of Information Management, 20, 349–367.
doi:10.1016/S0268-4012(00)00028-1
Armstrong, M. (1994). Requirements for the de-
velopment of GIS-based group decision support
systems. Journal of the American Society for Infor-
mation Science American Society for Information
Science, 45(9), 669–677. doi:10.1002/(SICI)1097-
4571(199410)45:9<669::AID-ASI4>3.0.CO;2-P
Web-Based Geospatial Services
Balram, S., & Dragic’evic, S. (2006). Modeling
collaborative GIS processes using soft systems
theory, UML and object oriented design. Transac-
tions in GIS, 10(2), 199–218. doi:10.1111/j.1467-
9671.2006.00253.x
Balram, S., Dragicevic, S., & Meredith, T. (2003).
Achieving effectiveness in stakeholder participa-
tion using the GIS-based collaborative spatial
Delphi methodology. Journal of Environmen-
tal Assessment Policy and Management, 5(3),
365–339. doi:10.1142/S1464333203001413
Chang, Y.-S., & Park, H.-D. (2006). XML Web
service-based development model for Internet
GIS applications. International Journal of Geo-
graphical Information Science, 20(4), 371–399.
doi:10.1080/13658810600607857
Chatterjee, S., & Webber, J. (2004). Developing
enterprise Web services: an architect’s guide. Up-
per Saddle River, NJ: Prentice Hall PTR.
Egenhofer, M., & Herring, J. R. (1991). High-
level spatial data structures for GIS. In D. J.
Maguire, M. F. Goodchild & D. W. Rhind (Eds.),
Geographical information systems: principles
and applications (Vol. 1, pp. 227-237). London:
Longman Scientific Publications.
Eldrandaly, K. A. (2006). A COM-based expert
system for selecting the suitable map projection
in ArcGIS. Expert Systems with Applications,
31(1), 94–100. doi:10.1016/j.eswa.2005.09.008
EPA. (2007). Brownfields cleanup and redevelop-
ment. Retrieved January 2007, from http://www.
epa.gov/brownfields/
Goodchild, M. F., Hanning, R., & Wise, S. (1992).
Integrating GIS and spatial data analysis: problems
and possibilities. International Journal of Geo-
graphical Information Systems, 6(5), 407–423.
doi:10.1080/02693799208901923
Hu, S. (1999). Integrated multimedia approach
to the utilization of an Everglades vegetation
database. Photogrammetric Engineering and
Remote Sensing, 65(2), 193–198.
Huse, S. M. (1995). GRASSLinks: A new model
for spatial information access for environmental
planning. Retrieved from http://www.regis.berke-
ley.edu/sue/phd/
Jankowski, P., & Nyerges, T. (2001). Geo-
graphic information systems for group decision
making:Towards a participatory geographic in-
formation science. New York: Taylor and Francis.
Jones, M., & Taylor, G. (2004). Data integration
issues for a farm decision support system. Trans-
actions in GIS, 8(4), 459–477. doi:10.1111/j.1467-
9671.2004.00196.x
Kingston, R., Carver, S., Evans, A., & Turton,
I. (2000). Web-based public participation geo-
graphical information systems: An aid to local
environmental decision-making. Computers,
Environment and Urban Systems, 24(2), 109–125.
doi:10.1016/S0198-9715(99)00049-6
MacEachren, A. M. (2001). Cartography and
GIS: Extending collaborative tools to support
virtual teams. Progress in Human Geography,
25, 431–444. doi:10.1191/030913201680191763
MacEachren, A. M., Pike, W., Yu, C., Brewer, I.,
Gahegan, M., & Weaver, S. D. (2006). Building
a geocollaboratory: Supporting human-environ-
ment regional observatory (HERO) collaborative
science activities. Computers, Environment and
Urban Systems, 30(2), 201–225. doi:10.1016/j.
compenvurbsys.2005.10.005
MapInfo. (2007). miAware2.0 documentation
library. Retrieved May 2007, from http://refer-
ence.mapinfo.com/common/docs/mapxtend-dev-
web-none-eng/
MapServer. (2007). MapServer. Retrieved May
2007, from http://mapserver.gis.umn.edu/
63

Nyerges, T., & Jankowski, P. (2001). Geographic
Information Systems for Group Decision Making.
London: Taylor & Francis.
OGC. (2007). OpenGIS. Geography markup
language (GML) implementation specification.
Retrieved May 2007, from http://opengis.net/gml/
OGC. Web map service interface (No. 03-109r1):
OGC.
Peng, Z.-R., & Tsou, M.-H. (2003). Internet GIS:
Distributed geographic information services for
the Internet and wireless networks. Wiley.
Peng, Z.-R., & Zhang, C. (2004). The roles of
geography markup language (GML), scalable
vector graphics (SVG), and Web feature service
(WFS) specifications in the development of
Internet geographic information systems (GIS).
Journal of Geographical Systems, 6(2), 95–116.
doi:10.1007/s10109-004-0129-0
Putz, S. (1994). Interactive information services
using World-Wide Web hypertext. Retrieved 2006
from http://www2.parc.com/istl/projects/www94/
mapviewer.html
Rao, M., Fan, G., Thomas, J., Cherian, G.,
Chudiwale, V., & Awawdeh, M. (2007). A web-
based GIS decision support system for manag-
ing and planning USDA’s conservation reserve
program (CRP). Environmental Modelling &
Software, 22(9), 1270–1280. doi:10.1016/j.env-
soft.2006.08.003
Sayar, A., Pierce, M., & Fox, G. (2006). Integrat-
ing AJAX approach into GIS visualization Web
services. In Telecommunications, 2006. AICT-
ICIW ‘06. International Conference on Internet
and Web Applications and Services/Advanced
International (pp. 169-169).
Schueler, T. (1999). Microbes and urban water-
sheds. Watershed Protection Techniques, 3(1),
551–596.
64
Web-Based Geospatial Services
Schuurman, N., & Leszczynski, A. (2006). Ontol-
ogy-based metadata. Transactions in GIS, 10(5),
709–726. doi:10.1111/j.1467-9671.2006.01024.x
Sikder, I., & Gangopadhyay, A. (2002). Design and
Implementation of a Web-based collaborative spa-
tial decision support system: Organizational and
managerial implications. Information Resources
Management Journal, 15(4), 33–47.
Sikder, I. U., & Gangopadhyay, A. (2003). Dis-
tributed data warehouse for go-spatial services.
In G. Grant (Ed.), ERP & data warehouse in
organizations: Issues and challenges (pp. 132-
145). IRM Press.
Sikder, I. U., & Gangopadhyay, A. (2004). Col-
laborative decision making in Web-based GIS.
In M. Khosrow-Pour (Ed.), Advanced topics in
information resources management (Vol. 3, pp.
147-162). Idea Group Publishing.
Siyuan, F., Griffiths, T., & Paton, N. W. (2007).
GML for Representing data from spatio-historical
databases: A case study. Transactions in GIS, 11(2),
233–253. doi:10.1111/j.1467-9671.2007.01043.x
Stoimenov, L., & Djordjevic-Kajan, S. (2005).
An architecture for interoperable GIS use in a
local community environment. Computers &
Geosciences, 31(2), 211–220. doi:10.1016/j.
cageo.2004.09.017
Tait, M. G. (2005). Implementing geoportals:
applications of distributed GIS. Computers,
Environment and Urban Systems, 29(1), 33–47.
Tsou, M. H., & Buttenfield, B. (2002). A dynamic
architecture for distributing geographic informa-
tion services. Transactions in GIS, 6(4), 355–381.
doi:10.1111/1467-9671.00118
Tuchyna, M. (2006). Establishment of spatial data
infrastructure within the environmental sector
in Slovak Republic. Environmental Modelling
& Software, 21(11), 1572–1578. doi:10.1016/j.
envsoft.2006.05.014
Web-Based Geospatial Services

Worboys, M., & Deen, S. M. (1991). Seman-
tic heterogeneity in distributed geographic
databases. SIGMOD Record, 20(4), 30–34.
doi:10.1145/141356.141366
Zipf, A., & Jost, M. (2006). Implementing adaptive
mobile GI services based on ontologies: Examples
from pedestrian navigation support. Computers,
Environment and Urban Systems, 30(6), 784–798.
doi:10.1016/j.compenvurbsys.2006.02.005

This work was previously published in Web Engineering Advancements and Trends: Building New Dimensions of Information
Technology, edited by Ghazi I. Alkhatib and David C. Rine, pp. 281-297, copyright 2010 by Information Science Reference
(an imprint of IGI Global).

65
66

Chapter 7
A Brief Cartography of
Smart Cameras:
Proactive Surveillance and Control

Fernanda Bruno
Universidade Federal do Rio de Janeiro, Brazil

ABSTRACT
This chapter carries out a brief cartography of the so-called “intelligent” video surveillance systems.
These systems are programmed to accomplish real time automated detection of situations considered
irregular and/or suspicious in specific environments, in order to predict and prevent undesirable events.
Three aspects of the smart cameras are focused in this cartography. First, the author explores its regime
of visibility and note how it prioritizes the capture of irregularities in the body’s movements in urban
space. Second, the author shows how the type of monitoring and profiling of bodies and behaviors in
these systems generally acts at the visible, surface and infra-individual level of human conduct. Finally,
he analyzes the temporality of smart cameras, especially in its proactive dimension that intends to
foresee and intervene, in real time, in future events. The analysis of these three aspects of the intelligent
video surveillance identifies and highlights discourses, processes and operations that are common to
the exercising of power and surveillance in contemporary societies – more specifically, those which are
included in the realm of control devices.

INTRODUCTION activity patterns exist in the monitored area, and

We care what you do, not who you are. We aim to
analyze and model the behavior patterns of people
and vehicles moving through the scene, rather than
attempting to determine the identity of people...
We are trying to automatically learn what typical
DOI: 10.4018/978-1-4666-2038-4.ch007
then have the system look for atypical patterns that
may signal a person of interest—perhaps some-
one engaging in nefarious behavior or a person
in need of help.” (James W. Davis, a computer-
science and engineering professor at Ohio State
University, describing the construction project of
smart-camera networks that automatically track
suspect behavior, 2009:30)

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A Brief Cartography of Smart Cameras
It is not uncommon for the more enlightened
statements with regard to control mechanisms to
be stated by the people who design them, rather
than by social critics. Part of what we will deal
with in this chapter has already been revealed,
albeit in a condensed and controversial way, by
the words of the inventor of a camera-based se-
curity system. This system is capable of tracking
suspect or risky behavior by integrating multiple
cameras which have high-resolution panoramic
vision, geographic information systems (GIS),
and “machine learning methods. This enables the
computer to perform the kind of visual recognition
that seems effortless for humans”1. According to
the researcher:
We envision our research being applied to com-
mon urban surveillance tasks. To date, we have
been focusing on the monitoring and analysis of
pedestrian movement and activity...Other related
applications we are considering include persistent
tracking of vehicles through complex downtown
environments and analyzing the traffic flows to
look for atypical patterns. (Davis, 2009: 30)
Such a system is part of a new “generation” of
so-called “smart” video surveillance2 that heralds
ways of monitoring behavior in an automated
fashion. In most cases, the intention is for such
cameras to recognize and differentiate regular
patterns (which are considered to be safe) of con-
duct and occupation of spaces from irregular ones
(which are categorized as suspect, dangerous, or
simply non-functional). These “smart cameras”
or “intelligent video surveillance” consist of soft-
ware3 that can be added on to cameras in order to
filter or read images according to algorithms and
point out individuals, objects, and/or attitudes that
should be the focus of attention of the “scene”.
This is carried out in accordance with pre-defined
applications in the system. For example, a body
stopped for a certain period of time very near
the safety line beside the tracks at a subway sta-
tion should be automatically pointed out on the
surveillance screen so that intervention can take
place in time to impede the possible lethal leap
of a potential suicide victim. This same device
can also automatically pinpoint the following
on the screen: an object that was left in the sta-
tion, individuals or groups of people behaving
suspiciously, bodies moving against the flow, or
any situation previously categorized as needing
to be highlighted in the machine’s and/or camera
operators’ field of attention. In technical terms:
Intelligent visual surveillance systems deal with
the real-time monitoring of persistent and transient
objects within a specific environment. The primary
aims of these systems are to provide an automatic
interpretation of scenes and to understand and
predict the actions and interactions of the observed
objects based on the information acquired by sen-
sors (Velastin & Remagnino, 2006: 1).
Computational vision, pattern analysis, ar-
tificial intelligence, and data management are
some of the subjects and techniques involved in
the construction of such systems. They rely on
applications in the most varied sectors: safety,
administration, transportation, commerce, and
entertainment, among others. The tasks they can
perform, and which are always in real time, are
also varied: detection, classification, and track-
ing of objects, people, behavior, movements, and
human activities; classification and interpreta-
tion of human behavior; detection of changes in
the context or the pattern of how individuals or
crowds occupy spaces; identification, etc. This
broad range of functions and applications is vis-
ible in products currently on the market of smart
video surveillance. ADVISOR (Annotated Digital
Video for Surveillance and Optimised Retrieval),
for example, which focuses on surveillance in the
area of public transportation, includes automatic
incident detection, content-based annotation of
video recordings, and individual and crowd
behavior pattern analysis. ISCAPS (Integrated
Surveillance of Crowded Areas for Public Secu-
67

rity) develops biometric techniques for identifying
people in populous areas. BEHAVE, DETER, and
PRISMATICA focus on the analysis and detec-
tion of suspect behavior in public places, while
CAVIAR (Context Aware Vision using Image-
based Active Recognition) is aimed at analyzing
and recognizing objects, contexts, and situations
for the purposes of urban surveillance and to allow
for the planning of marketing strategies.
What is at stake? Or better yet, what is it that
these smart cameras bring up for discussion? Most
certainly, the question is not to simply describe
this technological innovation and its effectiveness.
Rather, it is a question of analyzing these devices
as socio-technical systems (Latour, 1993 and 1999;
Lianos, 2001) that modernize a number of aspects,
discourses, and processes specific to surveillance
and the control of bodies and behaviors in the
produced and protected spaces of contemporary
societies. In the reference chapter of this section,
David Lyon (2010) discusses how our mundane,
everyday mobility has become more and more
transparent and traceable by way of technologies
that are, themselves, mobile and locational (cell
phones, GPS, RFID labels). Knowing “‘where
you are’ matters increasingly from the point of
view of contemporary regimes of governance,
whether of commerce, policing or government”
(Lyon, 2010: 4-5). With one single movement,
the potentialities of these technologies multiply
for the surveillance - “as observing, monitoring,
tracing” - of citizens, consumers, employees, etc.
Such potentialities do not only represent threats
to the privacy of individuals; they also articulate
together with ways of governing that act in fluid
and virtual spaces through “coded information
to monitor, predict and direct the behaviour of
individuals...” (Yar, 2003: 257 apud Lyon, 2010).
Intelligent video surveillance systems make up
these ways of governing. Within the realm of image
analysis and perception and attention automation,
these systems perform functions that are similar
to the technologies of observation, monitoring,
and tracing mentioned by Lyon. However, they
68
A Brief Cartography of Smart Cameras
operate by inspecting and monitoring what you
do or what you are doing when in a determined
location, rather than simply where you are.
In this chapter, we will carry out a brief cartog-
raphy of these socio-technical systems. Mimick-
ing the devices in question, which visualize their
“territories” by sweeping over surfaces – without
going into greater depth, we will map their most
prominent traits in three domains. Initially, we
will deal with the regime of visibility present in
these surveillance devices, as well as their policies
and orderings of that which is visible, especially
with regard to perception and attention. We will
note that such a regime focuses on picking up ir-
regularities in the algorithmic visualization of the
movements of bodies and their general patterns of
conduct. Next, we will attempt to show that this
regime of visibility and surveillance goes hand in
hand with a specific type of power and knowledge
over individuals’ bodies that generally acts at the
infra-individual, visible, and surface level of the
action and the conduct in order to incite, guide, or
contain these actions and conducts. Finally, we will
focus on the device’s temporality, demonstrating
that there is surveillance in force in this device
that is intended to be proactive and which ideally
coordinates the real time of the vision and the
action in the present with a memory or archive
of indexes and patterns of conduct that project a
future needing to be controlled.
The perspective that guides the overview of
these three aspects of smart video surveillance
systems identifies and highlights discourses,
processes, and operations that are characteristic of
the way power and surveillance are exercised in
contemporary societies – more specifically, those
which are included in the realm of control devices
(Deleuze, 1992; Lianos, 2001; Rose, 2000). As a
strategy of analysis, we will compare two models
of power in which surveillance and visibility play
a central role: discipline, as Foucault (1995) un-
derstood and analyzed it; and control, for which
Deleuze (1992) traces more general lines and to
which other analyses are added that have greater
A Brief Cartography of Smart Cameras
or lesser affinities (Rose, 2000; Lianos, 2001)
with this author. Such a comparison should not be
understood according to principles of exclusion
or historic overcoming. Although we recognize
the strong implications of each of these models
for their eras, there are currently zones of com-
monality between them. Discipline is not lodged
in a past that has been completely left behind, just
as control is not the current, exclusive model of
power. However, the zones of commonality do
not minimize the strong transformational process
taking place, which involves the emergence of
new surveillance processes. In the analysis of
intelligent video surveillance, we will privilege
the aspects that attest to its affinities with control
and contrasts with discipline. We will note that
elements which were already indispensable to
disciplinary power – such as the knowledge of
interiority and the processes of interiorization,
the detailed analysis of each individual, and the
careful production of their identities – certainly
do not disappear. However, these elements are no
longer at the forefront of the operations of power
and knowledge in most of these devices.
It should be noted that the cartography traced
out here is as brief as it is incomplete. We not only
leave out a series of aspects, but also remain on the
surface of the processes analyzed. We do not go
into the necessary depth that such topics deserve
in order to be understood in their complexity.
However, the device in question is well-suited for
the type of analysis that is intended to be carried
out, considering the fact that tasks which humans
are not able to fully accomplish are intended to
be delegated to the device. It thus embodies part
of the ideals of surveillance and control that are
present in the socio-technical systems of our time.
Moreover, these ideals are made especially explicit
because they deal with automated systems that
need to be programmed. Hence, our map serves
as a first reading guide of these automated and
“intelligent” surveillance devices.
VISIBILITY, PERCEPTION,
AND ATTENTION
What does a smart camera see or perceive? Or
better yet, what regime of visibility is put to work
by this video surveillance that claims to be intel-
ligent? One of the most recurring arguments in
research and industries that design such artifacts
is that they surpass the perceptive and attentional
limitations present in monitoring carried out by
conventional video surveillance systems. Human
attention is excessively limited when dealing
with the volume and monotony of surveillance
images. After only 20 minutes, the attention of
most individuals falls to a level below that which
is acceptable for an efficient monitoring activity
(Hampapur et alli, 2003 apud Gouaillier & Fleu-
rant, 2009). A guard is not capable of attentively
observing from nine to twelve cameras for more
than fifteen minutes. Some studies show that the
proportion of the number of guards in relation to
the number of screens in some cases can actually
vary from 1:4 to 1:78 (Hearing et alli, 2008 apud
Gouaillier & Fleurant, 2009).
Moreover, according to such research and
industries, this is one of the main reasons for the
inefficiency of video surveillance in providing
safety – cameras are more effective at producing
after-the-fact proof than at preventing incidents
or crimes. Therefore, adding an “intelligent” or
cognitive layer to the cameras has the intention
of automating perception and attention in a way
that points out only the indexes of threat or danger
or any other situation that is worth highlighting
in the images, in accordance with the defined
purpose. The system should automatically dif-
ferentiate and recognize that which is meaningful
in a scene and that which is irrelevant, that which
is irregular and that which is regular. Herein lies
its “intelligence”. In other words, it not only
deals with seeing or capturing, transmitting, and
archiving images, but also with “interpreting”,
in accordance with pre-defined categories, that
which is “viewed” in a scene. The system thus
69

incorporates cognitive functions – more specifi-
cally perceptive and attentional functions – in an
automated and pre-programmed way.
Let us take DETER (Detection of Events for
Threat Evaluation and Recognition) as an example.
Its focus is to detect and report unusual patterns of
pedestrian and vehicle movement in external areas.
For this system to work, it must recognize and
differentiate the objects in the scene (pedestrians
and vehicles), as well as the usual and unusual
patterns of movement of each object, in order to
recognize threats. According to a research report
(Morellas; Pavlidis; Tsiamyrtzis, 2003), DETER
has been successful in, for example, detecting the
following patterns of movement that the system
considered to be potentially alarming in certain
external areas, such as parking lots: overspeeding
vehicles, overrun of pedestrian pavement, running
pedestrians, multi-vehicle attack (vehicles enter-
ing simultaneously from different entries in the
parking lot and converging on a common parking
lot lane) and M pattern (this refers to pedestrians
following a serpentine course. It is more often
than not a suspicious pattern associated with
vehicle break-ins). Upon detecting such potential
threats, the system sounds an alarm in the guard’s
attentional field, thus allowing for intervention at
the moment of or before the predicted incident.
Through this delegation of perception and at-
tention (Latour, 1994), not only are technical and
administrative parameters of efficiency defined
and reiterated; regimes of visibility, surveillance,
and control of the space and of human behavior
are as well. A regime of visibility does not consist
so much of what is seen. Rather, it consists of that
which makes what is seen possible. Machines,
practices, discourses, and rules that are linked to
the formation of knowledge and games of power
(Foucault, 1995) participate in these conditions
of visibility (Deleuze, 1992). Intelligent video
surveillance systems make evident that which
is currently defined as needing to be visible and
noteworthy in the field of surveillance. In the
context of an excess of images and a lack of at-
70
A Brief Cartography of Smart Cameras
tention, which is the case of video surveillance
operators as well as contemporary societies in
general, there is a need to filter out that which
is relevant. In these systems, only that which is
irregular or unusual is worth seeing. The flowing
regularity of bodies in movement remains in the
attentional and perceptual background. However,
it should be added that irregularity is an index
of threat or suspicion – and that is why it should
occupy the foreground. Moreover, the irregularity
in question is no longer confused with the focal
points of the disciplinary regimes of visibility,
which read an abnormal soul through behavioral
deviation. Here, behavioral deviation is an index
of imminent risk and undesired or threatening
action. The psychological motives or traits that
underlie the action are of little importance.
This task of catching irregularities or even
foreseeing them is already present in conventional
video surveillance and in many other forms of
human or mechanical inspection. In this case,
however, paying attention to that which is irregular
is already incorporated or programmed into the
viewing device itself. Thus, the type of ordering
of that which is visible becomes more explicit
in this model of surveillance. In this ordering,
there is a particular way of observing everyday
activities. We know to what extent we can make
surveillance models correspond to regimes of
visibility of everyday activities. We also know
that disciplinary technologies make an entire field
of everyday gestures and actions, which used to
remain far from the investments of power, visible:
activities; gestures; exercises; the following of
rules, regulations, and schedules. The panoptic
scale of disciplinary surveillance is spoken of a
great deal. A whole political anatomy of detail,
which extends to the lay world and the ordinary
individual that which was already important to
theology and asceticism, is relegated to a second
level:
The mystique of the everyday is joined here with
the discipline of the minute... The meticulousness
A Brief Cartography of Smart Cameras
of the regulations, the fussiness of the inspections,
the supervision of the smallest fragment of life
and of the body will soon provide, in the context
of the school, the barracks, the hospital or the
workshop, a laicized content, an economic or
technical rationality for this mystical calculus
of the infinitesimal and the infinite (Foucault,
1995: 140).
The meticulous observation of detail makes a
series of small things, which used to be ignored
by the great moral and judicial systems because
of their relative indifference, noticeable in the
sense of becoming meaningful and visible: de-
lays, absences, interruption of tasks, negligence,
rude gestures, immodest gestures, indecencies.
This optic scale, which was at the same time
both very detailed and all-encompassing, also
depended on the constitution of meticulous files
for each individual and the sharp observation of
exams, in which the overlapping of the relations
of power and knowledge assumed “all its vis-
ible brilliance” (Idem, p.190). The daily exam
rituals in disciplinary institutions correspond to
the individual entering the field of human and
“clinical” sciences, through which he or she “may
be described, judged, measured, compared with
others, in his very individuality” (Idem, p.191).
The Intelligent video surveillance systems of
today differ from this disciplinary optic scale.
They deal with an observation that tracks bodies
and spaces with a sweeping motion. This obser-
vation is not really concerned with the minutiae
of everyday activities. It also does not concern
itself with an inventory and objectivization of
the individual differences that allow one to see
an interiority – through gestures, activities, and
performances – that is no less rich in details. These
systems deal with an algorithmic view that pays
less attention to details and depth than to superfi-
cial patterns and movements of bodies, which are
understood more in their general contours than in
their individuality. Some software even uses the
term “silhouette” in order to designate the patterns
that correspond to certain body movements.
This scrutinization of that which is visible is
intimately linked to its normative effects. Such
cameras are not so much for introducing or estab-
lishing normality in the heart of deviant popula-
tions (as in the case of some panoptic institutions).
Rather, they are for capturing or catching a break
in the flowing order. It is worth pointing out that
order and regularity are maintained. However, their
occurrence is not because of an internalization
of values that guide an identity or biographical
project, as was predicted by normalizing, disciplin-
ary jurisdictions and their regimes of visibility.
These cameras have a “formal,” “pragmatic,” and
“utilitarian” normative effect in which “seeming
normal” is more decisive than “being normal.”
Normality is, above all, a surface effect, an adher-
ence to protocols of action and the use of managed
spaces, and a belonging to a particular “ethos”
– the supermarket or bank customer, the airline
passenger, the subway user, etc. (Cf. Lianos, op.
cit., p.126). Deviance, as we have already stated,
is an index of threat or suspicion. However, the
focus is on the action more than on the agent.
What is being dealt with here is not seeing spe-
cifically or even seeing through. Rather, it has
to do with recognizing in an automated way and
seeing into the future (making behavioral patterns
correspond to patterns of pixels in the image). It
is a view that has foresight and also allows for
action in real time, thus frustrating or averting
that which was predicted. In the following topics
we will explore the emphasis given to the surface
of the bodies and their correlated categorization
of monitored behaviors in greater detail, as well
as the temporality of this proactive surveillance.
MOBILE BODIES, ACTION,
AND INFORMATION
The body has always been of interest to power.
As a vast tradition of studies on the topic reveals,
71

there are many ways for power to be interested
by bodies: the body of habits and customs (Elias,
1969); of sexuality (Brown, 1988; Foucault,
1976); of life, health, and illness (Canguilhem,
2009; Foucault, 1963; Rose, 2006); of work; of
care with daily cleanliness and beauty (Vigarello,
1987 and 2004), among others. As was seen in the
previous topic, there are also many ways for power
to “gaze upon” the body, to make it available for
visibility, and to make it an object of perception,
attention, description, knowledge, classification,
record (Crary, 1992 and 2001; Sekula, 1986).
We have recently seen a recurring diagnosis
arise in post-industrial societies with regard to
the investments of power over the body. On the
one hand, we are living an intensification of the
most varied forms of biopower (Foucault, 2004;
Agamben, 1998) – in other words, of the different
entries of the body and of life into the dominion
of the explicit calculations of power, from bio-
technologies to biosocialities (Rabinow, 1992) and
to the penetration of power in multiple spheres
of social, cognitive, affective, genetic life, etc.
(Negri & Hardt, 2000; Lazzarato, 2004). At the
same time, it has been certified that the body, in its
materiality and physical force, interests power less
and less. Power is more interested in the body’s
dimensions that are informational, immaterial,
cognitive, etc. (Lazzarato, 2004). Indeed, we see
distinct forms of relationship between body and
power that coexist in contemporary societies. This
can also be seen in surveillance technologies,
which invest in biometric and genetic data, as well
as in various ways of monitoring bodies and their
activities from a distance – be it in physical space
or in informational space. Social and affective
dimensions are not left out of these investments,
given that more and more of these surveillance
devices exist, not only in control and security
circuits, but also in entertainment, sociability,
show circuits, and, above all, in communication
technologies and networks.
In intelligent video surveillance, the monitor-
ing of bodies in spaces reflects some important
72
A Brief Cartography of Smart Cameras
aspects of the current forms of distance control
and surveillance (Lianos & Douglas, 2000). As
has been noted, bodies are inspected in their daily
mobility and movement without interrupting the
flow of displacement in urban spaces, except in the
event of suspicion. Disciplinary devices needed to
stop and confine the bodies or register them in a
predefined system of activities in order to observe
them, know them, and normalize them. Video
surveillance systems, however, are interested in
the flow of moving bodies within the course of
their lives and daily actions. In this sense, they
are part of a set of contemporary technologies that
makes it possible to track the everyday mobility
and displacement of individuals in physical space
(Wood and Graham, 2006; Lyon, 2010) and in
cyberspace (Andrejevic, 2004). As Lyon states
in the reference chapter of this section, “tracing
parolees by satellite would have sounded like
science fiction, and posting invitations by text
message to potential customers passing near to
Starbucks like a marketer’s fantasy. All these
things now happen.” (Lyon, op.cit., p.2-3).
Nevertheless, the monitoring of mobility is
not recent. The demarcation of borders and ter-
ritorial protections, the control of migrations,
and the circulation of people, goods, diseases,
etc., remind us of just how old the crossovers
are between surveillance and mobility (Salter &
Zureik, 2005; Foucault, 2007). On the one hand,
surveillance practices and techniques have his-
torically acted as mobility containment. On the
other hand, and especially since the constitution
of modern cities and states, surveillance devices
do not simply run counter to mobility. Rather,
these devices accompany the different forms of
mobility and order them, not only in the sense of
containing or immobilizing them, but also in the
sense of capitalizing on them and potentializing
them for their service. This is especially apparent
in the urbanization processes that began in the
18th century, in which surveillance and security
devices (the constitution of police forces, health
and population administration policies, etc.)
A Brief Cartography of Smart Cameras
sought not only to establish limits, borders, and
localizations, but also to permit and guarantee the
circulation of people, goods, etc. (Foucault, 2007).
This dual relationship between surveillance and
mobility has become more acute and more com-
plex in the contemporary world (Lyon, 2002 and
2009). As has already been seen, the observation
of mobile bodies by smart cameras is not simply
to know where you are (Lyon, 2010), but rather
to know what you are doing when in a determined
place. As researcher James W. Davis pointed out
at the beginning of this chapter, it is not always
“you” that is important. Although some systems
potentialize identification procedures, the first
observational focus of many of them is not on
the identity or the identification of the individuals
but on their actions and behavior. It remains to be
understood in what way this automated and “intel-
ligent” visual monitoring of bodies and behaviors
is associated with the formation of knowledge and
power over individuals.
Disciplinary systems certainly made use
of careful analyses of behaviors, actions, and
performances. However, all of this was to be
accompanied by a technology of the soul (which
was correlated to the political investments in the
body) where knowing and producing, examining
and normalizing, measuring and reforming were
sought in one single movement. This incorporeal
but very real subjectivity was, as we know, deci-
sive in circulating and articulating the effects of
a certain type of power and knowledge that made
up the disciplinary apparatus and its regime of
visibility. According to Lianos (2001), in a highly
institutionalized and media-influenced world,
the mechanisms of control organize themselves
around the action itself, instead of focusing on
the supports of a socially-filtered conscience
that sustains the action. In a model of control and
government that uses simulation and incitation
more than interdiction and coercion in order to
accomplish and obtain results, the investments and
calculations of power grant privilege to autonomy,
initiative, surpassing, and continual choice (Rose,
199). The active user and consumer are the privi-
leged subjects of contemporary control. In such
a scenario, control is often carried out by being
mixed with the very contexts of consumption and
use of goods and services and with the very systems
and technologies of training and coordinating of
individual choices, thus providing protocols for
action and contexts for interaction. In a word,
technologies of empowerment of individual choice
and technologies of control blend more and more
into contemporary managed spaces (Rose, 2000;
Lianos & Douglas, 2000).
In a parallel manner, a whole culture and rheto-
ric of risk and security (Garland, 2001) permeates
the ordering of contemporary socio-technical
spaces, bodies, and environments. It is not enough
to just incite the action. It is also necessary to
foresee and avoid all behavior that represents risk
or danger. This focus on the action and behavioral
patterns in order to construct algorithms capable
of foreseeing events is recurrent in other control
devices in the fields of health, crime, and labor.
Such devices act more on the illness than on those
who are ill (Rose, 2006), on the crime rather than
the criminal (Garland, 2001), on the performance
rather than the training (Ehrenberg, 1999).
Thus, the functions of management, control,
and security overlap more and more in contem-
porary socio-technical environments. From shop-
ping centers to subways and electronic Internet
sites, comfortable and safe conditions for actions
and choices are created, as are mechanisms for
predicting and preventing actions considered to
be suspicious or risky.
All of these elements are in some way present
in intelligent video surveillance systems. They
also aim to be diverse in their application by
monitoring and intervening in behaviors and ways
of occupying spaces with the most varied ends:
consumption, commerce, management, security.
Among their various applications are: counting
clients; analyzing consumption habits, patterns of
movement, and transportation; identifying intru-
sions in controlled or prohibited zones or areas of
73

risk; detecting abandoned objects and suspicious
behavior; monitoring and analyzing populational
or crowd flows (crowd density estimation, face
detection and recognition in crowds, and crowd
behavior monitoring). The diverse nature of smart
cameras is explicit in the advertising materials
of certain manufacturers: “Whether the need is
real-time security and surveillance or gathering
business intelligence, ObjectVideo is the key to
enabling truly intelligent solutions” (http://www.
objectvideo.com/solutions/).
In the great majority of cases, what is being
dealt with is the monitoring of routine behaviors
in order to extract patterns with regard to the
daily gestures of bodies located in their ordinary
environments so as to incite the bodies to act in
a certain way (consumption, movement, atten-
tion) or to recognize and impede irregular and/
or suspect patterns of conduct. With its differing
visual and algorithmic layers, a single system
monitors, analyzes, categorizes, and intervenes
in the behaviors. This process of categorization
of bodies, movements, and routine gestures is
intimately linked to the ordering and codifying
of urban spaces, which in turn are converted
into “automated socio-technical environments”
(Lianos & Douglas, 2000).
Smart cameras and their sensors possess a
type of power and knowledge over bodies that
transforms their actions within a physical space
into information (Manovich, 2006). This informa-
tion permits the extraction of regular and irregular
conduct patterns and thus makes interventions,
planning, and calculations on the ordering of
bodies and their flows possible. Intelligent video
surveillance, in a way, goes hand in hand with
the so-called intelligent environments that add
informational layers to spaces. This therefore
broadens the possibilities of interacting with
the bodies present and allows them to be called
upon in a contextual manner by the surrounding
environment itself. In the words of Manovich,
74
A Brief Cartography of Smart Cameras
By tracking the users – their mood, pattern of
work, focus of attention, interests, and so on
– these interfaces acquire information about
the users, which they then use to automatically
perform the tasks for them. The close connection
between surveillance/monitoring and assistance/
augmentation is one of the key characteristics
of the high-tech society (Manovich, 2006: 222).
In this way, intelligent video surveillance pro-
duces a typology or taxonomy of the daily move-
ment of bodies in spaces, codifying their patterns
of conduct and allowing for the calculating and
planning of actions on these bodies. It is part of
the capture model proposed by Agre (1994) and
revisited by Kitchin & Dodge (2006):
The capture model acknowledges that the mecha-
nisms by which information is gathered is increas-
ingly an integral part of the same system that they
seek to monitor and regulate (e.g., a computer
operation system logs its own use by an individual)
and that these mechanisms in turn re–define and
re–configure that system (e.g. change workplace
practices), quite often in real–time. Agre (1994)
argues that these modes of informational capture
is possible because a grammar of action (for-
malized rules) has been imposed on a system. A
grammar of action is a means of systematically
representing aspects of the world, an organized
language for modeling human behaviors (Kitchin
& Dodge, 2006: 6).
But of what does this typology consist? In what
way is it similar to or different from the modern
techniques of categorizing corporal data? If we
compare this typology to the known historical
methods of classifying corporal characteristics
that also used vision technologies, such as pho-
tography, we notice that modernity’s inventories
of corporal patterns were focused on specific traits
and details. We need only remember the catalogs
of ear types designed by Bertillon (Sekula, 1986).
As in that time, we are now dealing with the exer-
A Brief Cartography of Smart Cameras
cise of political and scientific power over bodies
through a classification and recording system that
converts the image into convincing information
and permits the analysis of indications and the
recognition of corporal patterns, their norms,
and their deviances. However, in the place of
the rows of ears, little fingers, and forearms,
the taxonomies of intelligent video surveillance
depend on algorithms that “represent” patterns
of conduct and movement of bodies in a given
context. Instead of categorizing specific traits
and fragmented corporal details, the emphasis is
on contexts of the broadest gestures and action
of the bodies. These contexts are then converted
into information and detected by smart camera
sensors and algorithms.
Modern taxonomies were also focused on the
erudite and meticulous eyes of specialists and
authorities who were to combine instruments and
knowledge, such as photography, physiognomic
description, anthropometry, and statistics. The
categorizations of smart cameras, in their turn, are
to be recognized in an automated fashion by the
machine itself, which depends on computational
vision, pattern analysis, artificial intelligence, and
data management. Another significant difference
is the main purpose of each of the systems. The
visual diagnosis of a criminal personality hidden
within bodily details and measurements (present,
above all, in the composite portraits of Galton)
and the identification and recognition of specific
criminals (the first objective of Bertillon’s records)
in the heart of an anonymous mass that suddenly
appeared in the confused territory of cities oc-
cupied the forefront of concerns and purposes of
the modern typology of bodies, their measure-
ments, and images. In the typologies with which
we are dealing here, as has already been seen, we
are no longer dealing with diagnosing a criminal
personality through bodily images and details.
Rather, we are concerned with foreseeing – on
the surface of the image and in the movements of
the bodies in a given context – a possible crime
or incident that is about to happen. In one case,
the body-image is the index of something that lies
beneath or that comes before; in the other, it is an
index of an action that is yet to come.
The differences between these indexing mod-
els and their respective temporalities are better
understood if we consider the proactive nature of
intelligent video surveillance, as will be discussed
in the next topic.
REAL TIME, PREDICTION,
AND PROACTIVITY
The last trait in relief in our cartography is the
temporality of smart cameras. In the descriptions
and advertisements of these devices, the main
emphasis is given to the “real time” of the moni-
toring, detecting, and intervening in the contexts
under surveillance. In other words, in addition to
transmitting and observing the image in real time,
which is already present in video surveillance
controlled by human operators, the intelligent
systems operate an automated and “instant”
analysis of the scene under observation. This
allows for any intervention to occur at the very
moment in which the incident is taking place. In
theory, possible delays or lapses are eliminated,
be they because of human perception or attention
or because of action at the opportune moment. It
should be noted that real time here is not that of the
observation, detection, or action taken as isolated
elements. Rather, it is that of the almost immediate
or instantaneous connection between these three
normally diachronic, separate “moments,” with
the interval between them being reduced to the
extreme. The image, which already has a real time
of observation, now has a real time of reaction as
well. In order to understand the reach or extent
of this continuously alert and active present, it is
necessary to understand the past or the memory,
which serves as a parameter, as well as the future,
which is projected in the present.
The memory of the device is not simply, as
in “conventional” video surveillance, its ability
75

to register. Smart cameras are equipped with an
algorithmic memory, which consists of a repertoire
of patterns that the camera should recognize as
regular – and thus remain indifferent – or as ir-
regular – and therefore intervene. In this way, the
memory serves as a parameter or a framework for
the analysis of the scene being monitored. If we
consider the device’s whole sequence of elabora-
tion, this “intelligent” memory, which serves as a
guide for the camera’s “glance” and its perceptive
and attentional filters, is constituted through the
observation and archiving of an immense mass of
images submitted for analysis in order to extract
patterns. These archives can be elaborated in or-
der to build an intelligent system, or they can be
put together by reusing and analyzing previous
records, which in large part do not have any use
or end. The gigantic and humanly unsearchable
archives of video surveillance images become
a useful memory, according to the creators of
intelligent video surveillance. As researchers
and manufacturers Frost & Sullivan state, video
analytics “is the next step to video surveillance
where the intelligence of the system is utilized
to enable the end user to identify key trends and
significant changes in recorded images. The end
user will be alerted when there is an unusual event.
Video analytics is the key to help end users with
the abundance of recorded information.” (http://
www.frost.com/prod/servlet/svcg.pag/AEID)
The greatest promise of this memory or of
these intelligent archives, as can already be noted,
is not to save and recover the past. Rather, it is
to anticipate and avoid undesirable future events,
above all when surveillance is for the purpose of
security. We are talking about a record of patterns
that is to serve as an index of the future. Interven-
tion in real time is, in this case, an action in the
present in order to avoid the occurrence of another
action in the future. It is for this reason that we
suggest discussing proactive video surveillance
here, which should come before the foreseen action
and, in this case, avert it. We now return to the
type of indexing present in this model of surveil-
76
A Brief Cartography of Smart Cameras
lance. It does not have to do with the index of the
past, nor is it solely the index of the present and
the force of evidence that is characteristic of the
real time of the transmission and observation of
the image. It also mainly deals with the index of
an occurrence which is yet to come. An extreme
example of this proactivity is provided by the
FAST (Future Attributable Screening Technol-
ogy4) software, which uses sensors and starts at a
distance of around two meters in order to capture
a series of body signals of individuals, such as
skin temperature, blood-flow patterns, perspira-
tion, and heart and breathing rates, which serve
as indications of a hostile act yet to come.
This index and this “yet to come” have a curi-
ous, paradoxical, disturbing meaning, however.
The index of the past has the facticity and relative
stability of “something happened” (Barthes, 2006),
and the index of the present has the supposed live
referential that “something is” or “something is
happening.” The index of the future, however, is
uncertain by nature. In this case, it is also “unveri-
fiable”, given that the proactive aspect should im-
pede or avert that which the body-image indicates
is about to happen. The uncertainty and the many
possible ramifications of what could happen in the
course of time are therefore reduced. However, this
reduction does not take place because of accuracy
in the prediction. Rather, it occurs through the
effectiveness or performance of the anticipation.
We will never know if the body that stops for too
long at the edge of the subway tracks would have
jumped to its death or not because the future is
anticipated. In this case, the future is anticipated
in order to be averted. Even though it is uncertain
by nature, it becomes certain at the moment that
it is foreseen, and yet it cannot be carried out.
The same is true for the repertoire of suspicious
behaviors that serves as an index of crimes or
infractions that are yet to occur or that perhaps
will not be carried out. They become suspect at
the moment they are predicted, thus transforming
the body-images into potential criminals.
A Brief Cartography of Smart Cameras
The future projects itself onto the present and
thus limits the stock of possibilities that belong
to the present. When smart cameras are focused
on optimizing the use of space, consumption, or
the efficiency of services, this colonization of the
present by anticipating the future is less ostensible.
Instead of averting an undesired action, there is
an attempt to incite desirable behavior in specific
contexts. At any rate, and above all in the case of
proactive surveillance with the purpose of security,
the “birth of the management of probability as
a possible reality as valid as the present reality
places the problematic of power in a qualitatively
different context; it frees power from the limit
of a reference to the occurrence and allows it to
make that which has yet to exist (and which will
perhaps never exist) into a ‘happening of power’”5
(Lianos, 2001: 135).
Smart cameras, through the combining of real
time, foresight, and proactivity, produce a type of
power over bodies, images, and times that is guided
by an indexing of suspicion and the proactivity
of surveillance. They therefore come to augment
the already abundant assortment of predictive
and proactive surveillance devices, such as data
monitoring and profiling, in the most diverse fields:
consumption, health, security, efficiency, etc. Video
surveillance surpasses the dominion of the proof of
that which happened and promises to proactively
intervene in that which has not yet occurred.
REFERENCES
Agamben, G. (1998). Homo Sacer: Sovereign
Power and Bare Life. Stanford, CA: Stanford
University Press.
Agre, P. E. (1994). Surveillance and capture: two
models of privacy. The Information Society, 10(2),
101–127. doi:10.1080/01972243.1994.9960162
Andrejevic, M. (2004). Monitored mobility in the
era of mass customization. Space and Culture,
6(2), 132–150. doi:10.1177/1206331203251256
Barthes, R. (2006). La Chambre Claire: Note Sur
LA Photographie. Paris: Gallimard.
Brown, P. (1988). The Body and Society. New
York: Columbia University Press.
Canguilhem, G. (2009). Le normal et le
pathologique. Paris: PUF.
Crary, J. (1992). Techniques of the observer.
Cambridge, MA: MIT Press.
Crary, J. (2001). Suspensions of Perception.
Cambridge, MA: MIT Press.
Davis, J. W. (2009). Research Report Article: Eye
in the Sky. OSU. The Alumni Magazine, (May/
June): 30.
Davis, J. W. (n.d.). James W. Davis Home Page,
from http://www.cse.ohio-state.edu/~jwdavis/
Deleuze, G. (1992). Postscript on the Societies
of Control. October, 59, Winter, 3-7.
Ehrenberg, A. (1999). Le culte de la performance.
Paris: Hachette.
Elias, N. (1969). The Civilizing Process: Vol. I.
The History of Manners. Oxford: Blackwell.
Eye in the Sky. (2009). Research Report Article,
May/June, 30. Retrieved from www.ohiostatea-
lumni.org
Foucault, M. (1963). Naissance de la clinique.
Paris: PUF.
Foucault, M. (1976). Histoire de la sexualité I - La
Volonté de savoir. Paris: Gallimard.
Foucault, M. (1995). Discipline & Punish: The
Birth of the Prison. New York: Vintage Books.
Foucault, M. (2004). Naissance de la biopolitique.
Paris: Gallimard/Seuil.
Foucault, M. (2007). Security, Territory, Popula-
tion: Lectures at the Collège de France 1977-78.
London: Palgrave Macmillan. Garland, D. (2001).
The Culture of Control. Chicago: University of
Chicago Press.
77

Gouaillier, V., & Fleurant, A.-E. (2009). La vi-
déosurveillance intelligente: promesses et défis:
rapport de veille technologique et commerciale.
Montréal: Centre de recherche informatique de
Montreal (CRIM)/ Technopôle Défense et Sé-
curité, 2009.
Hampapur, A., Brown, L., Connell, J., Pankanti, S.,
Senior, A., & Tian, Y. (2003). Smart Surveillance:
Applications, Technologies and Implications.
In Joint Conference of the Fourth International
Conference on Information, Communications &
Signal Processing and Fourth Pacific-Rim Con-
ference on Multimedia, 2 (pp. 1133- 1138). New
York: December.
Hardt, M., & Negri, A. (2000). Empire. Cambridge,
MA: Harvard University Press.
Hearing, N., Venetianer, P. L., & Lipton, A. (2008).
The evolution of video surveillance: an over-
view. Machine Vision and Applications, 19(5-6),
279–290. doi:10.1007/s00138-008-0152-0
Kitchin, R., & Dodge, M. (2006). Software and
the mundane management of air travel. First
Monday, 11(Nr. SI-7).
Latour, B. (1993). We Have Never Been Modern.
Cambridge, MA: Harvard University Press.
Latour, B. (1994). On technical mediation:
philosophy, sociology, genealogy. Common
Knowledge, 2(3).
Latour, B. (1999). On recalling ANT. In Law, J.,
& Hassard, J. (Eds.), Actor Network Theory and
After. Oxford: Blackwell.
Lazzarato, M. (2004). Les revolutions du capital-
isme. Paris: Les Empêcheurs de Penser en Rond.
Lianos, M. (2001). Le nouveau contrôle social.
Paris: L’Harmattan.
Lianos, M., & Douglas, M. (2000). Dangerization
and the end of deviance. The British Journal of
Criminology, 40, 261–278.
78
A Brief Cartography of Smart Cameras
Lyon, D. (2002). Surveillance Studies: under-
standing visibility, mobility and the phenetic fix.
Surveillance & Society, 1(1), 1–7.
Lyon, D. (2010). Why where you are matters:
mundane mobilities, transparent technologies,
and digital discrimination. In Firmino, R., Duarte,
F., & Ultramari, C. (Eds.), ICTs for Mobile and
Ubiquitous Urban Infrastructures: Surveillance,
Locative Media and Global Networks. Hershey,
PA: IGI Global.
Manovich, L. (2006). The poetics of augmented
space. Visual Communication, 5(2), 219–240.
doi:10.1177/1470357206065527
Morellas, V., Pavlidis, I., & Tsiamyrtzis, P. (2003).
DETER: Detection of events for threat evaluation
and recognition. Machine Vision and Applications,
15(1), 29–45. doi:10.1007/s00138-003-0121-6
Rabinow, P. (1992). Artificiality and enlighten-
ment: from sociobiology to biosociality. In Crary,
J., & Kwinter, S. (Eds.), Incorporations (pp.
234–252). New York: Zone Books.
Rose, N. (2000). Power of Freedom: Reframing
Political Thought. Cambridge, UK: Cambridge
University Press.
Rose, N. (2006). Politics of life itself: biomedicine,
power and subjectivity in the twenty-first century.
Princeton, NJ: Princeton University Press.
Sekula, A. (1986).The Body and the Archive.
October, 39, 3-64.
Velastin, S. A., & Remagnino, P. (Eds.). (2006).
Intelligent distributed video surveillance systems.
London: Institution of Engineering and Technol-
ogy.
Vigarello, G. (1987). Le Propre et le Sale.
L’hygiène du corps depuis le Moyen Age. Paris:
Seuil.
Vigarello, G. (2004). Histoire de la beauté, le
corps et l’art d’embellir de la renaissance à nos
jours. Paris: Seuil.
A Brief Cartography of Smart Cameras
Wood, D., & Graham, S. (2006). Permeable bound-
aries in the software-sorted society: surveillance
and the differentiation of mobility. In Sheller, M.,
& Urry, J. (Eds.), Mobile Technologies of the City
(pp. 177–191). London: Routledge.
ADDITIONAL READING
Agre, P. E. (2001). Changing Places: Contexts
of Awareness in Computing. Human-Com-
puter Interaction, 16, 177–192. doi:10.1207/
S15327051HCI16234_04
Andrejevic, M. (2003). Tracing space: moni-
tored mobility in the era of mass customiza-
tion. Space and Culture, 6(2), 132–150.
doi:10.1177/1206331203251256
Andrejevic, M. (2005). Nothing comes between
me and my CPU: smart clothes and “ubiquitous”
computing. Theory, Culture & Society, 22(3),
101–119. doi:10.1177/0263276405053723
Beer, D. (2007). Thoughtful territories: imagining
the thinking power of things and spaces. City, 11(2),
229–238. doi:10.1080/13604810701395845
Bogard, W. (1996). The Simulation of Surveil-
lance: Hypercontrol in Telematic Societies. New
York, NY: Cambridge University Press.
Cole, S. A. (2001). Suspect Identities: A History
of Fingerprinting and Criminal Identification.
Cambridge: Harvard University Press.
Crang, M., & Graham, S. (2007). Sentient cities
ambient intelligence and the politics of urban space.
Information Communication and Society, 10(6),
789–817. doi:10.1080/13691180701750991
Dodge, M., & Kitchin, R. (2005). Code and the
transduction of space. Annals of the Association
of American Geographers. Association of Ameri-
can Geographers, 95(1), 162–180. doi:10.1111/
j.1467-8306.2005.00454.x
Dodge, M., & Kitchin, R. (2005). Codes of life:
Identification codes and the machine–readable
world. Environment and Planning. D, Society &
Space, 23(6), 851–881. doi:10.1068/d378t
Dodge, M., & Kitchin, R. (2007). Outlines of a
world coming into existence: pervasive comput-
ing and the ethics of forgetting. Environment and
Planning. B, Planning & Design, 34(3), 431–445.
doi:10.1068/b32041t
Dodge, M., & Kitchin, R. (2009). Software, ob-
jects and home space. Environment & Planning
A, 41(6), 1344–1365. doi:10.1068/a4138
Galloway, A. (2004). Intimations of ev-
eryday life: Ubiquitous computing and the
city. Cultural Studies, 18(2), 384–408.
doi:10.1080/0950238042000201572
Graham, S. (2005). Software-sorted geographies.
Progress in Human Geography, 29(5), 562–580.
doi:10.1191/0309132505ph568oa
Greenfield, A. (2006). Everyware: The Dawning
Age of Ubiquitous Computing. Berkeley, CA:
Peachpit Press.
Law, J., & Hassard, J. (Eds.). (1999). Actor-
Network Theory and After. Oxford: Blackwell.
Lessig, L. (1999). Code. New York: Basic Books.
Lovink, G., & Gerritzen, M. (Eds.). (2002). Mobile
Minded. Corte Madera, CA: Gingko Press.
Lyon, D. (Ed.). (2003). Surveillance as social
sorting: Privacy, risk, and digital discrimination.
London: Routledge.
Lyon, D. (2003). Surveillance After September
11. Cambridge, UK: Polity Press.
McCue, C. (2005). Datamining and predictive
analytics: battlespace awareness for the war on ter-
ror. Defense Intelligence Journal, 13(1/2), 47–63.
79

McCullough, M. (2004). Digital Ground: Archi-
tecture, Pervasive Computing and Environmental
Knowing. Boston, MA: MIT Press.
Mitchell, W. (2004). ME++: The Cyborg Self and
the Networked City. Cambridge, MA: MIT Press.
Monahan, T. (forthcoming). Dreams of control at a
distance: Gender, surveillance, and social control.
Cultural Studies - Critical Methodologies, 8(4).
Shields, R. (2004). Visualicity. Visual Culture in
Britain, 5(1), 23–36.
Virilio, P. (1994). The Vision Machine. Bloom-
ington: Indiana University Press.
Wright, D., Gutwirth, S., Friedewald, M., Vildjiou-
naite, E., & Punie, Y. (Eds.). (2008). Safeguards in
a World of Ambient Intelligence. London: Springer.
KEY TERMS AND DEFINITIONS
Control: In this paper, the term control is
used in the sense proposed by Gilles Deleuze: a
system of power characteristic of contemporary
societies, which differs from the system of disci-
plinary power, as conceived by Michel Foucault
in modern societies. In addition to Gilles Deleuze,
this text uses the contributions of Michalis Lianos
on social control in Automated Socio-Technical
Environments.
Human Behaviour Monitoring: Regular,
systematic and focused monitoring of human
behavior in order to control, influence or prevent
specific actions and/or events.
Intelligent Video Surveillance: Intelligent
visual surveillance systems are defined as real-
time monitoring of persistent and transient objects
within a specific environment. The main objec-
tives of these systems are to provide an automatic
interpretation of scenes and to understand and
predict the actions and interactions of the observed
This work was previously published in ICTs for Mobile and Ubiquitous Urban Infrastructures: Surveillance, Locative Media
and Global Networks, edited by Rodrigo J. Firmino, Fabio Duarte and Clovis Ultramari, pp. 257-271, copyright 2011 by
Information Science Reference (an imprint of IGI Global).
80
A Brief Cartography of Smart Cameras
objects based on the information acquired by sen-
sors (Reference: Velastin & Remagnino, 2006: 1).
Proactive Surveillance: Procedures and/or
technologies that seek to anticipate danger or risk
in order to act preventively to avoid the events
anticipated to take effect.
ENDNOTES
1
For a technical description of the project, Cf.
Karthik Sankaranarayanan; James W. Davis
A Fast Linear Registration Framework for
Multi-Camera GIS Coordination. IEEE In-
ternational Conference on Advanced Video
and Signal Based Surveillance, September
2008.
2
Video surveillance systems are usually di-
vided into three categories or generations:
operator-controlled video surveillance, basic
automated video surveillance, and smart
video surveillance.
3
Such software, usually called video analytics,
seeks to automatically analyze and identify
specific objects, behavior or attitudes in the
video sequences. The software transforms
the video into data that will be transmitted
or archived to allow the video surveillance
system to act, be it by activating a mobile
camera in order to obtain more precise data
of the scene or by simply sending an alert
signal to the system’s operators.
4
Software that is part of the Project Hostile
Intent, developed by The Human Factors
Division of America’s Department of Home-
land Security (DHS). Cited in the article
“If looks could kill”, The Economist print
edition, Oct 23rd 2008.
5
In the original, in French, “événement de
pouvoir” (Lianos, 2001, p. 135).
 81

Chapter 8
GIS, Grid Computing
and RFID in Healthcare
Information Supply Chain:
A Case for Infectious Disaster Management

Yenming J. Chen
National Kaohsiung First University of Science and Technology, Taiwan

EXECUTIVE SUMMARY
Several healthcare disasters have occurred in the past decade, and their occurrence has become more
frequent recently due to one natural catastrophe after another. The medical application requirement for
such a disaster management system includes effective, reliable, and coordinated responses to disease
and injury, accurate surveillance of area hospitals, and efficient management of clinical and research
information. Based on the application requirements, this case study describes a grid-based system in a
health information supply chain that monitors and detects national infectious events using geographi-
cal information system (GIS), radio-frequency identification (RFID), and grid computing technology.
This system is fault-tolerant, highly secure, flexible, and extensible, thus making it capable of operation
in case of a national catastrophe. It has a low cost of deployment and is designed for large-scale and
quick responses. Owing to the grid-based nature of the network, no central server or data centre needs
to be built. To reinforce the responsiveness of the national health information supply chain, this case
study proposes a practical, tracking-based, spatially-aware, steady, and flexible architecture, based on
GIS and RFID, for developing successful infectious disaster management plans to tackle technical is-
sues. The architecture achieves a common understanding of spatial data and processes. Therefore, the
system can efficiently and effectively share, compare, and federate—yet integrate—most local health
information providers and results in more informed planning and better outcome.

DOI: 10.4018/978-1-4666-2038-4.ch008

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 GIS, Grid Computing and RFID in Healthcare Information Supply Chain
ORGANIZATION BACKGROUND
The outbreak of severe acute respiratory syndrome
(SARS) in Southeast Asia in 2003 had a serious
impact and proved disastrous for the entire society,
healthcare practitioners, healthcare institutions
and public works personnel worldwide (Esswein
et al., 2004). In recent years, the numerous natural
disasters have always entailed a massive preva-
lence of contagious diseases. For example, after a
devastating earthquake in Haiti in 2010, there was
reported an outbreak of a cholera-like disease that
had killed hundreds of people. In order to better
manage such disasters in the future, much atten-
tion was given to interoperability of a nationwide
health information network (Chau & Yip, 2003).
Inherent in the massive collection of data, there is
expected to be trade-offs between quick response
and accuracy that arises with mass surveillance
systems. The application requirements for such
a system include an effective and coordinated
response to disease and injury, accurate surveil-
lance of area hospitals, and efficient management
of clinical and research information.
A nationwide health information network or
supply chain can be extremely complex, since
it needs to integrate geographically-distributed
healthcare providers and other units with distinct
functions and mutual dependencies. In this case
study there has been discussed an enhanced health
information network that dealt with natural disaster
caused by mass epidemic outbreak.
Due to the dynamic nature of propagation of
epidemics, many researches investigate, follow-
ing various dynamic approaches, the causes and
consequent behaviour patterns in the outbreak of
infectious diseases (Forys, 2002; Scheffer et al.,
2001). System dynamics modelling allows the inte-
gration of multiple political, environmental, social
and structural variables into a single model. It also
analyses the behaviour of all the variables in the
system, allowing policies to be tested repeatedly
(Forrester, 1961). The system dynamics model-
ling methodology has been applied to the health
82
sector many a time and proven itself in resolving
complex, systemic issues (Yousefi & Lauridsen,
1998; Flessa, 1999; Ritchie-Dunham & Galv´an,
1999; Lane, Monefeldt & Rosenhead, 2000).
This case study tries to reinforce the respon-
siveness of the national health information supply
chain with a practical, tracking-based, spatially-
aware, steady-to-use, and flexible architecture
for developing successful infectious disaster
management plan and implementing GIS and
RFID to tackle technical issues. The proposed
architecture will achieve a common understanding
of spatial data and processes, which will enable
this system to efficiently and effectively share,
compare, and federate yet integrate most of the
local health information providers and results for
more informed planning and better outcomes.
SETTING THE STAGE
Early intervention is a key to stop a massive out-
break of diseases and new technologies can help
in this. A mass epidemic outbreak caused by a
natural disaster drew much attention recently. In
response to such events, bodies related to biomedi-
cal, public health, defense, as also intelligence
communities, are developing new approaches
for real-time disease surveillance in an effort to
augment existing public health surveillance sys-
tems. The term ‘syndromic surveillance’ refers
to methods relying on detection of clinical case
features that are discernible before confirmed
diagnoses are made (Forslundet al., 2004). In
particular, even before a laboratory confirms an
infectious disease, sick people may show certain
behavioural patterns, symptoms or signs, or there
may be certain laboratory findings, that can be
tracked through a variety of data sources. New
information infrastructure and methods to support
timely detection and monitoring, including the
discipline of syndromic surveillance, are evolving
rapidly (Homer et al., 2004; Hoard et al., 2005).
GIS, Grid Computing and RFID in Healthcare Information Supply Chain
CASE DESCRIPTION
Technology Concerns
The applications of spatial data analysis and min-
ing are widespread. But in a syndromic surveil-
lance system, geographical data have not been
associated with finding local clusters of diseases.
Environmental health studies often rely on GIS
software to map areas of potential exposure and
locate where people live in relation to these areas.
An overview of geographical health issues has
been given by Boulos (2004, 2005). Information
on long-term entomological and epidemiological
surveillance conducted in an infected area should
be manipulated and presented using spatial analy-
sis tools available in GIS. This is important to
predict impending epidemics in order to use limited
resources in a cost-effective and efficient manner
and to control any epidemic of infectious diseases.
Technology Components
A geographical information system (GIS) can
be applied to develop maps, presenting spatial
distribution of infected areas, infected popula-
tions and infected factors which are important for
spatial and temporal dynamics of transmission of
infectious diseases (Ramljaket al., 2003). A few
studies (Reiskind et al., 2001) discussed the use
of GIS and spatial analysis for spread of dengue
around the world.
Radio frequency (RF) tags also may be attached
to responding personnel, vehicles, equipment or
supplies. An RFID system represents a major
advantage over handwritten or bar-code-based
identification systems, because it supports the
simultaneous collection of information from
thousands of RF-tagged people or objects and
does not depend on line-of-sight contact between
the receiver and the tag. Moreover, some of
these systems enable data stored on the RF tags
to be updated or expanded (e.g., permitting the
revision of triage scores on RF tags attached to
individual victims). In addition, RF-tagged victims
or resources may be located via GIS to produce a
real-time map of the entire population (Yu & Lim,
2004). RFID technologies are found to be more
reliable and durable than bar codes in different
application and functionality (Li et al., 2004).
Grid technologies have been widely adopted in
scientific and technical computing (Foster, 2004).
Grid technologies and infrastructures support the
sharing and coordinated use of diverse resources
in dynamic, distributed virtual organizations
(Chervensak, 2001). The creation of virtual com-
puting systems, from geographically distributed
components operated by distinct organizations
with differing policies, is sufficiently integrated
to a health information supply chain. A grid may
be explained as a set of heterogeneous systems
federated over a wide-area network (Foster et al.,
2001). In contrast to the general Internet, such
systems are usually interconnected using special
high-speed, wide-area networks in order to get the
bandwidth required for their applications.
While the notion of grid computing is simple
enough, the practical realization of grids poses a
number of challenges. Key issues that need to be
dealt with are security, heterogeneity, reliability,
application composition, scheduling, and resource
management (Buyya, 2002). We try to elaborate
cross-platform federation without sacrificing
flexibility, scalability, reliability and extensibility.
Management and
Organizational Concerns
There are several concerns involved in the man-
agement of infectious disease control. A suitable
infectious disaster control and management sys-
tem has to be versatile. Disasters are commonly
defined as emergencies of severity and magnitude
resulting in deaths, injuries, illness, and/or prop-
erty damage that cannot be effectively managed
by the application of routine procedures or re-
sources (Landesman,2001). A properly-designed
emergency management system involves a set of
83
 GIS, Grid Computing and RFID in Healthcare Information Supply Chain
arrangements, procedures, resources, personnel
and relations that can reduce the impact of hazards,
emergencies and disasters (Abrahams, 2001).
Natural disasters such as floods, earthquakes
and hurricanes pose a greater risk to populations
than other disaster events (Landesman, 2001).
Since each disaster has different characteristics,
it is important for emergency planners to take an
adjusted approach to disaster preparedness. An
enhanced health information network dealing
with natural disasters caused by mass epidemic
outbreak is discussed in this case study. The Severe
Acute Respiratory Syndrome (SARS) epidemic,
from November 2002 to June 2003, came with
much public attention and left swiftly, resulting
in more than 8,000 probable cases worldwide and
774 deaths (WHO, 2003b).
The World Health Organization (WHO) praised
the role of GPHIN (Global Public Health Intel-
ligence Network) for early detection of SARS.
GPHIN provided some of the earliest alerts to
the November outbreak in China (WHO, 2003a).
A part of WHO’s Global Outbreak Alert and
Response Network, it was developed and is oper-
ated by Health Canada’s Centre for Emergency
Preparedness and Response. It is essentially an
Internet crawler specializing in detecting news
articles indicating unusual events relevant to
public health. The network continually scans
more than 400 international sources for news of
any outbreaks of 31 communicable diseases, as
well as articles about natural disasters and drug-
resistant pathogens, rather than relying on official
reports from government sources.
CURRENT CHALLENGES
FACING THE ORGANIZATION
Rapid detection of outbreaks caused by natural
disasters requires immediate response to minimize
health issues and other social impacts. Public
health departments routinely conduct surveillance
of naturally occurring diseases within the commu-
84
nity, with the aim of early detection and treatment.
An effective syndromic surveillance system can
provide timely information to detect outbreak and
adverse public health events by aiding in earlier
intervention (Luckett & Grossenbacher, 2003).
The requirement for availability and effec-
tiveness of syndromic surveillance data locally,
regionally, and nationally argues for a common
infrastructure that would enable use at all levels
of decision making. This argues very strongly for
a common communication vehicle for syndromic
surveillance. The resulting network of surveillance
systems needs to be able to be queried for infor-
mation about what they are seeing both locally
and regionally, and also beyond. This requires a
significant level of interoperability and data inte-
gration, including a common functional behaviour.
The emergence of electronic healthcare data
systems has resulted in the ability to monitor
health data almost in real time, presenting a new
opportunity to monitor the health status of the
community. Implementing syndromic surveillance
based on automated acquisition of clinical data
requires both the development of secure, reliable
information systems and the use of those systems
in public health practice. The rapid collection
and dissemination of information, the sharing of
that information and a comprehensive reporting
mechanism could be thought of as permitting
earlier intervention in outbreaks. Being more sen-
sitive to underlying changes in the data, outbreak
signals could be detected sooner, thus mitigating
the effects of large scale disease outbreak.
As highly contagious diseases pose many
challenges, having efficient monitoring devices
for patient-tracking is essential to the success of
early intervention in epidemic surveillance. Wire-
less communication may support the widespread
and instantaneous tracking of victims or resources
through barcode or RFID systems. It may eliminate
manual data entry and introduce the potential for
automated processes to increase productivity even
under difficult working environments.
GIS, Grid Computing and RFID in Healthcare Information Supply Chain
SOLUTIONS AND
RECOMMENDATIONS
An integrated information infrastructure with a
nationwide health information network would
provide the means to access all available clinical
information, at the corporate, regional, national or
even international level, and to meet challenges
posed by patient mobility and the fact that an indi-
vidual’s health data may reside at many geographi-
cally dispersed information systems (Forslund &
Kilman, 2000). In addition to providing spatially
distributed information, the system may be a valu-
able tool in basic and clinical research, medical
decision making, epidemiology, evidence-based
medicine, and in formulating public health policy.
Such a system is fundamentally about bringing
timely health-related information to, and aiding
communication among, those making health deci-
sions for themselves, their families, their patients
and their communities (White, 1994; Tsiknakis,
Katehakis & Orphanoudakis, 2002).
Very large-scale information systems are
typically deployed as centralized or distributed
architectures. In a centralized architecture, a single
server is responsible for processing all user requests.
Centralized architectures simplify administration
and coordination of a service, and they often require
fewer resources. However, centralized architectures
also have several disadvantages. A centralized
server represents a bottleneck in processing that
can delay response time. It may represent a single
point-of-failure, and for adversaries, a single
point-of-attack, without which the entire service
fails. Finally, even if the server is resourceful and
available, the routing function provided by the
communications layer may fail to provide a path
from some clients to the server.
Grid federation is gaining more and more
attention during the past few years in a variety
of R&and (research and development) efforts.
Any federated approach towards the creation of a
health information network environment should be
capable of providing uniform ways for accessing
authentic, physician-generated information that is
physically located in different clinical information
systems (Brailsford et al., 2004). Grid computing
makes extreme demands on distributed program-
ming because they are typically large-scale, and
they exploit wide-ranging networks consisting
of a variety of protocols and systems that may
span organizational boundaries. A description
of one kind of health information supply chain
architecture based on grid technologies is follows.
The national health supply chain architecture
is shown in Figure 1. There are four federated
services in this architecture. Each federated service
acts as a gatekeeper to actual information provid-
ers. Federated syndromic surveillance service pro-
vides outbreak detection and mitigation services.
Federated GIS atlas service provides map services.
Federated health record service provides medical
records. Federated patient tracking repository
service act as a huge data warehouse containing
massive data of RFID recordings.
The grid client shown in Figure 1 need not be
an end-user system, but could be an agent acting
on its behalf, and there may be many such clients
or agents, acting independently with no central
control over the components shown. To make use
of resources, a client first uses sources of informa-
tion located within service brokers to discover those
resources needed for execution of a task. Multiple
sources may need to be consulted to locate all the
resources needed for a computation. Assuming the
discovery and allocation steps were successful, the
client then sends the input data and executables,
receiving a reference to the execution in return.
These actions may be accomplished in several
stages or as one consolidated action, depending on
the nature and complexity of the task. As resourc-
es are allocated, the resource manager may need
to update the information in the registry to enable
reasonable bids for resource allocation from other
clients. Lastly, the client monitors execution of the
task using the reference it previously received. It
can retrieve the results, or be sent status of the task
as it progresses.
85
 GIS, Grid Computing and RFID in Healthcare Information Supply Chain
Figure 1. National health information supply chain architecture
A federated service provides access to static and
dynamic information of resources. There are three
major components within this federated service:
Grid Resource Information Service (GRIS), Grid
Index Information Service (GIIS) and Lightweight
Directory Access Protocol (LDAP) Provider.
Figure 2 represents the conceptual view intercon-
nection of the federated service. As illustrated, the
resource information is obtained by the informa-
tion provider and is passed on to a GRIS. The
GRIS registers its local information with a GIIS,
which, in its turn, registers with another GIIS,
and so on. Federated service clients can get the
resource information directly from a GRIS (for
local resources) and/or a GIIS (for grid-wide re-
sources). The federated services use LDAP, which
provides decentralized maintenance of resource
information. This information contains the objects
managed by federated services, which represent
components resources, bothstatic and dynamic.
A GRIS is the repository of local resource
information derived from information providers.
It is able to register its information with a GIIS,
86
but it itself does not receive registration requests.
The local information maintained by a GRIS is
updated when requested, and cached for a period
of time known as the time-to-live. If no request
for the information is received by it, the informa-
tion would time out and be deleted. If a later request
for the information is received, the system would
call the relevant information providers to retrieve
the latest information.
A GIIS is a repository that contains indexes
of resource information registered by a GRIS and
other GIISs. It can be seen as a grid-wide infor-
mation server. It has a hierarchical mechanism,
structurally like DNS, and each GIIS has its own
name. This means client users can specify the
name of a GIIS node to search for information.
The local health information providers translate
the properties and status of local resources to the
format defined in the schema and configuration
files. In order to add local health information re-
sources to a GRIS of federated services, specific
information providers need to be created in a
predefined format.
GIS, Grid Computing and RFID in Healthcare Information Supply Chain

Figure 2. Components for a federated service provider

The broker discovers resources, establishes
their cost and capability, and then prepares a
schedule to map requesting jobs to resources. It
first identifies a service broker and assigns a job
to it. A job has a task specification that specifies
a list of operations to be performed (Figure 3).
To process a job, the broker dispatcher deploys
its agent on that resource, who executes a list of
commands specified in the job’s task specification.
A typical task specification contains necessary

Figure 3. Components for a service broker

87
 GIS, Grid Computing and RFID in Healthcare Information Supply Chain
commands and medical records from one of the
agent resources, execution of the commands, and
finally copying results back to the user. It can
also contain special commands for accessing the
records from the remote database. The data broker
looks at the replica catalogue for a list of sites
providing federated services, checks the status of
those sites and selects a suitable site. The infec-
tious distribution fetch command can then request
the federated service provider for a patient record
and sent to outbreak detection federation. They
should be deployable in forms flexible enough
to meet the needs of incidents of various scopes
and complexities, and provide useful and reliable
information. Moreover, technologies, shared data
and terminology are standardized from a variety
of agencies.
REFERENCES
Abrahams, J. (2001). Disaster management in
Australia: The national emergency management
system. Emergency Medicine, 13(2), 165–173.
doi:10.1046/j.1442-2026.2001.00205.x
Boulos, M. N. K. (2004). Towards evidence-based,
GIS-driven national spatial health information in-
frastructure and surveillance services in the United
Kingdom. International Journal of Health Geo-
graphics, 3(1), 1–50. doi:10.1186/1476-072X-3-1
Boulos, M. N. K. (2005). Research protocol: EB-
GIS4HEALTH UK—Foundation evidence-based
and ontology-based framework of modular, reus-
able models for UK/NHS health and healthcare
GIS applications. International Journal of Health
Geographics, 4(2), 1–14.
Brailsford, S., Lattimer, V., Tarnaras, P., & Turn-
bull, J. (2004). Emergency and on-demand health
care: Modelling a large complex system. The
Journal of the Operational Research Society,
55(1), 34–42. doi:10.1057/palgrave.jors.2601667
88
Buyya, R. (2002). Economic-based distributed
resource management and scheduling for grid
computing. Doctoral dissertation, Monash Uni-
versity, Australia.
Chau, P., & Yip, P. (2003). Monitoring the severe
acute respiratory syndrome epidemic and assess-
ing effectiveness of interventions in Hong Kong
special administrative region. Journal of Epide-
miology and Community Health, 57, 766–774.
doi:10.1136/jech.57.10.766
Chervenak, A., Foster, I., Kesselman, C., Salisbury,
C., & Tuecke, S. (2001). The data grid: Towards
an architecture for the distributed management
and analysis of large scientific. Journal of Net-
work and Computer Applications, 23, 187–200.
doi:10.1006/jnca.2000.0110
Esswein, E., Kiefer, M., Wallingford, K., Burr, G.,
Lee, L.-H., & Wang, J.-D. (2004). Environmen-
tal and occupational health response to SARS,
Taiwan, 2003. Emerging Infectious Disease.
Retrieved from http://www.cdc.gov/ncidod/EID/
vol10no7/03-0728.htm
Flessa, S. (1999). Decision support for malaria
control programmes: A system dynamics model.
Health Care Management Science, 2(3), 181–191.
doi:10.1023/A:1019044013467
Forrester, J. (1961). Industrial dynamics. Cam-
bridge, MA: MIT Press.
Forslund, D., Joyce, E., Burr, T., Picard, R., Wok-
oun, D., & Umland, E. (2004). Setting standards for
improved syndromic surveillance. Engineering
in Medicine and Biology Magazine, 23(1), 65–70.
doi:10.1109/MEMB.2004.1297176
Forslund, D. W., & Kilman, D. G. (2000). The
impact of the global, extensible electronic health
record (pp. 3–13).
Forys, U. (2002). Marchuk’s model of immune
system dynamics with application to tumour
growth. Journal of Theoretical Medicine, 4(1),
85–93. doi:10.1080/10273660290052151
GIS, Grid Computing and RFID in Healthcare Information Supply Chain

Foster, I., & Kesselman, C. (Eds.). (2004). The Luckett, S., & Grossenbacher, K. (2003). A critical
Grid: Blueprint for a new computating infrastruc- systems intervention to improve the implementa-
ture (2nd ed.). Morgan Kaufmann. tion of a district health system in KwaZulu–Natal.
Systems Research, 20, 147–162. doi:10.1002/
Foster, I., Kesselman, C., & Tuecke, S. (2001).
sres.535
The anatomy of the grid: Enabling scalable virtual
organizations. International Journal of High Ramljak, D., Puksec, J., Huljenic, D., Koncar, M.,
Performance Computing Applications, 15(3), & Simic, D. (2003). Building enterprise informa-
200–222. doi:10.1177/109434200101500302 tion system using model driven architecture on
J2EE platform. In 7th International Conference
Hoard, M., Homer, J., Manley, W., Furbee, P.,
on Telecommunications (pp. 521–526).
Haque, A., & Helmkamp, J. (2005). Systems
modeling in support of evidence-based disaster Reiskind, M., Baisly, K., Calampa, C., Sharp, T.,
planning for rural areas. International Journal Watts, D., & Wilson, M. (2001). Epidemiologi-
of Hygiene and Environmental Health, 208(1-2), cal and ecological characteristics of past dengue
117–125. doi:10.1016/j.ijheh.2005.01.011 virus infection in Santa Clara Peru. Tropical
Medicine, 6(3), 212–215. doi:10.1046/j.1365-
Homer, J., Hirsch, G., Minniti, M., & Pierson, M.
3156.2001.00703.x
(2004). Models for collaboration: How system
dynamics helped a community organize cost- Ritchie-Dunham, J. L., & Galv’an, J. F. M.
effective care for chronic illness. System Dynam- (1999). Evaluating epidemic intervention
ics Review, 20(3), 199–222. doi:10.1002/sdr.295 policies with systems thinking: A case study
of Dengue fever in Mexico. System Dynam-
Landesman, L. Y. (2001). Public health manage-
ics Review, 15(2), 119–138. doi:10.1002/
ment of disasters: The practice guide. Washington,
(SICI)1099-1727(199922)15:2<119::AID-
DC: American Public Health Association.
SDR163>3.0.CO;2-G
Lane, D., Monefeldt, C., & Rosenhead, J. (2000).
Scheffer, M., Carptenter, S., Foley, J. A., Folke,
Looking in the wrong place for healthcare im-
C., & Walker, B. (2001). Catastrophic shifts
provements: A system dynamics study of an ac-
in ecosystems. Nature, 413(6856), 591–595.
cident and emergency department. The Journal
doi:10.1038/35098000
of the Operational Research Society, 51, 518–531.
Tsiknakis, M., Katehakis, D. G., & Orphanou-
Li, C.-J., Liu, L., Chen, S.-Z., Wu, C. C., Huang,
dakis, S. C. (2002). An open, component-based
C.-H., & Chen, X.-M. (2004). Mobile healthcare
information infrastructure for integrated health
service system using RFID. In Proceedings of
information networks. International Journal
the 2004 IEEE International Conference on Net-
of Medical Informatics, 68, 3–26. doi:10.1016/
working, Sensing and Control (pp. 1014–1019).
S1386-5056(02)00060-6
Taipei, Taiwan.
White, S. (1994). Total quality management in
Lipsitch, M., & Bergstrom, C. T. (2004). Invited
the public sector: Issues for implementation.
commentary: Real-time tracking of control mea-
Health Information Management Journal, 24(3),
sures for emerging infections. American Journal
115–121.
of Epidemiology, 160(6), 517–519. doi:10.1093/
aje/kwh256

89
 GIS, Grid Computing and RFID in Healthcare Information Supply Chain

WHO. (2003a). Severe acute respiratory syndrome
(SARS): Status of the outbreak and lessons for the
future. Retrieved June 2005 from http://www.who.
int/csr/media/sars wha.pdf
WHO. (2003b). Summary of probable SARS cases
with onset of illness from 1 November 2002 to
31 July 2003, September, 2003. Retrieved June
2005 from http://www.who.int/csr/sars/country/
table2003 09 23/en
Yousefi, S., & Lauridsen, J. (1998). Socio-spatial
formalism and nonlinear interaction in the dif-
fusion of epidemics in regional space. Technical
Report ersa98p406, European Regional Science
Association. Retrieved from http://ideas.repec.
org/p/wiw/wiwrsa/ersa98p406.html
Yu, H., & Lim, E.-P. (2004). LTAM: A location-
temporal authorization model. In M. P. Willem
Jonker (Ed.), Secure Data Management: VLDB
2004 Workshop, SDM 2004, LNCS 3178, (pp.
172–176). Springer-Verlag GmbH.

This work was previously published in Cases on Supply Chain and Distribution Management: Issues and Principles, edited by
Miti Garg and Sumeet Gupta, pp. 256-268, copyright 2012 by Business Science Reference (an imprint of IGI Global).

90
 91

Chapter 9
On Modeling and Analysis
of Multidimensional
Geographic Databases
Sandro Bimonte
Laboratoire d’Informatique en Images et Systèmes d’information, France

ABSTRACT
Data warehouse and OLAP systems are tools to support decision-making. Geographic information
systems (GISs) allow memorizing, analyzing and visualizing geographic data. In order to exploit the
complex nature of geographic data, a new kind of decision support system has been developed: spatial
OLAP (SOLAP). Spatial OLAP redefines main OLAP concepts: dimension, measure and multidimen-
sional operators. SOLAP systems integrate OLAP and GIS functionalities into a unique interactive and
flexible framework. Several research tools have been proposed to explore and the analyze spatio-multi-
dimensional databases. This chapter presents a panorama of SOLAP models and an analytical review
of research SOLAP tools. Moreover, the authors describe their Web-based system: GeWOlap. GeWOlap
is an OLAP-GIS integrated solution implementing drill and cut spatio-multidimensional operators, and
it supports some new spatio-multidimensional operators which change dynamically the structure of the
spatial hypercube thanks to spatial analysis operators.

INTRODUCTION OLAP tools provide the ability to interactively

A Data Warehouse is a centralized repository
of data acquired from external data sources and
organized following a multidimensional model
(Inmon, 1996) in order to be analyzed by On-
Line Analytical Processing (OLAP) applications.
DOI: 10.4018/978-1-4666-2038-4.ch009
explore multidimensional data presenting detailed
and aggregated data. The results of analyses are
the basis of strategic business decisions.
It has been estimated that about 80% of data
in databases contains geo-referenced informa-
tion (Franklin, 1992). Geographic Information
Systems (GISs) allow memorizing, analyzing and
visualizing geographic data.

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 On Modeling and Analysis of Multidimensional Geographic Databases
OLAP systems reduce geographic data to
simple textual data. Therefore, they do not pres-
ent any cartographic representation of geographic
data. On the contrary, maps are fundamental for
the spatial decision making process because they
stimulate user’s cognitive process, and reveal hid-
den precious geospatial information. Therefore,
some solutions, called Spatial OLAP, which
integrate GIS cartographic visualization and
interaction functionalities into OLAP systems,
have been developed. Spatial OLAP (SOLAP) is
“a visual platform built especially to support rapid
and easy spatio-temporal analysis and exploration
of data following a multidimensional approach
comprised of aggregation levels available in carto-
graphic displays as well as in tabular and diagram
displays” (Bédard, 1997). Spatial OLAP systems
integrate advanced OLAP and GIS functionalities
(Rivest et al., 2005; Kouba et al., 2000). They
visualize measures on maps at different spatial
granularities revealing relations between facts and
dimensions (Bédard, et al., 2001). Moreover, maps
allow triggering spatio-multidimensional opera-
tors through simple mouse clicks, also. Different
SOLAP models have been proposed. They address
various aspects of geographic information allow-
ing to model different spatio-multidimensional
applications. SOLAP applications can address
several and different domains: environmental
studies, marketing, archaeology, epidemiology,
etc.. SOLAP models define the concepts of spa-
tial/geographic dimension, spatial/geographic
measure, and spatio-multidimensional operators.
The integration of spatial data into multi-
dimensional models and systems rises several
theoretical and implementation issues. Therefore,
in this chapter, we introduce main OLAP and
GIS concepts. Then, a detailed review of SO-
LAP models, architectures and research SOLAP
tools is presented. The chapter describes our
Web-based prototype for the analysis of spatio-
multidimensional databases (GeWOlap) (Bimonte
et al., 2006; Bimonte et al., 2007a, Bimonte et al.,
2007b). We describe main architectural features,
92
and we present spatio-multidimensional and GIS
operators using a study case concerning pollution
in French cities. Main outcome and limits of our
approach as regards to existing SOLAP tools are
detailed, also. Finally, future research directions
in spatio-multidimensional visualization and
interaction are discussed.
BACKGROUND
Data Warehouse and OLAP Systems
Data warehouse and OLAP systems are business
intelligence tools intended to support multidimen-
sional analysis of huge datasets. Data are modelled
according to the multidimensional model, which
is based on the concepts of dimensions and facts
(Inmon, 1996). Dimensions represent analysis
axes. They are organized in hierarchies’ schemas.
Facts, described by numerical values (measures),
are subjects of analysis. Measures are analyzed
at different granularities corresponding to dimen-
sion hierarchies’ levels, and they are aggregated
by means of SQL aggregation functions. The
instance of a multidimensional model is the hy-
percube. It is a set of cells representing measures
at all combinations of dimensions’ levels. OLAP
operators permit to navigate into the hypercube.
Most common operators are drill and cut operators.
Drill operators (i.e., Roll-Up and Drill-Down) let
navigating into dimension hierarchies aggregat-
ing measures. Cut operators (i.e. Slice and Dice)
permit to reduce the analysis space, by selecting
a sub-set of dimensions members.
Usually, data warehouse and OLAP systems
are based on a three-tier architecture. The first tier
is the data warehouse, where data, coming from
external heterogeneous sources, are uniformed
and stored according to the multidimensional
model. The second tier is the OLAP Server. It
implements OLAP operators and pre-computes a
set of multidimensional queries to grant effective
query response times. The OLAP server imple-
On Modeling and Analysis of Multidimensional Geographic Databases
ments other advanced functionalities also, such
as control accesses, multidimensional calculation
engine, etc. Finally, the OLAP client allows visual-
izing and formulating OLAP queries in a unique,
interactive and user-friendly visual interface
composed of graphic displays and pivot tables.
Geographic Information Systems
A Geographic Information System encompasses a
set of tools for organizing, memorizing, analyzing
and visualizing geographic information (Longley
et al., 2001).
Geographic information is the representation
of geo-referenced real phenomena. It is described
by a spatial component (the shape and the posi-
tion on earth surface) and a set of alphanumerical
attributes and (spatial) relationships.
Main GIS functionality is the integration
and memorization of geographic data. Data are
organized in layers. A layer represents a set of
geographic data of the same type (i.e. roads,
buildings, departments, etc.). Each element of the
layer (geographic object or feature) is described
by geometry and some alphanumeric attributes.
Figure 1. Spatial SQL a) Spatial table creation b) Spatial query c) Geographic object creation d) Tabular
and cartographic representations
Layers are stored in GIS using Spatial DBMSs,
which extend classical DBMS with spatial types
(i.e. point, line, etc.) and spatial functions (i.e.
overlap, etc.). As an example, the spatial table
representing French departments, and a spatial
query selecting departments with an area greater
than 5000 Km2, using Oracle’s Spatial SQL, are
shown on Figure 1a and Figure 1b, respectively. A
tuple creation and its cartographic representation
are shown in Figure 1c and Figure 1d, respectively.
GIS’s geographic data can be analyzed by
means of spatial analysis methods. They allow
estimating, predicting and understanding spatial
data and phenomena by means of (spatial) statis-
tic (i.e. centroid, etc.), transformation (i.e. buffer,
etc.) and querying techniques. These methods are
supported by visual representation of maps’ fea-
tures. Indeed, visualization plays a central role
for the spatial analysis process. Map feature’s
geometries are represented using a two dimen-
sional space. Alphanumeric attributes are visual-
ized with graphic displays (i.e. bars, pies charts,
etc.), coded using visual variables (size, value,
colour, shape, orientation and grain) (Bertin &
Bonin, 1992). GISs make possible to change map’s
93
 On Modeling and Analysis of Multidimensional Geographic Databases
scale, reduce visualized features, access to alpha-
numeric information using the map, etc. Some-
times, maps provide also an interface to advanced
(visual) analysis techniques. Users can explore
alphanumeric data, perform complex analysis
methods, and obtain personalized and complex
visual representations of data by simply interact-
ing with map’s features (interactive maps). Inter-
active maps are the basis of geovisualization
systems (MacEachren, et al., 2004). Such systems
integrate scientific visualization and image
analysis techniques, and GIS tools into an interac-
tive, flexible and user-friendly framework in order
to explore and analyze spatial data. Interactive
capabilities are mandatory for spatial decision-
making process (MacEachren & Kraak, 1997).
SPATIO-MULTIDIMENSIONAL
DATABASES: MODELS AND TOOLS
Spatio-Multidimensional Models
The more natural and common manner to integrate
spatial information into multidimensional models,
is to use it as a dimension. As defined in Bédard et
al. (2001), a spatial dimension can be “spatial non
geometric” (i.e. with text only members), “spatial
geometric” (i.e. with members with a cartographic
representation) or “mixed spatial” (i.e. combining
cartographic and textual members). Malinowsky
Figure 2. Multidimensional model with spatial dimension
94
& Zimányi (2005) define a spatial dimension as a
set of spatial hierarchies. A “spatial hierarchy” is
a hierarchy with at least one level with the spatial
attribute (spatial level). Topological intersection or
inclusion relationships exist between members of
different spatial levels. Bimonte (2008) proposes
the concept of “Geographic Dimension” enrich-
ing spatial dimensions with “Map Generalization
Hierarchies”. “Map Generalization Hierarchy”
represents geographic information at different
scales or according secondary themes where the
members of a level are the result of map generaliza-
tion operators applied to the members of the lower
level. Map generalization is the process used to
derive data and maps for secondary scales and/or
themes, preserving a good and clear representation
focused on the goal of the map (Weibel & Dutton,
2001). Map generalization provides a simpli-
fied vision of the spatial phenomenon enriching
spatio-multidimensional analysis capabilities and
improving SOLAP clients’ visualization.
An example of SOLAP application using a spa-
tial dimension is a study for pollution supervision
in French cities. This multidimensional application
presents three dimensions: “Time”, “Pollutants”,
and “Location” (spatial dimension), and a nu-
merical fact, “Pollution”. This fact is depicted by
three measures giving minimum, maximum and
average pollution values (see Figure 2) (Bimonte
et al., 2007a). This multidimensional application
answers questions like “What are the average,
On Modeling and Analysis of Multidimensional Geographic Databases
min and max values per month, and pollutant for
departments with population above 2M?”
A very different way to introduce spatial in-
formation in data warehouses is using it as an
analysis subject, i.e. as a fact. Different definitions
of the spatial measure can be found in literature:
a collection of geometries (spatial objects) (Ste-
fanovic et al., 2000; Rivest et al. 2001), geometries
or numerical values resulting from spatial (i.e.
topological and metric) operators (Malinowsky
& Zimányi, 2004), and/or spatial members
(Marchand et al., 2003). Spatial aggregations (i.e.
union, intersection, etc.) replace SQL SUM, MIN,
MAX, AVG, and COUNT functions. Maps, then,
are the cells of the hypercube.
Let us take the spatio-multidimensional model
given in Figure 3a. The spatial attribute of the
Figure 3. Spatio-multidimensional models a) Spatial measure, b) Geographic measure
“City” level of application of Figure 2 is now used
as spatial measure. The spatial measure is aggre-
gated using the topological union. Pollution values
grouped by 5mg/l are used as analysis dimension.
This model analyzes polluted French cities ac-
cording to time, pollutants and pollution values.
In this model, the user should be able to deduce
information about the influence of geographical
location of cities in the pollution problem.
Descriptive attributes of geographic data could
be useful to the spatio-multidimensional deci-
sional process. Thus, Bimonte (2008) introduces
the concept of “Geographic Measure”. A “Geo-
graphic Measure” is a geographic object described
by alphanumeric and spatial attributes. Moreover,
it could belong to one or many hierarchy schemas.
95
 On Modeling and Analysis of Multidimensional Geographic Databases
This establishes a complete symmetry of geo-
graphic measures and geographic dimensions.
Replacing the spatial measure of the multidi-
mensional application in the previous example
with the geographic measure representing cities
(see Figure 3b), it should be possible to answer
queries like: “What cities, their population, and
their socio-economic types, are polluted by CO2
per month?” Indeed, a city is a geographic object
described by geometry and two alphanumeric
attributes: population and socio-economic type.
Note that a (spatial) aggregation function is ap-
plied to each (spatial) attribute of the geographic
measure (i.e. topological union for geometry,
list for name, sum for population and a ratio for
socio-economic type).
Spatio-multidimensional operators extend
drill and cut OLAP operators. “Spatial Roll-up”
and “Spatial Drill-down” authorize to navigate
into spatial dimensions by the simple interaction
with the map component of the SOLAP user-
interface (Rivest et al., 2005). “Spatial Slice”
makes possible cutting the spatial hypercube by
selecting directly spatial members through SO-
LAP user-interface (Rivest et al., 2005), using
spatial/alphanumeric predicates (Sampaio et al.,
2006) or spatial analysis operators such as buffer
(Scotch & Parmanto, 2006). Exploiting the sym-
metrical representation of geographic dimensions
and measures, Bimonte (2008) proposes two
operators, “Permute” and “Measure navigation”.
“Permute” allows exchanging dimension and
geographic measure. This operator dynamically
modifies the structure of the spatial hypercube.
“Measure navigation” allows navigating into
the geographic measure’s hierarchy, changing
granularity of the measure on the fly. For instance,
since cities belong to departments (Figure 3b),
“Measure navigation” operator permits to analyze
polluted French departments (instead of cities)
along time and pollutants dimensions. Moreover,
to make the spatio-multidimensional paradigm
more flexible and being closer to spatial analysis
process, Bimonte et al. (2007b) propose a new
96
kind of operators which change the structure of
the spatial hypercube through the introduction of
new spatial members into geographic dimension
thanks to spatial analysis operators.
Spatial OLAP Tools: A Panorama
Bédard et al., (2005) identify three typologies
of Spatial OLAP tools (OLAP dominant, GIS
dominant and OLAP-GIS integrated), according
to their OLAP and GIS functionalities. Some
prototypes have been implemented. They allow
an effective spatio-multidimensional analysis of
spatial data warehouses thanks to advanced OLAP
and GIS functionalities. However, an overall set
of mandatory functionalities has not been defined,
yet. Indeed, the following SOLAP research tools’
panorama reveals that spatial data mining tools,
spatial analysis methods and geovisualization
techniques should enrich spatio-multidimensional
operators.
GIS Dominant Solutions
GIS dominant solutions simulate OLAP server
thanks to a relational database modelled using
the star or the snow-flake schema. As shown
in Bédard et al. (2005), these solutions provide
all GIS functionalities: storage, analysis and
visualization. However, since they lack of an
OLAP Server, they do not implement advanced
OLAP functionalities, such as derived measures,
complex hierarchies, etc.. Consequently, GIS
dominant solutions limit spatio-multidimensional
analysis capabilities. To best of our knowledge,
the only GIS dominant solution is CommonGis
(Voss et al., 2004). It is a geovisualization system
extended to support multidimensional databases.
It provides multi-criteria functionalities, and spa-
tial analysis and visualization techniques for the
analysis of spatio-temporal data, using temporal
series. CommonGIS has been adapted to analy-
sis of spatio-multidimensional datasets, where
spatial information is used as analysis axes. The
On Modeling and Analysis of Multidimensional Geographic Databases
user interface is flexible and interactive. It offers
spatio-multidimensional drill and cut operators,
and some advanced geovisualization techniques.
OLAP Dominant Solutions
OLAP dominant solutions provide all advanced
multidimensional analysis functionalities thanks
to an OLAP system. On the contrary, GIS func-
tionalities are reduced to simple cartographic
visualization and selection of geographic objects
(Bédard et al., 2005). OLAP dominant solutions
can be grouped into two classes: tools using static
maps (Stolte et al., 2003; Colonnese et al., 2005;
Mohraz, 2000) and tools using interactive maps
(Silva et al., 2006; Sampaio et al., 2006; Shek-
har et al., 2001; Han et al., 1997; Pourabbas &
Rafanelli, 2002).
Polaris system (Stolte et al., 2003) allows the
visualization of alphanumeric measures, using
non-interactive maps incorporated into the cells
of the pivot table. PostGeOLAP (Colonnese et al.,
2005) is an open source SOLAP tool supporting
numerical measures and spatial dimensions. It
provides a set of methods to create spatio-mul-
tidimensional databases and materialize spatial
views. In Mohraz (2000) the OLAP system SAP’s
Business Information Warehouse is integrated
with a cartographic visualization tool.
Silva et al. (2006) present a Web-based SO-
LAP solution, whose principal feature is the use
of geographic Web services for the definition of
GeoMDQL. GeoMDQL is a new query language
for spatial data warehouses. It extends the OLAP
language of Microsoft (MDX) with spatial types.
The prototype is based on OLAP Server Mondrian,
which has been modified to handle GeoMDQL
queries, and OLAP client JPivot, which is coupled
with an interactive map. In Sampaio et al. (2006),
the authors describe a Web-based SOLAP sys-
tem which handles spatial measures and allows
querying spatial data warehouses using drill
and cut operators on spatial dimensions. This
solution presents a Web interface composed of a
cartographic component and a text zone to define
spatio-multidimensional queries.
Finally Shekhar et al. (2001) and Han et al.
(1997) introduce data mining techniques into
OLAP systems, and Pourabbas & Rafanelli (2002)
use visual languages. In particular, Shekhar et
al., (2001) develop the cube operator extending
aggregation of numerical data to spatial data. The
Web-based prototype is based on this operator, and
it is especially conceived for the observation and
the discovery of spatio-temporal trends. In Pourab-
bas & Rafanelli (2002) a visual language is used
to query spatial data warehouses. This approach
allows the user to formulate slice operations using
an iconic language.
OLAP-GIS Integrated Solutions
OLAP-GIS integrated solutions combine GIS
and OLAP functionalities. GIS analysis and vi-
sualization functionalities enrich and complete
OLAP navigation and visualization functionalities,
allowing a real and effective analysis of spatial
data warehouses. Some OLAP-GIS integrated
solutions have been developed (Rivest et al., 2005;
Scotch & Parmanto, 2006; Matias & Moura-Pires,
2005; Escribano et al., 2007).
These solutions store spatial multidimensional
data using a Spatial DBMS. Spatial data warehous-
es allow modelling complex spatial hierarchies,
topological intersection relationships, and spatial/
geographic measures. Star and snowflake schemas
are modified to handle spatial data (Stefanovic et
al., 2000; Malinowsky & Zimányi, 2007).
(S)OLAP Servers implement spatio-multidi-
mensional operators explicitly taking into account
spatial members’ geometry to define cutting
spatial predicates and to perform drill operators.
They allow defining spatial members, hierarchies,
measures and derived measures.
SOLAP clients improve OLAP clients thanks
to cartographic visualization and interaction
functionalities. They combine and synchronize
tabular and graphic visualization components with
97
 On Modeling and Analysis of Multidimensional Geographic Databases
interactive maps (Rivest et al., 2005). SOLAP
operators are available through the simple inter-
action with maps. Each spatial level corresponds
to a cartographic layer. Performing (S)OLAP op-
erators on the pivot table and/or on the map will
result in the calculation and displaying of a new
clickable map (interactive map), merging differ-
ent layers according to the different granularities
selected for the spatial members, and displaying
aggregated and/or detailed measures with visual
variables and/or graphic displays (i.e. bars, pies,
etc.) (Rivest et al., 2005).
For instance, considering the application of
Figure 2, one should be able to show and to inter-
act with a map displaying average and minimum
values for selected months and pollutants at the
granularity of cities for one department, and at
department level for the rest of the map.
In order to take advantage from human visual
perception capabilities, interactive maps should
be coupled with advanced geovisualization tech-
niques such as MultiMaps, Space-Time Cubes, etc.
(MacEachren et al., 2004). Finally, GIS operators
are available through the cartographic component.
Indeed, decision makers should be able to custom-
ize their maps, add layers, query the spatial data
warehouse, use spatial analysis tools, etc.
Rivest et al. (2005) present a SOLAP tool
which permits to enrich the elements of the spa-
tial data warehouse with multimedia documents
such as photos, videos, etc.. SOVAT (Scotch &
Parmanto, 2006) extends SOLAP functionalities
with (spatial) data mining tools (i.e. clustering,
etc.). It provides also statistic and spatial analysis
methods. The SOLAP tool described in Matias
& Moura-Pires (2005) allows overlapping two
hypercubes using their spatial dimensions. PIET
(Escribano et al., 2007) is a Web-based SOLAP
solution providing a new query processing method.
Spatial members are first organized into open
convex polygons, and then the overlay of these
polygons is used to answer to multidimensional
queries using user-defined window queries.
98
In the following paragraph we describe our
OLAP-GIS integrated solution.
GeWOlap: Spatio-
Multidimensional Analysis Tool
GeWOlap is a Web-based system for exploration
and analysis of spatio-multidimensional databases
(Bimonte et al., 2006; Bimonte et al., 2007a,
Bimonte et al., 2007b). It provides advanced (S)
OLAP and GIS functionalities through a flexible
interactive and user-friendly interface. GeWOlap
is based on three tier architecture.
The spatial data warehouse tier is implemented
using Oracle for its native support for spatial data
and its object-relational capabilities.
OLAP server Mondrian provides an easily
customizable, full featured (i.e. aggregate tables,
user-defined functions, complex hierarchies,
MDX support, etc.) OLAP server. It is important
to notice that no spatial extension has been inte-
grated into Mondrian. Spatial data is represented
using simple textual labels. In order to overcome
this limitation, GeWOlap handles spatial data in
the spatial data warehouse and client tiers, and it
provides a mapping mechanism between spatial
data and their textual representation in the OLAP
Server.
The client tier is implemented using the OLAP
client JPivot to provide JSP pages, and a Java applet
developed using MapXtreme Java (MapXtreme)
to support map visualization/interaction and GIS
functionalities. The client tier allows a visual rep-
resentation of spatio-multidimensional structures
[(geographic) dimensions and measures] through
pivot table, graphic displays and interactive map.
Figure 4 displays the visual interface of Ge-
WOlap for the application of Figure 2. The pivot
table represents pollution values for Ile de France
region and for its departments (Essone, Hautes de
Seine, etc.). The cartographic component shows a
thematic map representing departments’ pollution
values using pie charts.
On Modeling and Analysis of Multidimensional Geographic Databases
Figure 4. GeWOlap’s user interface
Spatio-multidimensional and GIS operators
are accessible through the simple interaction with
the pivot table and the interactive map of the cli-
ent using only few mouse clicks.
In particular, GeWOlap provides a set of
drill operators which are available through the
interaction with the pivot table and the map:
“roll-up drill-down replace”, “drill-down posi-
tion”, “expand-all”, “drill-through” (Bimonte et
al., 2007a).
For instance, let us suppose that the user wishes
to “see” the measures for departments of the Ile
de France region. By pointing the mouse on that
region she/he can apply the drill-down position
operator (Figure 5a). As a result, the pivot table
displays average pollution values for the Ile de
France region, and for other departments (Figure
5b). Several synchronization problems rise from
the topological inclusion relationships between
spatial members of different levels, and the num-
ber of measure values that must be displayed. For
example, unlike the pivot table, the map cannot
display at the same time a region and its depart-
ments, and visualize the pollution values for each
pollutant granting a good cartographic readability.
The user can cut the hypercube by using the
“Cube Navigator” tool provided by JPivot. The
Cube Navigator provides a tree representation of
dimension members which can be used to custom-
ize pivot table axes and select a sub-set of mem-
bers. Moreover, GeWOlap extends OLAP cut
operators by introducing two new cut operators:
“Slice Predicate” and “Slice Position” (Bimonte
et al., 2007b). Thanks to “Slice Predicate”, the
user can select spatial members by directly click-
ing on the interactive map. “Slice Predicate” allows
cutting the hypercube through Spatial SQL queries.
Let us suppose the user is interested in pollu-
tion values for departments crossed by from the
Seine river. Then, thanks to the Slice Predicate’s
wizard she/he defines its spatial query:
select * from MEASURES_VIEW, RIVERS
where (MEASURES_VIEW.NAME = ‘Essone’)
OR… OR (MEASURES_VIEW.NAME = ‘Yve-
lines’) AND SDO_GEOM.RELATE(MEASURES_
VIEW.GEOM, ‘ANYINTERACT’,RIVERS.
GEOM, 0.5)= ‘TRUE’ AND RIVERS.NAME =
‘Seine’
99
 On Modeling and Analysis of Multidimensional Geographic Databases

Figure 5. Spatial Drill-down position, a) Input: Ile de France region, b) Output: Departments of Ile de
France

As a result, the pivot table and the map show
only the departments satisfying the spatial predi-
cate.
Cut and drill operators allow to explore the
spatial hypercube using, directly or not, the spa-
tial attribute of the spatial members. The spatial
hypercube does not change its structure. As under-
lined in Bimonte, et al. (2007b), it represents an
important limit, as the spatial analysis process is
iterative and flexible, or, in other words, the user
can change and transform spatial data all along the
spatial decisional process. GeWOlap fits buffer,
overlay and dissolve spatial analysis operators to
the multidimensional paradigm. Indeed, GeWOlap
provides three new spatio-multidimensional
operators which dynamically create new spatial

100
On Modeling and Analysis of Multidimensional Geographic Databases
Figure 6. Overlay operator
members thanks to spatial analysis operators and
calculate their associated measures using MDX
formula and/or Java user-defined functions. The
“Dissolve” operator merges adjacent spatial
members having an alphanumeric attribute with
the same value. This attribute is chosen by the
user through the Dissolve’s wizard. The “Buffer”
operator (Bimonte, et al., 2007b) creates a buffer
region around one spatial member selected by
the user through a mouse click on the interactive
map. The distance of the region buffer is chosen
by the user thanks to the Buffer’s wizard. The
GIS overlay operator creates a new layer whose
features are obtained using the intersection op-
erator on two input layers as shown in Figure 6.
GeWOlap adapts this operator to create n new
spatial members (Bimonte, et al., 2007b).
We present here an example of the “Dissolve”
operator as it is representative of this class of
operators. An example is shown in Figure 7. Start-
ing from the query represented in Figure 7a, which
shows pollution average values for Ile de France’s
departments, the analyst chooses the department
type attribute. Since “Essone” and “Seine-et-
Marne” are adjacent and their type is “Commer-
cial”, then they are merged into one new region.
A new spatial member (“Essone-Seine-et-Marne”)
is created, and its measure is calculated using a
weighted average on the surface (Figure 7b).
Previous examples show spatio-multidimen-
sional operators applied to geographic dimensions
with numerical measures. Similarly, GeWOlap
supports geographic measures. In this case, pivot
table’s cells contain identifiers of geographic
objects and the interactive map shows geograph-
ic measure dynamically chosen by the user. More
details about aggregation and visualization of
geographic measures, and the implementation of
“Permute” operator can be found in Bimonte et
al. (2006) and Bimonte (2007).
In addition to spatio-multidimensional operators,
GeWOlap provides also pure GIS functionalities:
“Zoom in/out”, “Pan”, “Retrieve”, “Map print”,
“Map export”, “Rule tool” and “Control Layer”.
“Pan” allows moving the map using the mouse,
“Map print” permits to print the map and “Map ex-
port” permits to save map in JPG and PDF formats.
“Rule Tool” is a metric tool that calculates distance
between two points selected by the user. Finally,
“Control Layer” provides different functionalities.
It allows customizing the visual representation of
visual variables: colour, size, etc., backgrounding
the spatio-multidimensional application by adding
raster and/or vector layers, and querying the spatial
data warehouse using Spatial SQL.
In conclusion, GeWOlap is a full-featured
OLAP-GIS integrated solution, which supports
geographic dimension and numerical measures,
and implements drill and cut multidimensional
operators. Moreover, GeWOlap enriches exist-
ing SOLAP tools functionalities supporting
measures defined as geographic objects (geo-
graphic measures) and implementing three new
spatio-multidimensional operators which change
the structure of the spatial hypercube thanks to
spatial analysis operators. However, some limits
remain to overcome: the integration of ad-hoc
indexing techniques (Papadias, et al., 2002) to
improve the performance of our system, and the
implementation of mechanisms for automatically
detecting and computing spatial materialized
views (Colonese et al., 2005).
101
 On Modeling and Analysis of Multidimensional Geographic Databases

Figure 7. Dissolve operator, a) Input: Ile de France’s departments, b) Output: Departments of Ile de
France and the new spatial member Essone-Seine-et-Marne

FUTURE RESEARCH DIRECTIONS Map generalization is mandatory for SOLAP

Spatial OLAP technologies have reached matu-
rity. They allow an effective multidimensional
analysis of spatial data warehouses. However,
the introduction of complex spatial data raises
several problems.
systems (Bimonte, 2008). On the other hand,
map generalization implies the reformulation of
classical spatial hierarchies by introducing multi-
association relationships and taking into account
imprecise measures. (Un)Fortunately, visualiza-
tion of multidimensional imprecise (spatial) data
is an open issue (Pang, 2008).

102
On Modeling and Analysis of Multidimensional Geographic Databases
Spatial OLAP models and tools are based on
the vector representation of spatial data. Field
data, which represent spatial data as a regular grid
whose cells are associated with alphanumeric at-
tributes, could be used for spatio-multidimensional
analysis. Indeed, Ahmed & Miquel (2005) provide
a continuous representation of spatial dimensions,
but the introduction of field data as analysis subject
remains unexplored. Aggregation of field data
could be supported by Map Algebra operators. Map
Algebra defines a set of operations on field data
(local, focal and zone operators) (Tomlin, 1990).
Adaptation of Map Algebra to multidimensional
data structures, definition of pre-aggregation and
visualization/interaction techniques for continu-
ous measures are challenges to overcome for an
effective SOLAP tool supporting field data.
The integration of trajectories data into data
warehouses raises several problems because
classical multidimensional models are based on
discrete facts and dimensions and they do not
take into account spatial predicates. This problem
has been investigated by some works (Wan et al.,
2007; Orlando et al., 2007) in the last years. How-
ever, the definition of a SOLAP client to visually
query and analyze trajectory data warehouses is
an unexplored important challenge.
Bertolotto et al. (2007) affirm that visual
analysis of spatial data mining results is improved
by exploiting the third dimension of spatial data
through an interactive 3D geovisualization system.
The integration of advanced geovisualization
techniques within OLAP clients in order to support
multidimensional 3D spatial data is an interesting
research direction.
Finally, several semiology problems have to
be solved for the correct and relevant visualiza-
tion of measures. Measures can be displayed with
labels but it is sometimes worthwhile to use more
expressive, significant visual components. The
way measures will be displayed on the map must
depend on several criteria: nature of the measure
(quantitative or qualitative measure), number
of measures to be displayed, and current repre-
sentation of the spatial dimension (point, line or
polygon). Moreover, GIS users are usually specific
knowledge domain decision makers. GIS takes
into account their profiles and preferences in order
to provide well-suited cartographic visualization
(Vangenot, 2001). Thus, the ambition is to define
a method to automatically find out the most ap-
propriate cartographic representation of SOLAP
queries results. The visual variables (size, colours,
etc.) and the graphic representation (i.e. bar, pie,
etc.) used to represent measures on maps can be
automatically deduced thanks to SOLAP query
patterns taking into account number, type and
current representation of dimensions, measures
types, aggregation functions involved in the query
and user profile and/or preferences (Bellatreche,
et al., 2005).
CONCLUSION
Spatial OLAP refers to the introduction of spatial
data into data warehouse and OLAP systems.
SOLAP enhances decision analysis capabilities
of OLAP systems allowing exploiting the com-
plex nature of geographic information. SOLAP
redefines main OLAP concepts. It defines spatial/
geographic dimensions as dimensions with spa-
tial attributes, spatial measures as a collection of
spatial objects or the result of spatial operators,
and geographic measures as geographic objects
belonging to hierarchy schemas. SOLAP extends
multidimensional navigation operators defining
spatial drill and cutting operators which allow
navigating into spatial/geographic dimensions and
cutting the hypercube thanks to spatial and non-
spatial predicates. Other spatio-multidimensional
operators permit to change the structure of the spa-
tial hypercube thanks to spatial analysis operators,
to permute dimensions and geographic measures,
and to navigate in the hierarchy of the geographic
measure. SOLAP tools are based on the integration
of OLAP and GIS functionalities. In this work, we
have provided an analytical overview of SOLAP
103
 On Modeling and Analysis of Multidimensional Geographic Databases
tools grouping them according to their architecture
approach. We described our Web-based prototype
(GeWOlap) for the multidimensional analysis of
huge spatial datasets. GeWOlap combines SOLAP
and GIS operators into a unique interactive and
user-friendly framework. GeWOlap supports geo-
graphic dimensions and geographic measures, and
it introduces some new spatio-multidimensional
operators which allow to change the structure
of the hypercube thanks to dissolve, buffer and
overlay GIS operators.
REFERENCES
Ahmed, T., & Miquel, M. (2005). Multidimen-
sional structures dedicated to continuous spatio-
temporal phenomena. In M. Jackson, D. Nelson,
& S. Stirk (Eds.), Proceedings of the 22nd British
National Conference on Databases (Vol. 3567,
pp. 29-40). Berlin, Germany: Springer.
Bédard, Y. (1997). Spatial OLAP. In Proceedings
of the 2nd Forum annuel sur la R-D, Géomatique
VI: Un monde accessible, Montréal, Canada.
Bédard, Y., Merrett, T., & Han, J. (2001). Funda-
ments of spatial data warehousing for geographic
knowledge discovery. In H. Miller & J. Han (Eds.),
Geographic data mining and knowledge discovery
(pp. 53-73). London: Taylor & Francis.
Bédard, Y., Proulx, M., & Rivest, S. (2005). Enrich-
issement du OLAP pour l’analyse géographique:
Exemples de réalisation et différentes possibilités
technologiques. In Revue des Nouvelles Technolo-
gies de l’Information, Entrepôts de données et
l’Analyse en ligne (pp. 1-20).
Bellatreche, L., Giacometti, A., Laurent, D., &
Mouloudi, H. (2005). A personalization frame-
work for OLAP queries. In I. Y. Song & Y. Trujillo
(Eds.), Proceedings of the ACM 8th International
Workshop on Data Warehousing and OLAP (pp.
9-18). New York: AMC Press.
104
Bertin, J., & Bonin, S. (1992). La graphique et
le traitement graphique de l’information. Paris:
Flammarion.
Bertolotto, M., Di Martino, S., Ferrucci, F.,
& Kechadi, T. (2007). Towards a frame-
work for mining and analyzing spatio-tem-
poral datasets. International Journal of Geo-
graphical Information Science, 21(8), 1–12.
doi:10.1080/13658810701349052
Bimonte, S. (2007). Vers l’intégration de
l’information géographique dans les entrepôts de
données et l’analyse en ligne: De la modélisation
à la visualization. Unpublished doctoral disserta-
tion, INSA Lyon, France.
Bimonte, S. (2008). Des entrepôts de données,
l’analyse en ligne et l’information géographique.
Journal of Decision Systems.
Bimonte, S., Tchounikine, A., & Miquel, M.
(2007). Spatial OLAP: Open issues and a Web
based prototype. In M. Wachowicz & L. Bodum
(Eds.), Proceedings of the 10th AGILE Interna-
tional Conference on Geographic Information
Science.
Bimonte, S., Tchounikine, A., Miquel, M., &
Laurini, R. (2007). Vers l’intégration de l’analyse
spatiale et multidimensionnelle. In M. Batton-Hu-
bert, T. Joliveau, & S. Lardon (Eds.), Proceedings
of the Colloque International de GEOmatique et
d’Analyse Spatiale.
Bimonte, S., Wehrle, P., Tchounikine, A., &
Miquel, M. (2006). GeWOlap: A Web based spa-
tial OLAP proposal. In R. Meersman, Z. Tari, &
P. Herrero (Eds.), Proceedings of the Workshop
on Semantic-Based Geographical Information
Systems (Vol. 4278, pp. 1596-1605). Berlin,
Germany: Springer.
Colonese, G., Manhaes, R., Montenegro, S., Car-
valho, R., & Tanaka, A. (2005). PostGeoOlap: an
open-source tool for decision support. In Proceed-
ings of the 2nd Simpósio Brasileiro de Sistemas
de Informação.
On Modeling and Analysis of Multidimensional Geographic Databases
Escribano, A., Gomez, L., Kuijpers, B., & Vais-
man, A. (2007). Piet: A GIS-OLAP implementa-
tion. In I. Y. Song & T. B. Pedersen (Eds.), Pro-
ceedings of the ACM 10th International Workshop
on Data Warehousing and OLAP (pp. 73-80). New
York: ACM Press.
Franklin, C. (1992). An introduction to geographic
information systems: Linking maps to databases.
Database, 15(2), 13–21.
Han, J., Kopersky, K., & Stefanovic, N. (1997).
GeoMiner: A system prototype for spatial data
mining. In J. Peckham (Ed.), Proceedings of the
ACM SIGMOD International Conference on
Management of Data (pp. 553-556). New York:
ACM Press.
Inmon, W. H. (1996). Building the data warehouse
(2nd ed.). New York: John Wiley & Sons.
Kouba, Z., Matousek, K., & Miksovsky, P. (2000).
On data warehouse and GIS integration. In M.
Ibrahim, J. Kung, & N. Revell (Eds.), Proceedings
of the 11th International Conference on Database
and Expert Systems Applications (Vol. 1873, pp.
604-613). London: Springer.
Longley, P., Goodchild, M., Maguire, D., & Rhind,
D. (2001). Geographic information systems and
science. New York: John Wiley & Sons.
MacEachren, A., Gahegan, M., Pike, W., Brewer,
I., Cai, G., Lengerich, E., & Hardisty, F. (2004).
Geovisualization for knowledge construction and
decision support. IEEE Computer graphics and
application, 24(1) 13-17.
MacEachren, A., & Kraak, M. (1997). Exploratory
cartographic visualization: Advancing the agenda.
Computers & Geosciences, 23(4), 335–343.
doi:10.1016/S0098-3004(97)00018-6
Malinowsky, E., & Zimányi, E. (2004). Represent-
ing spatiality in a conceptual multidimensional
model. In D. Pfoser, I. Cruz, & M. Ronthaler
(Eds.), Proceedings of the 12th ACM International
Workshop on Geographic Information Systems
(pp. 12-22). New York: ACM Press.
Malinowsky, E., & Zimányi, E. (2005). Spatial
hierarchies and topological relationships in Spa-
tialMultiDimER model. In M. Jackson, D. Nelson,
& S. Stirk (Eds.), Proceedings of the 22nd British
National Conference on Databases (Vol. 3567,
pp. 17-28). Berlin, Germany: Springer.
Malinowsky, E., & Zimányi, E. (2007). Logical
representation of a conceptual model for spatial
data warehouses. Geoinformatica.
Marchand, P., Brisebois, A., Bédad, Y., & Edwards,
G. (2003). Implementation and evaluation of a
hypercube-based method for spatio-temporal ex-
ploration and analysis. Journal of the International
Society of Photogrammetry and Remote Sensing,
59(1), 6–20. doi:10.1016/j.isprsjprs.2003.12.002
Matias, R., & Moura-Pires, J. (2005). Spatial
on-line analytical processing: A tool to analyze
the emission of pollutants in industrial installa-
tions. In C. Bento, A. Cardoso, & G. Dias (Eds.),
Proceedings of the 5th International Workshop on
Extraction of Knowledge from Databases and
Warehouses.
Mohraz, K. (2000). Geographical analysis in SAP
business information warehouse. In K. Li, K.
Makki, N. Pissinou, & S. Ravada (Eds.), Proceed-
ings of the 8th ACM Symposium on Advances in
Geographic Information Systems (pp. 191-193).
Washington: ACM Press.
Orlando, S., Orsini, R., Raffaetà, A., Roncato, A.,
& Silvestri, C. (2007). Spatio-temporal aggrega-
tions in trajectory data warehouses. In I. Y. Song,
J. Eder, & T. M. Nguyen (Eds.), Proceedings of the
9th International Conference Data Warehousing
and Knowledge Discovery (Vol. 4654, pp. 66-77).
Berlin, Germany: Springer.
Pang, A. (2008). Visualizing uncertainty in
natural hazards. In A. Bostrom, S. French, & S.
Gottlieb (Eds.), Risk assessment, modeling and
decision support: Strategic directions series: Risk,
governance and society (pp. 261-294). Berlin,
Germany: Springer.
105
 On Modeling and Analysis of Multidimensional Geographic Databases
Papadias, D., Tao, Y., Kalnis, P., & Zhang, J.
(2002). Indexing spatio-temporal data ware-
houses. In Proceedings of the 18th International
Conference on Data Engineering (pp. 166-175).
Los Alamitos, CA, USA: IEEE Computer Society.
Pourabbas, E., & Rafanelli, M. (2002). A pictorial
query language for querying geographic databases
using positional and OLAP operators. SIGMOD
Record, 31(2), 22–27. doi:10.1145/565117.565121
Rivest, S., Bédard, Y., & Marcand, P. (2001). To-
wards better support for spatial decision-making:
Defining the characteristics of spatial on-line
analytical processing. Journal of the Canadian
Institute of Geomatics, 55(4), 539–555.
Rivest, S., Bédard, Y., Proulx, M., Nadeaum, M.,
Hubert, F., & Pastor, J. (2005). SOLAP: Merging
business intelligence with geospatial technology
for interactive spatio-temporal exploration and
analysis of data. Journal of International Society
for Photogrammetry and Remote Sensing, 60(1),
17–33. doi:10.1016/j.isprsjprs.2005.10.002
Sampaio, M., Sousa, A., & Baptista, C. (2006).
Towards a logical multidimensional model for
spatial data warehousing and OLAP. In Y. Song &
P. Vassiliadis (Eds.), Proceedings of the 9th ACM
International Workshop on Data Warehousing
and OLAP (pp. 83-90). New York: ACM Press.
Scotch, M., & Parmanto, B. (2006). Develop-
ment of SOVAT: A numerical-spatial decision
support system for community health assess-
ment research. International Journal of Medical
Informatics, 75(10-11), 771–784. doi:10.1016/j.
ijmedinf.2005.10.008
Shekar, S., Lu, C., Tan, X., Chawla, S., & Vat-
savai, R. (2001). Map cube: A visualization tool
for spatial data warehouses. In H. Miller & J. Han
(Eds.), Geographic data mining and knowledge
discovery (pp. 74-109). London: Taylor & Francis.
106
Silva, J., Times, V., & Salgado, A. (2006). An open
source and Web based framework for geographic
and multidimensional processing. In H. Had-
dad (Ed.), Proceedings of the ACM Symposium
on Applied Computing (pp. 63-67). New York:
ACM Press.
Stefanovic, N., Han, J., & Kopersky, K. (2000).
Object-based selective materialization for efficient
implementation of spatial data cubes. IEEE Trans-
actions on Knowledge and Data Engineering,
12(6), 938–958. doi:10.1109/69.895803
Stolte, C., Tang, D., & Hanrahan, P. (2003). Multi-
scale visualization using data cubes. IEEE Trans-
actions on Visualization and Computer Graphics,
9(2), 176–187. doi:10.1109/TVCG.2003.1196005
Tomlin, D. (1990). Geographic information sys-
tems and cartographic modeling. Upper Saddle
River, NJ: Prentice Hall.
Vangenot, C. (2001). Supporting decision-making
with alternative data representations. Journal of
Geographic Information and Decision Analysis,
5(2), 66–82.
Voss, A., Hernandez, V., Voss, H., & Scheider, S.
(2004). Interactive Visual exploration of multidi-
mensional data: Requirements for CommonGIS
with OLAP. In Proceedings of the 15th Interna-
tional Workshop on Database and Expert Systems
Applications (pp. 883-887). Los Alamitos, CA:
IEEE Computer Society.
Wan, T., Zeitouni, K., & Meng, X. (2007). An
OLAP system for network-constrained moving
objects. In Y. Cho, R. Wainwright, H. Haddad, S.
Shin, & Y. Koo (Eds.), Proceedings of the ACM
symposium on Applied computing (pp. 13-18).
New York: ACM Press.
Weibel, R., & Dutton, G. (2001). Generalizing
spatial data and dealing with multiple representa-
tions. In P. Longley, M. Goodchild, D. Maguire,
& D. Rhind (Eds.), Geographic Information
systems and science (pp. 125-155). New York:
John Wiley & Sons.
On Modeling and Analysis of Multidimensional Geographic Databases

KEY TERMS AND DEFINITIONS Spatial Decision Support System: Interac-

Geographic Information System: Informa-
tion system for capturing, storing, analyzing,
managing and presenting data that are spatially
referenced.
Multidimensional Model: Conceptual model
for the multidimensional analysis of huge datasets,
based on the concepts of dimensions, facts and
measures.
Spatial Data Warehouse: Subject-oriented,
non volatile, time variant and integrated reposi-
tory of spatial data that is designed to facilitate
reporting and spatial analysis.
tive, computer-based system designed to support
a user or group of users in achieving a higher
effectiveness of decision making while solving
a semi-structured spatial problem.
Spatial OLAP: Visual platform built espe-
cially to support rapid and easy spatio-temporal
analysis and exploration of data following a mul-
tidimensional approach comprised of aggregation
levels available in cartographic displays as well
as in tabular and diagram displays.

This work was previously published in Data Warehousing Design and Advanced Engineering Applications: Methods for Com-
plex Construction, edited by Ladjel Bellatreche, pp. 96-112, copyright 2010 by Information Science Reference (an imprint of
IGI Global).

107
108

Chapter 10
Technology and the
Multipolar Global Economy:
Implications for European Competitiveness

Steven McGuire
Aberystwyth University, UK

ABSTRACT
The European Union has engaged in several attempts to increase the level of innovation in member
states – largely without success. This chapter argues that a combination of global and regional issues
explain this failure. What was once a process dominated by Europe and America, innovation is “going
global,” with many countries now developing national innovation systems. European firms are conduct-
ing more of their research in emerging markets, and in Europe, firms seem unable to raise their share of
overall research and development spending. The Commission, for its part, lacks the necessary regulatory
and legislative powers to forge a new regional innovation system and is instead reliant on a policy mix
of overarching objectives, some directed funding, and comparative analysis of member state policies.

INTRODUCTION based capitalist system remains the best way to

The globalization of production that developed in
the latter parts of the last century is now augmented
by the globalization of innovation. This process
is at once welcome and potentially destabilizing.
It is welcome in the sense that it represents a
victory for the western liberal economic tradi-
tion advocated by the United States and Europe:
that a comparatively open, predictable and rule-
DOI: 10.4018/978-1-4666-2038-4.ch010
spur economic growth and raise living standards.
However, in the United States and the European
Union there is increasing anxiety that the disper-
sion of innovative capabilities represents, if not a
zero-sum competition between economies, then
at least an episode of relative economic decline.
While the Kindle is sold by Amazon (an Ameri-
can company) and marketed as an example of
American ingenuity, the product itself could not
be manufactured in the US. This is not because of

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Technology and the Multipolar Global Economy
cost; rather, the problem is that the US economy no
longer has the technological capacity to produce
the item (Munro, 2010). Aside from the ink the key
manufactured elements of the Kindle rely on Asian
manufacturing firms. For now, the US dominates
the design process, but extensive academic work
on innovation suggests that design and production
are, in the medium term, tightly interwoven. The
tacit knowledge acquired in the production process
eventually underpins new innovations at the design
phase. Europe shares many of the anxieties seen
in the United States. The European Commission
has made several high-profile efforts to address the
perceived decline in European competitiveness.
The Lisbon Agenda, launched in the early 2000’s
to much fanfare, was effectively abandoned late in
the last decade and the Commission attempted to
relaunch a European drive to increase the region’s
position in high technology industries through
the Competitiveness and Innovation Framework
Program (CIP). The programme had three ele-
ments: first, support for entrepreneurs; second, a
set of policies designed to generate greater use of
information technology in European firms; and,
finally, an initiative to promote innovation in
sustainable energy supplies. In 2010 a new effort,
Europe 2020, was launched, featuring a broader
assault on the perceived failings of Europe in the
areas of innovation, new business creation, tertiary
education and job creation.
Innovation is often envisaged as the lone in-
ventor working in his laboratory, hoping for that
crucial breakthrough. Though serendipity can be
important, the key to the success of Europe and
American innovation was that both actors cre-
ated systems, comprised of firms, government
and other stakeholders such as universities and
research institutes. European universities – par-
ticularly German institutions – led the world in
the developing institutions devoted to scientific
research, and linked with German firms, who
were among the first to develop in-house R&D
facilities (Dosi, et al, 2005:21). American firms
were also early adopters of a model where firms
devoted money to research, rather than leave new
product development to opportunistic acquisition.
American universities were also well funded by
a variety of sources – federal and state govern-
ment, students and philanthropy – and played a
crucial role in the ascent of the US to dominance
of innovation in the 20th century. Both Europe
and America adopted what has been described as
the ‘linear’ model, which presumed a relatively
straightforward process of discovery, develop-
ment and sales, with many held within the firm,
or within its control (Chesbrough, 2003). As the
name suggests, the model envisaged a more or
less linear relationship between basic research
conducted at universities, leading to applied work
at corporate research and development centres
leading to successful commercialisation.
So successful were Europe and America that
follower states developed their own innovation
systems, borrowing and adapting European and
American practice as needed. Japan of the 1980s
was the most successful follower state but the
new millennium has seen the field of top-quality
innovating nations become much more crowded.
China is merely the most obvious gainer, with
Singapore, India and Taiwan other states that have
developed top quality innovation systems. Just as
production was globalised in the last century, it
seems likely that the 21st century will see the true
globalization of knowledge, as many economies
develop the capabilities to innovate at the very
frontier of a variety of technologies – and are
simultaneously dependent on knowledge cre-
ated elsewhere. This ‘networked’ technological
system poses challenges to state policies oriented
around keeping innovative activities confined to
the home base.
In both Europe and America, this development
has been met with apprehension and even hostility
in some quarters. This is because the high-skill,
‘knowledge’ economy was supposed to be a safe
haven for developed economies faced with the
loss of manufacturing jobs. Public policy in both
economies has focused on the need to enhance the
109

innovative capacity of the economy, via improved
education, government support for research and
development and increased use of innovation by
firms for competitive advantage. The unstated as-
sumption was that emerging economies like China,
though they were good at cheap manufacturing,
would not be able to duplicate European and
American efforts in technology intensive sectors.
Though there is evidence that a significant
skills premium remains, the assumptions about
the security of even graduate-level, managerial
and technical are proving misguided. The key
distinction insofar as jobs are concerned is not
between knowledge workers and manual labour,
but between jobs that can be traded (even if elec-
tronically) and those that are non-tradable (Blinder,
2005). To the extent that emerging economies
develop talents in tradable, highly skilled goods
and services, they will compete directly with
European and American scientists, engineers, as
well as management consultants and accountants.
China is dramatically increasing its innovative
capacity, with increased research and develop-
ing spending actually increasing faster than the
overall rate of economic growth (OECD, 2006a).
Though a note of caution must always apply to
Chinese statistics, China’s overall effort in R&D,
gross expenditure on research and development
(GERD) is, at 1.31 per cent of GDP, not far off
the European average (Zhou and Leydesdorff,
2006: 84). China is now the 5th largest patent-
ing nation on Earth, though with a share barely
a quarter of Japan and America’s (WTO, 2006).
China has put particular emphasis on nanotechnol-
ogy, thus emulating several states in viewing this
technology as a key driver of future innovative
activity. China is now second only to the US in
terms of single country production of published
scientific research on nanotechnology (Zhou and
Leydesdorff, 2006: 94). The overall quality of
China’s innovation infrastructure can be called into
question. Though money is pouring into Chinese
universities, laboratory equipment and academic
training remain relatively poor by European and
110
Technology and the Multipolar Global Economy
American standards. However the trajectory
and pace of China’s advance, to say nothing of
renewed efforts in Singapore and Malaysia and
the expansion of India is impressive. The notion,
popularized by numerous politicians, that Europe
and America could find an ‘economic haven’
among high-skilled, high technology sectors is
proving gravely mistaken.
Exactly why this diffusion is happening is
not widely considered in the literature. A 2007
European Commission document suggested that
this process has resulted from ‘the growing S&T
[science and technology] capacity of China and
India and the expanding global knowledge pro-
duction chains inside MNEs’ (2007: 45). Athreye
and Cantwell (2008) similarly highlight the role of
internationalized supply chains and the diffusion
of more knowledge-intensive functions within
these supply networks. However, this explanation
merely highlights the process; why did developed
state firms feel able to diffuse these functions in
the first place? The political economy literature
of the late 1990s and 2000s, far from envisaging
the diffusion of innovation, fretted that the inno-
vation ladder was being pulled up from develop-
ing states. A particular target for criticism in this
regard was the WTO’s Trade Related Aspects of
Intellectual Property (TRIPs) (Sell, 1996). The
world envisaged in this literature was one where
developing states remained trapped in low skilled,
low value-added activities. Why this expectation
has been confounded, not least in China but also
India, Brazil and other emerging states, needs
consideration.
Multilateral Trade, R&D, and
the Diffusion of Innovation
The globalization of innovation is now clearly
underway, though like other aspects of global-
ization the density and reach of this process is
uneven. Just as Richard Florida’s (2006) work on
economic growth has identified patterns of both
flatness and spikiness in the global economy, so too
Technology and the Multipolar Global Economy
for research and development. This globalization
has also been accompanied by an increase in the
number of states actively engaged in technology
policies in support of emergent industries. These
policies involve an array of research and develop-
ment initiatives designed to both tempt foreign
firms to locate R&D facilities in the locality and
to enhance the innovative capabilities of domestic
firms. Trade liberalization under the WTO has
facilitated this, but not in a straightforward man-
ner. Trade liberalization has two potential effects
on firm incentives to undertake R&D: first, it can
create positive incentives to invest, as liberaliza-
tion expands the market (scale effect) but: second,
it can dampen incentives if new competitors are
created (competition effect) (Teteryaknikova,
2008). As with any other subsidy, research and
development incentives risk being captured by
organized producer interests (Brou et al., 2009),
but they can also address the market failures as-
sociated with risky investments, and in so doing be
welfare-enhancing. Thus, multilateral disciplines
on subsidies need to accommodate two demands
that are in tension with each other. Unsurprizingly
then the Subsidies and Countervailing Measures
(SCM) agreement in the WTO is comparatively
complex and, when first negotiated, sought to
accommodate a range of permissible subsidies
– as well as outlawing export subsidy (McGuire,
Figure 1. China’s patenting activity in the USPTO, 1991-2005 (Source: Hu and Matthews, (2008))
2002). As both Evenett and Kang note, WTO
disciplines on R&D subsidies are comparatively
weak, whereas intellectual property provisions
under TRIPs are comparatively strong (Evenett,
2009; Kang 2000). This partly explained by the
desire of developed world (or OECD) firms
pressing their governments to gain multilateral
protection for their existing stock of intellectual
property. TRIPs has been controversial because
of the unfairness implicit in the protection of
developed world knowledge, thus restricting the
ability of developing states to catch up. Recent
experience has, however, called this into question,
as Asian countries have demonstrated an ability
to develop a range of emergent high technology
sectors. Forero-Pineda (2006), for example, argued
that TRIPs has increased incentives to pursue
niche technology policies in several developing
states. Similarly, Hu and Matthews (2008) note
that innovative activity in China soared after that
country’s accession to the WTO in 2001 (Figure
1). They argue that foreign firms had, in effect,
been waiting until China joined the WTO before
investing in higher value-added activities there.
The SCM Code, in contrast to TRIPs, has not
proved problematic for emerging economies, aside
from a few high-profile cases such as Embraer’s
subsidization by the Brazilian government (Gold-
stein and McGuire, 2004). The language of the
111

SCM – what Evenett refers to as its incomplete-
ness – offers considerable scope for states to
develop national programs under the aegis of
environmental technologies or regional eco-
nomic development. During the Uruguay Round
negotiations, the United States succeeded in per-
suading other states of the merits of a specific
research and development exemption (ie green
light) under the SCM code (GSI, no date). The
recent economic crisis has merely added to the
view that the SCM is not a significant obstacle to
governments seeking to promote or protect in-
dustries. As Falke (2010) notes, the US bail-out
for the automobile sector was the most costly
subsidy in US corporate history. Europe did not
retaliate, in large part because of several Euro-
pean states went ahead and introduced similar
subsidy measures designed to preserve their auto
sectors. The European Union essentially sus-
pended its anti-subsidy enforcement activities,
fearing a political storm should the Commission
be seen as sacrificing member state economies to
WTO-inspired regulations (Evenett, 2009). The
wide scope of the Subsidies Code has been a boon
to nations like India, which have been able to
pursue national innovation policies in key areas
such as pharmaceuticals and biotechnology (Ra-
mani, 2002).
European Innovation Policy1
European innovation policy is complicated by
the simple fact that the EU is not a state, but
a regional economic integration project, and
so policymaking is that much more complex.
Moreover, in contrast to other areas of policy, the
Commission has relatively few instruments at its
disposal, and so relies to a much greater extent on
the willingness of member states to cooperate with
initiatives. The result has been a comparatively
disappointing performance, particularly in light
of the success of technology programs in several
emerging economies such as Korea and China.
Nonetheless, the relative failure is not new.
112
Technology and the Multipolar Global Economy
Jean-Jacques Servan-Schreiber articulated
European fears about American domination of the
global economy over forty years ago, yet his work
speaks to current anxieties about Europe’s ability
to compete. Writing in the 1960s, he observed the
fantastic growth of US multinationals and feared
that Europe’s future was to be a mere branch
plant for American multinationals, who had com-
pletely supplanted European firms across sectors
(Servan-Schreiber, 1968). Indeed, for pessimists,
the European Union has missed not one but two
major waves of innovation: the 1980s innovation
of computers and related industries; and the 1990s
service-based innovation that built on the earlier
wave. The figures on comparative expenditure on
R&D paint a depressing picture. Overall European
Union expenditure on research has remained stable
at just under 2 per cent of GDP for several years, in
spite of the Lisbon Agenda and a number of other
EU-level initiatives designed to improve this per-
formance. Though expenditures were also stable
in the US, America spends considerably more on
innovation (2.6 per cent of GDP) and has done so
for many years (Eurostat, 2008). In an updated
analysis performed in the 2010, the Commission
noted that both the United States and Japan had
maintained their lead over the European Union
in a range of innovation indicators, most notably
in the area of business-level R&D expenditure
(European Commission, 2011).
Of course, a note of caution is in order: some
European states are very competitive in the new
technology and service economies. Sweden and
Finland, for example, regularly feature among
the top in any ranking of competitive economies.
At one level then Europe’s ‘problem’ is one of
uneven performance, where lagging member
states tend to create the perception that all EU
members have innovation gaps. In virtually no
other policy area is there as wide a dispersion
in performance between the best and worst EU
member state as in innovation policy (Archibugi
and Coco 2005: 443).
Technology and the Multipolar Global Economy
The EU lags both in total economic resources
devoted to innovation, and particular in business-
level research and development expenditure. The
percentage of R&D funded by business in the EU-
27 is 55%, compared with 64% in the US and 75%
in Japan (European Commission, 2007). Its share
of world patenting appears to have stalled. Though
the situation is improving, European firms show a
lower propensity to use IT to increase productivity
than American firms. European universities are
poorly funded by American standards, though in
fact they employ a higher percentage of research-
ers (37%) than the United States 15% (European
Commission, 2007). A National Science Founda-
tion study of immigrant scientists and engineers
noted at strong belief among European emigrants
that the US offered better research infrastructure
and job opportunities (NSF, 2007a). As a study for
the European Commission noted, the problem is
‘that Europe has structural lags in top level science
and innovative performance vis-à-vis the US…’
(Dosi et al. 2005:1, emphasis added). On a range
of university-related indicators of innovation,
such as the numbers of PhDs awarded and citation
rates of scientific publications, the United States
has maintained it lead (European Commission,
2011). The problem is not simply one of com-
mercialization – where excellent science work is
left underexploited by firms – but rather one of
weakness in both basic research and its applica-
tion to the creation of goods and services (Dosi et.
al, 2006). Though European universities account
for a significant percentage (about 22 percent) of
all research and development expenditures – and
produce significant numbers of technicians and
engineers – this does not translate into innovative
success for firms (OECD, 2006b: 209).
Europe has been here before, with the superior
performance of other economies provoking con-
cern about European competitiveness. European
policymakers fastened on the scale advantages
offered by the large American market, and the
role of an interventionist state in Japan, to argue
for the creation of national champions in vari-
ous sectors. These firms, usually created by the
government-sponsored merger of smaller firms,
would enjoy a dominant position in a given na-
tional market and thus reap economies of scale
benefits. However, the protection afforded these
firms gave the wrong series of incentives; the
policy produced inefficient, non-innovative firms
which were unable to cope with the dynamism of
Japanese or American competitors. Outside the
aerospace sector (and perhaps nuclear power) the
government-led national champion policies were
notable for failure rather than success. Policy fail-
ure highlighted the need to expose European firms
to greater competition; this could be achieved in
the first instance by creating a proper European
market. The Single Market Programme was the
liberal response to the poor competitiveness of
European firms. Creation of the unified market
would allow European firms to reap the same scale
economies that the American market provided to
US firms decades earlier. For reasons which are
not yet clear, Europe’s innovative capacity did
not accelerate after the SMP: indeed as indicated
above, it deteriorated.
The Failure of the Lisbon Strategy
The Lisbon Summit in 2000 committed EU
member states and the Commission to making
Europe, ‘the most competitive and dynamic
knowledge-based economy in the world by 2010’.
The declaration was essentially an admission that
efforts in the 1990s had either failed or needed
to be intensified and refocused. Two measures of
economic performance were cited by European
policymakers in defense of new strategies: the
deterioration of per capita GDP growth relative
to the US; and the stagnation of European pro-
ductivity. European per capita GDP relative to
the United States climbed steadily in the years
following the Second World War, peaking in the
1970s at about three quarters of US levels. The
trend, if anything, has been downward and in
2006, the gap was a worrying large 32 per cent
113

(Fassbender, 2007). Some eight years after the
launch of the Lisbon strategy, R&D intensity in the
European Union has stagnated at approximately
1.84% of GDP (EU, 2009) http://eur-lex.europa.
eu/LexUriServ/LexUriServ.do?uri=COM:2009:
0558:FIN:EN:PDF#page=2.
The second problem, in a sense, gave rise to the
first. EU labor productivity growth hardly moved
during the 1990s, whilst in the same period, the US
was undergoing a so-called productivity miracle
(Daveri, 2002). Exactly why Europe lagged behind
has not been adequately answered, though various
attempts had been made. One suggestion was that
the still incomplete Single Market process was to
blame; as long as firms could exist in relatively
uncompetitive national markets, they lacked the
incentive to increase their investment in produc-
tivity enhancing processes and assets (McGuire,
2006). Another explanation drew on social and
cultural attributes, with the observation that Eu-
ropeans produce less than Americans because of
a rational decision to devote more time to leisure
(Fassbender, 2007). A more recent study has like-
wise suggested that, at least in certain parts of the
EU, demand for innovative products and services
remains low, with entrepreneurs thus unwilling
to risk entering the market (EU Yearbook, 2009).
One of the aims of the CIP programme is to ad-
dress this by making early stage finance more
available, thus overcoming what appears to be a
market failure.
There is a view that the demands of innova-
tion policy in the 21st century are increasingly
in conflict with entrenched political and social
practices in the European Union. Throughout the
world, innovation policies emphasize the process
of ‘creative destruction’ where older firms and
industries are swept and as labor and capital are
reallocated to new activity (Schumpeter, 1942).
Advocates of the Schumpeterian perspective see
the globalization as the very essence of creative
destruction. Competitiveness policies ought to
assist in the process by encouraging new indus-
tries and allowing other sectors to wither away. In
114
Technology and the Multipolar Global Economy
contrast, some Europeans see in the competitive-
ness agenda an effort to unpick the social market
protections that citizens have enjoyed for decades.
It does not follow that the need for creative de-
struction requires a small state and low taxes; in-
deed, government activity and public expenditure
might be necessary for innovation in many fields.
There is thus dispute over the orientation of in-
novation policy. Japan and the United States are
both highly innovative and successful economies
with considerable numbers of competitive firms.
Yet, both operate very different mechanisms for
generating this innovation. Even within Europe,
Nordic countries are reasonably successful at rec-
onciling high taxation with appropriate incentives
for firms to invest and innovate. Some European
states, including the UK, Ireland and Hungary,
are very reliant on foreign firms for research and
development expenditures, whilst others, such as
Germany, retain significant national capabilities
(OECD, 2006b). In short, very real and impor-
tant elements of variation of technology policies
exist, and European policymakers have found it
difficult to articulate a clear and cohesive strategy
for Europe in the face of conflicting evidence
and societal pressures. In a clear admission that
Europe is not a ‘single innovation space’ the Com-
mission dramatically revised its methodology for
analyzing member state performance in 2010.
The result of the Innovation Union Scoreboard
was that there are four distinct groups or clusters
of countries within the EU. Denmark, Finland,
Germany and Sweden are ‘Innovation Leaders’
with internationally competitive innovation sys-
tems and performance. The next group, Innovation
Followers, have strong innovative capabilities
but without the outstanding performance of the
leading group. The next two groups, ‘Moderate
Innovators’ and ’Modest Innovators’ are mainly
small and newly-acceded member states without
significant innovative capabilities (European
Commission, 2011)
One difficulty the EU has is the multilayered
nature of governance, which makes the develop-
Technology and the Multipolar Global Economy
ment and implementation of innovation policies
difficult and contested. In the innovation policy
literature, the EU proves troublesome for analysts.
The EU is distinct in that it is not a national sys-
tem of innovation, yet neither is it a fully-fledged
regional system (Kuhlmann, 2001). In innovation
management, as with other areas, EU policies
and institutions must be understood as a complex
negotiated order among national, supranational
and local actors (Smith and Elgstrom, 2000). The
Commission’s involvement has been governed
by the principle of subsidiarity: that the role of
Brussels was to augment and catalyze efforts at
the regional or nation state level (Kuhlmann,
2001:963). In sum, in marked contrast to issue
areas like trade and competition policies, the
Commission enjoys relatively little power or au-
tonomy. The Commission lacks effective tools to
force through policy changes, even if agreement
could be generated on the nature of the problem
and the response. This has left the Commission
in the position of advocating networking and
mobility among research institutes and firms, not
necessarily because networking and mobility do
much to aid innovation, but because such limited
policy tools are about the only ones that the Com-
mission has (Dosi, et al. 2006).
EU R&D policy thus continues to operate at
the margins of influence. This limits its ability to
develop and enforce policy directions and instead
makes it prone to backsliding. The Kok Report
observed that European states cannot develop solu-
tions in isolation – and only co-ordinated action
will benefit Europe. Yet in a 2007 working paper,
Commission officials conceded that member states
focus their innovation strategies on domestic
concerns and work most closely with domestic
constituents. UNICE bemoaned the inability of
member states to meet their Lisbon commitments,
noting that the temptation to use public funds for
objectives other than enhancing competitiveness
seemed hard to resist (UNICE, 2006).
In 2010, the Commission announced the re-
placement for the Lisbon Strategy: Europe 2020.
Whilst retaining some of the key performance
indicators of the Lisbon process, including a
Union-wide target for R&S spending, the 2020
process concentrates more on enabling processes
such as human capital formation and on employ-
ment generally. A more sector-based approach
to innovation is also taken, with the Commis-
sion seeking to capitalize on existing European
strengths in low-carbon energy technologies
(European Commission 2011b)
The Direction of European R&D
Evolutionary perspectives on innovation draw
attention to the uncertainty surrounding innova-
tion (McGuire, 2006). The innovative process is
uncertain (that does not mean it has no organi-
zational logic) and characterized by incremental
improvements and developments that are, from
time to time, interrupted by dramatic, discontinu-
ous change. Schumpeter referred to a ‘process of
industrial mutation…that incessantly revolution-
izes the economic structure from within.’ (Drejer,
2004: 556). Schumpeter’s early interest in indi-
viduals as inventors and entrepreneurs gave way
in time to a recognition that creative destruction
could – and was – carried out in larger organiza-
tions. Mowery and Rosenberg (1989) described
the American economy of the late 19th and early
20th centuries as a period where, in a somewhat
paradoxical way, innovation was made routine by
institutionalizing activities underpinning inven-
tion. These included the spread of secondary and,
later, higher education, as well as the creation
of in-house research and development units in
American firms. Thus, the innovative process
may be uncertain and non-linear, but it is not
random. Government policies should be able to
provide incentives (where needed) to encourage
the appropriate conditions for innovation.
If innovation is as non-linear as evolution-
ary perspectives suggest, then targeted national
champion policies are unwise for a different
reason: since a state cannot be certain where and
115

how innovations will arise, the best policy is to
encourage innovative activity across a range of
sectors. Moreover, a key benefit of innovation lies
in its diffusion through the economy to medium
and low technology sectors. This diffusion ought
to make less technologically advanced sectors
more productive (Navarro, 2003: 12; Patel 2007).
Thus, targeted policy that comes at the expense
of basic, blue-skies research seems problematic
(Brusconi and Guena, 2003; Dosi, Llaverena and
Labini, 2005) Some evidence in support of an
emphasis on supporting wide ranging research
comes from a comparison of the early develop-
ment of biotechnology in the United States and
Germany. Both countries invested considerable
sums in this new technology, with German gov-
ernment support dating from the 1960s. Yet, by
1997, the United States biotechnology sector was
larger by a factor of ten (Giesecke, 2000: 207).
The crucial factor is that diffuse US support for
basic research allowed the organic development
of the sector. German efforts, by contrast, were
hampered by the dominant pharmaceutical firms,
which resisted greater competition and weak
incentives in the university system to transform
basic research into patentable innovations (ibid).
By and large the Commission’s pronounce-
ments on the trajectory of research and develop-
ment policy reflect an evolutionary economics
perspective, stressing the need for basic research.
Central to the Lisbon Agenda is raising the amount
spent on R&D by Member States to 3 per cent
of GDP, though as some academics note, most of
this increase is to come from industry (Archibugi
and Coco, 2005). By 2004, there were several
states spending at this level, notably the Nordic
states, Germany and France. The Commission’s
own funding streams, principally the Framework
Programs and Structural Funds, were also tilted
toward basic research and designed to catalyze
innovation. However, these efforts are small in
comparative terms: the € 10.5 billion of Structural
Funds earmarked for innovation from 2000-2006
were roughly equivalent to Sweden’s annual
116
Technology and the Multipolar Global Economy
R&D expenditure (Potocnik, 2007). An additional
obstacle was the tension between allocating re-
search money on the basis of the excellence of
the proposed research, and awarding money to
less technologically capable partners in poorer
member states. Plans for the European Research
Council to centralize and co-ordinate long term
basic research were developed and accepted in
2006. In a departure from previous practice, ‘sci-
entific merit’ would be the overriding criterion for
awarding funding (European Commission, 2004c:
21). There is a persistent tension between cohe-
sion and innovation underpinnings of EU R&D
policy (Sharp, 1998). The former implied the need
to direct R&D support to poorer member states
even if this was unlikely to produce cutting-edge
research; the latter suggested that the best ideas
needed to win out in a process of competition if
EU science was to progress. By 2004, the Kok
Report, a high-level review of EU innovation
policy chaired by former Dutch prime minister
Wim Kok, offered indications that the latter view
was prevailing (European Commission 2004c).
This plan was generally welcomed in member
states with strong university systems that could be
expected to win the greatest share of money, but
other actors preferred to retain a ‘developmental’
aspect to EU-level research policies.
All that said governments and the Commis-
sion are only partly responsible for the problem:
European businesses invest less in innovation
than competitors. Business R&D investment –
particularly in the major European economies
– remains low compared to the United States
and Japan. In Europe, businesses conduct 64 per
cent of R&D expenditures, whereas in Asia and
the US the figure is 70 per cent (NSF 2007d).
European firms are more reliant on public funds
for innovative activities – and this leaves them
more vulnerable to political processes that favor
expenditure in other areas. European firms make
less use of technicians and scientists than American
or Japanese companies. Corporate research staffs
represent 80 per cent of all researchers in the US,
Technology and the Multipolar Global Economy
while the comparable figure in the EU (25) is 49
per cent (OECD, 2006b: 36). What is the problem
with lagging business investment? In the United
States, high-technology industries, which are the
main spenders on research and development, are
a larger part of the economy (European Commis-
sion 2007c). For some, the problem remains a lack
of pump-priming of the small and medium sized
business sector, the sector seen as a key driver of
new technologies. The Kok Report, called for the
creation of new tax incentives for SMEs to engage
in research and development (2004c). By 2010,
there is precious little evidence that this had any
effect (European Commission 2011a)
Another fear, however, is that EU-based firms
are expending less of their own funds because
they are spending more money in foreign markets.
The diffusion of innovative activities has been
noted in the academic literature but in recent
years has made an appearance in public debates.
As countries like India and China develop, there
is a strong business case for relocating research
and development activities closer to market. Ger-
man multinationals illustrate the trend in R&D
offshoring. By 2007, nearly 20 per cent of all
German corporate R&D was conducted outside
of Germany, more than twice the figure of two
decades earlier (DB Research, 2011). Though
most of this offshoring went to companies located
in neighboring EU states or the United States, an
unmistakable trend in using emerging countries
was also notable. Fully 29 per cent of German
patents filed during 2008-2008 had at least one
emerging market co-filer (DB Research, figure
5). The assumption of the linear model of innova-
tion – that innovative activities are centralized the
core and product produced in the peripheries – is
being severely undermined by corporate activity.
As the EU notes, the US has increased its share
of foreign affiliates’’ spending on R&D, to a large
extent by taking a greater share of research from
Europe (EU, 2005: 36). Between 1997-2003,
the EU ran an ‘R&D deficit’ – where net R&D
outflows exceeded inflows – with the United
States and Japan of about $7 billion (European
Commission 2007). Though innovative activities
are diffusing throughout the globe, Europe’s rela-
tive attractiveness to foreign firms is declining.
A Global Innovation Space?
For most of the past 100 years, innovation was
dominated by the Triad of Europe, the United
States and Japan. The application of science and
technology to economic activity first arose in
German and American firms early in the 20th
century and was part of the set of characteristics
that allowed the transatlantic economy to domi-
nate the global system for most of the past 200
years. Though governments devote considerable
resources to capturing the benefits of innovation
for their own firms through a variety of policies,
in fact innovative activity does tend to diffuse
across borders. Firms not only create their own
intellectual property, but also scan the competitive
environment for knowledge that might yield a com-
petitive advantage. Firms were diversifying their
innovation activities some years before it became
commonplace to talk about the globalization of
research. American firms invested abroad earlier
than many foreign competitors, with much of this
investment going to Europe. For much of the past
twenty years, the ‘globalization’of innovation was,
in fact, the regionalization of research and devel-
opment among the US, the EU and Japan. These
three actors accounted for the bulk of global R&D
and when the research was conducted outside the
home country, it was located in one of the other
two economies. In 2002, US firms placed 61 per
cent of their foreign research and development
investment in the European Union, worth about
$15 billion (OECD, 2006b:123). Over half of
this investment went to two sectors: automobiles
and pharmaceuticals. The United States was a net
exporter of research and development activities
to Europe until the turn of the millennium, when
European firms expanded their activities in the US.
Trade in ideas, represented by flows of royalties,
117

fees and license income from industrial processes,
was remarkably balanced, with the US generally
running only a modest surplus on its trade with
Europe (NSF, 2007d).
The US was, by the early 2000s, the favored
location for foreign OECD-based multinationals,
many keen to capitalize on US leadership of the in-
formation technology sector. The US attracted the
bulk of all foreign R&D activity conducted in the
OECD (42 per cent). During the 1980s and 1990s
was the degree to which foreign firms began to
locate their research facilities in the United States,
perhaps to take advantage of basic research done at
US universities. The percentage of patents granted
to foreign firms was about 45 per cent by the mid-
1990s (Mowery, 1998). Access to the American
science base was particularly important for sectors
like pharmaceuticals, with European firms com-
ing to rely on their US operations for much of the
basic research work (Brusoni and Geuna, 2003).
Indeed, fully half of EU research and develop-
ment investment in the US was concentrated in
chemicals and pharmaceutical sectors (OECD,
2006b: 123). Indeed, in the pharmaceutical sector,
it is not an exaggeration to describe US-EU links
as a fully-fledged transatlantic innovation area.
Such corporate cooperation was underpinned by
broad congruence between the EU and the US on
various regulatory matters, not least contract law
and intellectual property.
Innovation is now even more globalised and,
as outlined above, Europe and America are both
experiencing a relative decline in their dominance
of the area. China has embarked on a remarkably
successful process of technological upgrading,
allow the country to act not merely as a host for
foreign R&D activity but to generate significant
amounts of indigenously-created research (Fu and
Gong 2011). It is notable that Europe is becom-
ing relatively less important as an R&D hub for
American firms, which are increasing the amount
of research work in Asia. In 1995, Europe ac-
counted for 70 per cent off all US firms’ overseas
research and development expenditures: by 2002,
118
Technology and the Multipolar Global Economy
this had dipped to 58 per cent (NSF, 2007b, Figure
22). Though American firms did increase their
R&D expenditures in Europe during the 1998-
2002, the rate of growth (4.8%) was dwarfed by
the rate of growth of American R&D investment
in Asia (28.6%) (NSF, 2007b). It is not hard to
understand the interest of American firms in this
regard. China has produced a stunning increase
in its innovative capacity, with R&D intensity
(research spending as a percentage of GDP) ris-
ing from 0.6 per cent in 1995 to 1.3 per cent in
2006. China’s pool of researchers grew 77 per cent
to 926,000 in the same period (OECD, 2006a).
Smaller developing states are also accounting
for more innovative activity, with the share of
global patenting accounted for by Brazil, China,
India and South Africa increased from 0.15 per
cent in 1991 to 0.58 per cent in 2004 (OECD,
2006a). A small part of the total to be sure, but
a dramatic increase and an indication of creative
talent that has not gone unnoticed by European
and American multinationals. What is good for
firms, and indeed the overall economy, may not
be good for individual European and American
workers however, hence the keen political interest
in the globalization and relocation of research and
development activities. Innovation policy may yet
come to have a political dynamic like that in trade
policy: the diffuse benefits of open trade attract
little attention compared to the concentrated losses
in industries. Consumers gain from innovation,
wherever it is created, but skilled workers are
not as mobile as their jobs and fear losing out to
competitors. Both Europe and America are strug-
gling to devise policy responses that accommodate
these conflicting pressures.
CONCLUSION
Europe is again in the midst of a crisis of confi-
dence about its ability to compete in knowledge-
intensive sectors of the globalized economy. Some
perspective needs to be kept: Germany and the
Technology and the Multipolar Global Economy
Nordic states enjoy strong innovation systems, and
the UK enjoys a significant exporter of ideas and
basic research. However, Europe’s market share
in high technology industries is under pressure,
and the continent is generally over-reliant on
a comparatively small set of very competitive
companies to maintain this market share. There
is a paucity of smaller firms with a significant
R&D presence.
European angst about this situation may yet
turn to something more troublesome. European
and American expectations and experiences were
crucial in shaping the international innovative
environment that allowed states such as Korea,
China and India to develop significant research
and development programs. Thus far, charges
of neomercantilism remain muted, but that may
change. Germany and Spain, for example, are
two European states with a significant stake in
both solar and wind energy generation. Should
the perception grow that China has acted unfairly
in developing its own industries, the resultant
trade conflict could undermine an already fragile
multilateral system.
REFERENCES
Archibugi, D., & Coco, A. (2005). Is Europe
becoming the most dynamic knowledge econ-
omy in the world? Journal of Common Market
Studies, 43(3), 433–459. doi:10.1111/j.0021-
9886.2005.00564.x
Athreye, S., & Cantwell, J. (2007). Creating
competition? Globalization and the creation of
new technology producers. Research Policy, 36,
209–226. doi:10.1016/j.respol.2006.11.002
Brou, D., Campanella, E., & Ruta, M. (2009, 25
November). The value of domestic subsidy rules
in trade agreements. (Staff Working Paper ESRD-
2009-12). World Trade Organization.
Chesbrough, H. (2003). The era of open innova-
tion. MIT Sloan Review, 44(3), 35–41.
Daveri, F. (2002). The new economy in Europe,
1992-2001. Oxford Review of Economic Policy,
18(3), 345–362. doi:10.1093/oxrep/18.3.345
Deutsche Bank. (2011). International division
of labor in R&D: Research follows production.
Deutsche Bank Economics Paper 82, February
2011.
Dosi, G., Llerena, P., & Labini, M. S. (2005).
Evaluating and comparing the innovation per-
formance of the United States and the European
Union. TrendChart Policy Workshop.
Drejer, I. (2004). Identifying innovation in sur-
veys of services: A Schumpeterian perspective.
Research Policy, 33, 551–562. doi:10.1016/j.
respol.2003.07.004
European Commission. (2007). The European
research area: New perspectives. Commission
Staff Working Document SEC (2007) 412/2.
European Commission. (2009). Annual report
on research and development. Retrieved May 3,
2010, from http://eur-lex.europa.eu/LexUriServ/
LexUriServ.do?uri=COM:2009:0558:FIN:EN:P
DF#page=2
European Commission. (2011a). Innovation union
scoreboard (IUS) 2010. Brussels, Belgium: Di-
rectorate for Enterprise and Industry.
European Commission. (2011b). Europe 2020.
Retrieved July 24, 2011, from http://ec.europa.
eu/europe2020/index_en.htm
Evenett, S. (2009). What can be learned from
crisis-era protectionism? An initial assessment.
Business and Politics, 11(3), article 4.
Falke, A. (2010). Much ado about nothing:
Obama’s trade policy after year one. Unpublished
manuscript, Freidrich-Alexander Universitiet,
Erlangen-Nuermberg, January 2010.
119

Fassbender, H. (2007). Europe’s productivity
challenge. The McKinsey Quarterly, 2.
Florida, R. (2002). The rise of the creative class.
New York, NY: Basic Books.
Forero-Pineda, C. (2006). The impact of stronger
intellectual property rights on science and tech-
nology in developing countries. Research Policy,
35, 808–824. doi:10.1016/j.respol.2006.04.003
Fu, X., & Gong, Y. (2011). Indigenous and
foreign innovation efforts and drivers of techno-
logical upgrading: Evidence from China. World
Development, 39(7), 1213–1225. doi:10.1016/j.
worlddev.2010.05.010
Giesecke, S. (2000). The contrasting roles of gov-
ernment in the development of biotechnology in-
dustry in the US and Germany. Research Policy, 29,
205–223. doi:10.1016/S0048-7333(99)00061-X
Global Subsidies Initiative. (n.d.). The WTO and
subsidies. International Institute for Sustainable
Development. Retrieved May 17, 2010, from
http://www.globalsubsidies.org/en/media-portal/
the-wto-and-subsidies
Goldstein, A., & McGuire, S. (2004). The political
economy of strategic trade policy and the Brazil-
Canada subsidies saga. World Economy, 27(4),
541–566. doi:10.1111/j.0378-5920.2004.00613.x
Hart, D. (2001). Antitrust and technological in-
novation in the US: Ideas, institutions, decisions
and impacts, 1890-2000. Research Policy, 30(6),
923–936. doi:10.1016/S0048-7333(00)00165-7
Hu, M.-H., & Matthews, J. (2008). China’s na-
tional innovative capacity. Research Policy, 37,
1465–1479. doi:10.1016/j.respol.2008.07.003
Kang, M. (2000). Trade policy mix under the
WTO: Protection of TRIPs and R&D subsidies.
KIEP Working Paper 00-11, Korea Institute for
International Economic Policy, December 2000.
120
Technology and the Multipolar Global Economy
McGuire, S. (2002). Between pragmatism and
principle. The International Trade Journal, 16(3).
McGuire, S. (2006). No more Euro-champions:
The interaction of EU industrial and trade poli-
cies. Journal of European Public Policy, 13(6),
887–905. doi:10.1080/13501760600838573
McGuire, S., & Smith, M. (2008). The European
Union and the United States: Competition and
convergence in the global arena. Basingstoke,
UK: Palgrave.
Munro, M. (2010). Amazon’s Kindle: Symbol of
America’s decline? The Brookings Institution,
February 25 201. Retrieved July 24, 2011, from
http://www.brookings.edu/opinions/2010/0225_
tech_muro.aspx
Ramani, S. (2002). Who’s interested in biotech?
R&D strategies, knowledge base and market sales
of Indian biotechnology firms. Research Policy,
31, 381–398. doi:10.1016/S0048-7333(01)00115-9
Schumpeter, J. (1942). Capitalism, socialism and
democracy. New York, NY: Knopf Publishers.
Smith, M., & Elgstrom, O. (2000). Introduction:
Negotiation and policy-making in the European
Union - Processes, system and order. Journal
of European Public Policy, 7(5), 673–683.
doi:10.1080/13501760010014894
Teteryaknikova, M. (2008). R&D in the net-
work of international trade: Multilateral versus
regional trade agreements. Staff Working Paper
ESRD-2009-03, World Trade Organization, 10
November 2008.
UNICE. (2007, 16 November). UNICE position
paper on cohesion policy and competitiveness.
Brussels, Belgium: UNICE. Retrieved January 7,
2007, from http://www.unice.org/DocShareNo-
Frame/Common/GetFileURL.asp?FileURL=F_4
Technology and the Multipolar Global Economy

Zhou, P., & Leydesdorff, L. (2006). The emer- ENDNOTES

gence of China as a leading nation in science.
Research Policy, 35, 83–104. doi:10.1016/j.
1
This section draws on McGuire and Smith,
respol.2005.08.006 2008

This work was previously published in Disruptive Technologies, Innovation and Global Redesign: Emerging Implications, edited
by Ndubuisi Ekekwe and Nazrul Islam, pp. 331-344, copyright 2012 by Information Science Reference (an imprint of IGI Global).

121
122

Chapter 11
Coupling GPS and GIS
Mahbubur R. Meenar
Temple University, USA

John A. Sorrentino
Temple University, USA

Sharmin Yesmin
Temple University, USA

ABSTRACT
Since the 1990s, the integration of GPS and GIS has become more and more popular and an industry
standard in the GIS community worldwide. The increasing availability and affordability of mobile GIS
and GPS, along with greater data accuracy and interoperability, will only ensure steady growth of this
practice in the future. This chapter provides a brief background of GPS technology and its use in GIS,
and then elaborates on the integration techniques of both technologies within their limitations. It also
highlights data processing, transfer, and maintenance issues and future trends of this integration.

INTRODUCTION a certain time, a series of places over time, or a

The use of the Global Positioning System (GPS)
as a method of collecting locational data for Geo-
graphic Information Systems (GIS) is increasing
in popularity in the GIS community. GIS data is
dynamic – it changes over time, and GPS is an
effective way to track those changes (Steede-Terry,
2000). According to Environmental Systems
Research Institute (ESRI) president Jack Dan-
germond, GPS is “uniquely suited to integration
with GIS. Whether the object of concern is moving
or not, whether concern is for a certain place at
DOI: 10.4018/978-1-4666-2038-4.ch011
place with no regard to time, GPS can measure it,
locate it, track it.” (Steede-Terry, 2000).
Although GIS was available in the market
in the 1970s, and GPS in the 1980s, it was only
in the mid-1990s that people started using GPS
coupled to GIS. The GPS technology and its
analogs (Global Navigation Satellite System or
GLONASS in Russia and the proposed Galileo
system in Europe) have proven to be the most
cost-effective, fastest, and most accurate meth-
ods of providing location information (Longley
et. al, 2005; Trimble, 2002; Taylor et. al, 2001).
Organizations that maintain GIS databases – be
they local governments or oil companies – can

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Coupling GPS and GIS
easily and accurately inventory either stationary
or moving things and add those locations to their
databases (Imran et. al, 2006; Steede-Terry, 2000).
Some common applications of coupling GPS and
GIS are surveying, crime mapping, animal track-
ing, traffic management, emergency management,
road construction, and vehicle navigation.
BACKGROUND
Need for GPS Data in GIS
When people try to find out where on earth they
are located, they rely on either absolute coordi-
nates with latitude and longitude information or
relative coordinates where location information is
expressed with the help of another location (Ken-
nedy, 2002). GIS maps can be created or corrected
from the features entered in the field using a GPS
receiver (Maantay and Ziegler, 2006). Thus people
can know their actual positions on earth and then
compare their locations in relation to other objects
represented in a GIS map (Thurston et. al, 2003;
Kennedy, 2002).
GIS uses mainly two types of datasets: (a)
primary, which is created by the user; and (b)
secondary, which is collected or purchased from
somewhere else. In GIS, primary data can be
created by drawing any feature based on given
dimensions, by digitizing ortho-photos, and by
analyzing survey, remote sensing, and GPS data.
Using GPS, primary data can be collected accu-
rately and quickly with a common reference system
without any drawing or digitizing operation. Once
the primary data is created, it can be distributed
to others and be used as secondary data. Before
using GPS as a primary data collection tool for
GIS, the users need to understand the GPS tech-
nology and its limitations.
The GPS Technology
The GPS data can be collected from a constellation
of active satellites which continuously transmit
coded signals to receivers and receive correctional
data from monitoring stations. GPS receivers
process the signals to compute latitude, longitude,
and altitude of an object on earth (Giaglis, 2005;
Kennedy, 2002).
A method, known as triangulation, is used
to calculate the position of any feature with the
known distances from three fixed locations (Le-
tham, 2001). However, a discrepancy between
satellite and receiver timing of just 1/100th of
a second could make for a misreading of 1,860
miles (Steede-Terry, 2000). Therefore, a signal
from a fourth satellite is needed to synchronize
the time between the satellites and the receivers
(Maantay and Ziegler, 2006; Longley et. al, 2005;
Letham, 2001). To address this fact, the satellites
have been deployed in a pattern that has each one
passing over a monitoring station every twelve
hours, with at least four visible in the sky all the
times (Steede-Terry, 2000).
The United States Navigation Satellite Timing
and Ranging GPS (NAVSTAR-GPS) constella-
tion has 24 satellites with 3 spares orbiting the
earth at an altitude of about 12,600 miles (USNO
NAVSTAR GPS, 2006; Longley et. al, 2005;
Steede-Terry, 2000). The GLONASS consists of
21 satellites in 3 orbital planes, with 3 on-orbit
spares (Space and Tech, 2005). The proposed sys-
tem GALILEO will be based on a constellation of
30 satellites and ground stations (Europa, 2005).
The NAVSTAR-GPS has three basic segments:
(1) the space segment, which consists of the satel-
lites; (2) the control segment, which is a network
of earth-based tracking stations; and (3) the user
segment, which represents the receivers that pick
up signals from the satellites, process the signal
data, and compute the receiver’s location, height,
and time (Maantay and Ziegler, 2006; Lange and
Gilbert, 2005).
123

Data Limitations and Accuracy Level
Besides the timing discrepancies between the
satellites and the receivers, some other elements
that reduce the accuracy of GPS data are orbit
errors, system errors, the earth’s atmosphere, and
receiver noise (Trimble, 2002; Ramadan, 1998).
With better attention to interoperability between
the GPS units, hardware, and software, some of
these errors can be minimized before the data are
used in GIS (Thurston et. al, 2003; Kennedy, 2002).
Using a differential correction process, the
receivers can correct such errors. The Differential
GPS (DGPS) uses two receivers, one stationary
and one roving. The stationary one, known as
the base station, is placed at a precisely known
geographic point, and the roving one is carried by
the surveyor (Maantay and Ziegler, 2006; Imran
et. al, 2006; Thurston et. al, 2003; Kennedy, 2002;
Taylor et. al, 2001; Steede-Terry, 2000). The base
station sends differential correction signals to the
moving receiver.
Prior to 2000, the GPS signal data that was
available for free did not deliver horizontal posi-
tional accuracies better than 100 meters. Data with
high degree of accuracy was only available to U.S.
government agencies and to some universities.
After the U.S. Department of Defense removed
the restriction in May 2000, the positional accu-
racy of free satellite signal data increased to 15
meters (Maantay and Ziegler, 2006). In September
2002, this accuracy was further increased to 1 to
2 meters horizontally and 2 to 3 meters vertically
using a Federal Aviation Administration funded
system known as Wide Area Augmentation Sys-
tem (WAAS). WAAS is available to the public
throughout most of the continental United States
(Maantay and Ziegler, 2006).
Depending on the receiver system, the DGPS
can deliver positional accuracies of 1 meter or
less and is used where high accuracy data is re-
quired (Maantay and Ziegler, 2006; Longley et.
al, 2005; Lange and Gilbert, 2005; Taylor et. al,
2001). For example, the surveying professionals
124
Coupling GPS and GIS
now use Carrier Phase Tracking, an application
of DGPS, which returns positional accuracies
down to as little as 10 centimeters (Maantay and
Ziegler, 2006; Lange and Gilbert, 2005).
INTEGRATION OF GPS AND GIS
The coupling of GPS and GIS can be explained
by the following examples:
• A field crew can use a GPS receiver to
enter the location of a power line pole in
need of repair; show it as a point on a map
displayed on a personal digital assistant
(PDA) using software such as ArcPad from
ESRI; enter attributes of the pole; and fi-
nally transmit this information to a central
database (Maantay and Ziegler, 2006).
• A researcher may conduct a groundwater
contamination study by collecting the co-
ordinates and other attributes of the wells
using a GPS; converting the data to GIS;
measuring the water samples taken from
the wells; and evaluating the water quality
parameters (Nas and Berktay, 2006).
There are many ways to integrate GPS data in
GIS, ranging from creating new GIS features in
the field, transferring data from GPS receivers to
GIS, and conducting spatial analysis in the field
(Harrington, 2000a). More specifically, the GPS-
GIS integration can be done based on the following
three categories – data-focused integration, posi-
tion-focused integration, and technology-focused
integration (Harrington, 2000a). In data-focused
integration, the GPS system collects and stores
data, and then later, transfers data to a GIS. Again,
data from GIS can be uploaded to GPS for update
and maintenance. The position-focused integration
consists of a complete GPS receiver that supplies
a control application and a field device application
operating on the same device or separate devices.
In the technology-focused integration, there is
Coupling GPS and GIS
no need for a separate application of a device to
control the GPS receiver; the control is archived
from any third party software (Harrington, 2000a).
Figure 1 provides an example of a schematic
workflow process of the GPS-GIS integration by
using Trimble and ArcGIS software. In short, the
integration of GPS and GIS is primarily focused
on three areas - data acquisition, data processing
and transfer, and data maintenance.
Data Acquisition
Before collecting any data, the user needs to de-
termine what types of GPS techniques and tools
will be required for a particular accuracy require-
ment and budget. The user needs to develop or
collect a GIS base data layer with correct spatial
reference to which all new generated data will be
referenced (Lange and Gilbert, 2005).
Figure 1. Example workflow process of GPS-GIS
integration
The scale and datum of the base map are also
important. For example, a large-scale base map
should be used as a reference in a site specific
project in order to avoid data inaccuracy. While
collecting GPS data in an existing GIS, the datum
designation, the projection and coordinate system
designation, and the measurement units must be
identical (Kennedy, 2002; Steede-Terry, 2000). It
is recommended that all data should be collected
and displayed in the most up-to-date datum avail-
able (Lange and Gilbert, 2005).
The user may create a data dictionary with
the list of features and attributes to be recorded
before going to the field or on-spot. If it is created
beforehand, the table is then transferred into the
GPS data collection system. Before going to the
field, the user also needs to find out whether the
locations that will be targeted for data collection
are free from obstructions. The receivers need a
clear view of the sky and signals from at least
four satellites in order to make reliable position
measurement (Lange and Gilbert, 2005; Giaglis,
2005). In the field, the user will check satellite
availability and follow the manuals to configure
GPS receivers before starting data collection.
GIS uses point, line, and polygon features, and
the data collection methods for these features are
different from one another. A point feature (e.g.,
an electricity transmission pole) requires the user
to remain stationary at the location and capture
the information using a GPS device. For a line
feature (e.g., a road), the user needs to record the
positions periodically as s/he moves along the
feature in the real world. To capture a polygon
feature (e.g., a parking lot) information, the posi-
tions of the recorder are connected in order to form
a polygon and the last position always connects
back to the first one. The user has to decide what
types of features need to be created for a GIS map.
In a small scale map, a university campus can be
shown as a point, whereas in a detailed map, even
a drain outlet can be shown as a polygon.
GPS coordinates can be displayed in real time
in some GIS software such as ESRI ArcPad, In-
125

tergraph Intelliwhere, and Terra Nova Map IT. In
the age of mobile GIS, users can go on a field trip,
collect GPS data, edit, manipulate, and visualize
those data, all in the field. While GPS and GIS
are linked, the GPS receiver can be treated as the
cursor of a digitizer. It is linked to the GIS through
a software module similar to a digitizer control-
ler where data are saved into a GIS filing system
(Ramadan, 1998; UN Statistics Division, 2004).
In real-time GPS/GIS integration, data may be
collected and stored immediately for future use in
a mapping application, or data may be discarded
after use in a navigation or tracking application
(Thurston et. al, 2003).
For example, Map IT is a new GIS software
designed for digital mapping and GPS data capture
with a tablet PC. The software connects a tablet pc
to a GPS antenna via a USB port. While conduct-
ing the field work, the user may use the software
to: (a) display the current ground position on the
tablet PC’s map display in real time; (b) create new
features and add coordinates and other attributes;
(c) edit or post-process the data in real time; and
(d) automatically link all activity recorded in the
field (including photographs, notes, spreadsheets,
and drawings) to the respective geographic posi-
tions (Donatis and Bruciatelli, 2006).
Although the integration of GIS and GPS can
in general increase accuracy and decrease project
costs and completion time, it can also create new
problems, including creation of inaccurate data
points and missing data points (Imran et. al, 2006).
Sometimes a handheld GPS navigator may not
be able to acquire a lock on available satellites
because of natural conditions like dense forest
canopies, or human-made structures like tall
buildings or other obstacles (Lange and Gilbert,
2005; Thurston et. al, 2003). Data collection with
GPS also might get affected by any equipment
malfunction in the field.
126
Coupling GPS and GIS
Data Processing and Transfer
Once the data are collected, they can be down-
loaded, post-processed, and exported to GIS
format from the field computer to the office
computer. Where real-time signals are needed but
cannot be received, the post-processing techniques
can be applied to re-process the GPS positions.
Using this technique, the feature positions can
be differentially corrected to the highest level of
accuracy. The users who integrate GPS data into
their own applications need to consider how and
when they should apply differential corrections.
Real-time processing allows recording and cor-
recting a location in seconds or less, but is usually
less accurate. Post-processing allows the surveyor
recording a location as much time as s/he likes,
and then differentially corrects each location back
in the office. This technique is used in mapping
or surveying (Steede-Terry, 2000; Thurston et.
al, 2003). Instead of relying on real-time DGPS
alone, the users should enable their applications
to record raw GPS data and allow post-processing
techniques to be used either solely or in conjunc-
tion with real-time DGPS (Harrington, 2000b).
Most GPS receiver manufacturers have their
own data file format. GPS data is stored in a
receiver in its own format and later can be trans-
lated to various GIS formats (Lange and Gilbert,
2005; Ramadan, 1998). Data can be transferred
in a couple of ways. One simple way is collecting
coordinates and attributes in a comma delimited
file from the GPS device storage. The other more
preferable way is converting the data from GPS
storage to the user-specific database interchange
format using a data translation program (Lange
and Gilbert, 2005). Such a program allows the
user to (1) generate metadata; (2) transform the
coordinates to the projection, coordinate system,
and datum of the user’s choice; and (3) translate
GPS data into customized formats that the GPS
manufacturers could never have anticipated
(Lange and Gilbert, 2005).
Coupling GPS and GIS
A number of file interchange protocols are
available to exchange data between different
brands and types of receivers. One widely used
interchange protocol is the Receiver Independent
Exchange Format (RINEX), which is supported
by most satellite data processing software (Yan,
2006). Another commonly used interface standard
is a standard released by the National Marine
Electronics Association (NMEA). Most GPS
receivers support this protocol and can output
NMEA messages, which are available in ASCII
format (Yan, 2006).
Data Maintenance
For data revisions or data maintenance, GIS data
is transferred back to the field computer and can
be verified or updated in the field. The user can
relocate features via navigation, verify the position
and attribute features, and navigate to locations
to collect new attribute data. The user may select
features and examine them in the field, modify
attributes, and even collect new features if desired.
Using receivers such as Trimble, any feature that
has been added or updated is automatically marked
to determine which data needs to go back to GIS
(Trimble, 2002).
FUTURE TRENDS
The future trends of GIS-GPS integration will be
focused on data accuracy, interoperability, and
affordability. In order to make the WAAS level of
precision available to users worldwide, the Unites
States is working on international agreements to
share similar technologies available in other parts
of the world, namely Japan’s Multi-Functional
Satellite Augmentation System (MSAS) and
Europe’s Euro Geostationary Navigation Overlay
Service (EGNOS) (Maantay and Ziegler, 2006).
In addition, the European satellite positioning
system, Galileo, will be dedicated to civilian ac-
tivities which will further increase the availability
of accurate data to general users.
New applications of GIS-GPS integration are
constantly becoming popular and widespread. The
latest developments in GPS technology should
encourage more use of such integration in the
future. Reduction in cost and personnel training
time of using GPS technology with high data ac-
curacy will eventually provide a cost-effective
means of verifying and updating real time GIS
mapping in the field (Maantay and Ziegler, 2006;
UN Statistics Division, 2004).
CONCLUSION
In today’s market, the mobile GIS and GPS devices
are available with greater accuracy at a reduced
cost. The data transfer process from GPS to GIS
has become faster and easier. GIS software is get-
ting more powerful and user friendly, and GPS
devices are increasingly getting more accurate
and affordable. The integration of GIS and GPS
has been already proven to be very influential in
spatial data management, and it will have steady
growth in the future.
REFERENCES
Donatis, M., & Bruciatelli, L. (2006). Map it:
the GIS software for field mapping with tablet
PC. Computers & Geosciences, 32(5), 673–680.
doi:10.1016/j.cageo.2005.09.003
Europa web site. http://www.eu.int/comm/dgs/
energy_transport/galileo /index_en.htm, accessed
on December 12, 2005
Giaglis, G. (2005). Mobile Location Services. In
M. Khosrow-Pour (Ed.), Encyclopedia of Infor-
mation Science and Technology, 4, 1973-1977.
Pennsylvania: Idea Group Reference.
127

Harrington, A. (2000a). GIS and GPS: Technolo-
gies that Work Well Together. Proceedings in the
ESRI User Conference, San Diego, California.
Harrington, A. (2000b). GPS/GIS Integration:
What Can You Do When Real-Time DGPS Doesn’t
Work? GeoWorld, 13(4). Available online at http://
www.geoplace.com/gw/2000/0400/0400int.asp,
accessed on August 25, 2006.
Imran, M., Hassan, Y., & Patterson, D. (2006).
GPS-GIS-based procedure for tracking vehicle
path on horizontal alignments. Computer-Aided
Civil and Infrastructure Engineering, 21(5),
383–394. doi:10.1111/j.1467-8667.2006.00444.x
Kennedy, M. (2002). The Global Positioning
System and GIS: An Introduction. New York:
Taylor and Francis.
Lange, A., & Gilbert, C. (2005). Using GPS for
GIS Data Capture. In Geographic Information
Systems: Principles, Techniques, Management,
and Applications (pp. 467-476). NJ: John Wiley
& Sons, Inc.
Letham, L. (2001). GPS made easy. Washington.
The Mountaineers, 5(12), 183–186.
Longley, P., Goodchild, M., Maguire, D., & Rhind,
D. (2005). Geographic Information Systems and
Science. New Jersey: John Wiley & Sons, Inc.
(pp. 122-123, 172-173).
Maantay, J., & Ziegler, J. (2006). GIS for the Urban
Environment. California: ESRI Press, 306-307.
Nas, B., & Berktay, A. (2006). Groundwater
contamination by nitrates in the city of Konya,
(Turkey): a GIS perspective. Journal of Environ-
mental Management, 79(1), 30–37. doi:10.1016/j.
jenvman.2005.05.010
Ramadan, K. (1998). The Use of GPS for GIS
Applications. Proceedings in the Geographic
Information Systems: Information Infrastructures
and Interoperability for the 21st Century Informa-
tion Society, Czech Republic.
128
Coupling GPS and GIS
Space and Tech web site. http://www.spacean-
dtech.com/spacedata/constellations/glonass_con-
sum.shtml, accessed on December 12, 2005
Steede-Terry, K. (2000). Integrating GIS and
the Global Positioning System. California: ESRI
Press.
Taylor, G., Steup, D., Car, A., Blewitt, G., &
Corbett, S. (2001). Road reduction filtering for
GPS-GIS navigation. Transactions in GIS, 5(3),
193–207. doi:10.1111/1467-9671.00077
Thurston, J., Poiker, T., & Moore, J. (2003). In-
tegrated Geospatial Technologies – A Guide to
GPS, GIS, and Data Logging. New Jersey: John
Wiley & Sons, Inc.
Trimble Navigation Limited. (2002). TerraSync
Software – Trimble’s Productive Data Collection
and Maintenance Tool for Quality GIS Data.
California: Trimble Navigation Limited.
UN Statistics Division. (2004). Integration of GPS,
Digital Imagery and GIS with Census Mapping.
New York: United Nations Secretariat.
USNO NAVSTAR GPS web site. http://tycho.
usno.navy.mil/gpsinfo.html, accessed on August
26, 2006.
Yan, T. (2006). GNSS Data Protocols: Choice and
Implementation. Proceedings in the International
Global Navigation Satellite Systems Society IG-
NSS Symposium, Australia.
KEY TERMS AND DEFINITIONS
Coordinate System: A reference framework
used to define the positions of points in space in
either two or three dimensions.
Datum: The reference specifications of a
measurement system, usually a system of coor-
dinate positions on a surface or heights above or
below a surface.
Coupling GPS and GIS

Differential GPS (DGPS): The DGPS is Projection: A method requiring a systematic

used to correct GPS signal data errors, using two
receivers, one stationary (placed at a precisely
known geographic point) and one roving (carried
by the surveyor). The stationary receiver sends
differential correction signals to the roving one.
GPS Segment: GPS consists of three seg-
ments: (i) space segment – the GPS satellites, (ii)
user segment – the GPS handheld navigator, and
(iii) ground control segment – the GPS monitor-
ing stations.
mathematical transformation by which the curved
surface of the earth is portrayed on a flat surface.
Scale: The ratio between a distance or area
on a map and the corresponding distance or area
on the ground, commonly expressed as a fraction
or ratio.
Wide Area Augmentation System: (WAAS):
The WAAS is a system that can increase the GPS
signal data accuracy to 1 to 2 meters horizontally
and 2 to 3 meters vertically.

This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 277-284,
copyright 2009 by Information Science Reference (an imprint of IGI Global).

129
130

Chapter 12
Using Geospatial Techniques
to Analyze Landscape
Factors Controlling Ionic
Composition of Arctic Lakes,
Toolik Lake Region, Alaska
Prasad A. Pathak
University of North Carolina at Greensboro, USA

Stephen Whalen
University of North Carolina at Chapel Hill, USA

ABSTRACT
The impacts of climate change on landscapes in arctic Alaska are evident in terms of permafrost melting,
frequent thermokarst activity, and the occurrence of more broadleaf vegetation. These changes may alter
natural biogeochemical cycles of ions along with major nutrients and affect ionic compositions of lakes,
as they are connected with the landscapes. However, the nature of the connectivity between lakes and
landscapes in this region is not yet explored. The authors propose that geospatial analysis of landscape
properties along with observed lake ion concentrations will enable an understanding of the currently
existing landscape controls over ion inputs into the lakes. For the watersheds of 41 lakes in the Arctic
Foothills region of Alaska, spatial properties of natural vegetation communities expressed in terms of
percentage, shape complexity, and patch density metrics were derived using satellite data. Regression
analyses were performed for concentration of ions as well as conductivity in lake water where the spa-
tial metrics along with lake physical properties, lake order, and glacial till age categories were used
as predicting variables in the regression. Landscape metrics for major land covers i.e., Percentage of
Moist Acidic Tundra (MAT) and Moist Non-acidic Tundra (MNT) were the major predicting variables
for concentration of several ions.

DOI: 10.4018/978-1-4666-2038-4.ch012

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
INTRODUCTION
The ionic composition of lakes plays a significant
role in controlling their trophic structure. Anions
and cations are important for many aquatic pri-
mary producers as they control various cellular
processes. Further, they are crucial to maintain the
pH of surface waters, creating favorable condi-
tions for several prokaryotes and plants to grow
(Smith, 1995).
In closed drainage basins, the ionic composi-
tion is controlled by the evaporation rate and pre-
cipitation (Lesack & Melack, 1991). Specifically
within low arctic watersheds, it has been observed
that transport of Na+ and K+ was governed by
precipitation whereas Ca2+ and Mg2+ concentra-
tions in lakes were related to chemical weathering
(Cornwell, 1992). For lakes in the coastal region,
atmospheric inputs were considered to be im-
portant. However, weathering of parent material
(i.e., soil and rock) is considered to be the major
source of ions for lakes (Wetzel, 2001). Weather-
ing processes are affected by several landscape
factors, with land use and land cover being the
most prominent (Drever, 1994).
Likens et al. (1977) and Drever (1994) em-
phasized that natural vegetation aids chemical
weathering of parental material via natural biota
in the root zone. In particular, the mineralization
and immobilization processes carried out by these
microorganisms are crucial for recycling ions
(Frost et al., 2006; Xiao & Ji, 2007; Canham et
al., 2004). It has been further explained that these
processes along with soil moisture and hydrologi-
cal events determine influx of ions into surface
water bodies. Therefore, it is necessary to derive
information about land covers to understand
their influence on water bodies. With the advent
of remote sensing technology, satellite datasets
have been used to obtain such information at
various scales (Jensen, 2005). Limnological stud-
ies, which have adopted remote sensing data, are
considered to be constrained only to above ground
information and not really inferred about root
zone interactions (Gergel et al., 1999; Griffith et
al., 2002).In contrast, King et al. (2005) argued
that if remote sensing data is converted into spa-
tial and structural information about land cover
researchers will be able to understand indirectly
root zone interactions and weathering processes.
Such spatial and/or structural information, when
expressed in terms of landscape metrics, allows
quantitative analyses for lake-landscape interac-
tions (O’Neill et al., 1997).
For example, Stewart et al. (2001) studied
Proportion and Fragmentation metrics related to
riparian vegetation. The authors considered ripar-
ian zones as regions of high water fluctuations and
correlated landscape metrics for the riparian zones
with the trophic structure of lakes. Similarly, Patch
Density of a land cover is another such metric
useful for investigating water quality (Johnson
et al., 1997). Patch Shape Complexity metrics,
mostly used in wildlife studies (Forman & Gordon,
1986), are also adopted in limnological studies
to understand growth stages of land covers. For
instance, prominent land covers within basins
will occupy most of the area and exhibit simple
shapes. On the other hand, emerging land covers
will display more complex shapes. It is likely that
the dominant land cover and the emerging land
cover may exhibit different root zone properties
and influence export of ions into water bodies ac-
cordingly. Using the same principle, Kearns et al.
(2005) have demonstrated that shape complexity
metrics can be applied to understand landscape
controls over ionic composition of lakes.
The arctic lakes also should be studied care-
fully for their ionic composition and associated
landscape controls, particularly when the climate
change is altering the dynamic between landscapes
and lakes. Arctic landscapes are highly sensitive
to the contemporary air temperature increase and
are therefore, experiencing severe alterations
(ACIA, 2005). Permafrost is rapidly melting due
to rising temperatures, which are also inducing
more frequent thermokarst activity. Both of these
processes could release stored ions within soils
131
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
into streams and lakes. Hence, the ion concentra-
tions of surface water may be altered in the near
future due to terrestrial inputs (Keller et al., 2007).
Previous studies quantified ions within particu-
lar vegetation communities (Marion et al., 1989;
Schimel et al., 1989; Giblin et al., 1991). These
studies indicated that soil moisture influences the
concentration of ions and vegetation communities
having higher amount of moisture in the soil of
their root zones exhibited higher leaf concentra-
tions of common ions. For example, moist soils
hosting broadleaf shrubs had higher potassium
contents compared to other vegetation types
with lower soil moisture (Marion et al., 1989).
Similarly, vegetation communities having deep
root structures were found to have more access
to ions and soil micronutrients. Therefore, these
communities were thought to be sinks for ions,
prohibiting release to surfaces waters (Schimel
et al., 1996).
Only recently, a more comprehensive study
about geochemical weathering and its effects on
soil and streams has been carried out by Keller et
al. (2007). In the Arctic Foothills region of Alaska,
the authors compared the relative concentrations
of Ca2+, K+, and P within certain streams, soils,
permafrost layer and bed rocks. They concluded
that permafrost has relatively higher concentra-
tions of these elements compared to active mineral
soil layer. By comparing the same bed rock mate-
rial on glacial tills of different ages, the authors
confirmed that differences in concentration of
these elements among the till surfaces was a
function of age. They concluded that weathering
plays a more important role than parent material.
Moreover, they predicted that with climate change,
active soils will receive an increased amount of
the elements stored in permafrost leading to higher
concentrations in the streams.
It is evident that these studies adopted a non-
spatial approach; however, they were helpful in
identifying important landscape factors and pro-
pose the underlying mechanism of lake-landscape
interactions, which is the focus of the current
132
research. In this study, a more comprehensive ap-
proach has been adopted to investigate currently
existing relationships between landscapes and
concentration of major ions in arctic lakes. We
analyzed lake water chemistry data previously
collected by our research group to evaluate rela-
tionships between common ions and vegetative
characteristics of the surrounding watershed.
Opting for geospatial techniques, we proposed
that spatial properties of different vegetation
communities derived from remote sensing data,
at the watershed level and within the vicinity of
water channels, would be useful indicators of the
delivery of common ions from landscapes to lakes.
The spatial knowledge of landscape factors was
derived from a satellite image and other second-
ary datasets. Age of the till surfaces forming the
watersheds and order of lakes (Riera et al., 2000)
along with certain physical properties of the lakes
were also included in the analyses as predicting
landscape factors.
Establishment of a set of landscape factors for
arctic watersheds with respect to the concentra-
tions of common ions was the primary aim of our
analysis. With adaptation of a geospatial approach
we aimed to create a broadly applicable methodol-
ogy so that other arctic regions could be analyzed
in the similar fashion.
METHODS
Study Area
The area under consideration was the Toolik Lake
region (68o 38’ N/ 149o 36’ W), situated in the
Arctic Foothills Province of Alaska, just north
of the Brooks Range (Figure 1). Regionally, lake
formation is by ice scour, thermokarst activity and
melting of ice blocks in glacial drift and outwash
plains (kettle lakes). The average local relief of
the area is 750 m. Hillocks, exposed barren areas,
and moraines characterize the entire landscape.
Water tracks, streams, and rivers, dissect the area.
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

This region has experienced multiple glaciations
in the mid-Pleistocene and late Pleistocene eras.
Study lakes are located in both the Sagavanirktok
and Toolik areas and therefore sited on glacial
drifts of varying age. Age of Itkillik I advance
obtained by carbon dating is 53,000 B.P. Follow-
ing that, the Itkillik II advance ranged from 25
to 11.5 ka. The Sagavanirktok River glaciation
is from Mid-Pleistocene time period, having an
estimated age ranging from 780,000 to 125,000
years B.P. (Hamilton, 1986). Age of these glacial
drifts has played an important role with respect
to soil pH, with older drifts showing more acidic
soils (Walker et al., 1994).
Data
Spatial Datasets
a. A SPOT (Satellite Pour L’Observation de la
Terre; 5 HRVIR) image acquired on July 25th,
2005 is used for this research. The image is
provided after merging the original panchro-
matic and multispectral images obtained at
the same time. The image has a pixel size
of 5 meters X 5 meters and it has 3 spectral
bands: Band1 (0.50 – 0.59 μm), Band2
(0.61 – 0.68 μm) and Band3 (0.79 – 0.89
μm). The SPOT Corporation did not provide
any specifications about pre-processing
and merging technique used to derive the
product. Traditionally such merged products
are used only for visual inspection, and they
are not used for digital classification as they
do not conserve original spectral proper-
ties. However, new techniques similar to
wavelet based fusion and High Pass filter
fusion retain original spectral properties of
the multispectral image used in the merging
process, allowing digital classification of
the end product (Gangkofner et al., 2008).
This study is based on the merged satellite

Figure 1. Location map of the study area and SPOT 5(2005) image snapshot (False Color image with
green, red, near infra-red bands) with lake groups under consideration

133
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

image so the nature of spectral properties
of major land covers was verified in the
ERDAS Imagine using the Spectral Profile
tool as well as NDVI ranges for land covers
compared with earlier studies. The image
encompassed approximately a 60 x 60 square
kilometer area surrounding the study region.
A Universal Transverse Mercator (UTM)
projection, zone 6, is applied to geometrically
correct the image (Figure 1).
b. Digital Elevation Model (DEM): A 5 meter
X 5 meter DEM of the study area is used to
delineate individual watersheds for each lake
using hydrology tools in ArcGIS. The DEM
data was collected by Intermap Technologies
using an airborne Star 3i X-Band radar sys-
tem. The horizontal accuracy of the data was
2.5 meter RMSE with a 1 centimeter verti-
cal resolution and 3 meters RMSE (Nolan,
2003).
c. Lake properties: Lake Order depicting
their landscape position was derived using
topographic maps of the area. The Strahler
stream order of the outlet streams for lakes
was followed to assign the order to each
lake. Lake area, perimeter, and shoreline
development factors were derived from
satellite data.
inorganic carbon [DIC; predominately HCO3-])
as well as specific conductance.
Derivation of Landscape Factors
Landscape factors were derived using following
steps:
1. Watershed Delineation: Hydrology tools
and Conditional tools in the Spatial Analyst
extension of ArcGIS 9.3 were used to derive
watersheds. For each lake an approximate
portion of the entire DEM was initially se-
lected. Using the portion of DEM as input,
Flow direction and Flow Accumulation
surfaces were derived. Based on the Flow
Accumulation surface and the SPOT image,
pour points for individual watersheds were
determined. The pour point and the respec-
tive Flow Direction surface were used to
delineate the watershed for every lake.
2. Digital Classification: These watersheds
were used to divide the SPOT image. The im-
ages obtained for each watershed then were
utilized for calculation of the Normalized
Difference Vegetation Index (NDVI) layer,
which represents vigor of vegetation but it
is also characteristic for different vegetation

Non-Spatial Datasets
Table 1. Lakes and sampling dates
Maximum lake depth, light attenuation coef- Lake # Sampling Date
ficients within the water column, euphotic zone 115, 116, 117, 118 29 June 2001
depth (depth of penetration of 1% surface irradi- 119, 120, 121, 122 7 July 2001
ance), and till surface age category were observed 127, 128, 129, 130 8 July 2002
during field campaigns. 131, 132, 133, 134 10 July 2002
135, 136, 137, 138 12 July 2002
Concentrations of Ions
139, 140, 141, 142 28 June 2003
143, 144, 145, 146 3 July 2003
Forty one lakes (Table 1) were sampled over three
147, 148, 149, 150 14 July 2003
different thaw seasons (2001-2003) to determine
151, 152, 153, 154 24 July 2003
concentrations of common anions and of cations
E4 28 July 2003
(Ca2+, Mg2+, Na+, K+, SO42-, Cl- and dissolved
123, 124, 125, 126 4 Aug 2003

134
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

communities (Jensen, 2005). The layer of
NDVI was added to spectral bands of each
watershed prior to classification with the
aim of improving the accuracy of the digital
classification.
The scheme for major land cover types for
this study was a combination of classification
systems used by Walker et al. (1994) and field
modifications done by Stine and Ray (Table 2).
The main categories were as follows.
The ISODATA clustering method for unsuper-
vised classification was run on the NDVI combined
image of each watershed. Initially 100 spectral
classes were derived and finally merged to achieve
the desired land cover categories. The Snowbed
complex category was derived after initial unsu-
pervised classification. A knowledge-based clas-
sification approach was used for identifying
snowbeds (Appendix B, Figure 5). As field con-
ditions described by Walker et al. (1994), Snow-
beds occur over gentle north facing slopes. The
NDVI and spectral value ranges observed for
Snowbed complexes in Stine et al. (2010) were
adopted and modified for this study. Using these
criteria, rules were developed in the Knowledge
Engineer available with ERDAS Imagine (Ap-
pendix B, Figure 5). The resultant Snowbed
complex layers were merged with the initially
classified image of each watershed. During the
summer season of 2008, six watersheds (Table 3)
were visited to collect ground truth information
at the points generated by a stratified random
method. The data was used to assess accuracy of
classification of these watersheds. The average
accuracy obtained was approximately (88.6%),
which was above the suggested minimum accu-
racy level for ecological studies (Congalton et al.,
1991). Lake GTH 153 and the nearby 3 watersheds
were affected by shadows of clouds. Even though
the shadow effect reduced accuracy for this group
of watersheds, classifying at watershed level
provided much better control to minimize the
effect.

Table 2. Vegetation communities and species composition

Class Description
Barren complex Barren surfaces, sparsely vegetated, rocks covered with lichens
Moist Acidic Tussock Tundra (MAT) Eriophorum vaginatum, Carex bigelowii, Betula nana, Salix pulchra, Sphagnum spp.
complex
Moist Non-acidic Tussock Tundra Salix reticulate, Saxifraga oppositifolia, Cares bigelowii, Carex membranacea, Dryas integri-
complex folia, Ledum decumbens, Equisetum ravens
Shrub complex Betula spp., Salix spp., Sphagnum spp.
Riparian complex Eriophorum anguistifolium, Salix pulchra, Salix alaxensis, Salix richardisonii
Fen Complex Carex rariflora, Carex rotundata, and mosses like Sphagnum spp., Carex chordorrhiza, Carex
aquitilis, and Tomentypnum nitens
Heath complex Festuca altaica, Empetrum hermaphroditum, Loiseleuria procumbens, Dryas octopetala, Cas-
siope tetragona, Salix phlebophylla
Snowbed complex Cassiope tetragona, Salix rotundifolia, Arnica frigida
Mountain Meadow complex Carex biglowii, Cassiope tetragona, Salix chamissonis
Aquatic vegetation complex Similar to fen but on lake-fringes
Water Lakes, streams, and rivers
Cloud
Shadows

135
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

Table 3. Overall accuracy values for six water-

Percentage
sheds visited during summer of 2008

Watershed Overall Accuracy (%) Kappa Once every watershed was completely classified,
GTH 120 95 0.92
an Area column was calculated. The same column
GTH 133 86.36 0.82
was used to calculate percentage of each land
GTH 135 82.75 0.78
cover within watersheds.
GTH 144 85.29 0.75
Patch Density
GTH 149 93.75 0.89
GTH 153 88.88 0.85
Patch Density (PD) is expressed in terms of number
of patches of a vegetation type per 100 hectares.
It was calculated at the class level. To analyze if
Derivation of Buffer Zones Along Water
overall spatial distribution properties of any par-
Channels
ticular land cover class would be controlling ion
inputs to lakes, Patch Density was incorporated
To extract land covers within the proximity of
in this research:
water channels, it was necessary to first derive
the water channels. Using Con tool in the Hydrol-
ni
ogy toolset of ArcGIS, water channels for each PD =
A
(10000)(100)
watershed were identified. A cell value of 500
for the Flow Accumulation surface was used for
the process. where, ni represents the number of patches and A
The value was determined based on visual is total area of the landscape. This value is multi-
inspection of the Flow Accumulation surface and plied by 10,000 and 100 to convert to hectares.)
observed tracks on the SPOT image. Based on our
field experience regarding soil moisture around Edge Density
water tracks, a buffer distance of 20 meters along
the channels was utilized to extract land covers. Edge Density (ED) is the number of edges of the
These land covers were assumed to experience given class per unit area (meters per hectare). The
higher soil moisture and therefore a greater prob- greater the Edge Density, the more fragmented
ability of ion exchange within their root zones the class under consideration. Similar to Patch
(Giblin et al., 1991). Density, the Edge Density index indicates the
extent of fragmentation for a land cover class
Fragstats/ArcGIS within the given watershed.

m
As explained in Table 3, landscape metrics for
proportion, fragmentation, and shape complexity ED =
∑ e
k =1 ik
(10, 000)
A
were calculated using Fragstats 3.3. Thematic
maps of the watersheds were used as inputs for
this process, where Spatial Analyst extension of Here, eik represents sum of total edge in meters.
ArcGIS was necessary for recognizing inputs in A is the area of the landscape in square meters and
IMG format. is multiplied by 10,000 to convert to hectares.

136
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
Landscape Shape Index
Landscape Shape Index (LSI) depicts the total edge
of a particular class (if this index is calculated at
class level), divided by the minimum possible
edge length for that class. Overall it is a more
direct measure of disaggregation of the land cover
class than Patch Density or Edge Density index.
E soil. For example, riparian complex zones would
LSI =
min E
Shape Index (Mean)
Mean Shape Index (MSI) measures the average
patch shape, or the average perimeter-to-area ratio
for a particular patch type (class) or for all patches
in the landscape. SHAPE = 1 when the patch is
circular (vector) or square (raster) and increases
without limit as the patch shape becomes more
irregular.
(0.25) p ij
MSI =
aij
For raster format of inputs pij represents sum
of the patch perimeter, aij represents patch area,
0.25 is a constant to adjust for a square standard.
Fractal Dimension Index (Mean)
Units: None. Range: 1 < FRACT < 2.
A fractal dimension greater than 1 for a 2-dimen-
sional patch indicates a departure from Euclidean
geometry, i.e., an increase in shape complex-
ity. FRACT approaches 1 for shapes with very
simple perimeters such as circles or squares, and
approaches 2 for shapes with highly convoluted,
plane-filling perimeters.
2 ln (0.25 pij )
FRACT =
ln aij
Here, pij represents patch perimeter and aij
represents patch area.
Shape indices were included in the current re-
search to obtain indirect information about active
soil depth and probable moisture content of the
occur only near major streams and have greater
soil depth compared to other vegetation classes
(Giblin et al., 1995). Structural properties such
as very small patches (Mean Patch Size index) of
riparian complex would indicate that only small
areas near streams have deeper active soil zone
and may either be source or sink of nutrients.
It is well established that many landscape
metrics provide redundant information (Turner et
al., 1989). To identify highly correlated metrics,
we initially performed correlation analysis. It was
found that the Edge Density metric was highly
correlated with the Patch Density metric; hence,
it was omitted from the main analysis. To suit
the analysis, the shape complexity metric was
converted into percent change from 1, which is
the numerical expression for the simplest shape.
ANALYSIS
Various statistical methods such as Correlation
Analysis, Analysis of Variance (ANOVA), Prin-
ciple Component Analysis (PCA), and Multiple
Regression Analysis are used in similar limnologi-
cal studies (Johnson & Guage, 1997). Advanced
methods such as Path Analysis are also useful
(Liu et al., 1997). However, regression analysis
is the most widely used of the analytical meth-
ods, because it is easy to interpret and replicate
(Webster et al., 1996).
In the current research, the backward elimi-
nation method of regression was adopted. The
landscape metrics along with non spatial attributes
137
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

of lakes such as lake order, euphotic zone depth
and maximum depth were treated as predicting
variables. Separate regression models were devel-
oped for estimates of each major ion as well as
estimates of specific conductance. The standard-
ized regression coefficients obtained for individual
predicting variables were used as indicators of
their relative contributions to respective models.
RESULTS
The statistically significant models obtained at
the confidence interval of 95% are represented in
Table 4. Appendix A (Table 6) provides explana-
tion about the acronyms utilized in these models.
The regression model for specific conductance
estimates explained 80% of the variability in the
data using the landscape metrics percentage of
MNT complex, Euphotic Zone Depth, and Patch
Density of Shrub complex observed within 20
meter buffers of water channels. The standardized
coefficient obtained for the variables (Prct_MNT
[0.772], EZD [0.364], and PD_B_Shrub [0.329]),
indicated that Prct_MNT has more influence over
the specific conductance estimates compared to
the other variables.
Approximately 75% of the variance within the
calcium estimates was attributed to Percentage of
MNT complex, the Euphotic Zone Depth, and the
Landscape Shape Index of MM complex. Accord-
ing to the standardized coefficients for the predict-
ing variables, the contributions from Prct_MNT
(0.571) and EZD (0.496) were roughly equal, but
LSI_MM (0.28) had a lesser contribution.
The model developed for magnesium was able
to explain 74% of variance within its estimates
using Percentage of MAT complex, presence
of the Till Age 2 category, and Euphotic Zone
Depth. We obtained standardized coefficients
for three of the predicting variables (Prct_MAT
[-0.546], Till Age Category 2 [0.3120], and EZD

Table 4. Landscape metrics

Type Metrics Level

Proportion Percentage Watershed, Buffer
Fragmentation Patch density, Edge density, Fractal Dimension, Patch size Watershed, Buffer
Shape Complexity Mean Shape Index, Landscape Shape Index Watershed, Buffer
Note: Basics concepts of landscape metrics were obtained from Fragstats documentation (McGarigal et al., 2002).

Table 5. Regression models for ions and conductivity

Dependent Variable Regression Model Obtained r2

Conductivity Conductivity = -29.426 + 2.246 (Prct_MNT) + 7.17 (EZD) + 0.195 (PD_B_Shrub) 0.80
Calcium Ca = -2.202 + 0.368 (Prct_MNT) + 2.13 (EZD) + 0.317 (LSI_MM)
2+
0.75
Magnesium Mg2+ = 2.415 – 0.037 (Prct_MAT) + 0.983 (Till Age 2) + 0.117 (EZD) 0.74
Sodium Na = 0.309 – 0.006 (Prct_MAT)
+
0.50
Potassium K = 0.212 – 0.003 (Prct_MAT) + 0.019 (EZD) + 0.114 (Till age 1)
+
0.63
Sulfate SO42- = 4.351 + 0.059 (PD_MM) – 0.407 (LSI_B_Fen) + 0.215 (LSI_ Shrub) 0.37
Chloride Cl = 0.186 + 0.042 (Prct_Heath) – 0.304 (Till age 2) + 0.001 (PD_ Fen) + 0.023 (MS_B_Fen)
-
0.60
+ 0.023 (MS_B_MM_Prct)
DIC DIC = 22.966 + 0.893 (Prct_MNT) + 27.86 (Till Age 2) – 5.169 (FD_B_Shrub) + 3.146 (EZD) 0.90
+ 0.596 (LSI_MM)

138
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
[0.227]). As the contribution is estimated by the
value and not by its sign, Prct_MAT was the most
significant variable in the model and EZD was
the least significant.
Only 50% of its variance was explained by the
regression model developed for sodium. The only
landscape factor included in the regression model
was the Percentage of MAT complex.
In contrast, more than 63% of variation with
potassium estimates was explained by the regres-
sion model. The Percentage of MAT complex,
Euphotic Zone Depth, and the presence of till
age 1 category were the predicting variables in
this model. Similar to the models for sodium and
magnesium, the amount of MAT complex within
the watershed was the most significant predicting
variable. The standardized coefficients obtained
for the predicting variables in this model were
Prct_MAT (-0.511), EZD (0.393), and Till Age
Category 1 (0.331).
Among the anions included in this study, the re-
gression model for sulfate was developed based on
Patch Density of MM complex, Landscape Shape
Index of Fen complex present within buffer zone,
and Landscape Shape Index of Shrub complex.
This model was able to explain only 37% of the
total variance within sulfate estimates. The Land-
scape Shape Index metrics for Fen complex within
buffer zone was the most significant contributor
to this model with a standardized coefficient of
-0.512. For comparison, the standardized coef-
ficients for PD_MM and LSI_Shrub were 0.401
and 0.398, respectively.
Percentage of Heath, Till Age Category 2, Patch
Density of Fen complex, Mean Shape Index of
Fen complex occurring within buffer zones, and
percent change from a value of 1 in the Mean
Shape Index of MM complex were the landscape
factors found to be significant for the regression
model developed for chloride. While depicting
60% of the variance within chloride estimates,
the standardized coefficients for predicting vari-
ables (Prct_Heath (0.568), Till_age2 (-0.322),
PD_Fen (0.212), and MS_B_MM_Prct (0.364)),
indicated that Prct_Heath contributed more than
other variables.
The regression model for DIC was obtained
using Percentage of MNT, Till Age category 2,
Fractal Dimension of Shrub within buffer zones
of channels, Euphotic Zone Depth, and Landscape
Shape Index of MM complex. These landscape
metrics were responsible for 65% of total variance.
The standardized regression coefficients obtained
for these variables (Prct_MNT [0.493], Till Age
Category 2 [0.372], FD_B_Shrub_Prct [-0.241],
EZD [0.256], and LSI_MM [0.224]) indicated that
the most significant contribution was displayed
by the Prct_ MNT.
DISCUSSION
The MAT complex and the MNT complex are the
major land cover types within the Arctic Foothills
Province of Alaska (Walker et al., 1994). It has
been observed that the MAT complex is prevalent
over geologically relatively older tills whereas
as MNT complex prominently occupies younger
tills (Walker et al., 1994). According to Walker
et al. (2009), the older glacial surfaces have ac-
cumulated wind-blown glacial silt leading to more
peat formation with higher soil moisture. All these
processes were responsible for introduction of
Sphagnum moss leading to acidic soils on older
tills. It also allowed development of well-drained
soil surfaces resulting in weathering of calcium
rich parent material, adding to acidity of the
soils. Similarly, Keller et al. (2007) indicated that
younger tills have relatively higher concentra-
tions of calcium and other ions, which are more
susceptible to physical weathering. The glacial
geology maps of this region are available only at
a 1:250,000 scale, which is coarse with respect to
the watershed level analysis in the current research
(Hamilton, 1980). Also, on the ground, we have
observed that the MAT and MNT complexes in
ArcGIS are often intermixed and co-occur on
older as well as younger tills. However, they
139
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
follow the generalized rule of till age and tundra
type dominance.
Hence, we suggest that landscape metrics
related to MAT and MNT complex are more in-
dicative of till surface properties than the actual
contribution by the land covers. For example,
the positive relationship between calcium and
Prct_MNT indicates that more calcium could be
released from the landscape dominated by the
MNT complex. Thus, geologically young tills,
which are active ion stores, would release more
calcium. Similarly, magnesium, potassium, and so-
dium were negatively correlated with Prct_MAT,
suggesting that older tills which are dominated by
MAT have already experienced weathering and
most of the ions have leached and not contribut-
ing towards contemporary lake water chemistry.
MAT and MNT complexes are major land
covers within the Arctic Foothills Province of
Alaska, with the MAT complex is prevalent over
glacially older tills and the MNT complex domi-
nant over relatively younger tills (Walker et al.,
1994). Wind-blown glacial silt has allowed older
glacial tills to have a more active soil depth and
increased soil moisture. As a result, well-drained
surfaces could be observed over older tills. These
tills, therefore, have been chemically weathered
and have lost calcium and other ions. At the same
time, older tills have an accumulation of higher
amounts of peat accompanied by increased growth
of Sphagnum moss rendering soils more acidic
(Walker et al., 2009). These factors lead to the
establishment of the MAT complex over older
till surfaces. The MAT complex has vegetation
species with deep root structures and with rela-
tively higher productivity rates. Therefore, we
considered the presence of MAT as an indicator
of ion contribution via breakdown of organic soil
content and transportation of ions through surface
and sub-surface water to lakes rather than an index
of actual weathering.
In contrast, younger tills have a shallow ac-
tive soil layer and do not display well developed
drainage system. The permafrost layer, which is
140
repository of cations and anions, is near to the sur-
face. Therefore, the MNT vegetation type, which
is prominently observed over younger till surfaces,
could be indicative of the possible release of ions
from weathering and thawing of permafrost layer
during the growing season. This logic is also sup-
ported by reportedly higher concentrations of Ca2+,
Fe3+, K+, and Na+ over Sagavanirktok glaciation
surfaces, which are relative younger than Itkillik
III surface (Keller et al., 2007).
Hence, we suggest that the positive relationship
of MNT to ions such as calcium is more indicative
of the relative abundance of ions in the younger
till surfaces. In contrast, the negative correlation
exhibited by magnesium, potassium, and sodium
with Prct_MAT could be indicative of retention
of these ions by the MAT complex along with
lower concentration of ions within active soil on
older tills. This is illustrated by considering lakes
GTH 119 and GTH 153 which show 8% and 43%
MAT complex within their respective watersheds
(Figure 2). The Na+ concentration in GTH 119 was
2.1 mg L-1, but was only 0.4 mg L-1 in GTH 153.
Lake GTH 119 is located over young till (Itk II),
whereas the GTH 153 watershed is situated on
an older glacial till. Thus, the higher concentra-
tion of Na+ within GTH 119 could be the result
of weathering and release from the permafrost
layer, unlike GTH 153 where MAT would utilize
all nutrients and ions for plant growth.
The broad leaf vegetation communities such
as Shrub complex and MM complex exhibited
their influence over ionic composition via either
Patch Density or other shape metrics. Patch Den-
sity of Shrub complex within buffer zones was
positively correlated with specific conductance,
indicating that if a watershed has a high number
of Shrub patches, it would exhibit higher spe-
cific conductance. Specific conductance of the
water is a representation of the overall ionic
strength of the water, irrespective of the relative
contributions of dissolved constituents. Hence,
in this study the regression model developed for
specific conductance was able to explain more
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
Figure 2. GTH 119 and GTH 153 MAT showing patches
variance than any regression model developed
for individual ions.
Vegetation species found in the Shrub complex
have deeper root systems, pointing to the presence
of a relatively greater soil depth (Giblin et al.,
1991). Sited within the buffer zones of seasonal
streams, the Shrub complex would experience
more water flux, aiding ion release from the root
zones to the nearby surface waters. To further il-
lustrate this concept, two representative watershed
examples are provided in Figure 3. GTH 125 has
a very few number of patches within buffer zones
of streams, therefore, having low patch density
(7.4 per 100 hectares). In the watershed of GTH
138, however, more patches of Shrub complex
were present within the buffers leading to higher
patch density (52.7 per 100 hectares). Therefore,
for the GTH 125 watershed, the lower number of
patches may indicate a relatively lower contribu-
tion of Shrub complex towards ion release from
root zones into streams and eventually into the
lake. The result could be the lower conductance
observed for GTH 125 (93.5μS cm-1). In GTH
138, however, the higher patch density could
have increased ion release from root zones of the
Shrub complex into seasonal streams, leading to
a specific conductance of 193.4μS cm-1 of the
lake water.
The same mechanism at watershed level could
be responsible for positive correlation between
Landscape Shape Index of Shrub complex and
sulfate estimates. According this relationship, as
the patches of Shrub complex are becoming in-
creasingly irregular in shape, more sulfate is
observed in the lakes. Complexity of patches
would be higher when patches are either growing
or subjected to fragmentation processes. Hence,
it could be inferred that complex patches would
enhance the release of sulfate ions from soils into
surface waters. We also observed a negative cor-
relation between fractal dimensions of Shrub
patches occupying buffer zones around water
channels indicating that complex patches of Shrubs
within the buffer zones would act as sinks for DIC
being transported to lake waters.
Shape complexity of MM complex expressed
in terms of Landscape Shape Index, exhibited a
positive relationship with calcium and DIC. MM
complexes are responsible for physical weathering
due to their root growth (Marion et al., 1989). This
weathering activity makes carbonates available
for exchange from parent material. MM complex
and lichens are responsible for primary physical
and biochemical weathering of parent material
on higher slopes (Fedina, 1978). The currently
observed relationship may be suggesting us that
the changes in shape complexity of MM complex
patches will have greater impact on weathering
parent material and eventually concentration and
transport of Ca2+ and DIC to lakes.
141
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
Figure 3. GTH 125 and GTH 138 showing Shrub patches
To support our argument, the following two
watersheds are helpful (Figure 4). The GTH 125
watershed has a lower number of contiguous
patches of MM complex, hence a low LSI index
(16.7). In contrast, the MM complex in the GTH
118 watershed has been dissected by Shrub com-
plex and other land covers leading to an increased
number of patches distributed over a large area
resulting in a higher LSI index of 33.1. As noted
earlier, the contribution of MM complex in physi-
cal weathering of parent material would be lower
in GTH 125, resulting in lower carbonates and
lake waters show a DIC concentration of 65.9
mg L-1. In comparison, the GTH 118 watershed
shows a, relatively higher possibility of physical
weathering by MM complex and a higher DIC
concentration of 81.2 mg L-1 in lake water.
We found that the Shape complexity of Fen
communities was related to sulfate and chloride
concentrations. More complex Fen patches
142
(higher Landscape Shape Index) within buffer
zones of water channels displayed a negative
correlation with sulfate concentrations in the lakes,
indicating that complex patches of Fen commu-
nity within buffers may act as sinks for this anion.
While studying carbon mineralization in the Fen
complex within the Upper Peninsula of Michigan,
Keller and Bridgham (2007) noted that sulfate
reduction is the most likely mechanism of or-
ganic carbon mineralization in anaerobic condi-
tions within vegetative units similar to the Fen
complex. This observation supports the proposed
role of Fen complex as sink for sulfate. W GTH
133 and similar watersheds showed a very few
contiguous patches within buffers of streams
(LSI_B_Fen 12.1), whereas some watersheds like
GTH 120 had many small patches distributed
within the buffer zones of streams (LSI_B_Fen
41.05). The measured sulfate concentration in
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
Figure 4. GTH 125 and GTH 118 showing MM complex patches
lake GTH 133 was 20.6 mg L-1 compared with 2
mg L-1 in lake GTH 120.
In contrast, the positive relationship between
Mean Shape Index of Fen complexes occurring in
buffers and chloride estimates indicates that more
complex patches of Fen act as a source of chloride
for Arctic Foothill lakes. Fen complexes are fed
by ground water and/or permafrost melt (Cooper
& Wolf, 2006), indicating that the water chemistry
of Fen complexes is governed by parent material
of the surface. There could also be a relationship
between pH and variations in chemical properties
of Fen complex (Walker et al., 1994). Detailed
investigation of chemical parameters within Rich
Fen and Poor Fen communities may help reveal
the underlying mechanism of cations and anions
transport to and from Fen communities.
Even though not the most significant predictor
by itself, till age played a major role in explaining
variance for some of the cations and anions in the
study. We observed that Till Age Category 2, which
was assigned to the Itk2 surface, was positively
correlated with magnesium and DIC estimates.
The younger category (Till Age Category 1)
representing the Itk2 readvance was positively
correlated with potassium concentrations in the
lakes. These observations follow the findings by
Keller et al. (2007), who stated that streams on
younger surfaces have higher concentrations of
cations, especially K+. However, we could not find
any logical reason behind the negative correlation
between chlorides and Till Age Category 2.
Euphotic Zone Depth exhibited a positive
correlation with specific conductance, calcium,
magnesium, potassium, and DIC estimates. These
relationships were likely fortuitous because eu-
photic depth plays a direct role for photosynthetic
activity (Wetzel, 2001) rather than controlling
concentration of cations and anions in the lakes.
Overall, landscape metrics were successfully
able to explain more than 50% of the variance for
most of the ions. Usually the ionic composition of
lakes is directly controlled by geological settings
and weathering phenomena. However, there was
no detailed spatial geological information avail-
able for the study region. Landscape metrics,
especially, Prct_MAT and Prct_MNT were able
to overcome this limitation and served as proxies
of geological conditions or till surfaces. Shape
Complexity metrics were useful in inferring prob-
able root zone conditions and were also useful in
inferring weathering connections with concen-
trations of ions within lakes. At the same time,
complexity of patches helped to understand the
role of vegetation communities as either a source
or sink of various ions. However, more specific
studies are required to understand thoroughly the
landscape controls over ion concentration in arctic
143
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
lakes. Our study could be considered a primary
initial effort in this regard.
This research could be enhanced with availabil-
ity of detailed geological maps. Currently, avail-
able maps for Toolik lake region are at 1:250,000
scale and do not include lake groups such as GTH
153 and surrounding lakes. At the same time, there
are certain factors which might have impacted our
findings. For example, the lake chemistry data was
collected during seasons (Table 1) for various lake
groups. As mentioned by Cornwell (1992), each
season might have experienced different patterns
of precipitation and insolation, resulting in the
observed ion concentrations. Moreover, all the
lake chemistry data was collected only once during
the thaw season. Thus, the current research does
not capture within season variation as attempted
by Cornwell (1992) and Marion et al. (1989). The
satellite data used for this research was from a
different date than the lake water sample collec-
tion. However, with then available access to the
region along with funding and time limitations,
our research represents a unique attempt to link
landscape factors with lake-ion chemistry.
To further enhance the study, we propose to
include metrics related to only MAT, MNT, and
broad leaf vegetation categories as the current
results suggest their strong correlation with ionic
compositions of lakes. Expanding our data set to
include additional lakes would also improve the
predictive power of regression models. Further, if
additional data were available, it would be useful
to adopt more advanced analytical techniques such
as Classification And Regression Trees (CART).
CONCLUSION
This study shows that the spatial and structural
properties of MAT complex, MNT complex, and
broad leaf vegetation such as Shrub complex
and MM complex are useful indicators of ionic
composition of lakes in the Arctic Foothills region
144
of Alaska. These landscape factors suggest that
vegetation types within watersheds are a result
of climatic conditions, dominant weathering
processes, and they influence the ion recycling
via interactions with soil moisture and root zone
processes.
In the near future, arctic lakes will experience a
changing influence of factors governing ion inputs
due to climate change. It is expected that the natural
transformation of MNT into MAT with aging of
tills will be accelerated with rising temperatures
(Walker et al., 2001). Also, there will be increase in
the amount of broad leaf vegetation such as Shrub
complex in arctic watersheds (ACIA, 2005). The
previous studies of Pienitz et al. (1997), Marion
et al. (1989), Oechel (1989), and Cornwell (1992)
have highlighted various probable mechanisms
for transportation of nutrients from land surfaces
into surface water bodies. With implementation of
geospatial technology and remote sensing datasets,
we were able to establish a set landscape metrics
useful and repeatable for other arctic watersheds.
This research could also be expanded to study
watersheds influenced by human activities. A
combination of human induced processes such as
deforestation and natural factors would make a
key set of predicting landscape factors for surface
water quality (Steedman, 2000).
ACKNOWLEDGMENT
This research was funded by the National Science
Foundation (Grant 051604).
REFERENCES
ACIA. (2005). Arctic climate impact assessment:
Scientific report. Cambridge, UK: Cambridge
University Press.
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
Canham, C. D., Pace, M. L., Papaik, M. J.,
Primack, A. G. B., Roy, K. M., & Maranger, R.
J. (2004). A spatially explicit watershed-scale
analysis of dissolved organic carbon in Adirondack
Lakes. Ecological Applications, 14(3), 839–854.
doi:10.1890/02-5271
Congalton, R. G. (1991). A review of assessing
the accuracy of classifications of remotely sensed
data. Remote Sensing of Environment, 37, 35–46.
doi:10.1016/0034-4257(91)90048-B
Cooper, D. J., & Wolf, E. C. (2006). Fens of the
Sierra Nevada, California. Retrieved April 14,
2011, from http://www.rigelstuhmiller.com/evan/
CooperWolfSierraFensFinalReport2006.pdf
Cornwell, J. C. (1992). Cation export from Alaskan
arctic watersheds. Hydrobiologia, 240, 15–22.
doi:10.1007/BF00013448
Drever, J. I. (1994). The effect of land plants on
weathering rates of silicate minerals. Geochi-
mica et Cosmochimica Acta, 58(10), 2325–2332.
doi:10.1016/0016-7037(94)90013-2
Fedina, A. E. (1978). The structure of landscape
altitudinal zonality on the northern slopes of the
Greater Caucasus in the Baksan Basin, USSR.
Arctic and Alpine Research, 10(2), 437–429.
doi:10.2307/1550774
Forman, R. T. T., & Gordon, M. (1986). Landscape
ecology. New York, NY: John Wiley & Sons.
Fragstats Conceptual Background. (n.d.). Land-
scape pattern metrics. Retrieved December
22, 2007, from http://www.umass.edu/landeco/
research/fragstats/documents/Conceptual%20
Background/Background%20TOC.htm
Frost, P. C., Larson, J. H., Johnston, C. A.,
Young, K. C., Maurice, P. A., Lamberti, G. A., &
Bridgham, S. D. (2006). Landscape predictors
of stream dissolved organic matter concentration
and physicochemistry on a Lake Superior river
watershed. Aquatic Sciences, 68, 40–51.
Gangkofner, U. G., Pradhan, P. S., & Holcomb,
D. W. (2008). Optimizing the high-pass filter
addition technique for image fusion. Photogram-
metric Engineering and Remote Sensing, 74(9),
1107–1118.
Gergel, S. E., Turner, M. G., & Kratz, T. K. (1999).
Dissolved organic carbon as an indicator of the
scale of influence on lakes and rivers. Ecological
Applications, 9(4), 1377–1390. doi:10.1890/1051-
0761(1999)009[1377:DOCAAI]2.0.CO;2
Giblin, A. E., Nadelhoffer, K. J., Shaver, G. R.,
Laundre, J. A., & McKerrow, A. J. (1991). Biogeo-
chemical diversity along a riverside toposequence
in arctic Alaska. Ecological Monographs, 61,
415–436. doi:10.2307/2937049
Griffith, J. A., Martinko, E. A., Whistler, J. L., &
Price, K. P. (2002). Interrelationships among land-
scapes, NDVI, and stream water quality in the US
central plains. Ecological Applications, 12, 1702–
1718. doi:10.1890/1051-0761(2002)012[1702:IA
LNAS]2.0.CO;2
Hamilton, T. D. (1980). Surficial geologic map of
the Killik River quadrangle, Alaska. 1:250,000.
Washington, DC: U.S. Geological Survey.
Hamilton, T. D. (1986). Late Cenozoic glaciation
of the Central Brooks Range. In Hamilton, T. D.,
Reed, K. M., & Thorson, R. M. (Eds.), Glacia-
tion in Alaska: The geologic record (pp. 9–49).
Anchorage, AK: Alaska Geological Society.
Jensen, J. R. (2005). Introductory digital image
processing: A remote sensing perspective (3rd ed.).
Upper Saddle River, NJ: Prentice Hall.
Johnson, L. B., & Gauge, S. H. (1997). Landscape
approaches to the analysis of aquatic ecosystems.
Freshwater Biology, 37, 113–132. doi:10.1046/
j.1365-2427.1997.00156.x
145
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
Johnson, L. B., Richards, C., Host, G. E., &
Arthur, J. W. (1997). Landscape influence on
water chemistry in Midwestern stream ecosys-
tems. Freshwater Biology, 37(1), 193–218.
doi:10.1046/j.1365-2427.1997.d01-539.x
Kearns, F. R., Kelly, N. M., Carter, J. L., & Resh,
V. H. (2005). A method for the use of landscape
metrics in freshwater research and management.
Landscape Ecology, 20, 113–115. doi:10.1007/
s10980-004-2261-0
Keller, J. K., & Bridgham, S. D. (2007). Path-
ways of anaerobic carbon cycling across an
ombrotrophic-minerotrophic peatland gradient.
Limnology and Oceanography, 52(1), 96–107.
doi:10.4319/lo.2007.52.1.0096
Keller, K., Blum, J., & Kling, G. W. (2007). Geo-
chemistry of soils and streams on surfaces of vary-
ing ages in arctic Alaska. Arctic, Antarctic, and
Alpine Research, 39, 84–98. doi:10.1657/1523-
0430(2007)39[84:GOSASO]2.0.CO;2
King, R. S., Baker, M. E., Whigham, D. F., Weller,
D. E., Jordan, T. E., Kazyak, P. F., & Hurd, M.
K. (2005). Spatial considerations for linking
watershed land cover to ecological indicators in
streams. Ecological Applications, 15, 137–153.
doi:10.1890/04-0481
Lesack, L. F. W., & Melack, J. M. (1991). The
deposition, composition, and potential sources of
major ionic solutes in rain of the central Amazon
Basin. Water Resources Research, 27, 2953–2977.
doi:10.1029/91WR01946
Likens, G. E., Bormann, F. H., Pierce, R. S., Eaton,
J. S., & Johnson, N. M. (1977). Biogeochemistry
of forested ecosystem. New York, NY: Springer.
doi:10.1007/978-1-4615-9993-7
Liu, Z., Hallberg, G. R., & Malanson, G. P. (1997).
Structural equation modeling of dynamics of
nitrate contamination in ground water. Journal
of the American Water Resources Association,
33(6), 1219–1225. doi:10.1111/j.1752-1688.1997.
tb03548.x
146
Marion, G. M., Hastings, S. J., Oberbauer, S.
F., & Oechel, W. C. (1989). Soil-plant element
relationships in a tundra ecosystem. Holarctic
Ecology, 12, 296–303.
McGarigal, K., Cushman, S. A., Neel, M. C., &
Ene, E. (2002). FRAGSTATS: Spatial pattern
analysis program for categorical maps. Retrieved
April 11, 2011, from http://www.umass.edu/lan-
deco/research/fragstats/fragstats.html
Nolan, M. (2003). Watershed-scale analyses of
an arctic drainage using a new high resolution
digital elevation model. Retrieved April 14, 2011,
from http://www.drmattnolan.org/olderprojects/
kuparuk/drmatt_kup_agu.htm
O’Neill, R. V., Hunsekar, C. T., Jones, K. B.,
Riiters, K. H., Wickham, J. D., & Schwartz, P.
M. (1997). Monitoring environmental quality at
the landscape scale. Bioscience, 47(5), 513–519.
doi:10.2307/1313119
Pienitz, R., Smol, J. P., & Lean, D. R. S. (1997).
Physical and chemical limnology of 24 lakes
located between Yellowknife and Contowoyto
Lake, Northwest Territories (Canada). Canadian
Journal of Fisheries and Aquatic Sciences, 54,
347–358. doi:10.1139/f96-275
Riera, J. L., Magnuson, J. J., Kratz, T. K., & Web-
ster, K. E. (2000). A geomorphic template for the
analysis of lake districts applied to the Northern
Highland Lake District, Wisconsin, USA. Fresh-
water Biology, 43, 301–318. doi:10.1046/j.1365-
2427.2000.00567.x
Schimel, J. P., & Clein, J. S. (1996). Microbial
response to freeze-thaw cycles in tundra and
taiga soils. Soil Biology & Biochemistry, 28,
1061–1066. doi:10.1016/0038-0717(96)00083-1
Smith, R. J. (1995). Calcium and bacteria. Ad-
vances in Microbial Physiology, 37, 83–133.
doi:10.1016/S0065-2911(08)60144-7
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition
Steedman, R. J. (2000). Effects of experimental
clearcut logging on water quality in three small
boreal forest lake trout (Salvelinus namaycush)
lakes. Canadian Journal of Fisheries and Aquatic
Sciences, 57, 92–96. doi:10.1139/f00-119
Stewart, J. S., Wang, L. Z., Lyons, J., Hor-
watich, J. A., & Bannerman, R. (2001). Influ-
ences of watershed, riparian-corridor, and
reach-scale characteristics on aquatic biota in
agricultural watersheds. Journal of the American
Water Resources Association, 37, 1475–1487.
doi:10.1111/j.1752-1688.2001.tb03654.x
Stine, R., Chaudhuri, D., Ray, P., Pathak, P., &
Hall-Brown, M. (2010). Comparing image clas-
sification techniques for Arctic tundra land cover,
Toolik Lake, Alaska. GI Science and Remote Sens-
ing, 47(1), 78–98. doi:10.2747/1548-1603.47.1.78
Turner, M. G. (1989). Landscape ecology: The ef-
fect of pattern on process. Annual Review of Ecol-
ogy and Systematics, 20, 171–197. doi:10.1146/
annurev.es.20.110189.001131
Walker, D. A., Bockheim, J. G., Chapin, F. S. III,
Eugster, W., Nelson, F. E., & Ping, C. L. (2001).
Calcium-rich tundra, wildlife, and the “Mam-
moth Steppe.”. Quaternary Science Reviews, 20,
149–163. doi:10.1016/S0277-3791(00)00126-8
Walker, D. A., Hamilton, T. D., Munger, C. A.,
& Maier, H. A. (2009). Glacial history and long-
term ecology in the Toolik Lake region. In Arctic
LTER (Ch. 3). Retrieved April 14, 2011, from
http://ecosystems.mbl.edu/StaffWeb/jhobbie/
arcbook/chapter3_environmental_settings.pdf,
draft chapter
Walker, M. D., Walker, D. A., & Auerbach, N. A.
(1994). Plant communities of a tussock tundra
landscape in the Brooks Range Foothills, Alaska.
Journal of Vegetation Science, 5, 843–866.
doi:10.2307/3236198
Webster, K. E., Kratz, T. K., Bowser, C. J., Mag-
nuson, J. J., & Rose, W. J. (1996). The influence
of landscape position on lake chemical responses
to drought in northern Wisconsin. Limnology and
Oceanography, 41(5), 977–984. doi:10.4319/
lo.1996.41.5.0977
Wetzel, R. G. (2001). Limnology: Lake and river
ecosystems (3rd ed.). New York, NY: Academic
Press.
Xiao, H., & Ji, W. (2007). Relating landscape char-
acteristics to non-point source pollution in mine
waster-located watersheds using geospatial tech-
niques. Journal of Environmental Management,
82, 111–119. doi:10.1016/j.jenvman.2005.12.009
147
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

APPENDIX A

Table 6. Acronyms for landscape factors

Landscape Metrics Vegetation Community Acronym

Aquatic Vegetation Complex Prct_AV
Moist Acidic Tundra Complex Prct_MAT
Moist Non-Acidic Tundra Complex Prct_MNT
Shrub Complex Prct_Shrub
Fen Complex Prct_Fen
Percentage
Heath Complex Prct_Heath
Mountain Meadow Complex Prct_MM
Riparian Complex Prct_Rip
Snowbed Complex Prct_SB
Barren/Thermokarst Prct_BT

Aquatic Vegetation Complex PD_AV

Moist Acidic Tundra Complex PD_MAT
Moist Non-Acidic Tundra Complex PD_MNT
Shrub Complex PD_Shrub
Fen Complex PD_Fen
Patch Density
Heath Complex PD_Heath
Mountain Meadow Complex PD_MM
Riparian Complex PD_Rip
Snowbed Complex PD_SB
Barren/Thermokarst PD_BT

Aquatic Vegetation Complex MS_AV

Moist Acidic Tundra Complex MS_MAT
Moist Non-Acidic Tundra Complex MS_MNT
Shrub Complex MS_Shrub

Mean Shape Index Fen Complex MS_Fen

(Percent change) Heath Complex MS_Heath
Mountain Meadow Complex MS_MM
Riparian Complex MS_Rip
Snowbed Complex MS_SB
Barren/Thermokarst MS_BT

continued on following page

148
Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

Table 6. Continued

Landscape Metrics Vegetation Community Acronym

Aquatic Vegetation Complex LSI_AV
Moist Acidic Tundra Complex LSI_MAT
Moist Non-Acidic Tundra Complex LSI_MNT
Shrub Complex LSI_Shrub
Fen Complex LSI_Fen
Landscape Shape Index
Heath Complex LSI_Heath
Mountain Meadow complex LSI_MM
Riparian Complex LSI_Rip
Snowbed Complex LSI_SB
Barren/Thermokarst LSI_BT

Aquatic Vegetation Complex FD_AV

Moist Acidic Tundra Complex FD_MAT
Moist Non-Acidic Tundra Complex FD_MNT
Shrub Complex FD_Shrub
Fen Complex FD_Fen
Fractal Dimension (Percent Change)
Heath Complex FD_Heath
Mountain Meadow Complex FD_MM
Riparian Complex FD_Rip
Snowbed Complex FD_SB
Barren/Thermokarst FD_BT

Aquatic Vegetation Complex MP_AV

Moist Acidic Tundra Complex MP_MAT
Moist Non-Acidic Tundra Complex MP_MNT
Shrub Complex MP_Shrub
Fen Complex MP_Fen
Mean Patch Size
Heath Complex MP_Heath
Mountain Meadow Complex MP_MM
Riparian Complex MP_Rip
Snowbed Complex MP_SB
Barren/Thermokarst MP_BT

Euphotic Zone Depth EZD

Lake Order Lake Order
Maximum Depth Max_Depth
Shoreline Development Factor SDF
Note: When these indices were calculated at buffer level, a suffix of “_B_” was added in their acronyms.

149
 Using Geospatial Techniques to Analyze Landscape Factors Controlling Ionic Composition

APPENDIX B

Figure 5. Expert classification rule used for snowbed complex

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 3,
edited by Donald Patrick Albert, pp. 37-57, copyright 2012 by IGI Publishing (an imprint of IGI Global).

150
 151

Chapter 13
Mobile Geographic
Information Systems
Yang Li
University of East London, UK

Allan J. Brimicombe
University of East London, UK

ABSTRACT
This chapter introduces the concept of Mobile Geographical Information Systems (Mobile GIS) as an
evolution of conventional GIS to being available on wireless mobile devices such as smart phones. The
evolution of the technology and its applications are charted in this chapter. The main elements of Mobile
GIS are then discussed. This focuses on: GIS servers; wireless mobile telecommunication networks;
wireless mobile devices; location-awareness technology; and gateway services. This is followed by a
discussion of the main features in terms of the services and usage of Mobile GIS: mobility; real-time
connectivity; location-awareness; broadened usage. Mobile Geographical Information Systems are an
important facilitating technology for Location-Based Services (LBS). A range of applications of Mobile
GIS for smart phones are described. The chapter closes with a discussion of the prospects and challenges
for Mobile GIS. Challenges derive from four broad areas: limitations that derive from the technologies
being used; areas of GIScience that still need to be adequately researched; users; and business models
for a sustainable presence.

INTRODUCTION With the fast development and increasingly wide

Mobile Geographic Information Systems (Mobile
GIS) can be viewed as an outcome of the conver-
gence of wireless mobile technologies and GIS
to have real-time and mobility characteristics.
DOI: 10.4018/978-1-4666-2038-4.ch013
usage of smart phones, Mobile GIS have become
embedded in many Location–Based Services
(LBS) applications, particularly in navigation
functions, which are provided via smart phones.
Mobile GIS can also be considered as a core part
of LBS. Mobile GIS integrates GIS and wireless
mobile communication technologies to offer a

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

real-time capability of interacting with external
software and remotely accessing and managing
data and information.
In this Chapter, the development of Mobile GIS
from conventional GIS is presented in the context
of GIS functionalities and mobile technologies and
smart phones. The principles and main features
of Mobile GIS are addressed in a systematic way,
covering topics regarding its real-time connectiv-
ity, mobility and location-awareness. Mobile GIS
takes advantage of mobile technologies (e.g. wire-
less networks, mobile devices, smart phones) to
offer the possibility of exchanging and analysing
spatial information in a real geographic world in
real-time. Mobile GIS also offers the potential for
users to employ GIS in more mobile and diverse
situations and for a range of applications through
smart phones. Mobile GIS are further presented
in the context of Location-Based Services (LBS).
LBS have been made possible due to the matura-
tion and convergence of a range of heterogeneous
technologies, such as mobile phones, the Internet
and the Web, global positioning system (GPS) and
GIS. LBS are defined as the delivery of spatial
data and information services where the content
of those services is tailored to the current or some
projected location and context of a mobile user
(Brimicombe & Li, 2009). LBS have been con-
sidered by many to have evolved professional GIS
applications to more public-centric services and
thus promoting greater ubiquity for GIS. Later in
this Chapter, a wide range of Mobile GIS applica-
tions via smart phones are covered, including areas
such as on-site data collection and monitoring;
navigation with real-time updates; wayfinding
for individuals to orientate and discover routes;
real-time tracking of contacts; mobile commerce
(m-commerce) with location-awareness; user-
solicited information for a range of business and
social purposes; coordinating emergency and
maintenance responses to accidents and essential
services; location-based artistic expression in the
community, mobile gaming where the players and
actions are location–based. The chapter ends with
152
Mobile Geographic Information Systems
a discussion of the challenges and research issues
raised by Mobile GIS, such as the influence of
fast development of mobile technologies.
FROM CONVENTIONAL
GEOGRAPHIC INFORMATION
SYSTEMS (GIS) TO MOBILE GIS
In mid-1960s, with the prospect of handling and
analysing spatial data digitally, the beginning of
geographic information systems (GIS) had been
explored in both professional and academic areas,
as shown in the time line in Figure 1. In Canada,
a prototype of GIS was started for the Canada
Land Inventory which aimed to map existing
land uses and analyse land capability for forestry,
agriculture, wildlife and recreation (Tomlinson,
1984). The output of this early stage of GIS was
regarded the cost-effective way for mapping the
whole land area of Canada at the time. This sys-
tem became fully operational in 1971. In 1964,
in the Laboratory for Computer Graphics and
Spatial Analysis, Harvard Graduate School of
Design, USA, the first digital mapping software,
named SYMAP, was created in 1964. SYMAP
used line printers to produce primitive maps to
visualise landscape themes of human and physical
phenomena for identifying spatial similarities and
groupings (McHaffie, 2000). These maps were
represented as coarse lines of equally spaced
characters and symbols. Evolving from SYMAP,
the GRID package with command-line user in-
terface and ODYSSEY as a line-based (vector)
prototype were developed as GIS software in the
Laboratory in 1970s. Also in mid-1960s, the US
Bureau of Census aimed to produce digital maps
of street blocks and census tracts for supporting
1971 census. This introduced explicit topology
into the data structure. These early developments
set the foundation of GIS, such as raster (grid) and
vector (line) data structures, the use of database
management for spatial data, the use of data lay-
ers (treating each theme as a layer of data), and
Mobile Geographic Information Systems
Figure 1. Time line of the development of mobile GIS and associated technologies
the use of topology in handling objects and their
spatial relationships (topological relationships of
points, lines and polygons).
The basic range of technological functionality
associated with GIS can be generally summed up
as: data entry and editing; transformation; query;
interpolation, cartographic processing, map alge-
bra and thematic mapping (Brimicombe, 2010).
Data entry can be either the importing of existing
digital data set(s), or digitising of secondary
sources (such as existing paper-based maps).
Spatial and attribute data from different sources
can be integrated in a GIS database on the basis
of geography. Data records can be geocoded, that
is have x,y co-ordinates attached to them, through
the use of geographic base files used for address-
153

matching and postcode-matching. Data can also
be edited and created within GIS. Transformation
in GIS mainly includes data conversion between
vector and raster, between different map projec-
tions, and between coordinate systems. Gener-
alisation can also be viewed as an important form
of a transformation in GIS, such as attribute re-
classification and spatial aggregation. Examples
of spatial aggregation can be re-representing
clustered points as zones, or joining a number of
small zones into a larger one. It can also be sim-
plifying lines or dissolving dividing lines to reduce
the complexity of objects. Query is an important
functionality in GIS. Query can be searching for
spatial objects according to specific attributes, or
conversely viewing the attributes of selected
objects or regions. Query also includes calculat-
ing the area of objects and distances between
them, as well as generating a range of statistics
of both the spatial objects and their attributes.
Interpolation in GIS includes point to area (such
as Theissen polygons), point to field (such as us-
ing computational techniques for contouring),
area to point (such as centroid of a polygon), area
to field (contouring from centroid), and area to
area (such as zone transformation). Another func-
tionality of GIS is cartographic processing,
mainly on vector layers, which consists of overlay
and buffering. Overlay is fusing a number of lay-
ers into a new layer containing rebuilt topology
and attributes of each contributing layer often
reclassifying using Boolean operators. Buffering
can be performed on points, lines or polygons.
Map algebra in GIS is mainly for raster data.
Raster layers can be added, subtracted, multiplied,
divided, and can also be weighted and compared
for maximum and minimum values. Thematic
mapping in GIS is for visualising data, with func-
tions such as choosing layer combinations, class
intervals, as well as colours, patterns, symbols,
and layout. Producing a good thematic map is a
very important aspect in GIS as it is the prime
means of communicating data and the outcomes
of analyses.
154
Mobile Geographic Information Systems
Geographic Information Systems (GIS) are
often regarded as software; however, the data
aspect of GIS is equally important as the software
and GIS also engage people and organisations
(Brimicombe and Li, 2009) in the production of
spatial data and the use of software. GIS involve
data collection and integration, processing, ana-
lysing and presenting in order to communicate
analyses of physical and social phenomena that can
assist in decision-making and problem-solving.
The problems and issues analysed, being both
scientific and practical, are overwhelmingly geo-
graphical in nature and focus on location. Thus
GIS was traditionally presented as a technology
to support science and problem solving; it has
come to be more than this (Longley, Goodchild,
Maguire, & Rhind, 2005). During the 1990s GIS
emerged as a research discipline in its own right,
often termed GIScience, to study substantive
questions arising out of the use of GIS tools and
spatial data. Longley et al. (2005) discuss some
of different views on GIS such as: a professional
user view of GIS as a tool and an analysis engine
to examine data and real patterns relationships
and anomalies; view from decision makers and
planners of GIS as a computerised tool to solve
geographic problems; a general public view of
GIS as a container of maps in digital form.
Starting in the early 1990s, new opportunities
became available to access and use maps on the
Internet. These quickly evolved to incorporate
basic GIS functionality of pan, zoom, and map
customisation to be followed later by wayfinding
tools for route selection. Specific products to de-
liver server side Internet GIS through a browser
interface started to be available by in the mid to
late 1990s, as seen in Figure 1. This represented a
fundamental shift whereby different levels of GIS
functionality and data sources could be accessed
and used online by remote users. Users could
have a choice of client side devices from desktop
computers, laptops or personal digital assistants
(PDA). GIS utilised through an Internet connection
are considered as Internet GIS, a phrase which
Mobile Geographic Information Systems
Figure 2. Basic elements of Internet GIS
was in use by the late 1990s (Peng, 1999). There
have been different terms to describe Internet
GIS, such as Web-based GIS (Doyle, Dodge, &
Smith, 1998), on-line GIS, distributed geographic
information (Plewe, 1997). Figure 2 illustrates the
main concept of Internet GIS. On the client side,
users send requests for GIS functions through a
Web browser via an Internet connection to the
server side where GIS software and data reside.
Requests have to be processed at server, and the
results or applets are then sent back to users via
the Internet. There are four basic parts in Internet
GIS, which are the user (the client), and the map
server, data server and Web server which reside
in the application server (Peng & Tsou, 2003).
Two different approaches can be adopted in
principle for Internet GIS, known as the server-
based approach and the client-based approach;
most commonly implementations are server-
based. In general, the ‘client’ is the side where
requests are made whereas the ‘server’ is the side
where requests are processed (Peng & Tsou, 2003).
Internet GIS using a server-based approach has
GIS software residing at the server site. Users
will access and employ GIS functionalities via
the Internet from their local computers or de-
vices. The Web browser is used as a graphic user
interface at the client side. The data processing
and analysis are performed by GIS software in-
stalled at server side. At the client side, users send
their requests through the Internet to the server.
When the request is received at the server side,
GIS software will be activated and relevant data
sets opened; the request is processed using GIS
functionality and the results sent back to the user,
usually as image files. This is the basic principle
of how the Internet GIS operate. There is a range
of technologies associated with the handling user
requests from browsers, which can be from desk-
top computers, laptops, hand-held devices to smart
phones; for parsing user requests to the GIS soft-
ware; and for data exchange between Web and
GIS software.
Apart from the professional usage of Internet
GIS such as for the use of mapping and data col-
lection in the field, popular applications accessed
by the public are Web-based mapping sites such
as Google maps (http://maps.google.com) which
through their API allow other users to embed maps
into their own Web pages. Access to and use of
these sites from smart phones is in many ways
very similar to the way these sites are accessed
from a PC or laptop as they take no account of
whether the user is static (at home or in an office)
or on the move.
The advantages of server-based Internet GIS
are mainly on its characteristics of centralisa-
tion. GIS hardware and software are managed
and updated centrally, which can not only be
maintained more easily than when decentralised,
but also allow users to have consistent access to
more up-to-dated functionalities. The server can
also have much more computing power and data
storage than mobile devices for example and are
155

able to process complex operations more quickly.
However, Internet GIS based on this approach
can face the challenge when a large number of
users send requests to a server simultaneously.
Since for all user requests the functionality has
to be preformed at the server side instead of at
the user (client) side, the system can slow down
or even crash in responding to large numbers of
user requests.
Another approach of Internet GIS is client-
based. It provides users at the client side the ability
to run GIS data processing and analysis locally.
GIS functionalities as users require are down-
loaded from the server onto user’s local devices
via the Internet, which can include different level
of GIS functions, required data and models. Small
executable application applets can be created to
carry out different tasks such as managing spatial
data, handling maps, performing basic analysis.
Such applets can be Java applets, ActiveX con-
trols, or other suitable plug-ins for mobile devices.
These application applets and plug-ins can be
downloaded at runtime to users’ local computers
or their mobile devices. The client-based approach
Internet GIS have the advantage of reducing the
server side burden of running GIS and data traffic
on the Internet. On the other hand, there will be less
centralised GIS function and data management.
Furthermore, more sophisticated and advanced
GIS functionalities may not be possible for users
to access and perform on client-based Internet GIS
due to lack of power and data storage.
The approach taken for Internet GIS implemen-
tation, therefore, depends on the criteria of appli-
cations, user volume, telecommunication network
capability, and so on. In principle, the Internet GIS
with server-based approach are more appropriate
to users with certain demands for advanced GIS
analysis and modelling functions. The volume of
such users accessing any one particular site is less
likely to be very high. Employing client-based
approach, Internet GIS would be more suitable
to applications with large user volume, but not
necessary requiring professional advanced GIS
156
Mobile Geographic Information Systems
functionality. Furthermore, distributed client/
server approach is being developed and employed
which can offer more benefit. There will be some
discussion in next section.
With the development of mobile telecommu-
nication network into its third generation (3G) in
2000s, the network offers the capability of both
voice and data communication with high speed
data transmission of 144Kpbs and latterly up to
8Mbps with the prospect of higher speeds. It can
support mobile Internet access and multimedia
applications. Mobile devices, particularly mobile
phones (cellular phones), have seen rapid devel-
opment from the early 1990s onwards. Mobile
phones have become one of the most widely used
electronic devices, largely viewed as a necessity
of daily life by many people. Smart phones have
the capability of always-on data connection, which
offers access to mobile Internet with aims of ‘any
time, any place and anywhere’ usage both for
voice and data services. There is a convergence
in mobile device development, which means the
boundary is becoming blurred between types of
data-oriented handheld devices (such as tablets
and PDA) and mobile phones. Smart phones can
be viewed as one such product. A further techno-
logical development is the miniaturisation of GPS
receivers and accelerometers to small chips so
that they can be integrated within mobile devices
making them location-aware. These technologi-
cal developments (shown in Figure 1) and high
levels of ownership of mobile devices provide
opportunities for the rise of wireless Mobile GIS.
THE PRINCIPLES AND
FEATURES OF MOBILE GIS
Wireless Mobile GIS are often shortened to wire-
less GIS or mobile GIS. In this chapter, we use
the term: Mobile GIS. Mobile GIS have resulted
from the convergence of wireless mobile tech-
nologies, GIS and GPS, offering users real-time
access while on the move using devices that are
Mobile Geographic Information Systems
themselves location-aware. This should not be
confused with the early usage of the term ‘mobile
GIS’ which were mainly referred to those GIS used
for working on site, in the field, through pocket
PCs or PC tablets. Those GIS systems could also
be connected to GPS to collect and manage spatial
data, but most operate as stand-alone GIS. Mobile
GIS using wireless technologies enable the real-
time connection. Thus, mobile devices (such as
smart phones) can synchronise their local data
with the database of the GIS on the server, and
access and manage data remotely in real-time.
Mobile GIS offer the capability of accessing
external GIS functionality and applications for
users on the move.
Numerous components are associated with
Mobile GIS. Here we focus on the following key
elements of Mobile GIS as illustrated in Figure 3:
GIS servers; wireless mobile telecommunication
networks (connectivity); mobile handheld devices
(such as smart phones); location-awareness tech-
nology; gateway services.
GIS servers provide GIS functionalities, data
and applications. The core of GIS, such as GIS
data structure and basic functionalities, follow
the same principles of conventional GIS. How-
ever, Mobile GIS are not stand-alone as are
Figure 3. Key elements in mobile GIS
conventional GIS. As discussed in the last section,
Internet GIS adopt a client/server approach to
delivery of GIS services to users. In Mobile GIS,
when based on client/server approach, a distrib-
uted component framework is often employed for
users to access GIS. This concept of distributed
component framework is different from conven-
tional client/server structure. Distributed compo-
nent technology allows clients to access hetero-
geneous servers dynamically (Tsou, 2004).
Utilising such technology, the applications on
either client or server side are broken up into a
range of components that can be interoperable
across networks, operating systems, applications,
tools, languages and hardware. Distributed com-
ponents can be structured as reusable components
which are able to manage complex requests from
client side. There are a range of benefits in adopt-
ing a distributed component approach in Mobile
GIS. One of them is the flexibility of client and
server roles in as much as a server can become a
client to another server or a client can act as
server depending on the distribution and func-
tionality of components. Another advantage of
using distributed component in GIS is that more
complex requests from the client can be managed
with sequence priority to offer more efficiency.
157

This is can be beneficial when a server is busy
when dealing with large numbers of client side
requests. Peng and Tsou (2003) discuss in detail
the advantages of distributed components in the
Internet GIS. Three major component technologies
used in the component-based application develop-
ment have been widely applied in developing
component-based GIS (Wu, Zhang, & Goddard,
2004). They are: Common Object Request Broker
Architecture (CORBA) specification from Object
Management Group (OMG); Distributed Com-
ponent Object Model (DCOM) from Microsoft
Corporation; and Java Remote Method Invocation
(RMI) from Sun Microsystem Inc. A distributed
component framework can be applied in Mobile
GIS to achieve these benefits.
Distributed Mobile GIS components can be
used to provide applications across networks.
Take the example of one of the popular distrib-
uted component technologies used for mobile
devices – the Java platform. Java platform is
able to offer support for distributed components
in Mobile GIS, particularly Java Micro Edition
(Java ME). Java ME emphases applications used
in mobile handheld devices, whilst Java SE (Java
Standard Edition) is for more general client side
applications. Java SE has been used for many
Internet GIS applications and Internet mapping.
On the other hand, Java Enterprise Edition (Java
EE) focuses on server side applications of GIS,
such as building distributed components that can
be reused for various services. Java ME, with
its focus on components for wireless handheld
devices, can be used to build applications for
Mobile GIS. The major components of Java ME
consist of mobile information device profiles,
connected device configuration and connected
limited device configurations, aiming to provide
utility programmes, security and connectivity for
small mobile devices
There are a number of advantages to the Java
platform. One is that it can offer a dynamic com-
ponent framework of Java applets for client side
and servlets for server side, which enables a more
158
Mobile Geographic Information Systems
dynamic and efficient client/server interaction.
Another advantage is Java’s ability of cross plat-
form performance. There is also Java’s dynamic
binding feature with downloadable Java applets.
However, one of the disadvantages of the Java
platform is that in comparison with machine level
programming with binary coding, running Java
programs is much slower due to its byte coding
nature. Also there is no standardised distributed-
object infrastructure in using Java. The reason we
put Java as an example is that Java is one of the
technologies that has been successfully deployed
for Internet GIS, and is still evolving.
For those architectures that rely on networks
for communication between the server and the
client, client/server framework and distributed
component framework are interrelated. Internet
GIS use Web browser as their user interface.
Distributed component applications can use Web
browser as the user interface, but also could use
their own GUI. Different protocols may be used for
distributed components in Mobile GIS other than
the HTTP protocol which is mostly used in Internet
GIS. In general, GIS with distributed components
do not have all of the system components in the
same physical location. These components could
be the processing, the database, the rendering or
the user interface. Mobile GIS adopt a distributed
component framework to support their services.
The second key element in Mobile GIS is
the mobile telecommunication network, which
provides wireless connectivity between Mobile
GIS users and the network where the main GIS
software, data and applications reside. In the early
1980s, the first generation (1G) analogue mobile
network was in operation to provide voice traffic.
Since then the mobile communication networks
have developed through the second generation
(2G) in 1990s with digital techniques and data
communication ability although limited, to the
third generation (3G). 3G networks provide high
speed data transmission, which can support fast
mobile Internet access and multimedia applica-
tions. Only at this phrase of development in mo-
Mobile Geographic Information Systems
bile wireless communication networks, Mobile
GIS can be realistic and deliver services. Such
networks enable users to access and use GIS, data
and services in mobile situation and in real-time.
However, geographical data are usual result in
large file sizes, therefore needing high bandwidth
and data transfer speeds to deliver. The bandwidth
and data speed of the network have direct impact
on services provided by Mobile GIS.
Another element of Mobile GIS is mobile
handheld devices. A mobile handheld device is
also known as mobile device, handheld device, or
handheld. It is a pocket-sized computing device
with wireless connection, typically having a display
screen with touch input and/or a miniature key-
board. Laptops are not viewed as mobile devices
due to their size. Mobile handheld devices have
developed over time, including handheld PDAs,
Ultra-mobile PCs, Tablet PCs and smart phones.
In order to access Mobile GIS applications and
services, the type of mobile device needs to be more
sophisticated in its connectivity, and computing
capability in comparison to basic mobile phones
focused on voice and text message communication
or conventional handheld computers with limited
connection. Smart phones belong to this type of
mobile handheld device, which is the emphasis of
this book. A smart phone can be regarded as a hand-
held mobile device that represents the convergence
of mobile phones with handheld computers. Such
mobile devices have the functions of mobile phones
with wireless telecommunication connection, and
operating systems to install and run advanced ap-
plications. Furthermore, operating system software
offers a platform to develop applications. As dis-
cussed early in the section, Java ME is one such
platform. More complex applications can also be
installed and run based on a specific platform in
a smart phone. Mobile handheld devices are con-
tinuously evolving and converging. Applications
of Mobile GIS need to be designed and developed
with characteristics of mobile handheld devices
in mind, such as relatively small screen size (GIS
visualisation usually associated with the display
of maps), processing power, battery consumption
and so on. Different types of Mobile GIS applica-
tions would emphasise different requirements of
mobile devices. Applications could be navigation,
finding nearest point of interest, real-time games,
to real-time data collection. Some might require
more fast and powerful processing capability and
screen quality, while others require more accurate
location-awareness technology.
Location-awareness technologies can be
viewed as a key element in Mobile GIS. Conven-
tional GIS is usually operated without concerning
where users are. It is similar for accessing Internet
GIS. Mobile GIS deliver services and applications
to users’ mobile handheld devices, often related
to the user’s mobile situation and location. A
typical application example can be providing a
user on the move the required route to a certain
destination based on where they are at that mo-
ment. The current location of the user is essential
for this service. There are a range of positioning
technologies that can be employed in mobile de-
vices, such as Global navigation Satellite Systems
(GNSS) including Global Positioning System
(GPS), mobile network positioning technologies,
wireless and RFID location technologies, sensor
positioning, visual-aided positioning and hybrid
indoor positioning. These positioning technologies
are discussed fully in various Chapters in this book.
In order to enable mobile handheld devices
(e.g. smart phones) to access GIS servers, mid-
dleware is required to link them up. This piece
of middleware is the gateway service, which is
another element of Mobile GIS. Gateway services
provide the capability of translating user requests
from mobile devices to GIS servers and format-
ting the output back to users at mobile device
side. Gateway services also perform tasks such
as protocol conversion.
Having discussed a number of key elements in
Mobile GIS, we now can look into the main fea-
tures in terms of the services and usage of Mobile
GIS. These features are mobility; real-time con-
nectivity; location-awareness; broadened usage.
159

Mobility is one important feature of Mobile
GIS. Mobile GIS has the ability to operate and
deliver functionalities, data and services without
necessarily requiring a fixed location or wired
connection. Mobility feature can also be seen as
portability from one aspect. Users can access GIS
via their handheld devices while on the move. The
mobility feature of Mobile GIS has three charac-
teristics, which are: mobile connectivity provided
by wireless telecommunication networks; mobile
handheld devices (e.g. smart phones) used to ac-
cess GIS functions and applications; GIS with
distributed component framework where com-
ponents could be the processing, the database or
the user interface at different physical locations
(servers) to deliver mobile services. Mobile GIS
enable data can be captured, stored, and man-
aged in remote locations and have the advantage
of real-time updating without storing duplicated
data. Data can be available from remote systems
to mobile users. Information can be uploaded and
downloaded between mobile devices and remote
systems. The feature of mobility brings GIS out of
fixed locations into the field and for people on the
move. Use of GIS in mobile situations brings more
emphasis on the interaction between the human,
mobile device and the environment in human-
computer interaction (HCI) research (Li, 2006).
Another main feature of Mobile GIS is the
real-time connectivity. With this feature, Mo-
bile GIS offer the possibility of exchanging and
analysing geographical information in real-time
in a real geographic world. The interactive ser-
vices can be provided with real-time refresh and
update. This allows more complex applications
such as augmented reality where new data will
be delivered as location is changed. It also opens
door to applications of push-services, such as
proximity advertising. For example, warning mes-
sages can be sent to users who are in such areas
where disruption and danger are or might occur.
The context of users can be taken into account
in delivery GIS services, which including user’s
160
Mobile Geographic Information Systems
physical surrounding, current situation, and their
personal preferences and requirements.
Location-awareness is another important fea-
ture of Mobile GIS, which differentiate Mobile
GIS from conversional and Internet GIS. Mobile
devices are position-enabled, such as GPS-enabled
smart phones; therefore the location of a mobile
device and their user can be identified. A range
of positioning technologies (as discussed previ-
ously and elsewhere in this book) is employed for
mobile devices, such as smart phones, to enable
them to be location-aware. Using such location-
aware mobile devices, more spatially-relevant
GIS applications and services can be available
to users according to their location. For example,
information can be received for fixed points of in-
terest such as nearest restaurant, petrol station and
hospitals, or for mobile subjects of interest such
as the whereabouts of friends and children. Such
devices also enable users’ location to be sent back
to servers, which allows Mobile GIS to manage
and process real-time data and provide up-to-date
applications and services more relevant to users.
Mobile GIS have broadened GIS user groups.
Conventional GIS focus more on professional users,
both in physical geography and in studying social
phenomena to assist decision making and problem
solving. GIS have been and continue to be utilised
for analysing spatial data, discovering patterns,
relationships and anomalies to support science
and problem solving. On the other hand, Mobile
GIS also provide functionalities and applications
to the wider public - services relevant to people’s
daily lives. GIS functionality would be accessed
and used by people having no GIS professional
knowledge. These services can range from: viewing
maps from users’ mobile phones, finding nearest
facilities, wayfinding, delivering location-related
messages, playing mobile games. Such a wide
range of user-centred services necessitates more
emphasis on the interaction between users, GIS
and the surrounding environment, which poses
challenges in Mobile GIS research. This will be
discussed further in the last section of this Chapter.
Mobile Geographic Information Systems
MOBILE GIS FOR LOCATION-
BASED SERVICES (LBS)
Mobile GIS have brought mobility, real-time
connectivity and location-awareness into GIS,
and also have broadened GIS usage and user
groups. Mobile GIS provide tools and services
not only for skilled professionals but also for a
wider public to use in their working and social
lives. Geographical data and information can thus
be accessed, extracted and displayed on user’s
mobile devices (e.g. smart phones) in real-time
while on the move. Mobile GIS can be seen as the
technology that facilitates the creation of mobile
services. This leads us into the area of Location-
Based Services (LBS).
Location-Based Services (LBS) are defined
as “the delivery of data and information services
where the content of those services is tailored
to the current or some projected location and
context of a mobile user” (Brimicombe and Li,
2009). LBS can be viewed as the outcome of the
convergence of a range of technologies, as dem-
onstrated in Figure 4. GIS provides geographical
science, technologies and data for LBS. New
Information and Communication Technologies
(NICTs) have evolved from ICTs with features of
mobility and location-awareness. NICTs include
a range of technologies associated with wireless
mobile telecommunication networks, other wire-
less networks, wireless handheld mobile devices
(e.g. smart phones) and positioning technologies
that integrated into mobile devices. NICTs also
have the important feature of having wide usage,
near ubiquity, in people’s lives. The convergence
of GIS, the Internet and NICTs has made LBS
possible. We also can see that LBS are tightly
coupled with Mobile GIS. LBS are therefore able
to offer data and information services strongly
related to user location in mobile situations, in
real-time wirelessly.
The focus of LBS is on delivering location
based data and information services that are mo-
bile user centred. Location plays a major role in
services provided by LBS. There are two main
elements in delivering LBS services: pull and
push. ‘Pull’ element in LBS generally employs
user location obtained via his/her mobile device
to offer services whereby the user can pull down
(access and retrieve) data and information relevant
to his/her current location and situation. ‘Pull’
element in LBS can be seen in applications such
as navigation and wayfinding, searching for near-
est point of interest. For the ‘Push’ element in
LBS, individual’s mobile location is used to push
(send) relevant information to users either as pre-
subscribed requests or independently of a query.
For example, information can be delivered to
users according to their current locations for all
kinds of business and social purposes. This may
take the form of proximity advertising and offers
of discounted goods, but could also be weather
forecast, traffic conditions, delays to trains and
flights, warning messages for potential hazards.
There is also the element of ‘tracking’, which is
monitoring the location of mobile devices con-
tinuously. ‘Tracking’ could be used to provide
services such as monitoring or managing vehicle
fleets, or in entertainment such as location-based
games. Furthermore, location related information
and services provided by LBS are expected to be
updated according to user’s current situation and
needs, and to be delivered in real-time. It is es-
Figure 4. Location-based services: An aspect of
the technological convergence
161

sential that services provide real-time, up-to-date
information pertinent to mobile users.
In order for LBS to provide needed information
and services to mobile users, context is important
aspect to be considered. The context in LBS mainly
refers to the current situation of a user that could
be relevant to tailoring location-based informa-
tion. Context can be information associated to
user personal situations that might be used in
setting up services particularly delivered to these
users. For example, large sized text or maps, or
louder ring tone could be extra features along
with the information provided to those users in
need. Context can also be surrounding situation at
certain location, such as noisy places, unfamiliar
areas. Other context may be time of day, speed
of travelling. All these context will have direct
effects in tailoring location-based information and
services for mobile users. This will be discussed
further later in this section.
Location-based geographical data and informa-
tion provided as services by LBS can take many
forms. One of the efficient ways to deliver geograph-
ical information is using maps, and is sometimes
narrowly viewed as what LBS are. However, LBS
should not be simply seen as delivery of maps to
users’ mobile devices. ‘Location’ in LBS concerns
more the geographical position of a user as the key
to tailoring information rather than the content of
any response to a query. Response to a query can be
just text or images, but such information has been
tailored to the user, who makes the query, accord-
ing to his/her mobile location and situation. For
example, if a user queries the closing time of a car
park where he/she is, the response sent back need
only be a simple text message. However, without
requiring the user to describe or state exact location
of the car park, LBS should be able to deliver the
right information in real-time by determining the
location through the user’s location-aware mobile
device. From this simple scenario, we can see that
in adapting information and services according to
location, GIS and the use of spatial data are fun-
damental aspects of LBS.
162
Mobile Geographic Information Systems
Information and services provided by LBS
can increase their commercial potential when
harnessing Mobile GIS. At the same time, LBS
also bring emphasis on a range of research ar-
eas in GIS and Mobile GIS. These have been
discussed in detail in the book “Location-Based
Services and Geo-Information Engineering” by
Brimicombe and Li (2009). Some of these key
areas are issues concerning the availability of
data, locating users and how to contextualise the
user, their environment and the technology for
delivering geographical information services.
As discussed, determining the location of a user
to whom services are delivered is essential for
tailoring information in LBS. The location can
be the current position of a user or some future
projected location. A number of positioning tech-
nologies are available to determine the location
of a mobile device, as discussed elsewhere in this
book. However, challenges still remain to obtain
user location with sufficient accuracy to enable
location-based information to be provided with
adequate detail and thus ensure utility. The level
of sufficient accuracy also depends on the nature
of services required. For example, the level of
positioning accuracy required is much higher for
responding say, to a car accident than in providing
weather information for a user location. In addi-
tion to determining user location, contextualising
the user and associated situation is important for
tailoring location-based services to mobile users.
Context of user’s geographical location and sur-
rounding environment could all need to be taken
into account. Take a wayfinding example. A route
will be given to a user according to his/her loca-
tion and requirement. The context such as time
of the day could directly influence the utility of
the information provided as being either as the
shortest route or the safest route at night. Informa-
tion associated with users and their preferences
is another aspect of context in providing tailored
geographical information. Technology is a further
source of context, being the type of mobile device
being used and its capabilities, quality of wireless
Mobile Geographic Information Systems
connection, or the availability of data resources
at the time of user request.
The key research areas in LBS and Mobile GIS
include dealing with the spatial query, communi-
cating the response and interoperability. Dealing
with spatial query and communicating responses
in Mobile GIS and LBS have their characteristics
as processing and providing information in real-
time, and providing geographical related services
to the wide public who should not be required
to have GIS skills. The interested reader can
refer to the book “Location-Based Services and
Geo-Information Engineering” for more in-depth
analysis of these areas. As LBS are the result of
the convergence of a range of heterogeneous
technologies, interoperability is one of important
research areas in GIS, Mobile GIS and LBS. Le-
gal/social issues and business models are issues
increasingly raised as GIS and LBS have wider
use in society, which will be referred to again in
the last section of this Chapter.
APPLICATIONS OF MOBILE
GIS FOR SMART PHONES
New opportunities for Mobile GIS applications
and services are rapidly growing with the rise of
smart phones. A 2010 comScore study indicates
14% of smart phone users (33.5 million) use mobile
mapping applications, a 44% increase from 2009.
This number is expected to grow further through
2011-12 (http://www.vce.com, viewed on August
2011). The availability of smart phones and the
widespread use of GIS consumer products on smart
phones have consequently aroused a substantial
anticipation of Mobile GIS applications for smart
phones (Renner et al., 2011).
However, Mobile GIS services remain in their
early stage. There is potential and also challenges
for mobile software development, as new hardware
is being released and various support platforms
co-exist. At present, most mobile GIS applications
and APIs for smart phones are built on three main
mobile operating systems (or smart phone plat-
forms) which are Google’s Android, Microsoft’s
Phone 7, and Apple’s iOS. Some applications
have recently extended to BlackBerry. While
individual users can use smart phone applications
for their daily life such as finding nearest points
of interest, organizations in many sectors can use
them to increase operational efficiency such as
improving workflow.
The range of applications of Mobile GIS for
smart phones are summarised here. Some of them
are still potential applications, which are expected
to emerge as Mobile GIS mature in the foresee-
able future. Others applications have already been
developed and are being used to a certain degree.
Relevant examples are presented under each
category to demonstrate different development
procedures and application types.
Applications Mainly for Accessing
and Using GIS Functionality
through Smart Phones
Basic GIS software functionalities can be imple-
mented on smart phones through Mobile GIS
applications, which are often extended to include
some location-based features, such as accessing
maps and data relevant to user location, collecting
location related data in real-time, real-time report-
ing of events with spatial tags. Such applications
provide users with smart phones access to basic
GIS services in real-time while on the move.
Here are some examples of this type of Mobile
GIS application. One of them is ArcGIS Apps for
Smart Phones developed by ESRI, which allows
users to navigate maps, collect and report data, and
perform GIS analysis via a smart phone. Such ap-
plications offer opportunities for smart phone users
to view maps and capture information related to
their current situation, access content from ArcGIS
Online or use functionalities from enterprise GIS,
and develop custom applications for user needs.
These apps could broaden GIS usage to a much
wider public. Another example is applications
163

built using open source software that are increas-
ingly available. It becomes practicably feasible
to build a specific single application on different
mobile platforms. For instance, a new Flex/AIR
GIS application for the smart phone BlackBerry
has recently developed by WebMapSolutions us-
ing Web-based Flex code (www.webmapsolution.
com, viewed on August 2011). Building Mobile
GIS functionality for smart phones has also been
aimed for Apple’s iOS platform using OpenScales
Flex sdk’s, ArcGIS Flex API and OpenLayers.
Also in 2008, a solution for accessing GIS
through BlackBerry smart phones started to be
designed (http://www.3-gis.com, viewed on Au-
gust 2011). With the introduction of an application
named Buzz for BlackBerry smart phones, users are
able to view and query GIS data with their smart
phones on the move. Developed to employ the
ESRI ArcGIS Server technology, Buzz is aimed at
services provided to industries such as the utility,
telecommunications and emergency management
firms; but also could be extended to other markets
with a geo-database. Currently, Buzz can run on
the BlackBerry® Pearl™ series, BlackBerry®
Bold™ and BlackBerry® Curve™ series smart
phones utilizing the AT&T, Sprint, T-Mobile and
Verizon Wireless telecommunication networks.
Mobile Applications for Public
and Business Services
In the early stage of mobile GIS applications on
smart phones, there is no clear boundary between
public and business services. However, increasing
numbers of applications are expected for various
specialised aspects of public and business ser-
vices in the future, as mobile technology itself is
evolving. Some of examples are given below to
illustrate mobile GIS applications used for public
and business services.
One example application is the free BAO (Busi-
ness Analyst Online) for Apple’s iOS platform
developed by ESRI, aiming to give access to key
demographic and market facts about any location
164
Mobile Geographic Information Systems
in the U.S. through iPhone, iPad, or iPod Touch.
This application provides information which can
be used to evaluate an area on-site in real-time and
assist businesses to understand an area in terms
of its geo-demographics. The knowledge of an
area can be: types of people living in the location,
including a brief demographic and market data
summary of a location (e.g. age, gender, socio-
economics, and spending habits); the difference
between people living in this location from other
areas by comparing demographic and market data
for two locations or one location vs. the entire
U.S. national level data. Such applications may
assist businesses to compare and choose prefer-
able locations for their services based on needs
and criteria. My Place History is another free
app from ESRI’s Geomedicine for Apple’s iOS
platform which links public health information
with user environmental experience. This appli-
cation provides services to assist users to better
understand their local environment in terms of
its effect on their health. U.S. street addresses
are used to create and manage a personal place
history. My Place History allows users to access
geographically relevant health information, such
as one’s proximity to environmental hazards or
exposures. The application of My Place History
can also link user’s personal place history with a
number of USA governmental databases such as
Dartmouth Atlas of Health Care, Toxic Release
Inventory (TRI) of the United States Environ-
mental Protection Agency, Haz-Map, and U.S.
National Library of Medicine.
Another example is a location-based service
developed for companies and state and local
governments by GCS Research, combining the
functionality of Mobile GIS with the GPS-based
sensor suite of an Android-enabled smart phone
(www.gcs-research.com, viewed on August 2011).
This is a thin-client application, aiming to be a
practical, cost-effective business tool that enables
mobile users to access geographic information
and geospatial functionality via their handheld
device. The Montana Parcel App, built by GCS
Mobile Geographic Information Systems
Research for this purpose, allows smart phone
users to remotely access information of real estate
in Montana. Users can use the phone’s built-in
GPS to get their current location in the state. They
also can zoom into a mapping interface to select
an area of interest by “double-tapping” the touch
screen. In real-time, the application can retrieve
location-specific property information for the se-
lected site from two Web-enabled databases. The
first database is a parcel boundary file maintained
as part of the Montana cadastral database. This
cadastral information is managed in an ArcSDE
geodatabase. The smart phone application can
trace the property boundaries in red on the phone’s
map display. The second database is the State of
Montana CAMA (Computer-Assisted Mass Ap-
praisal) System where it retrieves property details
such as owner name, parcel legal description, area
size, assessed value, and zoning code. Such an
application can create a mobile mash-up that uses
the Android mobile operation system to deliver
mobile services.
The advanced functionality of mobile operat-
ing systems have made it possible to bridge the
gap between the enterprise business GIS database
and the mobile phone, as we can see in the case
of the GCS Research application. Cost-effective,
location-based business tools can be created with
this technology. Users can download applications
such as Montana Parcel App for free on their
smart phones. Such mobile applications and other
Mobile GIS tools can be transferred to different
smart phone operating systems. The development
of location-based services, which converges a
range of technologies, can provide a wide range
of Mobile GIS applications for smart phones.
Applications for on-Site
Data Collecting, Monitoring,
and Management
Real-time communication and analysis play im-
portant roles to support efficient decision making.
On-site work can comprise of data collection,
monitoring and management as well as those ac-
tivities carried out in the field such as identifying
key variables contributing to a problem, taking
measurements, and inputting and handling data
residing locally or remotely. For a variety of
fieldwork operations, there has historically been a
lag between collecting data on-site and analysing
data back at the office, also in exchanging relevant
data for analysis and decision making. Mobile
GIS applications delivered via wireless networks
and mobile handheld devices (e.g. smart phones)
offer field operatives the opportunity to do on-
site data collection and monitoring while having
real-time connection with remote sites (or head
offices) to exchange data, to carry out analysis to
support decision making or to take action rapidly.
Furthermore, using mobile devices for fieldwork
has a number of extra advantages such as light
in weight and low in cost. Nevertheless there are
also a number of disadvantages to using a mobile
device in the field such as its comparatively low
CPU power for processing data and graphics,
small screen size and issues on its compatibility
when used alongside other equipment.
One example is a Mobile GIS application
aimed at providing and improving GIS ability for
field surveying work. SuperGIS Mobile Engine 3
is a Software Development Kit (SDK) developed
by SuperGeo (www.supergeotek.com, viewed
on August 2011), which can be used to build
Mobile GIS applications to meet demands for
effectively improving the surveyor’s productivity.
Additional GIS functionality can also be created
using SuperGIS Mobile Engine, and then embed
into the existing mobile application to provide
Mobile GIS functionality to users via their mobile
devices. It is aimed for user to develop Mobile
GIS applications through SuperGIS Mobile Engine
built-in controls, read-to-use features, to achieve
standard GIS functionality in mobile situations.
These mobile applications include map display,
GPS positioning, file editing, coordinate system
settings to meet user needs in the field. They can
be built for mobile devices supporting Micro-
165

soft®.NET Compact Framework, such as smart
phone, Pocket PC and Tablet PC, and thus reduce
the incompatibility problem that often occurs in
hardware and software.
Mobile GIS applications, similar to this, often
aim to offer users working on-site with capabili-
ties such as: collecting spatial data; editing and
updating data locally and remotely; accessing
maps to obtain up-to-date and relevant spatial data;
having positioning ability such as GPS; querying
and managing data to assist their fieldwork and
decision making. There is a range of examples of
such application. One is using mobile handheld
device for on-site monitoring of environmental
processes (Nurminen, Kruijff, & Veas, 2011).
Environmental parameters of certain areas are
observed continuously and measured regularly.
Spatial data as well as related data such as images
are communicated and exchanged between people
working on-site and decision maker who are in
different locations. Others include field map-
ping to update spatial data on-site with attribute
information, building and maintaining asset lists
with location and attributes, and real-time incident
recording into spatial database.
Another example for this category application
is BlackBerry Business Solutions (us.blackberry.
com/business/needs/field/, viewed on August
2011), which consists of a combination of ap-
plications, services and peripherals, including
printers, signature capture, GPS, digital forms
and barcode scanners. This application is designed
to provide mobile customer services. It provides
mobile field access to details to support cases and
customer activities which will raise support issues
in real-time. This mobile application can be used
to provide services such as updating service tick-
ets from different locations, triggering the billing
process and setting up feedback from the field.
Similar applications could provide services such
as checking availability of products at different
locations according to user’s current position and
placing orders, communicating and collaboration
166
Mobile Geographic Information Systems
with other business and organization using real-
time on-site information.
With a mobile traffic alert application de-
veloped in Ryerson University, Canada, road
geometry data can be extracted using GPS data
received from smart phones (Niu, Li, & Poursaeid,
2011). Based on road centrelines, this application
can be employed to assist capturing road network
data. The current conventional methods to obtain
such data, either by surveying or by using remote
sensing data, are time consuming and expensive.
In this application example, GPS data are crowd-
sourced through smart phones as an increasingly
popular way of collecting data that can be used
to provide real-time traffic information, identify
traffic patterns, and predicting traffic congestion.
The same type of data can be very useful for cost-
effective, fast updating of road network databases
due to its rich spatial and temporal coverage and
high data volume.
Applications for Navigation
and Wayfinding
A range of mobile applications are particularly
designed for navigation and wayfinding using
mobile devices such as smart phones. Navigation
applications are mainly developed for assisting
motorists to identify appropriately optimized
routes. For example, an optimised route can be
found and sent as a map to users’ mobile devices
when a vehicle is to be driven from one starting
point to one or more user-specified destination.
Furthermore, subscription services for real-time
traffic updates can be delivered to users’ smart
phones according to the user’s current location.
Mobile GIS applications that provide wayfind-
ing services are more focused on the means by
which individuals orientate themselves, discover,
and explore surrounding or certain areas of interest.
Individual users with their mobile devices could
use this type of application to discover points
of interest related to their current locations, to
find routes to certain destinations or places such
Mobile Geographic Information Systems
as tourist attractions, restaurants, and theatres.
Modes of transport and other spatially located
objects, landmarks and points of interest can
also be provided using location-aware mobile
devices. Such applications intend to deliver us-
ers a service in real-time according to their need
while on the move.
Other Mobile GIS Applications
on Smart Phones
A range of other applications can be identified,
many of which have been prototyped and in de-
velopment. These applications, along with those
applications for navigation and wayfinding, are
more closely related to LBS applications:
• Real-time tracking: such applications
can be real-time tracking of vehicle fleets,
business associates, social contacts (e.g.
to know when friends are nearby) or one’s
family members (e.g. tracking children
home by parents who are still at work).
• Mobile commerce: this would incor-
porate both transactions made by indi-
viduals on the move via a NICT and re-
ceiving (pushed) alerts (e.g. advertising,
notification of discounts) to opportunities
for transactions that are location-specific
(such as at a nearby shopping mall).
• User-solicited information: this type of
applications can be for all kinds of busi-
ness and social purpose, such as weather
forecast, traffic conditions, delays to trains
and flights, warning messages for potential
threats, film showing and ticket availabil-
ity, menus and local maps.
• Location-based tariffs: such as differen-
tials in road pricing, pay-as-you-go car in-
surance and similar schemes.
• Coordinating emergency and mainte-
nance responses to accidents, interrup-
tions of essential services and disasters.
• Artistic expression in the community
that might include digital graffiti, location-
based story lines and discovery trails.
• Mobile gaming where the players and ac-
tions are location-based.
• Social networks where location-aware
applications could be implemented with
wireless social network tools, such as twit-
ter. Users would share their location, and
what is on their mind through updating
their status, attaching a location to their
thoughts and activities. The updated status,
with the location graphically depicted by a
map, would be available on the screen of
smart phones.
The future trends of Mobile GIS applica-
tions for smart phones will be interwoven with
location-based services, as Mobile GIS, smart
phones, wireless telecommunication networks
and related technologies are further developed.
Mobile GIS enable more people without GIS
professional skills to engage with GIS-based ser-
vices in a more ubiquitous way. Smart phones are
expected to be used by more people on the move
as a powerful and readily available thin client for
various applications rather than just as phones.
More advanced real-time/on-site functionality
of Mobile GIS will become available on smart
phones due to the advancement of new informa-
tion and communication technologies (NICTs).
Therefore, the range of Mobile GIS application
will continue grow.
PROSPECTS AND CHALLENGES
OF MOBILE GIS
Mobile GIS have come a long way in a short time.
They have been facilitated by the rapid develop-
ment of mobile telecommunication networks to
deliver higher bandwidths, the rapidly evolving
design of mobile devices, and the convergence of
heterogeneous technologies into single devices
167

that are location-aware. As these technologies
further develop, so too will the possibilities for
Mobile GIS. However, Mobile GIS are technology
dependent and any shortcomings in the technol-
ogy will have knock-on effects for Mobile GIS
in the implementation and delivery applications
(Brimicombe and Li, 2009). Above all, by align-
ing GIS with mobility and mobile devices there
is the opportunity for engineering GI applications
that make GIS part of ubiquitous technologies and
at last achieve an embedded mainstream status.
The GIS community has long thought of itself
as focusing in two areas: systems and science. Once
the systems had adequately matured, the science
took off with research focusing on the substantive
questions posed by the use of the systems in solv-
ing real-world problems. The research agenda is
far from complete, particularly as new questions
arise from the pace of technological development,
the increasing quantity, diversity and granularity
of data, and the complexity of analyses that are
entailed. Nevertheless, GIScience is reaching a
maturity with a body of scientific and application
knowledge. Engineering is the systematic applica-
tion of scientific research to successfully and de-
pendably solve real world problems (Brimicombe,
2010). With the convergence of technologies shown
in Figure 4 and the potential for designing applica-
tions that are widely used, we see the opportunity for
progressing to GI Engineering, that is, “the design
of dependably engineered solutions to society’s use
of geographical information” (Brimicombe & Li,
2009). Examples already on the market are in-car
satellite navigation systems (SatNavs), similar ap-
plications on smart phones, and niche applications
of location-based services (LBS).
Whilst the prospects for Mobile GIS are good
with plenty of opportunities, there are challenges
along the road to ubiquity. These challenges
derive from four broad areas: limitations, even
threats that derive from the technologies being
used; areas of GIScience that still need to be ad-
equately researched; users; and business models
for a sustainable presence.
168
Mobile Geographic Information Systems
The rapid development in mobile telecom-
munication network and mobile device technolo-
gies will bring even faster data transfer speeds,
better connectivity and more advanced devices.
This bodes well for Mobile GIS. However, such
developments and changes in these industries
also raise issues concerning the reliability and
consistency of network infrastructure and devices
which Mobile GIS rely on to build, implement and
deliver applications. For example, the release of
the iPhone revolutionised smart phone technolo-
gies and required application designers to quickly
gear up to this new class of device. The design of
the devices (such as screen size and resolution,
keyboard/keypad, memory and optional additional
memory) is likely to be driven by the more popu-
lar uses such as texting, social networking and
tweeting and may remain sub-optimal for Mobile
GIS. For example, use of gestures across the face
of the screen is likely to make query definition
more difficult when compared to using a keypad,
but a pop-up keypad may obscure too much of
the maps being viewed. There is also the issue
of battery life, that is, the length of time between
charges that allows a device to remain serviceable
on the move without needing additional batteries.
Relatively short battery life has been an issue with
smart phones because of the power demands made
upon them. Continuity of connectivity and there-
fore of service has always been a problem with
wireless mobile devices due to blocked signals
in tunnels, underground systems, inside building
complexes and so on. Hand over from one cell to
another, particular at speed, can cause the service
to drop. Additionally, GPS in a smart phone for
location-awareness will not work and may even
give seriously erroneous results where a clear view
of three or more satellites is not available. Whilst
assisted GPS through the telecommunications
network helps overcome some of the effects, it
is not a complete solution. Location-awareness
may have to rely upon a number of technologies
including GPS, A-GPS, accelerometers (which are
Mobile Geographic Information Systems
now available as a chip), and RFID and Bluetooth
for locating through beacons.
Mobile GIS raise a number of research areas
for GIScience arising from the technology, data
requirements and the nature of applications. Im-
portant areas are:
• Distributed component systems: This
will be a fundamental departure for GIS
software which has tended to be mono-
lithic, that is, a large software package
resident on a single PC or server. The
smart phone and other mobile devices are
comparatively thin clients with relatively
low storage and processing power. An ad-
vanced desktop GIS package is unlikely
to fit on a smart phone and software for
mobile devices that are available on the
market are usually a sub-set of the func-
tionality geared to specific tasks. In order
to overcome such limitations Brimicombe
and Li (2009) have proposed a distributed
component approach to GIS using agent
technologies. A software agent can be de-
fined as an autonomous, problem-solving,
encapsulated entity in a dynamic and open
environment (Woolridge, 1997). The fun-
damental difference between a convention-
al software program and an agent is that
the former will run once and then terminate
until instructed to run again, whereas the
latter can sense the environment in which
it is embedded and act on it autonomously
in pursuit of its goals (Franklin & Gaesser,
1997). The advantage of this type of struc-
ture would be that an interface agent would
reside on the thin client mobile device and
would arrange for each task (such as lo-
cating the relevant data, conducting some
form of analysis and displaying the result)
to be carried out by other agent-based ser-
vices residing on servers on the Internet.
Functionality and data would be on an
on-demand basis. Agent technologies are
particularly suited to achieving interoper-
ability in heterogeneous computational
environments and would thus be a sound
structure for Mobile GIS.
• Scalability and response times: For
ubiquitous applications of Mobile GIS,
the technology will need to be scalable to
potentially millions of users without no-
ticeable degradation of service. GIS-type
queries are typically long transactions, that
is, querying spatial data takes much longer
than typical database queries of nonspatial
data. GIS queries tend to have both high
CPU cost and high I/O cost. Thus current
approaches to GIS software are unlikely to
be sufficiently scalable. Whilst open and
interoperable middleware will be impor-
tant, a large part of the solution rests in the
type of interoperable component systems
discussed above.
• Data: There has been a data explosion
over the last decade of unprecedented
proportions, an increasing proportion of
which is spatial data. Many areas of eco-
nomic endeavour have become data driven
and the focus on high resolution local data
has driven new services. Focusing in on lo-
cal areas leads to an exponential rise in the
amount of data required to satisfy informa-
tional needs (such as from Google Earth
to Google Street View). Zooming in also
requires data to be timelier with shorter up-
date cycles. Whilst the technology for this
exists, the costs can be high. But whilst
conventionally GIS have been deployed
for map data and their attributes, the main
mass markets will be for such data to be in-
tegrated into a multimedia approach to in-
formation delivery either through on-the-
fly mash-ups or by incorporating into GIS
non-traditional data that nevertheless have
spatial tags (e.g. tweets). This will add an
order of complexity to GIS.
169

• Cognition: This is central to how users
of Mobile GIS are going to able to inter-
act with the technology in relation to their
changing surroundings. Traditionally stud-
ies of spatial cognition have focused on how
users conceptualise, understand and use
map products. But in most interactive ap-
plications employing Mobile GIS, the user
initiates the query, the framing of which will
be based on spatial reasoning derived from
their perception and understanding of their
surroundings, desired goals and a range of
contextual variables. Spatial queries will
need to be framed as naïve natural language
or gesture-based requests instead of more
formal database queries. The formulation
of the query is all important in obtaining
the desired information and deriving utility
from the application.
Some important considerations of the user have
just been touched on. The design of the application
should provide for a compelling user experience
and most smart phones provide the functionality
and programming environments to achieve that. But
designers need to also consider the vulnerabilities
that the use of smart phones can impart to users.
In the UK there have already had to be public
service announcements on the potentially danger-
ous consequences of using mobile devices while
crossing the road. This and similar vulnerabilities
might be an important for Mobile GIS applications
for wayfinding. As a way of alerting the GI com-
munity to this issue, Brimicombe (in press) has
analysed the crime wave in the theft of portable
in-car satellite navigation devices (SatNavs) once
they became mass consumer products at the end of
2005. Many purchasers of these products (the first
mass consumer products to include GIS software)
will have been unaware of the vulnerability to
theft that these devices would impart to the owner.
Beyond gimmickry and status must-haves, Mobile
GIS applications will need to establish markets
for services that provide utility to users. That is,
170
Mobile Geographic Information Systems
the perceived benefits of a service must outweigh
the cost or effort in obtaining the service. This
requires business models that promote the utility
whilst being able to cover the costs of sustaining
the service (Tsalgatidou & Pitoura, 2001). Many
users of Internet and mobile services expect them
for free and consequently must put up with pushed
advertising and other inducements which will
cover the cost for the service. This type of busi-
ness model is likely to be restrictive for services
based on Mobile GIS because unless the volume
of users is very high, the click rate on advertise-
ments is unlikely to supply revenue that would
cover set-up and running costs. Simple applications
such as accessing Google Maps are free, but are
no different from mobile Internet. More complex
applications that provide analytical functions and
tailoring based on location-awareness are likely to
be subscription services. The challenge for these
services is to provide high levels of utility that are
error free in order to be sustained. Services based
on Mobile GIS are only at the beginning – this will
be an interesting space to watch.
CONCLUSION
Mobile GIS is the successful convergence of mo-
bile communication and GIS technologies. The
driving force of growth of Mobile GIS is the rising
demand for LBS applications from users of mobile
devices, where smart phone users have played a
key role. Wireless mobile devices, particular smart
phones, offer the possibility of querying, analysing
and visualising spatial information on-site and in
real-time. Mobile GIS provide core functionality in
the context of LBS, while LBS bring Mobile GIS
to various aspects of daily life. The research and
development of Mobile GIS are still at an early
stage, but great potential has been demonstrated
for a range of applications. With further progress
in wireless mobile technologies and GIS, Mobile
GIS will certainly make a substantial contribution
in the realisation of ubiquitous GIS.
Mobile Geographic Information Systems
REFERENCES
Brimicombe, A. J. (2010). GIS, environmental
modeling and engineering (2nd ed.). Boca Raton,
FL: CRC Press.
Brimicombe, A. J. (2012). Did GIS start a crime
wave? SatNav theft and its implications for
geo-information engineering. The Professional
Geographer, 64.
Brimicombe, A. J., & Li, C. (2009). Location-
based services and geo-information engineering.
Chichester, UK: Wiley.
Brimicombe, A. J., Li, Y., Al-Zakwani, A., & Li,
C. (2009). Agent-based distributed component
services in spatial modelling. In Gervasi, O. (Eds.),
Computational Science and its Applications -
ICCSA 2009 (pp. 300–312). Berlin, Germany:
Springer. doi:10.1007/978-3-642-02454-2_21
Doyle, S., Dodge, M., & Smith, A. (1998). The
potential of web based mapping and virtual reality
technologies for modelling urban environments.
Computers, Environment and Urban Systems, 22,
137–155. doi:10.1016/S0198-9715(98)00014-3
Franklin, S., & Graesser, A. (1997). Is it an agent,
or just a program? A taxonomy for autonomous
agents. In Jennings, N. R., Wooldridge, M. J., &
Müller, J. P. (Eds.), ECAI-WS 1996 and ATAL
1996, LNCS (Vol. 1193, pp. 21–35). Heidelberg,
Germany: Springer. doi:10.1007/BFb0013570
Li, C. (2006). User preference, information
transactions and location-based services: A study
of urban pedestrian wayfinding. Computers,
Environment and Urban Systems, 30, 726–740.
doi:10.1016/j.compenvurbsys.2006.02.008
Longley, P. A., Goodchild, M. F., Maguire, D. J.,
& Rhind, D. W. (2005). Geographical informa-
tion systems and science (2nd ed.). Chichester,
UK: Wiley.
McHaffie, P. (2000). Surfaces: Tacit knowl-
edge, formal language, and metaphor at the
Harvard Lab for computer graphics and spa-
tial analysis. International Journal of Geo-
graphical Information Science, 14, 755–773.
doi:10.1080/136588100750022778
Niu, Z., Li, S. N., & Poursaeid, N. (2011). Road
extraction using smart phones GPS. 2nd Interna-
tional Conference and Exhibition on Computing
for Geospatial Research and Applications. Wash-
ington DC. Retrieved on August 15, 2011, from
http://www.com-geo.org/
Nurminen, A., Kruijff, E., & Veas, E. (2011).
HYDROSYS - A mixed reality platform for
on-site visualization of environmental data. In
Proceedings of the 10th International Symposium
on Web and Wireless Geographical Information
Systems (pp. 159 - 175). Springer Lecture Notes
in Computer Science.
Peng, Z. R. (1999). An assessment framework for
the development of Internet GIS. Environment &
Planning B, 26, 111–132. doi:10.1068/b260117
Peng, Z. R., & Tsou, M. H. (2003). Internet GIS:
Distributed geographic information services for
the internet and wireless networks. Hoboken,
NJ: Wiley.
Plewe, B. (1997). GIS online: Information re-
trieval, mapping, and the internet. Santa Fe, NM:
OnWord Press.
Renner, R., Moran, M., Hemani, Z., Thomas, E.,
Pio, H. S., & Vargas, A. (2011). A comparison of
mobile GIS development options on smart phone
platforms. 2nd International Conference and Ex-
hibition on Computing for Geospatial Research
and Applications. Washington DC. Retrieved on
August 15, 2011, from http://www.com-geo.org/
Tomlinson, R. F. (1984). Geographic informa-
tion systems – A new frontier. The Operational
Geographer, 5, 31–35.
171

Tsalgatidou, A., & Pitoura, E. (2001). Business
models and transactions in mobile electronic
commerce: Requirements and properties. Com-
puter Networks, 37, 221–236. doi:10.1016/S1389-
1286(01)00216-X
Tsou, M. H. (2004, July). Internet GIS: Today’s
realities and a vision of the future. GIS Develop-
ment. http://www.gisdevelopment.net/magazine/
years/2004/july/
Wooldridge, M. (1997). Agent-based software
engineering. IEEE Proceedings in Software
Engineering, 114, 26–37. doi:10.1049/ip-
sen:19971026
Wu, X., Zhang, S., & Goddard, S. (2004). Devel-
opment of a component-based GIS using GRASS.
In Proceedings of the FOSS/GRASS Users Confer-
ence (pp. 11 - 20). Bangkok, Thailand.
ADDITIONAL READING
Braun, P. (2003). Primer on wireless GIS. Park
Ridge, IL: The Urban and Regional Information
Systems Association.
Brimicombe, A. J. (2008). Location-based services
and GIS. In Wilson, J. P., & Fotheringham, A. S.
(Eds.), Handbook of geographical information
science (pp. 581–595). Oxford, UK: Balckwell.
Brimicombe, A. J. (2009). GIS, environmental
modeling and engineering (2nd ed.). Boca Raton,
FL: CRC Press. doi:10.1201/9781439808726
Brimicombe, A. J., & Li, C. (2009). Location-
based services and geo-information engineering.
Chichester, UK: Wiley.
Brimicombe, A. J., & Li, Y. (2006). Mobile
space-time envelopes for location-based services.
Transaction in GIS, 10, 5–23. doi:10.1111/j.1467-
9671.2006.00241.x
172
Mobile Geographic Information Systems
Comer, J. C., & Wikle, T. A. (2008). World-
wide diffusion of cellular telephone. The
Professional Geographer, 60, 252–269.
doi:10.1080/00330120701836303
Drummond, J., Joao, E., & Billen, R. (2007). Cur-
rent and future trends in dynamic and mobile GIS.
In Drummond, J., Billen, R., Joao, E., & Forrest,
D. (Eds.), Dynamic and mobile GIS: Investigating
changes in space and time (pp. 289–300). Boca
Raton, FL: CRC Press.
Dueker, K. J., & Kjerne, D. (1989). Multi-purpose
Cadastre: Terms and definitions. Falls Church,
VA: American Society of Photogrammetry and
Remote Sensing and American Congress of Sur-
veying & Mapping.
Goodchild, M., Yuan, M., & Cova, T. (2007).
Towards a general theory of geographic rep-
resentation in GIS. International Journal of
Geographical Information Science, 21, 239–260.
doi:10.1080/13658810600965271
Heywood, I., Cornelius, S., & Carver, S. (2011).
An introduction to geographical information
systems (4th ed.). Harlow, UK: Pearson.
Kaasinen, E. (2002). User needs for location-aware
mobile services. In Personal and Ubiquitous
Computing (pp. 70–79). London, UK: Springer.
Karimi, H. A., & Hammad, A. (Eds.). (2004).
Telegeoinformatics: Location-based computing
and services. Boca Raton, FL: CRC Press.
Li, C. (2006). User preference, information
transaction and location-based services: A study
of urban pedestrian wayfinding. Computers,
Environment and Urban Systems, 30, 726–740.
doi:10.1016/j.compenvurbsys.2006.02.008
Li, C., & Maguire, D. (2003). The handheld
revolution: towards ubiquitous GIS. In Longley,
P., & Batty, M. (Eds.), Advanced spatial analysis:
The CASA book of GIS (pp. 193–210). Redlands,
CA: ESRI Press.
Mobile Geographic Information Systems
Li, C., & Willis, K. (2006). Modelling context aware
interaction for wayfinding using mobile devices.
In [Espoo, Finland.]. Proceedings of the Mobile-
HCI, 06, 97–100. doi:10.1145/1152215.1152235
Mountain, D., & Raper, J. (2007). Mixed reality
interfaces for mobile information systems. Aslib
Proceedings: New Information Perspectives, 59,
422–436. doi:10.1108/00012530710817618
Peng, Z. R., & Tsou, M. H. (2003). Internet GIS:
Distributed geographic information services for
the internet and wireless networks. Hoboken,
NJ: Wiley.
Raptis, D., Tselios, N., & Avouris, N. (2005).
Context-based design of mobile applications for
museum: A survey of existing practices. In [Sal-
zburg, Austria.]. Proceedings of MobileHCI, 05,
153–160. doi:10.1145/1085777.1085803
Reichenbacher, T., & Tollner, D. (2003). Design
of an adaptive mobile geovisualization services.
In Garter, G. (Ed.), LBS and tele cartography
(pp. 17–23). Geowissenschaftliche Mitteilungen.
Schmalsteig, D., & Reitmayr, G. (2007). The
world as a user interface: augmented reality for
ubiquitous computing. LBS and tele cartography
(pp. 369–391). Springer.
Shiode, N., Li, C., & Batty, M. (2004). The impact
and penetration of location-based services. In
Karimi, H. A., & Hammad, A. (Eds.), Telegeoin-
formatics: Location-based computing and services
(pp. 349–366). Boca Raton, FL: CRC Press.
Tsou, M. H., & Buttenfield, B. P. (2002). A dynamic
architecture for distributed geographic informa-
tion services. Transactions in GIS, 6, 355–381.
doi:10.1111/1467-9671.00118
Wagtendonk, A. J., & de Jeu, A. M. (2007).
Sensible field computing: Evaluating the use
of mobile GIS methods in scientific fieldwork.
Photogrammetric Engineering and Remote Sens-
ing, 73, 651–662.
Walker, M., Turnbull, R., & Sim, N. (2007). Fu-
ture mobile devices – An overview of emerging
device trends, and the impact on future converged
services. BT Technology Journal, 25(2), 120–125.
doi:10.1007/s10550-007-0035-3
Wilson, J. P., & Fotheringham, A. S. (Eds.). (2008).
The handbook of geographical information sci-
ence. Oxford, UK: Blackwell.
Zipf, A., & Jost, M. (2006). Implementing adap-
tive mobile GIS services based on ontologies:
Examples from pedestrian navigation support.
Computers, Environment and Urban Systems,
30, 779–784.
KEY TERMS AND DEFINITIONS
Challenges of Mobile GIS: Research chal-
lenges raised in Mobile GIS area including
limitations derived from the technologies being
used, areas of GIScience that still need to be ad-
equately researched, users, and business models
for a sustainable presence.
Geographical Information Systems (GIS):
Also known as the research discipline of Geo-
graphical Information Science. GIS can be defined
as a system of hardware, software, data, people,
organisations and institutional arrangements for
colleting, storing, analysing, visualising and dis-
seminating spatial information about areas of the
earth (Dueker and Kjerne, 1989; Brimicombe,
2010).
GI-Engineering: The design of dependably
engineered solutions to society’s use of geographi-
cal information.
Location-Based Services (LBS): The delivery
of data and information services where the content
of those services is tailored to the current or some
projected location and context of a mobile user.
Mobile Geographical Information Systems
(Mobile GIS): With features of mobility, real-time
connectivity and location-awareness, broader
user audience.
173
 Mobile Geographic Information Systems

Mobile GIS Applications: A range of soft- Wireless GIS: An alternative term for Mobile
ware applications with characteristics of wireless GIS.
communication, mobility and GIS functionality.

This work was previously published in Ubiquitous Positioning and Mobile Location-Based Services in Smart Phones, edited
by Ruizhi Chen, pp. 230-253, copyright 2012 by Information Science Reference (an imprint of IGI Global).

174
 175

Chapter 14
A Review of Geospatial
Information Technology for
Natural Disaster Management
in Developing Countries
Sam Herold
University of Ottawa, Canada

Michael C. Sawada
University of Ottawa, Canada

ABSTRACT
Disasters are deadly and destructive events, particularly in developing countries where economic, social,
political and cultural factors increase natural hazard vulnerability. The recent devastation of the Haiti
earthquake on January 12th, 2010 was a prime example of the human toll a natural disaster can take in
developing regions of the world. There is an imminent need to improve natural disaster management
capacity in developing countries to reduce disaster impacts. Given that disasters are spatial phenom-
enon, the application of geospatial information technology (GIT) is essential to the natural disaster
management process. However, in developing countries there are numerous barriers to the effective
use of GIT, especially at the local level, including limited financial and human resources and a lack
of critical spatial data required to support GIT use to improve disaster management related decision
making processes. The results of a thorough literature review suggests that currently available free and
open source GIT (FOS GIT) offers great potential to overcome some of these barriers. Thus, disaster
management practitioners in developing countries could harness this potential in an attempt to reduce
hazard vulnerability and improve disaster management capacity. The use of FOS GIT significantly
reduces software costs and can help build local level GIT knowledge/technical skills that are required
for successful GIT implementation.

DOI: 10.4018/978-1-4666-2038-4.ch014

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
INTRODUCTION
Although the United Nations designated the
1990’s as the International Decade for Natural
Disaster Reduction (IDNDR), there was a global
failure to reduce natural disaster impacts during
that time (IFRC, 2001). Ultimately contributing
to this trend are environmental degradation, rapid
urbanization and social marginalization (McEn-
tire, 1999), particularly in developing countries.
The increasing number of disasters suggests that
vulnerability to natural hazards is also rising and
so equates to changing the geography of risk.
By way of elaboration, more people are living
in low-lying coastal zones, seismically hazard-
ous areas and concentrated urban environments
(Amendola et al., 2008; Briceño, 2004; Burton et
al., 1993; El-Masri & Tipple, 2002). Vulnerable
populations will be at increased risk, for example,
as the geography and magnitude of hydrometero-
logical hazards that are historically associated with
some of the greatest disasters (Kondratyev et al.,
2002) change with global climate (Smith, 2004;
IPCC, 2007). Defining the geography of risk is
of a major concern in general and in particular in
developing countries, “where disasters jeopardize
important social development goals such as ad-
dressing poverty, ensuring adequate food, water,
and sanitation, and protecting the environment,”
(Levy 2005, p.375). Because natural disasters
have the greatest overall impact in developing
countries (Alexander, 1995; Bui et al., 2000;
IFRC, 2001), this is where geospatial information
technologies (GIT) have the greatest potential to
mitigate causalities.
The purpose of this paper is to examine the
use of GIT for natural disaster management,
with an emphasis on how these technologies, in
particular free and open source GIS (FOS GIT)
can be effectively utilized at the local level in
developing countries. Although natural disasters
cannot entirely be prevented, disaster losses (in-
cluding human, environmental and infrastructure/
personal property) can be minimized with effective
176
disaster management – the process of mitigation,
preparation, response and recovery. The field of
disaster management has greatly benefited from
recent advancements in computers and related
information technologies. Geospatial informa-
tion technologies (GIT), including geographic
information systems (GIS), remote sensing (RS),
global positioning systems (GPS) and Internet
GIS (IGIS) are currently being employed in a
variety of ways to support all phases of disaster
management. Since “each phase is geographi-
cally related to where people, places, and things
are spatially located” (Gunes & Kovel 2000, p.
136), the entire disaster management process can
be significantly enhanced through the effective
use of GIT (Goodchild, 2006). “Even though
the natural processes (e.g., floods, earthquakes,
landslides, etc.) that generate disasters might be
fundamentally different, the techniques to assess
and mitigate risk, evaluate preparedness, and assist
response have much in common and can share and
benefit from advances in geographic information
science (GIScience) (e.g., data acquisition and
integration; issues of data ownership, access,
and liability; and interoperability).” (Radke et
al. 2000, p. 15). We propose that currently avail-
able free and open source software (FOSS) can
fulfill many GIT requirements needed to improve
disaster management capacity at the local level.
While FOSS can create additional challenges
compared to commercial solutions (Camara &
Onsrud, 2004), with a clear understanding of the
barriers and benefits of FOSS from a developing
world perspective, FOSS is a capable and effec-
tive alternative.
In the first section of this review, we begin
by laying a set of brief contextual explanations
and definitions of natural hazards and disasters
while emphasizing their spatial components, and
then describe some of the factors that differenti-
ate developed and developing countries from a
natural disaster vulnerability standpoint. Included
in this section is a more detailed examination of
the most commonly cited GIT implementation
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
barriers faced by developing countries. This is
followed by a brief review of the phases that
comprise the disaster management cycle. In the
next section, which forms the bulk of this paper,
we review the extensive literature that describes
and explores the many uses of GIT in the field of
natural disaster management. We then examine
GIT-based FOSS, highlighting its potential as well
as limitations in terms of its ability to fulfill disas-
ter management related requirements. Finally, we
discuss and describe our vision of how FOSS can
greatly improve the ability of local level disaster
managers to implement GIT, and thus improve
overall disaster management capacity and reduce
vulnerability to natural hazards.
BACKGROUND
Natural Hazards and
Disasters: An Overview
To provide context for the pursuant issues, a brief
overview of the relation between a natural hazard
and a natural disaster is in order. Natural hazards
vary greatly in terms of frequency, duration,
scale, impact, etc., and these differences partly
determine the spatial data and technology needed
to effectively mitigate, prepare for, respond to and
recover from potential disasters that may result
from their occurrence.
Simply put, natural hazards are unpredictable
acts of nature, characterized by extremes in physi-
cal processes (Zerger & Smith, 2003). Examples
of natural hazards include earthquakes, tsunamis,
hurricanes, typhoons, droughts, wildfires, tropical
storms and floods. “The fundamental determinants
of natural hazards are location, timing, magnitude
and frequency,” (Alexander 2000, p. 7). The
spatial scale and duration of natural hazards can
vary greatly, which is important from a GIT per-
spective, and in particular from the perspective of
data requirements. Landslides, for example, have
a local impact, whereas major floods can affect a
large region. Earthquakes occur with little warn-
ing and last only a few seconds to minutes, while
a drought may build up over a period of months
over large regions and last even longer. Thus, a
distinction can be made between ‘rapid-onset’
natural hazards such as floods and earthquakes,
or slower ‘creeping crises’ hazards like drought or
disease (De Paratesi, 1989). However, Coppock
(1995) points out that those slowly developing
hazards have more in common with natural re-
source management, at least from a mitigation
perspective
McEntire (2001) describes “natural disasters
as the disruptive and/or deadly and destructive
outcome of triggering agents when they interact
with, and are exacerbated by, various forms of
vulnerability,” (p. 190). Simply put, when a haz-
ard intersects the zone of human use there is the
risk of disaster. The number of people affected
by disasters resulting from natural hazard events
of rapid-onset is increasing (Oloruntoba, 2005).
In particular, hydrometeorological related natural
disasters (e.g., floods, landslides/avalanches, for-
est/scrub fires, wind storms and waves/surges)
have more than doubled since 1996 and caused
over 90 percent of deaths from natural disasters
during the 1990’s (IFRC, 2001). Herein, our
main focus is on the role of GIT for mitigation
of rapid-onset natural hazards that cause the risk
of natural disaster.
Large-scale disasters represent a complex
and multidisciplinary problem for local disaster
managers and related organizations, as well as In-
ternational humanitarian/aid organizations. While
many natural disasters are characterized by short
reaction/response times, overwhelming damage
to property and infrastructure, and a strain on the
resources of the affected community, those less
frequent large-scale natural disasters are much
more deadly. Among the largest-scale globally
predominant natural disasters with 50,000 vic-
tims or more, three hazard types can be singled
out: earthquakes, tropical cyclones (with coastal
inundation), and river floods (Kondratyev et al.,
177
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
2002). These types of natural hazards have caused
the worst calamities both in the 20th century
and the entire history of humanity, namely: the
1970 flood and cyclone in Bangladesh (300,000
victims), the 1976 earthquake in China (242,000
victims), and the 1931 flood in China (140,000
victims). The 2004 earthquake that occurred off
the coast of Sumatra, Indonesia, and resulting tsu-
nami that devastated many countries surrounding
the Indian Ocean also ranks among the deadliest
natural disasters in history, with 283,000 reported
fatalities (Lay et al., 2005).
Disaster Vulnerability: Developed
vs. Developing Countries
Although natural hazards pose a considerable
threat to all countries, historically, developing
countries have been disproportionately affected
(Briceño, 2004). “About 95% of deaths caused by
natural hazards occur in developing countries,”
(Bui et al., 2000), and the loss of GNP due to
disasters is 20 times that of developed nations
(Alexander, 1995). These alarming statistics can-
not be attributed to a greater frequency of natural
hazards in the developing world, but can be partly
explained by differences in natural hazard vulner-
ability. The concept of vulnerability is complex
– it depends on a number of parameters – and
varies depending on the research orientation and
perspective (for example, see Cutter, 1996; Cutter
et al., 2000, 2003; McEntire, 2001; Morrow, 1999;
Weichselgartner, 2001). Thus, there is no single
accepted way to assess natural hazard vulnerability
(Simpson & Human, 2008). However, regardless
of how it is conceptualized or assessed, history
shows that developing countries are more vulner-
able to natural hazards than developed countries,
and as a result experience a greater number of
natural disasters.
Although location and proximity to natural
hazards is certainly an important factor, many
researchers propose that the scale of disaster
impact is more a function of human vulnerabil-
178
ity rather than of the physical magnitude of the
hazard (Hewitt, 1995; Quarantelli, 1998; Smith,
2001). In this sense, “natural disasters can more
accurately be seen as social phenomenon, where
the overall damage due to natural hazards is the
result of both natural events that act as ‘triggers’,
and a series of societal factors,” (Weichselgartner,
2001). Conceptualizing vulnerability in terms of
societal factors as opposed to biophysical fac-
tors helps to explain the “wide range of human
impact that can result from events of comparable
magnitude,” (Steiniger & Bocher, 2009, p. 1346).
For example, Montoya and Masser (2005) point
out that the “1988 Spitak earthquake in Armenia
(former USSR) and the 1989 Loma Prieta earth-
quake in California were of similar magnitude
and affected populations of comparable size;
however, the Armenian event killed 25,000 people
whereas the California earthquake killed 63,” (p.
495). Hurricanes “Hugo in 1989 and Andrew in
1992 caused less than 50 deaths each in the U.S.
On the other hand, cyclones in Bangladesh,” (p.
432) killed over half a million people in 1970 and
another 140,000 in 1991 (Bui et al., 2000). Clearly,
the extent of human (or societal) vulnerability is
a major factor that contributes to determining the
scale of a disaster, but why are people in develop-
ing countries so vulnerable?
The vulnerability of a geographic “area is
determined by its natural and human-made envi-
ronmental conditions, climatic patterns, and its
political, social and economic ability to withstand
and respond to natural hazard events,” (Jayara-
man et al. 1997, p. 294). Hazards tend to be
more destructive in developing countries where
political, social and economic instability lead to
poor organizational infrastructure and no adaptive
capacity as compared to most developed countries.
Henderson (2004), in examining the pervasive risk
of natural disaster faced by developing countries,
emphasizes that “conditions of poverty, poor
housing, lack of information about disaster risk,
poor telecommunications, and inadequate physical
infrastructure frequently exacerbate natural disas-
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
ters,” (p. 107). Insufficient disaster management
resources, the absence of enforced laws and the
shortage of trained experts all increase the diffi-
culty in coping with natural disasters (Guinau et
al., 2005). For example, the high cost and skills
required for the application of GIS technology has
hindered its utilization for disaster management
in developing countries (Rudyanto et al., 2001).
The lack of ability to effectively utilize relevant
technology is a societal factor that contributes to
increasing disaster vulnerability.
Developed countries have a greater awareness
and understanding of the importance of natural
disaster management. Bui et al. (2000) point out
that “they invest more in mitigation and preven-
tion, and have more resources available to enforce
legislation that may help reduce human vulnerabil-
ity,” (p. 432). In developed countries, “insurance
absorbs more than half of the economic losses from
natural disasters, in contrast, less than 2% of the
losses are insured in developing countries,” (Free-
man & Pflug 2003, p. 601). However, as a general
trend, in both developed and developing nations,
“disaster management becomes a more pressing
concern only after the disaster has struck,” (Cur-
rion et al., 2007). Perhaps the devastating impact
of recent large-scale disasters, such as the Indian
Ocean tsunami (2004), Hurricane Katrina (2005),
and the Haiti earthquake (2010) will encourage
disaster managers in both the developed and de-
veloping world to focus more on mitigation and
preparation, rather than response and recovery.
There are many factors that contribute to
determining disaster vulnerability, and it is not
our intention to discuss them all, but rather to
demonstrate that the concept is pervasive within
the disaster management literature, particularly
from a developing countries perspective. In ad-
dition, since many of the factors that determine
vulnerability vary spatially, vulnerability is well
suited to be assessed using GIT. For example,
McEntire (2001) considers vulnerability to be
“the dependent component of disaster that is
determined by the degree of risk, susceptibility,
resistance and resilience,”(p. 190). Risk is created
because of proximity to potential hazards, which
varies spatially; “resistance refers to the ability
of buildings and the infrastructure to resist the
strain exerted by natural hazards,” (McEntire
2001, p. 191) which also varies from one area to
another depending on building design, building
codes and the materials used, among other factors.
Susceptibility, “which is a product of social, politi-
cal, economic and cultural forces that determine
the proneness of groups and individuals to be
adversely affected by disasters,”(McEntire 2001,
p. 191) can also be analyzed in spatial terms. For
example, characteristics such as age, gender, race
and socioeconomic status are generally accepted to
be influential components of social vulnerability
(Cutter et al., 2003), and these characteristics have
a high degree of spatial variability, particularly
within large urban areas. Such characteristics are
typically captured in a census, and this informa-
tion can be analyzed and mapped at varying scales
using GIT, such as GIS.
Considering that assessing and understand-
ing hazard vulnerability is a key component of
disaster management, and that vulnerability varies
spatially, it is well suited to be analyzed using
GIT. Therefore, GIT must play a key role in any
comprehensive and effective disaster management
strategy, yet there are dramatic differences in GIT
use between developed and developing countries.
While each developing country is unique, there
exists a common set of well-identified barriers
that often limit the use of such technology, thus
increasing vulnerability to natural disaster. These
common barriers include:
1. A lack of financial resources
Generally, using geospatial technology is
costly: requiring a substantial investment in com-
puter software/hardware, spatial data, and human
resources (education and training). In many cases
disaster management responsibilities and duties
are decentralized to local governments “without
179
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
being accompanied by the necessary funds,”
(Montoya & Masser 2005, p. 494). Considering
that local governments have to provide many
essential “services (e.g., health, education and
infrastructure services such as water, electricity
and sewage, to name a few),” (Montoya & Masser,
2005, p. 495) they are “often not predisposed to
purchasing off-the-shelf commercial GIS software
due to other demands for funding,” (Renyi &
Nan, 2002, p. 1). It is difficult to justify spend-
ing limited budgets on technology to enhance
disaster mitigation and preparation when basic
needs have priority. In addition, software has no
benefits without the ability to purchase or collect
data. Thus, limited finances are a major barrier
and if GIT is going to play a role in local level
natural disaster management financially feasible
solutions such as FOSS can help.
2. A lack of local expertise/knowledge (human/
technical resources)
Effective use of GIT requires considerable
technical knowledge and skills, obtained through
education and training, as well as practical experi-
ence. Mohamed and Plante (2002) emphasize “the
lack of local expertise and capacity to operate and
maintain a GIS,” (p. 2286) in developing countries
and general awareness is strongly lacking (Murgia
et al., 2002). In contrast, the well established GIT
communities built up over time in developed
countries are an important resource for individual
practitioners to acquire new knowledge, methods
or techniques, discuss ideas and/or obtain support.
In some developing countries, practitioners may
be physically separated by hundreds of kilometres
and be without Internet or telephone (Britton,
2000). As such, local disaster management or-
ganizations would have trouble in trying to find
local personnel who could effectively manage the
various GIT components/aspects of a holistic natu-
ral disaster management strategy. When GIT has
been implemented at the local level in developing
countries, it is often the result of externally spon-
180
sored/funded technology development programs,
rather than through the gradual building of local
GIT capacity (Britton, 2000). “Local workforce
development and capacity building projects are
essential for,” (Mohamed & Plante 2002, p. 2286)
sustained GIT implementation and maintenance
that does not rely on external support or funding.
3. Institutional/political instability
Disaster management is dependent on the
“functional and effective operation of institutions,
whether formal or informal, and at the local level
where it matters most,” (Pande, 2006, p. 426).
The effectiveness of GIT and the ability to con-
vert information into action also depends on the
existence of supporting organizations, which are
generally lacking in developing countries (Cop-
pock, 1995). Political barriers have also been cited
as obstacles to local level GIT use (Montoya &
Masser, 2005). High resolution census data, for
example, contains a wealth of information that
could be utilized in a variety of GIT-based natural
disaster management related scenarios but this
data is often unavailable in developing countries
for various reasons. Governments in developing
countries can be unstable, and political terms
can be short. According to Murgia et al. (2002)
“politicians have their own circle of supporters
who will be appointed in public institutions, and
thus political instability is expanded to institutional
instability. Considering institutional instability,”
(p. 6). Ramasubramanian (1999) suggests that GIS
implementation in developing countries would
be more successful in the hands of “empowered
individuals and groups rather than organisational
structures,” (p. 359).
4. A lack of spatial data
“Accurate and comprehensive spatial data play
a critical role in all phases of disaster manage-
ment,” and are required for effective GIT use, yet in
many developing countries reliable spatial data are
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
a scarce resource (Brodnig & Mayer-Schönberger
2000, p. 3). Murgia et al. (2002) suggest that part
of the problem are “weak national data providers
such as mapping agencies, census organizations
and cadastres, and that many projects that have
been initiated are stand-alone and lack continuity
and consistent funding or are not embedded in
stable institutions or are politically misused,” (p.
6). In addition, governments in developing coun-
tries tend to have a very conservative approach
in terms of data and information management
(Shrestha, 1994). “For example, among the most
common obstacles to be found in Asia are official
restrictions on geospatial data for security reasons
as well as the rigidity and compartmentalization of
government bureaucracies, which consider certain
types of information as their property,” (Brodnig
& Mayer-Schönberger 2000, p. 3). In Costa Rica,
for example, access to census tract level data is
restricted (Montoya & Masser, 2005). Additional
factors that contribute to the lack of spatial data
include: (1) the remoteness of many areas; (2) the
lack of technical capacities to collect and assess
bio-physical and socio-economic data; and (3)
competing priorities in the fledging economies
(Brodnig & Mayer-Schönberger, 2000; Shrestha,
1994). Montoya and Masser (2005) emphasize the
need “to identify or develop cost-effective data
collection methods for producing spatially refer-
enced information in developing countries. The
role of spatial data in all areas of natural disaster
management cannot be underestimated; in fact,
Mansourian et al. (2004) propose the development
of a spatial data infrastructure (SDI) as a frame-
work to facilitate disaster management,” (p. 495).
Aside from the barriers to GIT implementation
just described, Ramasubramanian (1999) also
identifies cultural and language barriers in terms
of GIS implementation in developing countries
(for more on specific GIS implementation issues
in developing countries see Ramasubramanian,
1999). Given the aforementioned barriers, disas-
ter managers in developing countries are unable
to utilize GIT in the same ways as developed
countries, which in effect, further increases their
vulnerability to natural disasters. However, recent
growth in the domain of free and open source soft-
ware (FOSS), and in particular GIT-based FOSS,
provides new opportunities that were previously
unavailable. For example, the cost alone of com-
mercial GIT software (such as GIS) can prevent
it from being used at the local level in developing
countries (Ranyi & Nan, 2002); with FOSS, the
issue of cost is eliminated. For this reason, and
others that will be described later in this paper,
FOSS is increasingly being recognized within the
disaster management literature as a good fit for
developing countries.
Disaster Management
and the Role of GIT
Disaster management can be understood as a cycle
(Figure 1) which includes an effort to “mitigate
against, prepare for, respond to, and recover from
a disaster,” (Montoya 2002, p. 1). The core com-
ponents of this cycle are also commonly referred
to within the emergency management literature
(e.g., Cutter, 2003), and often in conjunction with
natural disasters (Cova, 1999). Thus, emergency
and disaster management are closely linked; after
Figure 1. Disaster management cycle (Adapted
from Alexander, 2000)
181
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
all, a disaster clearly constitutes an emergency, but
in this paper it will be referred to as the disaster
management cycle.
Mitigation refers to efforts that aim to eliminate
or reduce the risk to humans and/or property caused
by natural or man-made hazards (e.g., risk assess-
ment, insurance, engineering standards, land use
management, public education, etc.). Preparedness
refers to activities “necessary to the extent that
mitigation measures have not, or cannot, prevent
disasters,” (Johnson 2000, p. 2). This involves
developing operational capabilities for responding
to a sudden disastrous situation. In this phase
“governments, organizations, and individuals
develop plans to save lives and minimize potential
disaster damage,” (Johnson 2000, p. 2). This in-
cludes emergency planning, training exercises,
implementing hazard warning systems, evacuation
procedures and stock-piling of critical supplies.
In addition, “preparedness measures also seek to
enhance disaster response operations,” (Johnson
2000, p. 2). Response refers to those actions taken
just prior to, during, and immediately after a di-
saster that save lives, reduce property damage or
improve recovery. The most important aspect of
this phase involves providing emergency assistance
for victims (e.g., search and rescue, emergency
shelters, medical care and food/water). Disaster
responders “also seek to stabilize the situation and
reduce the probability of secondary damage,”
(ERSI 2006, p. 4). Recovery (often used in con-
junction with the word ‘relief’) includes those
activities that (1) restore vital life support systems
and (2) return the area/population to a pre-disaster
state. The former can be seen as part of the short-
term recovery plan while the latter as part of the
long-term, which may continue for a number of
years after a disaster. Reconstruction is closely
linked with mitigation, and is undertaken in ways
that aim to reduce vulnerability and improve pre-
paredness, thus disaster management occurs in a
cyclical fashion as depicted in Figure 1.
Geospatial information technology (GIT), and
its ability to acquire, interpret, analyze, map and
182
disseminate information, are essential in all areas
of natural disaster management. Because GIS is a
spatial decision support tool it is invaluable when
effectively used in a field like disaster manage-
ment that deals with critical spatial decisions
(Cova, 1999). For instance, GIT provides the
basis for estimating and mapping risk, planning
evacuation routes, determining suitable areas for
shelters, identifying disaster victims, and assign-
ing resources during recovery (Goodchild, 2006).
It follows then, that GIT must play a key role in
any comprehensive and effective disaster manage-
ment strategy. Figure 2 incorporates GIT within
the disaster management cycle to emphasize the
central role it plays during all phases.
Disaster management is complex, and involves
the participation and collaboration of many insti-
tutions/organizations/agencies operating at inter-
national, national, regional and local levels. Such
institutions/organizations/agencies use a range of
GIT depending on their specific information re-
quirements, financial resources and technical
capabilities. While a strong national disaster man-
agement initiative is important, Henderson (2004)
emphasizes that disaster management capabilities
should be decentralized to the regional or local
level given the wide variations in demographic,
socioeconomic, cultural and infrastructural condi-
tions within regions and local areas of a nation.
Regional and local authorities involved in disaster
management should ideally have the capacity to
effectively utilize GIT, yet even in developed
countries limited financial resources can prevent
such utilization within this administrative level
(Laben, 2002). Theoretically, this approach is
increasingly being recognized to be more effective
than the centralized approach, but as Montoya amd
Masser (2005) note, local authorities are often not
provided with the necessary financial resources
to develop and implement effective policies and
plans, which should incorporate GIT. Laben (2002)
emphasizes that it is extremely important to have
GIT alternatives available that can be used by all
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
Figure 2. A disaster management cycle incorporating GIT (adapted from Hussain et al., 2005)
levels of the emergency and disaster management
communities.
GEOSPATIAL INFORMATION
TECHNOLOGY AND
DISASTER MANAGEMENT
Geospatial Information Technology is used for
many disaster management functions, includ-
ing hazard and risk assessment (Dewan, Kabir,
Islam, Kumamoto, & Nishigaki, 2007; Ramli et
al., 2005), vulnerability assessment (Cutter et al.,
2000; Kienberger & Steinbruch, 2005; Weichsel-
gartner, 2001), vehicle dispatch and supply routing
(Dong, 2005), damage assessment (Chen et al.,
2005; Rivereau, 1995; Zhang et al., 2002) and
resource mobilization (Goodchild, 2006), among
many other essential tasks. In this section we ex-
amine the use of GIS, remote sensing and Internet
GIS in the area of natural disaster management.
Geographic Information
Systems (GIS)
A GIS is an “organized collection of computer
hardware, software, geographic data, and person-
nel designed to efficiently capture, store, up-date,
manipulate, analyze, and display all forms of
geographically referenced information” (ESRI,
1993, pp. 1-2). A GIS is a key component of any
effective and comprehensive disaster management
strategy, and are used to display, integrate, map,
analyse, and model data and information derived
from satellites, and other spatial data sources
(Kumar et al., 1999). GIS function primarily as
a support tool to help answer essential questions
and make informed, timely and appropriate deci-
sions that can help save lives. Perhaps the greatest
strength of GIS are their ability to integrate a wide
range of data types, including geographic, social,
economic, and political data into a single system
(Dash, 1997).
However, to utilize GIS not only requires the
software, but also hardware, data and trained per-
sonnel. Thus, a GIS capability involves not only
the software itself, but the necessary spatial and
183
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries

descriptive (attribute) data, computer hardware
and personnel who can effectively utilize it (Dash,
1997; Montoya, 2002). As a result, the use of GIS
for disaster management varies widely between
individual countries, and within disaster related
organizations and institutions at different levels
of government (Laben, 2002). These differences
in GIS capability are clearly reflected in the wide
range of literature that examines and describes
the use of GIS for natural disaster management.
This section of the paper will review the GIS
and disaster management literature in order to
identify examples of GIS capabilities (software,
data, hardware, and level of expertise required)
that are feasible for implementation in developing
countries. These are all important considerations
that will be emphasised and discussed further.
Since there are distinct differences between
using a GIS for pre- or post-disaster manage-
ment activities, this section has been divided in
half, one-half that deals with pre-disaster GIS
use and another that examines post-disaster use.
However, before these sections are presented it is
first necessary to discuss the importance of spatial
data, since it is the primary input for GIS and is
required for their use.
The Importance of Spatial Data
The success in utilizing GIS depends largely
on the availability of spatially-referenced data;
the quality of which is determined by locational
precision, the characteristics of the attribute data,
and by the extent to which standards are adopted
that allow for data transfer (Coppock, 1995). The
importance of good quality spatial data for use
with GIS to assist disaster management cannot
be underestimated. There are several types of
spatial information that are useful for disaster
management related decision making – the type,
scale and complexity of such data depend on the
type, scale and stage of the disaster (De La Ville
et al., 2002), as well as on the GIT capabilities
of the implementing institution/organization. As
an example, Table 1 describes some of the vari-
ous types of spatial data/information applicable
for disaster management that can be utilized in a
GIS environment.
Over the years, developments in spatial data
collection and use have led to the concept of
spatial data infrastructure (SDI). Musinquzi et al.
(2004) suggest that an ideal SDI includes “tech-
nologies, policies, standards, and human re-
sources to acquire, process, store, distribute, and
improve utilization of geo-spatial information”
(p. 790). Generally speaking, the “success of
disaster management largely depends on the avail-
ability, dissemination and effective use of infor-
mation,” (p. 3) much of which is spatial (Ven-
katachary et al., 2002). Accordingly, then, there
is a relationship between GIS, SDI and the extent

Table 1. Crucial spatial data for disaster management (adapted from Gunes & Kovel, 2000)

Data/Information Type Description

Disaster forecast Information concerning the extent of a particular hazard or disaster
Vulnerability analysis Information on critical facilities (hospitals, schools, shelters, police and fire facilities, dams, trauma centers,
industrial facilities, etc.);
Information regarding human vulnerability (age, gender, socioeconomic status, etc.)
Damage assessment Data/imagery of the actual impact of a hazard
Resource inventory Location information regarding supplies, equipment, vehicles, or other material resources
Infrastructure Shows transportation networks (roads, railroads, bridges, traffic control points, and evacuation routes) as
well as complete utility grids (electric, gas, water, and sewer)
Mass care/shelter status Monitors the movement of people to and from government or voluntary agency shelters by providing infor-
mation on capacity, availability, supplies, and suitability to victims’ needs

184
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
to which spatial information is utilized for disas-
ter management related purposes.
Most developed countries are at various stages
of developing national SDIs, and their success
can be linked to high levels of technology, avail-
ability of funds, trained personnel and political
support and stability (Musinguzi et al., 2004).
Many of these factors that contribute to success-
ful national SDI development are lacking in most
developing countries, and as a result, there are vast
differences in spatial data quality and quantity
between developed and developing countries. As
well, without a national SDI there is no effective
mechanism for coordinating data collection ef-
forts or for sharing spatial data among agencies
or departments. This often results in duplication
of data and wastage of human/financial resources.
Limitations in accessing and sharing spatial data
are particularly problematic for disaster manag-
ers, whose spatial data needs typically cut across
departmental/agency boundaries (Rego, 2001).
However, some developing countries are making
progress, including Sri Lanka, for example, who
just recently (2008) initiated a national SDI (http://
www.survey-dept.slt.lk/). Their policy statement
makes specific reference to disaster management,
and how a SDI is required to reduce existing
inefficiencies.
While there are distinct differences between
developed and developing countries in terms of
the quality and quantity of currently available
spatial data, efforts are being made to improve
the current situation. Nebert (2004) points out
that “many national, regional, and international
programs and projects are working to improve
access to available spatial data, promote its reuse,
and ensure that additional investment in spatial
information collection and management results
in an ever-growing, readily available and useable
pool of spatial information” (p. 6). Although such
initiatives may not be part of a formal SDI policy,
they undoubtedly help the global GIS community
and others who may use spatial data.
The Internet has also drastically increased
access to spatial data, as numerous websites
now offer free access to a wide range of data
(e.g., political/administrative boundaries, roads,
hydrology, digital elevation, land cover, etc.).
However, “despite the rapid growth in available
spatial data there has been little attention paid”
to aspects of its quality, including “currentness,
lineage, locational accuracy, completeness, and
overall usefulness,” (Engler & Hall 2007, p. 345).
This is unfortunate, because the quality of spatial
data is particularly important when it is used for
disaster management, when lives are potentially
at risk. Although satellite imagery is an essential
type of spatial data for disaster management, and
is abundantly available online, its discussion is
reserved for subsequent sections, and spatial data
in the present context refers to vector data only.
Scale can also limit the usefulness of online
spatial data. Currently, data available for develop-
ing countries is most often at the national scale
and derived from small-scale map sources. Such
data is inappropriate for regional or local level
use for a variety of reasons, including locational
inaccuracy and a lack of attribute information.
Even if the scale of the data is appropriate, some
data may require additional time investments to
clean, correct, or add the relevant attribute data
before they can be effectively utilized. In addi-
tion, problems can arise when trying to combine
data from disparate sources due to variations in
coordinate systems and map projections. Engler
and Hall (2007) emphasize that, in many cases, it
may be better to create or purchase the required
data elsewhere, as there may be too much work
to be done or uncertainty associated with using
data freely available from the Internet.
Spatial data are essential in all areas of disaster
management, yet good quality spatial data may
be hard to come by, especially at the local level in
developing countries. To improve and increase the
use of GIS for disaster management will require a
substantial effort in data collection, one that can
be facilitated with an effective SDI policy. Rego
185
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
(2001) stresses that the development of spatial
databases should be “built bottom up from the
lowest administrative unit in country (e.g., the
sub-district or district). The district databases
would then feed into the state/provincial database
and then into the national database,” (p. 2). A
bottom-up approach to spatial database produc-
tion is especially relevant for local-level GIT use
in developing countries.
Pre-Disaster
A GIS has much to offer in the pre-disaster man-
agement phases of mitigation and preparedness.
To effectively mitigate and prepare for a disaster
requires not only detailed knowledge/informa-
tion about the “expected frequency, character,
and magnitude of hazardous events in an area,
but also the vulnerability of the people, build-
ings, infrastructure and economic activities in
a potentially dangerous area,” (Van Westen &
Hofstee 2001, p. 442). This information forms
the cornerstone of preparedness planning and
helps determine appropriate mitigation strate-
gies. A GIS allows for the synthesis and analysis
of such data/information to help determine risk
levels, assess vulnerability, model scenarios, plan
evacuation routes, determine resource require-
ments, and create a variety of useful information
products to aid decision-making. Some specific
examples of GIS use in the pre-disaster phases are
particularly relevant for developing countries and
so we elaborate upon these as cases of effective
GIT implementation in pre-disaster management.
Van Westen and Hofstee (2001) present a
case study that used GIS to create a spatial data-
base of buildings, land parcels, roads and other
infrastructure by digitizing features from aerial
photographs. Subsequently, a field team of in-
vestigators collected specific attribute data for
each digitized land parcel and building, using a
pre-determined checklist. For example, land use
of the parcel (residential, industrial, commercial,
etc.), building material, building age, number
186
of floors and whether previous hazard damage
had occurred were among the attribute data col-
lected. The resulting database, when combined
with historical data on previous disasters, such
as flood depth, was used to generate a variety of
vulnerability and risk maps. The database and
maps that can be generated serve as a basis for
future development and planning that takes into
account both biophysical and human vulner-
abilities. This study was conducted for the city
of Turrialba, Costa Rica, which has a population
just over 30,000. Data collection and analytical
methods are relatively simple to repeat, making
this an excellent example of the type of GIS use
that is practical for local level disaster managers
in developing countries.
Cutter et al. (2000) present a county level
“hazards-of-place” (Cutter, 1996) based GIS
method for assessing hazard vulnerability in spatial
terms. Using twelve environmental threats and
eight social indicators (e.g., age, race/ethnicity,
income levels, gender, building quality, etc.) their
study demonstrates how a GIS can be used to
integrate both biophysical and social factors that
contribute to hazard vulnerability. Social charac-
teristics of the population are available from most
national censuses, and census data is often used
in a GIS to map human vulnerability to various
natural hazards. The research methodology and
conceptualization of hazard vulnerability provides
a template for others to follow, and helps fill a “void
in the literature on spatial analytical approaches to
vulnerability,”(Cutter 1996, p. 530) assessment. If
vulnerability to natural hazards can be identified,
appropriate steps can then be taken to reduce the
social and economic impacts of potential disasters.
However, the authors acknowledge that imple-
menting this approach at the “local level may
pose challenges in terms of availability of funds
for training and data acquisition,” (Cutter 1996,
p. 530). Furthermore, results are based on a data
intensive methodology, and while this approach
is possible in developed countries, it is unlikely
feasible in most developing counties due to a
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
lack of detailed spatial data, in addition to other
GIS implementation issues previously described.
However, it seems likely that this methodology
could be simplified and still produce useful hazard
vulnerability information.
Guinau et al. (2005) used a GIS to create a haz-
ard susceptibility map to help mitigate landslide
risk. The methodology applied focused solely on
the biophysical factors that cause landslides. The
authors digitized present and past landslides from
aerial photographs to create a landslide inventory,
which was then “overlaid on terrain data, such as
lithology, slope, soil characteristics and land use”
analysis of terrain conditions in areas affected by
landslides made it possible to determine zones
with similar characteristics,” (p. 320) and through
further analysis the delineation of low, medium
and high susceptibility zones. Perhaps most im-
portant about this study is that it demonstrates
a relatively simple GIS-based methodology to
assess landslide vulnerability that is feasible for
developing countries.
Dewan, Islam, et al. (2007) integrated GIS and
remote sensing techniques to analyze the flood
hazard and risk levels in Dhaka, Bangladesh. A
major impediment was unavailable digital geospa-
tial data, and as a result, a number of the required
data layers had to be created. “Flood-affected
frequency and flood depth were estimated from
multi-date Synthetic Aperture Radar (SAR) data,”
(Dewan et al. 2007a, p. 1602) (from RADARSAT),
based on previous flood events. Land-cover was
generated using a combination of approaches and
data sources, including on-screen digitization of
features from high resolution remotely sensed
imagery, recently produced topographic maps,
field surveys and the use of hand-held GPS units.
A geomorphic map was also developed using a
1999 LANDSAT TM image in conjunction with
available paper maps and field observations. El-
evation data in the form of a DEM was obtained
from the Institute of Water Modeling (IWM),
Bangladesh. All the vector data layers were then
converted to raster at identical resolutions. Us-
ing a relatively simple procedure that involved
assigning a weighted score to each data type (to
represent varying significance) and the use of
GIS overlay functionality, the authors were able
to create maps depicting flood hazard potential
and flood risk zones. The data collection methods,
analytical techniques and general approach used
in this study demonstrate innovative and adaptive
use of GIT in the context of a city-scale hazard
assessment for developing countries.
Uncontrolled and informal housing develop-
ment is a recurring problem in cities in developing
countries (Thomson & Hardin, 2000), and such
development practices contribute to increasing
natural hazard vulnerability. To address this prob-
lem, Thomson and Hardin (2000) used GIS and
satellite imagery (LANDSAT TM) to “identify
potential low income housing sites in the eastern
portion of the Bangkok Metropolitan Area, where
flood risk is a concern. Location, infrastructure,
land cover, and natural environment factors were
used to assess the characteristics appropriate for
public housing sites. Due to a shortage of spatial
data, all GIS coverage required were derived
from satellite imagery or digitized from map
sources, including land use, land parcels, roads
and drainage,” (p. 102). To derive land cover,
for example, a combination of techniques was
used, including unsupervised (to identify spectral
clustering) and supervised classification based on
local knowledge, and the use of aerial photographs
for verification. Major roads were digitized on-
screen in the GIS from the satellite imagery, and
the remaining data were generated using a variety
of low-tech methods, including visual image
interpretation (and digitization) and field work.
Using basic GIS functions, such as overlay and
buffering, a final map was produced that identi-
fied “sites possessing appropriate criteria for
low-income housing: non-built, greater than 25
hectares, within 1 km of a road, and least prone
to flooding,”(Thomson & Hardin 2000, p. 98).
The map can be effectively utilized at large and
medium scales (e.g., 1:20,000 – 1:50,000), and the
187
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
analysis methods and techniques are considered
feasible for cities in developing countries, in terms
of available data and required skill sets. A better
understanding of suitable locations for potential
housing sites could help guide development in
ways that reduce natural hazard vulnerability.
Knowledge of the spatial characteristics
of hazards and vulnerability before a disaster
strikes enables disaster management authorities
and emergency workers to identify areas that are
likely to be most affected, and thus efficiently
focus their resources. Such knowledge is critical
to the development of effective mitigation and
preparation strategies that will ultimately help
reduce the devastating impacts that can result
from natural disasters.
Post-Disaster
Disaster responders, emergency personnel, aid
workers and anyone involved in the response and
relief effort need timely and up-to-date informa-
tion, such as the extent of damage and location of
potential victims, the location of critical facilities
(e.g., shelters, hospitals, air strips, etc.), available
resources (e.g., food, water, blankets, medical
supplies, etc.), infrastructure conditions (e.g.,
damaged roads/bridges/utility lines, etc.), and
evacuation or supply drop off points. Much of
this information is spatial and is thus well suited
to be compiled and analyzed in a GIS and then
disseminated as maps. However, using GIS post-
disaster is different and poses greater challenges
than using it pre-disaster, since time becomes the
critical factor (Goodchild, 2006). Immediately
following a disaster, information must be quickly
collected, analyzed and assembled into useful
information products that can assist response
efforts. Thus, to the extent possible, GIS data
should be collected and/or analysed in advance of
an event, rather than having to put it all together
during the aftermath (ESRI, 2006). Bottlenecks
in information flow and dissemination can mean
the difference between life and death. Therefore,
188
having a database of the most critical and useful
spatial data before a disaster strikes is essential.
It allows for quick updates of the data to reflect
ground conditions and reduces the time it takes
to produce critical information products required
by disaster responders and related emergency
response personnel.
Following the initial response period, and once
the situation has stabilized, a GIS can be used to
analyze disaster impacts and help plan the reha-
bilitation process in a way that reduces potential
vulnerabilities. Development of this fashion is
termed ‘invulnerable development’ (McEntire,
2001). The following few examples demonstrate
the value and necessity of GIS during the post-
disaster management phases.
Impact analysis is one fundamental use of GIS
in the post-disaster phase. Impact analysis can
assist response efforts by identifying those areas
most in need, and can help guide reconstruction
efforts in a way that will minimize the potential
for future disasters; for example, through im-
proved land use planning that takes into account
local hazard vulnerabilities. In conjunction with
IKONOS panchromatic images, De La Ville et
al. (2002) used GIS to evaluate the distribution
of landslide erosion scars and their effects on
several urban areas situated among six mountain
catchments in Venezuela. The GIS was used to
analyze and map the distribution of scars and
deposition zones, and to study the factors (e.g.,
slope, geology, land cover, etc.) that contributed
to causing the various types of mass movements.
All the information derived from this study was
used by the corresponding government agencies
as the basis for the preparation of reconstruction
plans for the affected areas.
Following the Indian Ocean tsunami of 2004,
and after evoking the International Charter on
Space and Major Disasters to acquire pre- and
post-disaster high resolution satellite imagery,
Magsud et al. (2005) used a GIS for rapid damage
assessment of buildings in Galle, Sri Lanka. The
GIS was used primarily to combine different data
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
types and to support a visual analysis of building
damage. QuickBird multi-spectral imagery and a
1:5000 scale vector layer of buildings (obtained
from the Survey Department) were overlaid in a
GIS to accurately identify construction that existed
prior to the disaster. In addition, a ground survey
of 81 buildings was undertaken that identified the
level of damage following the disaster (“complete-
ly destroyed, partially damaged mainly inside,
partially collapsed with roof intact, and slightly
damaged,”(Magsud et al. 2005, p. 1)). GPS was
used in conjunction with ground photographs to
accurately record the location and building dam-
age, respectively. This allowed for a comparison
between the damage level in the photographs and
satellite imagery. Results indicated that heavily
damaged buildings can easily be identified, but
partial damage, particularly if the roof was still
intact, was difficult to determine from the satellite
imagery alone. The mapped results clearly show
the location of destroyed buildings, which is valu-
able information that could be used to prioritize
response operations. With an adequate amount
of technically trained personnel, the authors’
propose that a “near real time damage assess-
ment could be possible,” (Magsud et al. 2005, p.
2). The analytical methods and data inputs used
in this study are relatively simple, yet effective,
and are therefore considered appropriate/suitable
for disaster managers in developing countries
given that suitable data availability and delivery
methods are in place.
In general, GIS are well suited to provide valu-
able information to assist disaster response opera-
tions. For example, GIS has been used to determine
the extent of a disaster and estimate damage (Ranyi
& Nan, 2002), organize resource inventories and
their geographic distribution (Hussain et al., 2005),
monitor shelter/refugee camp status and the state
of transportation infrastructure (Gunes & Koval,
2000) and integrate disparate spatial data sources
that may be required to guide response (Amdahl,
2001). As such, GIS can help with search and res-
cue, providing medical services, debris removal,
sheltering, and infrastructure repairs. However, a
large number of spatial data layers are required
for planning and coordinating such operations,
and without them the value and usefulness of a
GIS decreases substantially, possibly to a point
where it may no longer be required. Therefore,
to some extent, the usefulness of GIS following a
disaster hinges upon the existence of the required
framework datasets like roads, critical facilities,
and population density.
Finally, Zerger and Smith (2003) emphasize
that “the suitability of GIS for planning verses
real-time applications is quite different. Results
from a test scenario indicated that the utility of
GIS for real-time decision making is questionable
owing to a number of practical and implementa-
tion impediments,”(p. 124) including a lack of
training “and the need for temporal resolution
rather than spatial detail,” (p. 124). Using GIS
for pre-disaster management functions, such as
vulnerability assessment or evacuation route plan-
ning, poses fewer challenges than trying to use the
technology post-disaster, when time is critical and
ground conditions may be constantly changing.
Nevertheless, assuming there is adequate data and
personnel to effectively utilize GIS, few would
question its ability to assist disaster response and
recovery operations.
Remote Sensing
The potential of remote sensing (RS) to provide
critical earth observation information for disaster
management (e.g., hazard assessment, disaster
mitigation, preparedness, response and recovery)
has repeatedly been emphasized (e.g., Becking,
2004; Chen et al., 2005; Jayaraman et al., 1997;
Mansor et al., 2004; Rivereau, 1995), and is
reflected in over 400 scientific articles between
1972 and 1998 (Showalter, 2001). For instance,
remotely sensed imagery can answer questions
such as: “what did/does the area look like pre-/
post-disaster? RS data can show the land cover
and topographic features in an area and can il-
189
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
lustrate infrastructure and population density,”
(Becking 2004, p. 1). Showalter (2001) provides
an in-depth review of the use of RS in hazard and
disaster research, and “concludes that the tech-
nique is primarily used to detect, identify, map,
survey and monitor existing hazards and/or their
effects;” secondary goals of RS focus “on damage
assessment, improved planning, or the provision
of data for disaster management functions,” (p.
22). Simonovic (2002), in examining the repeat
frequency, spatial resolution, and types of sensors
on-board, provides insight into the suitability of
specific satellites to different natural disasters.
Multispectral scanners (optical sensors) and radar
collection systems are probably the two most
widely recognized RS capabilities that are used
in support of disaster management (GDIN, 1997).
The applicability of remote sensing for disaster
management is perhaps best exemplified in the
case of flooding, and many researchers address
its use for this disaster type. Satellite imagery
can be used to assess the extent of past flood
events (Dewan, Islam et al., 2007) and aid in
the development of flood hazard potential maps.
Zhang et al. (2002) and Jayaraman et al. (1997)
highlight the potential of RS technology to drasti-
cally assist flood response and relief operations
by providing inundation mapping and damage
assessment. Disasters resulting from floods are
a logical choice for RS analysis because: (1)
floods generally cover large areas, and thus oc-
cur at spatial scales much larger than the spatial
resolution of most satellite imagery, and (2) water
has a unique spectral reflectance which makes it
clearly discernable from other ground features
(Showalter, 2001). In contrast, earthquakes, for
example, may cause significant damage to build-
ings and infrastructure, but without high resolution
imagery or change detection capabilities it can be
difficult or impossible to identify.
Remotely sensed digital elevation data (often
termed DEM (digital elevation model) or DTM
(digital terrain model)) are a digital representation
of surface topography and are frequently used in
190
the study of natural hazards. DEMs are a critical
data input for assessing landslide susceptibility
(Guinau et al., 2007), delineating flood risk po-
tential (Dewan, Kabir et al., 2007), flood hazard
mapping (Dewan, Islam et al., 2007; Sanyal &
Lu, 2003) and for a variety of coastal hazard (e.g.,
tsunami, storm surges, etc.) and disaster assess-
ment purposes (for example, see Chen et al. 2005).
DEMs are also commonly used to derive new
data that is required for specific types of disaster
management related analysis or visualization. The
types of new datasets that can be generated include,
but are not limited to: slope, aspect, contour lines,
flow direction, flow accumulation, watersheds,
solar insolation, viewsheds, hillshade visualiza-
tions, and many others. The most important factors
that determine the suitability of a DEM for any
particular disaster management related application
are the spatial resolution and vertical accuracy.
The potential of remotely sensed data to as-
sist disaster management is very clear, however,
there are some limitations and potential obstacles,
including image resolution, repeat frequency (tem-
poral resolution) and the suitability of particular
sensors. Perhaps the most significant obstacle
facing the use of satellite imagery for a number
of disaster management requirements is low pixel
resolution. In large urban centers, satellite image
resolution can be less than 1 meter, but in more
remote areas and many parts of the developing
world, only 15-30 meter Landsat or ASTER im-
agery are publicly available (Nourbakhsh et al.,
2006). Low image resolution can limit the range
of potential uses of RS data, including using it
as a source to create digital datasets of basic, yet
fundamental information for disaster management.
For example, Rüdenauer and Schmitz (2005) used
supervised classification techniques and found that
Landsat 7 ETM+ data (30m resolution) could be
used to identify and separate out settlements, but
that the detection of linear features such as roads,
railways and waterways was unsatisfying. Inte-
grating the panchromatic band (15 m resolution)
into the analysis failed to significantly improve
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
results. Thus, while Landsat imagery can be
used to identify settlements it cannot be used as
a source to accurately derive all spatial features
on the ground.
Repeat frequency is also an important con-
sideration, as time becomes a critical factor in
post-disaster use of RS imagery, some satellites
may not be available within an appropriate/suit-
able time frame (San Miguel-Ayanz et al., 2000).
For example, repeat frequency can vary from 12
hours (NOAA) to 35 days (ERS) (Simonovic,
2002). In addition, Cutter (2003) points out that
the “pre- and post-processing time for remote
sensing images may negate their use in immedi-
ate response activities,” (p. 442). Even if they
are able to pass over a disaster affected region,
cloud cover and the time of day (day vs. night)
can constrain the use of optical sensors for image
acquisition (Coppock, 1995). Therefore, the type
of sensor onboard a particular satellite as well as the
atmospheric conditions will partly determine the
suitability of satellites for a given disaster related
requirement. Even partial cloud cover can be a
significant obstacle that limits the use of optical
imagery. In cloudy conditions satellites employing
radar systems are required, but they too have their
limitations (Simonovic, 2002) and are at times not
suitable for acquiring the type of information that
may be required. Shadows can also be a limiting
factor in both optical and radar imagery. Magsud
et al. (2005) found that shadows prevented the
visual identification of damaged buildings using
high resolution QuickBird imagery following the
Indian Ocean tsunami of 2004. The same authors
also note that intact roof tops can be deceiving, as
initially certain buildings appeared to be damage
free, but upon ground truthing the building was
essentially hollowed out, with the roof top and
support beams being the only remnant. Accord-
ingly, caution must be used when drawing initial
conclusions based on visual imagery analysis,
and ground truthing should be considered before
any significant actions are taken.
Recognizing the importance of high resolu-
tion satellite imagery in disaster management,
many major space agencies and satellite operators
signed the International Charter on Space and
Major Disasters by the year 2001. This multilat-
eral agreement stems from the fact that no single
operator or satellite can match the data-related
challenges of natural disaster management. “The
International Charter aims at providing a unified
system of space data acquisition and delivery to
those affected by natural or man-made disasters,”
(Space Agencies 2005, p. 1). The charter has been
successfully evoked on numerous occasions by
countries all around the world – from Ecuador to
India to Russia – providing indispensable high
resolution imagery for no fee. For example, in
the aftermath of the Indian Ocean tsunami several
agencies and private companies provided RS im-
agery for response and relief work (ESRI, 2006).
DigitalGlobe provided 60-centimeter QuickBird
images, which were the highest resolution com-
mercial satellite images available. Although
developing countries cannot send satellites into
orbit, opportunities do exist for acquiring high
resolution remotely sensed imagery for disaster
management purposes. However, Magsud et
al. (2005) “point out that once satellite data are
received by the coordinating office,” (p. 8) it is
then up to the authorized user to effectively use
the data – there are few mechanisms in place to
help countries in need of support to use the data.
In developing countries this can pose a problem
because potential data users are not fully aware
of the capabilities of GIS and remote sensing, and
they require additional assistance in order to put
the data to good use. Moreover, the International
Charter does not provide free imagery for pre-
disaster GIT operations.
The Haitian earthquake of January 2010 was an
example of the International Charter on Space and
Major Disasters being invoked and imagery being
acquired from satellites including Japan’s ALOS,
CNES’s Spot-5, the US WorldView and QuickBird
platforms, Canada’s RADARSAT-2 and the ESA’s
191
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
ERS-2 and Envisat sensors. Moreover, the delivery
of these datasets was facilitated by Google (http://
www.google.com/relief/haitiearthquake/geoeye.
html) and GeoEye Inc. (www.geoeye.com) to le-
verage Google’s web-presence and bandwidth to
provide 100s of GBs of high resolution GeoEye 1
and IKONOS images. In addition, many Google
pages like the Google Earth Library (http://www.
gelib.com/haiti-earthquake.htm) offered critical
infrastructure data for direct viewing in Google
Earth formatted KML. The wide availability of
post-disaster KML datasets allowed for rapid
assessments by non-GIS experts, thereby, and it
is too soon to tell, increasing the effectiveness of
relief efforts. Overall, the geospatial response to
the Haitian disaster was quite rapid when compared
to the 2004 Indian Ocean tsunami, largely due to
the need for rapid geospatial assessments during
disasters. By way of illustration, in “December
2006, the United Nations General Assembly es-
tablished the United Nations Platform for Space-
based Information for Disaster Management and
Emergency Response - UN-SPIDER,” (http://
www.un-spider.org/haiti). A visit to UN-SPIDER
provides a plethora of metadata and links to di-
saster mapping for Haiti. Numerous other private
organizations like ESRI Inc. (http://www.esri.
com/haiti/) created a central earthquake response
in support of Haiti, providing maps and mapping
support for the relief phase. By contrast, the first
web-based mapping and delivery platform for
the 2004 Indian Ocean tsunami was the Tsunami
Disaster Mapping Portal launched on January
10th, 2005, and was led by DM Solutions Group,
collaborators including researchers and scientists
from the Laboratory for Applied Geomatics and
GIS Science (LAGGISS), University of Ottawa,
Asian Institute of Technology (AIT), Pathum
Thani, Thailand, Chulalongkorn University,
Bangkok, Thailand, and Media Center, Osaka
City University, Osaka, Japan. By comparison,
there were well over 45 map-based websites or
data delivery sites 14 days after the devastation
in Haiti. Our point here is that there has been a
192
significant amount of incremental knowledge
gained about the critical nature of geospatial data
needs in a time of disaster and this recognition has
led to a greater degree of cooperation in terms of
data availability, timeliness and mode of delivery
for different practitioner communities.
In sum, RS technology acts as a tool to gather
spatial information that can support many disaster
management functions, especially when combined
with other spatial data (points, lines and polygons)
in a GIS. Satellite imagery is critical for natural
disaster response in remote and inaccessible
regions, or regions where primary (or basemap)
geospatial data are non-existent or difficult to
collect. In such regions, satellite imagery often
offers the best means of obtaining the necessary
information (Kerle & Oppenheimer, 2002). This
is particularly true in many developing countries
which tend to lack digital spatial data infrastruc-
tures (Brodnig & Mayer-Schönberger, 2000;
ESRI, 2006).
Internet GIS
Internet GIS (IGIS), the integration of GIS and
the Internet, has quickly evolved over the last
decade and has achieved significant recognition
within the disaster and emergency management
community for two main reasons. First, “GIS
provides a way to centralize and visually display
critical information,” (ESRI 2006, p. 4) relevant
to a disaster, since “most of the data requirements
are of a spatial nature and can be located on map,”
(Johnson 2000, p. 3). Second, the Internet provides
an ideal medium for multiple users with a range
of backgrounds and skill sets to access spatial
information and mapping capabilities. Radke et
al. (2000) emphasize that “data acquisition and
integration may be the single largest contribu-
tion area needed for emergency and disaster
response,”(p. 15) – IGIS can help address both
these issues. During the response phase, access
to pertinent spatial data/information is among
the most essential requirements (Amdahl, 2001;
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
Jayaraman et al., 1997). Disaster managers, lo-
cal authorities, aid workers and the public need
up-to-date data/information to enable quick and
effective decision making in the immediate af-
termath of a disaster. Organizations such as the
Global Disaster Information Network (GDIN)
“provide evidence of the importance and value of
disaster-related information, as well as the need to
be able to obtain and share it effectively,”(http://
www.gdin.org).
The integration of GIS and the Internet first
began in the early 1990’s (Plewe, 1997), and has
grown rapidly since that time. This growth is
partly attributable to advancements in computer
and information technology and the building of
spatial data infrastructure (SDI) worldwide (Yang
et al., 2005). It is therefore pertinent in this review
to briefly define and describe the capabilities
of Internet GIS, discuss the implications of this
relatively new technology on the field of disaster
management, and examine its potential from a
developing countries perspective.
In this emerging field of Internet GIS, there is
no general agreement on the terms used to describe
GIS-based programs on the Internet. They are com-
monly referred to as Internet GIS, web-based GIS,
distributed GIS and On-line GIS, among others.
Peng and Tsou (2003) help to better differentiate
between these terms and classify the various types
of applications. The term ‘Internet GIS’ (IGIS)
will be used henceforth herein and is defined “as
network-based geographic information services
that can utilize wired or wireless Internet protocols
to access geographic information, spatial analysis
tools and GIS Web services,” (Peng & Tsou 2003,
p. 3). It is worth noting here that although the term
GIS is included with “Internet”, IGIS focus mainly
on displaying geographic information (in map
form) as well as data dissemination tasks but tend
to lack comprehensive GIS capabilities common
to most desktop software. Kraak (2004) points out
that most IGIS applications currently in use are
limited to (interactive) mapping (with zoom, pan,
measure distances, identify spatial features, etc.),
although some offer basic GIS functions such as
address matching, proximity searches, and route
planning like Google Maps for example. However,
more recent advancements are allowing for the
development of Internet distributed GIS services
with the capabilities to “interact with multiple
and heterogeneous systems and servers that sup-
port more advanced GIS functions,” (Dragićević
2004, p. 11). Such capabilities are reflected in the
rapidly developing standards for web-based GIS
services (http://www.ogc.org).
The characteristics and capabilities of IGIS
offer considerable potential in the field of disaster
management, particularly during the response
phase when access to spatial information is a
key requirement (Radke et al., 2000). Disaster
impacts can significantly alter the landscape
(natural and built environments), and often vary
across organizational, sociological, political, and
geographic boundaries. This is especially true fol-
lowing large-scale disasters. Information about the
disaster zone needs to be disseminated to many
stake-holders involved in the response, at local,
regional, national and international levels, both
public and private (GDIN, 1997). Much of this
information is spatial, such as the location of the
most devastated towns or villages, the location of
supply drop-off points and the extent of in-tact
transportation networks. IGIS are well suited to
fulfill these information requirements since (1)
they allow for the integration of various spatial
datasets and (2) they can be accessed from any
location with an Internet connection.
The often ad-hock nature of disaster response
involves people and organizations (local to inter-
national) that require spatial information relevant
to their own logistical operations. Comfort (2000)
reports that during disaster response there are
often dynamic and spontaneous actions taken by
responding organizations and individual groups
of people. These types of efforts are encouraged,
but could be more effective and better coordinated
with an improved spatial awareness of the disaster
zone, achieved through the use of appropriate
193
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries

spatial information and maps accessed using
IGIS-based websites. Becking (2004) states that
obtaining geographic awareness is one of the
most important steps in effectively understand-
ing a disastrous situation, and being able to make
appropriate decisions. Traditionally, geographic
information has been distributed using paper maps,
which are costly to produce and difficult to update
and distribute to all parties involved.
“Experience has shown that a top-down ap-
proach to data sharing is not entirely effective,”
(Radke et al. 2000, p. 23), “as disaster respond-
ers must gain access to a number of department
managers and organizations, their unique maps
and data,”(Johnson 2000, p.3). Knowing which
data is where and how to effectively access it can
be a significant barrier. Furthermore, this type of
approach often results in duplication of efforts.
For example, following the Indian Ocean tsunami,
many agencies created damage maps at almost the
same time (ESRI, 2006). An alternative option
is to utilize current IGIS architectures to create
a disaster information system with wide acces-
sibility, which allows multiple users to access
relevant spatial data/maps provided by a variety
of organizations through distributed servers/
sources. An IGIS-based mapping system that is
intuitive, user-friendly and easy to access could
help provide geographic awareness to disaster
responders, many of whom have limited or no
GIS experience but could certainly benefit from
access to spatial information and maps. Andre and
Smith (2003) note that disaster responders often
require simple cartographic products (for example,
the location of impassable roads identified on a
road network map) as opposed to products derived
from advanced spatial analysis.
Experience to date indicates that IGIS has
played a key role in the collection and dissemi-
nation of spatial information during the disaster
response and recovery of recent natural disasters,
including the Indian Ocean tsunami (2004) and
Hurricane Katrina (2005). However, the success/
failure of IGIS in that regard must be interpreted
within the context of the very young IGIS field.
With that caveat in mind, following these natural
disasters IGIS sites were established that integrated
a large number of spatial datasets, including
coastlines, satellite imagery, damage maps, trans-
portation networks, population centers and other

Figure 3. Examples of two IGIS-based natural disaster map viewers: a) South East Asia and Indian
Ocean Tsunami Response Map Viewer (produced by DM Solutions Group, http://www.dmsolutions.ca/
showcase/), and b) Hurricane Katrina Disaster Viewer (produced by ESRI Inc., Redlands California, USA)

194
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
data intended to provide a visual overview of the
regions impacted by the disasters. Figure 3 shows
images of the main interface of two such sites.
The user friendly map based information
systems created using IGIS can increase access
to relevant spatial information required during
disaster response. However, their design, develop-
ment and implementation require substantial GIS/
technical knowledge and significant human and
financial resources. As demonstrated earlier, in
many developing countries each of these require-
ments represents a challenge, and could prevent
them from creating IGIS for disaster response. In
fact, the two known IGIS sites created to assist
with the Indian Ocean tsunami response were not
created by any developing country that was af-
fected by the disaster; one was created by a Ca-
nadian software development company (DM
Solutions Group – http://www.dmsolutions.ca)
and the other by the Pacific Disaster Center, which
is based out of Hawaii. Both have significant GIT
resources and capabilities, and were able to es-
tablish these IGIS sites in a relatively quick pe-
riod of time. This is an important point, as it is
unlikely that GIT practitioners in developing
countries have the expertise required to effec-
tively develop an IGIS site in a reasonable time
frame. In addition, Internet connections in many
parts of the developing world can be slow and
patchy (Nourbakhsh et al., 2006), which reduces
the potential for accessing information using IGIS
sites.
In general, the ability to develop an IGIS for
disaster response requires excellent GIT skills and
a large amount of spatial data, which may exist at
differing scales and in varying formats. Currently,
the GIT requirements and skills associated with
creating IGIS are beyond those of disaster manag-
ers in developing countries, and we suggest that
trying to implement IGIS may not be the best use
of their limited resources.
GEOSPATIAL INFORMATION
TECHNOLOGY SOFTWARE
We have reviewed many uses and barriers of geo-
spatial information technology (GIT) to natural
disaster management. However, there has been
no discussion regarding specific aspects of GIT
software, including the difference between com-
mercial and free and open source (FOS) software
(FOSS), and the implications of these differences
for adoption within developing countries. Turning
our attention to that issue, we now discuss the
main differences between commercial/proprietary
software and FOSS, and explore the significance
of these differences from a developing country
perspective. We also briefly discuss the FOS
GIT software domain. Although there are some
well documented drawbacks to using FOSS, and
challenges to overcome, the potential benefits that
can be gained far outweigh any downside, and
thus FOSS should be considered a viable model
(Fitzgerald, 2004; Hoe, 2006).
Free and Open Source
Software (FOSS)
While some consider “free” and “open source”
software to be something close but not identical,
they are similar enough that for the purposes of
this paper they are considered the same. We can
lump them together because for the general soft-
ware user the difference is negligible, although
philosophical and legal differences do exist
(Rajani et al., 2003; Steiniger & Bocher, 2009).
Although there are a number of criteria that a soft-
ware package must meet to be considered FOSS,
three essential features capture the essence of the
semiofficial “Open Source Definition” (http://
www.opensource.org/docs/osd):
• “The source code must be distributed with
the software or otherwise made available
for no more than the cost of distribution.
195
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
• Anyone may redistribute the software for
free, without royalties or licensing fees to
the author.
• Anyone can modify the software or derive
other software from it, and then redistrib-
ute the modified software under the same
terms,” (Weber 2004, p. 5).
Licensing agreements such as the General
Public License (GPL) define the rights users have
over the software product (http://www.gnu.org/
licenses/gpl.html). Wu and Lin (2002) as well as
Cook and Horobin (2006) provide further infor-
mation about the various FOSS licensing models.
What makes FOSS different than proprietary
or commercial software is that the source code
is “free.” “In this context free means freedom,
not necessarily zero price,”(Weber, 2004, p. 5)
although almost all are available at no cost. Free
software implies that the user has the freedom to
run, view, copy, modify, and distribute a piece
of software, irrespective of financial limitations.
Thus, users can improve the source code, by either
Table 2. Key differences between proprietary and FOS software (modified after Steiniger & Bocher 2009)
Proprietary Software Free and Open Source Software
Advantages warranty of developing company on product
(holds for every computer)
components should work together
well documented software
regular release times for new versions.
regular service packs
customization at the API level
Disadvantages software price and maintenance fees
training costs
maintenance fees tied to specific licensed com-
panies, software options and time period
customised development can be difficult due to
available resources of vendors
support only as long as software company
exists
Some limitations on out-of-box functional-
ity where vendor partner’s are necessary for
upgraded functionality
reliance - retraining costs when software
versions change (e.g., ESRI’s ArcView3.2 to
ArcGIS 8.x) or data models change at vendor’s
whim or development cycle
196
enhancing its existing functionality or by adding
new functions. In contrast, proprietary software
indicates ownership, and has its source code closed
– it cannot be viewed, modified or redistributed, as
is stated in the EULA (End User License Agree-
ment) that you must agree to prior to installation.
Free, as in price, is perhaps the most significant
characteristic of FOSS that differentiates it from
commercial software, which is inherently sold
for profit. Another characteristic of FOSS is that
it can be accessed (downloaded) from anywhere
with an Internet connection; this is typically not
true of commercial/proprietary software. Table
2 outlines some of the key differences between
proprietary and FOS software products, in terms of
advantages and disadvantages. With certain types
of software, including GIT software, there are also
considerable differences in terms of functionality
and ease of use.
The FOSS movement is continually gaining
momentum, but has already garnered significant
attention in some areas. For example, “Apache
dominates the web server market, as over 65% of
no licence fees
unrestricted use (e.g., no limits for the number of installa-
tions)
no update enforcement
support of open standards
support usually available from several providers
customisation at source code level
platform independent
installation know-how necessary in many cases
training costs
interoperability issues between FOSS
quality (but self-correcting if actively used)
no responsible authority
support can lack for some packages
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
all active web sites use it,” (Weber 2004, p. 6).
Ramsey (2007) points out that its success can be
linked to a powerful user community that is com-
mitted to maintaining the Apache platform. Cor-
porate mainstays such as IBM and HP, government
agencies and academic contributors are all part
of the Apache community. Other FOSS projects
that have become very popular include office
suites such as OpenOffice, database systems like
MySQL, the Mozilla web browser, and the Linux
operating system. As of February 2009, there were
over 230,000 registered projects and over
2,000,000 registered users at Sourceforge.net, the
world’s largest open source software development
website. In fact, most users can find an application
that will exactly meet their specific needs (Wu &
Lin, 2001). According to Wheeler (2007), who
provides quantitative data, FOSS “reliability,
performance, scalability, security and total cost
of ownership are at least as good as or better than
its proprietary competition, and under certain
circumstances are superior,”(Anderson & More-
no-Sanchez 2003, p. 448). This evidence suggests
that FOSS can, and already is, rivaling certain
commercial software domains.
Although a strong case can be formed for the
FOSS model, it also has some often cited disad-
vantages compared to commercial software. For
starters, software support and technical assistance
is a major issue facing the adoption of FOSS.
However, FOSS proponents are quick to point
out that support may be found within the FOSS
community, in the form of user groups and “ar-
chives of past queries and answers available on the
Internet,” (Ramli et al. 2005, p. 522). Installation
and user documentation is also common among
the more mature FOSS products (Steiniger &
Bocher, 2009). Furthermore, “organizations that
deploy FOSS freely offer advice to one another,
sharing insights and lessons learned,” (Fitzgerald
2004, p. 93). As a result, solutions for many typi-
cal problems can often be found at no cost. When
they cannot, there is a growing FOSS support and
custom development industry that can be utilized
in such circumstances.
Another concern about FOSS relates to its
long-term survival. Fitzgerald (2004) reports
that “studies of Freshmeat.net and Sourceforge.
net (two popular FOSS development websites)
revealed that most projects have only one or two
developers, and that follow up studies reported no
change in version number or size of code base for
many listed projects several months later,” (p. 95).
This sort of vague analysis of the FOSS domain
can be deceiving, since there are so many FOSS
projects it is not surprising that most have only
one or two developers, and that new versions are
not coming out regularly, especially if the par-
ticular package is sufficient for its purpose. That
is the nature of the FOSS model, as it encourages
individuals or small teams to develop and share
software. It is up to potential users of FOSS to
examine individual products, consider potential
advantages and disadvantages, and choose what
best fits their needs. Câmara and Onsrud (2004)
examined FOSS GIT and identified many differ-
ences that exist – in terms of support, maturity and
functionality – between products led by a single
individual, products produced by small teams
and corporate led products. They conclude that
corporate led projects tend to be of better quality,
at least from these three standpoints.
Free and Open Source Geospatial
Information Technology Software
Turning our attention now to the specific do-
main of FOS GIT software, there is reason for
optimism. As opposed to questionable long-term
survival, Steiniger and Bocher (2009) emphasize
the increasing interest and development of FOS
GIT products. They point out in their overview
of FOS GIS (e.g., gvSIG, Quantum GIS, SAGA,
uDig, GRASS, etc.) that:
a. Four out of ten desktop projects examined
receive governmental funding support;
197
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
b. There is an increase in the download rate of
FOS GIS software; and
c. There is an increasing number of use cases
of FOS GIS.
Furthermore, Ramsey (2007) points out that
“existing products are now entering a phase of
rapid refinement and enhancement… (FOS) soft-
ware can provide a feature-complete alternative
to proprietary software in most system designs.”
Some commercial software manufacturers are
even starting to back FOS GIT initiatives, which
is encouraging especially from a user support and
longevity perspective. In late 2005 the software
industry giant Autodesk, in association with the
MapServer community and DM Solutions Group,
announced that it would support and promote open
source web mapping (a form of IGIS) through the
creation of the MapServer Foundation (Schutz-
berg, 2005). “The Foundation is expected to pro-
vide a stable infrastructure for the now extended
MapServer family’s code base and its growing
community,” (Schutzberg, 2005, p. 1).
The FOS GIT software community is steadily
growing, and since 2006 has been spearheaded by
the Open Source Geospatial Foundation (OSGeo)
(http://www.osgeo.org). “Their mission is to sup-
port and promote the collaborative development of
open geospatial technologies and data,” (OSGeo
2008, p. 1). The OSGeo hosts an increasing number
of software projects, publishes the OSGeo journal,
founded an education and curriculum committee,
and presents the annual Free and Open Source
Software for Geospatial (FOSS4G) international
conference (Steiniger & Bocher, 2009).
The comprehensive list of links to FOS GIT
related software projects/products available
at opensourcegis.org and freegis.org provides
evidence of an active development community.
Some notable GIT programs, which offer a range
of functionality from simple data/map viewing to
more advanced spatial analysis and Internet GIS
capabilities, include: Quantum GIS, DIVA-GIS,
OpenEV, uDig, gvSIG, GRASS (Geographic
198
Resources Analysis Support System), MapServer
and OSSIM (Open Source Software Image Map).
Other programs focus on more specific tasks, such
as data management, format processing, geosta-
tistical analysis and data visualization. Currently,
the freegis.org website contains information and
links to over 300 FOS GIT projects. Ramsey
(2007) provides an excellent review of some of
the more mature projects within the FOS GIT
software domain, categorized by development/
implementation languages, such as C, Java and.
Net. For a more in-depth overview of current
FOS desktop GIS, in terms of organizations,
software groups and functionality, see Steiniger
and Bocher (2009).
FOSS and Developing Countries
From the perspective of developing countries
the FOSS model is a particularly good fit, for
reasons that include: cost, freedom, accessibility,
customizability, compatibility, capacity develop-
ment and reducing the overall so-called ‘digital
divide’ (Hoe, 2006). The lower cost of FOSS is
definitely the most significant factor that makes
it attractive to developing countries (Rajani et al.,
2003), with the absence of licensing fees seen
as a major benefit (Hoe, 2006). The software
that comprises the commercial GIT domain is
particularly expensive. Table 3 provides a few
examples of some popular commercial GIS and
remote sensing software applications and their
general price ranges.
Even the most basic and required proprietary
software necessary for many computing needs,
such as Windows XP and MS Office, can be too
expensive to purchase in a developing country
(Roets et al., 2007). With that in mind, it is not
surprising that the highest software piracy rates
occur in developing countries such as Vietnam,
China and Indonesia, with piracy rates of 97, 94
and 89 percent, respectively (Rajani et al., 2003).
In such countries is not unusual to have new
computers come with a range of pre-installed
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
Table 3. Approximate costs of selected commercial GIS/RS software (the wide price ranges reflect pos-
sible variations in license type, software options [software extensions], purchasing agreements, etc.)
Application Price range (USD)
High end GIS/RS $10,000 to $50,000
Low end GIS/RS $1,000 to $3,000
pirated proprietary software (Weerawarana &
Weeratunge, 2004). This evidence demonstrates
that users in developing countries most often
don’t, and perhaps more importantly, cannot afford
to pay for computer software. Although piracy is
very common, it actually devalues the economic
benefits of FOSS by falsely reducing the price of
proprietary software (Weerawarana & Weera-
tunge, 2004). The zero cost of FOSS should be
considered a major advantage over proprietary
software in the case of developing countries, and
could also help reduce their reliance, and inad-
vertent support and encouragement, of the illegal
software market.
The freedom to access and study the source
code is another fundamental advantage of using
FOSS. As a result, the choice to use FOSS is
not only a software choice, but also a means of
acquiring knowledge about the software itself. In
developing countries this is important for “capacity
building of the local population, and can help them
better understand and deploy new technologies
successfully,” (Hoe 2006, p. 31). Mohamed and
Plante (2002) emphasize that “local workforce
development and capacity building are critical
for system maintenance and operation over the
long run,” (p. 2286). In addition, using FOSS
can contribute to the development of the local
information and communications technologies
(ICT) industry (of which GIT is a part of), as sup-
port, development and maintenance contracts can
be provided by local businesses that offer FOSS
Software
ArcGIS (ArcINFO)
Genasys
ENVI
ERDAS Imagine
ER-Mapper
MapInfo
ArcGIS (ArcView)
ArcView extensions
ERDAS Imagine Essential
services (Cook & Horobin, 2006). This will, in
turn, provide more jobs in the FOSS industry,
attract trained professionals, and help stimulate
the local economy.
The freedom to customize the source code is
also important from a developing country perspec-
tive. Krakowski (2006) points out that proprietary
software developed within the Western world is
designed to fit well with Western culture. As a
result, it may not be well-suited for use in other
regions of the world, where local customs and
practices may be quite different. With FOSS,
applications can be tailored to meet the specific
needs of the users, and take into account cultural
variations and social practices that deviate from
industrialized societies. Similarly, “the majority of
people in developing countries do not understand
English, yet proprietary software is often only
available in English,” (Hoe 2006, (p. 6); Nonogaki
et al., 2004). Clearly, interface and documentation
language represents a major obstacle to users who
do not understand English; however, FOSS can be
adapted to meet local language needs and in doing
so helps to tackle the digital divide. For example,
Ubuntu is an African adjusted Linux distribu-
tion that attempts to resolve language problems
(Krakowski, 2006). GNOME 2.22 – the default
desktop environment of Ubuntu – offers support
for 46 languages. Similarly, multi-language capa-
bilities have been incorporated in recent versions
of GRASS and MapServer (Nonogaki et al., 2004;
Raghavan et al., 2006) and gvSIG is available
199
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
in more than ten languages. Additional efforts
need to focus on translation of online help files,
manuals and so on.
Another advantage of FOSS is that they usually
do not require the newest or best computer hard-
ware in order to function efficiently and correctly.
In contrast, and especially in the GIT software
domain, proprietary companies are continually
releasing new versions that demand the newest and
most advanced hardware. In developing countries
most users do not have access to powerful, high-
end computers with the most current hardware
available (Caldeweyher et al., 2006). Therefore,
even if they could purchase sophisticated GIT
software they would also require computers with
sufficient RAM and CPU power to take advantage
of the advanced capabilities that are offered in
most commercial GIT packages.
A couple of final remarks about the potential
of FOSS for developing countries relate to acces-
sibility and unlimited installations. First, FOSS
can be downloaded (accessed) from anywhere
with an Internet connection; and where the In-
ternet is unavailable FOSS can be delivered on
a CD or provided via a data storage device such
as a USB or external hard drive. This makes
computer software accessible to potential users
who would have absolutely no ability to purchase
or otherwise obtain such software. Second, the
freedom to install FOSS on as many computers
as is desired is another fundamental benefit not
offered by proprietary software. If a single use
license is purchased from a proprietary company
only one computer can legally run the software,
and multi-use licenses are expensive and con-
sidered unaffordable for most software users in
developing countries, even governments (Roets
et al., 2007). This limits the accessibility of the
software, preventing it from being installed and
used on more than one computer. In an office en-
vironment, for example, this is likely to represent
a significant problem.
In sum, the literature that addresses issues re-
lated to FOSS and its adoption in developing coun-
200
tries, in terms of advantages and disadvantages, is
diverse and abundant. The preceding discussion
is meant to provide only a brief overview of some
well acknowledged and commonly discussed
advantages that FOSS provide and elaborate on
those specific to GIT. We acknowledge that there
are many more advantages to FOSS in general for
the developing world that are beyond the scope of
this paper to discuss and so recommend reviews
by Rajani et al. (2003) and Weerawarana and
Weeratunge (2004), or Cook and Horobin (2006)
who examine FOSS from a public administration
(government) perspective.
DISCUSSION
Considering that disasters are spatial phenomenon,
it would be reasonable to assume that GIT can
be effectively employed throughout all phases
to assist the disaster management process. While
this is true in the case of developed countries,
the same cannot be said for developing coun-
tries. The use of GIT for disaster management in
developing countries is limited by a number of
well documented barriers, including: a lack of
financial resources (Renyi & Nan, 2002), a lack
of local GIT expertise/knowledge (Mohamed &
Plante, 2002), institutional/political instability
(Pande, 2006), and a shortfall of spatial data
(ESRI, 2006; Murgia et al., 2002). The degree to
which each of these GIT implementation barriers
limits the use of GIT varies between developing
countries, and within each country depending on
the administrative level.
We have examined the literature and discussed
various aspects of research on natural hazards and
disaster management. Developing countries are
highly vulnerable to natural disasters, a vulner-
ability that could be addressed partially through
GIT. However, there are numerous barriers to
GIT implementation in the developing world.
We reviewed the application of GIT in natural
disaster management and focused on describing
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
the significant potential for GIS, RS and IGIS
to assist the disaster management phases. We
discussed the concept of FOSS, and argued that
they have reached a stage a maturity that makes
FOSS a viable alternative model over commercial
software, especially in developing countries. The
FOS GIT software domain is steadily growing
and there are already many mature projects that
currently offer a range of functionality (Raghavan
et al., 2006; Steiniger & Bocher, 2009).
Given the continued growth of FOS GIT, we
propose that its current availability offers increas-
ing potential to the local level disaster management
practitioner community in developing countries.
In this section we discuss why and how FOS
GIT can, and should, be implemented at the local
level in developing countries in order to improve
disaster management capacity, and propose future
research needs in this area.
The cosmopolitan existence of FOSS is a rela-
tively new phenomenon, as are the applications
that are emerging. Currently, within the literature
that addresses FOS GIT, there appears to be very
little that examines its use in the specific research
domain of natural hazards and disaster manage-
ment, and this finding may be related to the fact
that the entire FOS GIT domain is quite new
compared with proprietary GIT software. Table
4, containing search results from the SCOPUS
database, provides quantitative evidence of the
difficulty in finding literature that applies FOS
GIT in the field of natural disaster management
in developing countries. Although the concepts of
GIT have existed for decades, and are now reflected
in a large commercial GIT software market, those
concepts have only recently been transformed into
capabilities within FOS GIT, with the exception
being GRASS. However, GRASS was originally
developed by the U.S. Army Corp of Engineers in
the early 1980s and “is currently used in academic
and commercial settings around the world, and by
numerous government departments (e.g., USGS
and NASA) (http://grass.itc.it/intro/general.php).
As a result of its long use and development (which
is now multi-national) history it has evolved into
a sophisticated and powerful raster/vector GIS,
“used for geospatial data management and analy-
sis, image processing, graphics/map production,
spatial modeling, visualization and much more,”
(http://grass.itc.it/intro/general.php). Although
GRASS is very powerful, its main drawback is
its unfriendly user environment, as it functions
primarily through command line operations which
can be a challenge for non-expert users (Ramli et
al., 2005; Steiniger & Bocher, 2009).
Furthermore, a key word search of ‘hazard’
and ‘disaster’, independently, of all presentation
titles from the FOSS4G 2006 and 2007 http://
www.foss4g2006(2007).org/) conference yielded
only three that contained ‘hazard’ and one that
contained ‘disaster.’ This is surprising given the
increasing frequency of natural disasters and the
critical role that GIT plays within the various
management phases. However, although very few
presentation titles include the word ‘hazard’ and/
or ‘disaster’, it is likely that many of the topics
discussed do contain recent research or software
development relevant to the study and/or manage-
ment of natural hazards and disasters. For ex-
ample, addressing how FOS GIS software can
contribute to SDI development is relevant to di-
saster management, since the required spatial data
are often lacking, although an explicit connection
may not necessarily be made. Coppock (1995),
in a study of GIS and natural hazards research,
makes an analogous point, explaining that GIS
and natural hazards are multidisciplinary fields,
and that much of what has been written appears
in the literature of those disciplines.
While a review of the application of FOS GIT
in the area of natural hazards and disaster manage-
ment is desirable, the lack of available literature
(Table 4) makes this task nearly impossible.
However, when you remove the FOSS component,
there is an abundance of literature that utilizes
proprietary GIT software, as was demonstrated
in the previous section on GIT and disaster man-
agement, and in Table 4. More common from the
201
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries

Table 4. Literature themes search results from Scopus (http://www.info.scopus.com)

LITERATURE ADVANCED SEARCH PHRASES USED # OF RE-

THEMES (in: title-abstract-keyword) SULTS
Natural Hazards and
Disasters
((“disaster*” OR “natural disaster*” OR “natural hazard*” 254,109
OR “environment* hazard*” OR “flood*” OR “earth- 459, 881
quake*” OR “storm*” OR “hurricane*” OR “tsunami*” OR
“landslide*”))
+ “hazard*” (which includes hazards that are not of natural
origin)
Natural Hazards and Disasters ((“disaster*” OR “natural disaster*” OR “natural hazard*” 54,283
+ OR “environment* hazard*” OR “flood*” OR “earth-
Disaster Management quake*” OR “storm*” OR “hurricane*” OR “tsunami*”
OR “landslide*”) AND (“disaster* manage*” OR “hazard*
manage*” OR “mitigation” OR “prepare*” OR “respon*”
OR “relief”))
Natural Hazards and Disasters ((“disaster*” OR “natural disaster*” OR “natural hazard*” 4,994
+ OR “environment* hazard*” OR “flood*” OR “earth-
Disaster Management quake*” OR “storm*” OR “hurricane*” OR “tsunami*” OR
+ “landslide*”) AND (“disaster* manage*” OR “hazard* man-
Geospatial Information Technology (GIT) age*” OR “mitigation” OR “prepare*” OR “respon*” OR
“relief”) AND (“geo* info* sys*” OR “GIS” OR “geoinfo*”
OR “3S” OR “geospatial” OR “geo* info* tech*” OR “re-
mote sens*” OR “RS” OR “earth observation*” OR “EO”
OR “satellite” OR “spatial*”))
Natural Hazards and Disasters ((“disaster*” OR “natural disaster*” OR “natural hazard*” 769
+ OR “environment* hazard*” OR “flood*” OR “earth-
Disaster Management quake*” OR “storm*” OR “hurricane*” OR “tsunami*” OR
+ “landslide*”) AND (“disaster* manage*” OR “hazard* man-
Geospatial Information Technology (GIT) age*” OR “mitigation” OR “prepare*” OR “respon*” OR
+ “relief”) AND (“geo* info* sys*” OR “GIS” OR “geoinfo*”
Developing Countries OR “3S” OR “geospatial” OR “geo* info* tech*” OR “re-
mote sens*” OR “RS” OR “earth observation*” OR “EO”
OR “satellite” OR “spatial*”) AND (“develop* countr*” OR
“develop* region*” OR “develop* nation*” OR “transition*
countr*” OR “third world” OR “less develop*” OR “East*
Europe” OR “Africa” OR “Asia” OR “South America” OR
“Central America” OR “China” OR “India” OR “Pakistan”
OR “Indonesia” OR “Malaysia” OR “Thailand” OR “Ban-
gladesh”))
continued on following page

FOSS perspective is literature that emphasizes the
great potential it offers in areas such as disaster
management, without going into any more detail
about specific applications or examples of suc-
cessful usage. Thus, further research is required
regarding the current capabilities of FOS GIT, and
more specifically, how they can be used by the
disaster management community in developing
countries. Again, we emphasize that sufficient
time has not elapsed to allow for the creation of
a literature base in this research area, yet strongly
believe that more and more examples of successful
FOSS implementation will undoubtedly emerge
in the coming years. This belief is based on the
increasing interest in FOSS and overall growth of
the user and development communities, through
initiatives such the OSGeo and websites like
opensourcegis.org and freegis.org.
Among the broad set of Millennium Develop-
ment Goals that the United Nations established
in 2000, one stands out: “Make available the
benefits of new technologies, especially informa-

202
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries

Table 4. Continued

Natural Hazards and Disasters ((“disaster*” OR “natural disaster*” OR “natural hazard*” 15

+ OR “environment* hazard*” OR “flood*” OR “earth-
Disaster Management quake*” OR “storm*” OR “hurricane*” OR “tsunami*” OR
+ “landslide*”) AND (“disaster* manage*” OR “hazard* man-
Geospatial Information Technology (GIT) age*” OR “mitigation” OR “prepare*” OR “respon*” OR
+ “relief”) AND (“geo* info* sys*” OR “GIS” OR “geoinfo*”
Developing Countries OR “3S” OR “geospatial” OR “geo* info* tech*” OR “re-
+ mote sens*” OR “RS” OR “earth observation*” OR “EO”
Free and Open Source Software (FOSS) OR “satellite” OR “spatial*”) AND (“develop* countr*” OR
“develop* region*” OR “develop* nation*” OR “transition*
countr*” OR “third world” OR “less develop*” OR “East*
Europe” OR “Africa” OR “Asia” OR “South America” OR
“Central America” OR “China” OR “India” OR “Pakistan”
OR “Indonesia” OR “Malaysia” OR “Thailand” OR “Ban-
gladesh”) AND (“open source” OR “OS” OR “open source
software” OR “OSS” OR “free software” OR “FS” OR “free
and open source” OR “FOS” OR “free and open source
software” OR “FOSS” OR “F/OSS” OR “free/libre and open
source software” OR “FLOSS” OR “open source software/
free software” OR “OSS/FS” OR “FOSS4G” OR “GRASS”
OR “gvSIG” OR “Quantum” OR “SAGA” OR “uDig” OR
“ILWIS” OR “JUMP” OR “DIVA” OR “KOSMO”))

tion and communications technologies” (http://
www.un.org/millenniumgoals). GIT are within
the domain of ICT, yet peppered throughout this
current review are a number of barriers that can
limit GIT implementation in developing countries.
We propose that FOSS can reduce some of these
barriers, and in doing so, can also help in achiev-
ing the development goals established by the UN.
Primarily, FOSS eliminates the need to pur-
chase expensive software licenses and makes GIT
accessible to anyone, including local level disaster
managers operating on very small budgets. This
point is of fundamental importance when consid-
ering that in many cases “disaster management
responsibilities and duties are decentralized to
local governments without being accompanied by
sufficient funding,” (Montoya & Masser 2005, p.
494). Second, there exists a lack of local expertise
and knowledge that is required to implement and
maintain the use of GIT. FOS GIT software could
be used to learn about the technology itself, and
the potential ways in which it could be used for
natural disaster management purposes. Ragha-
van et al. (2006) emphasize that previous FOS
GIT technology transfer was an arduous task
for novice users, but recent developments in the
availability of packaged solutions, such as FW-
tools (see maptools.org), and commercial support
have helped reduce implementation obstacles. In
addition, many FOS GIT software products are
available as a single executable file, and once
downloaded, are installed in the same manner as
proprietary software (e.g., point and click). Albeit,
some software versions designed for Linux or
other FOSS operating systems do require specific
installation instructions and a certain amount of
computer knowledge. Furthermore, that FOSS
can be an effective educational/training tool is an
idea that is widely acknowledged, given that they
can be accessed at no cost and experimented with
in whatever way users see fit. Local workforce
development and capacity building projects are
essential for sustained GIT implementation and
maintenance that does not rely on external sup-
port or funding.
In general, disaster management is dependent
on the “functional and effective operation of
institutions, whether formal or informal, and at
the local level where it matters most,” (Pande,
2006, p. 426). However, government instability

203
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
is common in developing countries, which leads
to institutional instability (Murgia et al., 2002),
which can result in a lack of commitment to ad-
dress disaster management requirements, such as
providing adequate technology and training. As a
result of this instability, Ramasubramanian (1999)
suggests that GIS implementation in developing
countries is likely to be more successful if it re-
lies on the capacity of “empowered individuals
and groups rather than solely on organizational
structures,” (p. 359) such as public institutions.
To some extent, FOS GIT could be harnessed to
empower such groups, and allow them to benefit
from technology without having to rely only on
institutional initiatives and financial support.
Lastly, and perhaps most importantly, FOS
GIT offers considerable potential to help build
spatial data infrastructures required to fulfill basic
disaster management mitigation and preparation
planning, and following a disaster, speed response
and recovery efforts. It has already been shown
that many essential disaster management require-
ments hinge upon the existence of required spatial
datasets, such as critical facilities, transportation
infrastructure and population distribution (loca-
tion of towns and villages), yet many of these
key datasets are missing in developing countries.
Especially for “assessing the situation in more
remote regions, which are frequently only reached
days or even weeks after an event, a detailed
inventory of settlements and infrastructure may
save many lives and speed up recovery,” (Kerle
& Oppenheimer 2002, p. 142). Montoya and
Masser (2005) emphasize the need to identify or
develop “cost-effective data collection methods
for producing spatially referenced information
in developing countries,”(p. 495). Mansourian et
al. (2004) propose the development of a SDI as
a framework to facilitate disaster management,
and many currently available, and user friendly,
desktop FOS GIS (Steiniger & Bocher, 2009)
have the necessary functionality to assist the SDI
development process.
204
Many of the barriers to successful GIT imple-
mentation in developing countries can potentially
be addressed with FOSS. Currently, there are many
basic FOS GIT software available and others are
undergoing rapid development (Raghavan et al.,
2006). However, there still remains a need to iden-
tify particular software products, establish feasible
methodologies and workflows, and in general
document specific examples of FOSS usage. While
few would argue against the potential of FOSS,
there exists a research need to clarify the role of
FOS GIT during each of the phases of the disaster
management cycle. Regardless, a lack of attention
to the aforementioned GIT implementation barri-
ers “in any decision on whether to adopt GIT will
significantly increase the risk of system failures,
a matter of some significance in any environment
characterized by a lack of financial, technical and
human resources,” (Sliuzas 1999, p. 3).
This review suggests that the successful ap-
plication of GIT in natural disaster management
can require significant quality spatial data, GIT
knowledge and technical skills, and/or advanced
functionality. Each of these requirements consti-
tutes significant obstacles from a developing coun-
try perspective, especially at any administrative
level other than national. Another noteworthy fact
is that the use of proprietary GIT typically requires
significant computing resources. This represents a
huge obstacle in developing countries, which often
do not have access to modern (i.e., less than 2-3
years old) computer hardware (Caldeweyher et al.
2006, p. 885). GIS users in developing countries
are likely unable to take advantage of advanced
spatial analysis and data visualization features
prevalent in current proprietary GIS software. In
comparison, FOS GIS software generally require
less processing power, and are more suitable for
users who do not have modern computers. Thus,
while advanced GIT usage is feasible in developed
countries and within academic/research environ-
ments (including some in developing countries),
for the typical disaster management practitioner
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
in developing countries this type of high-end GIT
software usage is unachievable (Cutter, 2003).
Although we have argued that currently avail-
able FOS GIT have great potential, we do acknowl-
edge that proprietary GIT software, at this point,
offer superior overall functionality and provide
advanced analytical and visualization capabilities
that are beyond what current FOS GIT software can
provide, with GRASS being a possible exception.
However, GRASS is notoriously difficult to use
and requires an advanced GIT skill set, and as a
result, it is not suitable for use at the local level
in developing countries where such expertise is
in short supply, and where an inability to success-
fully use the software may result in frustration
and discouragement. To avoid such a result, we
suggest that simpler FOS GIT software could be
implemented and utilized more successfully, and
at the same time provide sufficient functionality
to improve current disaster management GIT
requirements at the local level. In addition, com-
bining a number of FOS GIT software together
can extend the functionality in the long term. GIT
usage must not be over ambitious in the early
stages and should focus on the core requirements
needed to improve existing disaster management
capabilities. Osti et al. (2008) make a similar
point, and suggest that “low-tech systems, which
are less cost intensive and can be handled by the
local population,” (p. 105) are of high interest in
the context of developing countries.
Within the hazard and disaster management
literature that addresses developing countries
and GIT implementation issues there are a couple
additional topics that deserve mentioning: the par-
ticipatory approach (e.g., PGIS) and the concept of
traditional/indigenous environmental knowledge.
Both of these topics are also partly embedded
within the technology transfer literature, and
although not within the scope of this paper, we
will briefly touch on these two topics since FOSS
has a potentially large role to play.
With technology transfer there is a tendency
to assume that development occurs when less
developed countries adopt the technologies and
methods of the more advanced countries (Britton,
2000). However, this approach is increasingly
being recognized as ineffective (Brewer et al.,
2005; Heeks, 2002). Although GIT implemen-
tation successes are often reported in “popular
magazines or trade journals, their sustainabil-
ity is seldom documented, and failures are not
reported or glossed over to avoid criticism,”
(Ramasubramanian 1999, p. 360). Heeks (2002)
explains that the “introduction of GIS in develop-
ing countries has been problematic, as they are
seen to incorporate a number of assumptions and
requirements that derive from Western rational-
ism,” (p. 107). One way to possibly improve the
success of technology transfer initiatives is through
participatory approaches. Within the GIT domain,
GIS has garnered considerable attention from a
participatory standpoint and thus will be used as
an example. In terms of developing countries,
Abbot et al. (1998) define PGIS as “an attempt
to utilise GIS technology in the context of the
needs and capabilities of communities that will
be involved with, and affected by, development
projects and programmes.” A central objective
of the participatory approach is to capture local
knowledge and combine it with more traditional
spatial information, and to facilitate GIS produc-
tion and use which are community-based. Creating
GIT environments that pull in new ideas, and pos-
sibly new spatial methods and techniques, rather
than having them pushed in from the outside,
appears to be a more effective solution (Britton,
2000). Developing countries could benefit from
international development programs that focus
on capacity building in the area of FOSS-based
participatory GIT implementation. Knowledge-
able practitioners in developing countries could
then further build local capacity, especially with
FOSS, which have been shown to be a good fit
for developing countries. If appropriately utilized,
Rambaldi et al. (2006) emphasize that participa-
tory approaches could exert “profound impacts on
205
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
community empowerment, innovation and social
change,” (p. 1).
Kienberger and Steinbruch (2005) applied
a PGIS approach in Mozambique, in which the
main objectives were to “acquire spatial data of
the communities through GPS surveys,”(p. 5) to
better visualize certain aspects of the community
to assist “disaster risk committees, and to identify
secure locations in the case of floods,” (p. 5). Re-
sults indicate that the participatory approach was
suitable “for the collection of spatial information
and to capture the perception of the local people
regarding disasters,”(Kienberger & Steinbruch,
2005, p. 5). A follow up of this study revealed that
in Ssome communities, and without outside/donor
influence, maps were also used at meetings to dis-
cuss issues with a spatial relevance,”(Kienberger
& Steinbruch, 2005, p. 5).
Generally speaking, participatory approaches
can help build local capacity to understand and
implement GIT to improve disaster management
capabilities. Using a participatory approach local
knowledge about natural hazards – their loca-
tion and frequency – can be ascertained from
community members and represented spatially
within a GIS. For example, using a GPS coor-
dinates of the location of previous floods can be
captured, along with elevation data. Within the
GIS the flood point locations can be analyzed in
conjunction with elevation data (e.g., a DEM or
contour lines), and potential flood scenarios can be
considered and prepared for accordingly. Comb-
ing locally derived hazard knowledge with other
spatial data such as transportation infrastructure,
critical facilities and towns/villages within a GIS
environment will help the local population to be
more prepared for and help mitigate potential
disaster impacts. Ultimately, the goal is to reduce
disaster vulnerability, and this can be achieved with
a better spatial understanding of the relationship
between natural hazards and the elements at risk
(e.g., people, infrastructure, etc.) (Montoya &
Masser, 2005).
206
In the past, there has been unwillingness to
use ‘non-scientist’ or ‘indigenous knowledge’ as
data. However, there is growing recognition of
the importance of indigenous knowledge, and the
need to better understand how it can be combined
with Western knowledge to reduce natural hazard
vulnerability (Mercer et al., 2007; Pande, 2006).
Local communities in developing countries rely
heavily on their environment; “they depend on the
subtleties of ecological cycles and patterns, and
have accumulated a body of wisdom commonly
referred to as ‘traditional environmental knowledge’
(TEK),” (Brodnig & Mayer-Schönberger 2000, p.
3). “TEK is in essence a geographical information
system derived from and embedded in the close
relationship of local people with their land and
natural resources...with members of the community
serving as repositories for different types and cat-
egories of data…” (Brodnig & Mayer-Schönberger,
2000). Raghavan et al. (2006) propose that FOS
GIT could help local practitioners incorporate
spatial knowledge, and thus integrate TEK with
technology use. Such an approach could prove
valuable for natural disaster management, since
local knowledge of environmental hazards and
traditional coping strategies could be very useful.
CONCLUSION
As the human population continues to grow, and
considering recent evidence of climate change that
might exacerbate meteorological related natural
hazards in particular, there is reason for concern
that natural disasters may occur in the future with
increasing frequency and consequence. As a result,
there is an immediate need to utilize available
technology to reduce natural hazard vulnerability
and in general to be more prepared to effectively
respond to disasters when they occur.
This article has reviewed and examined the
use of geospatial information technology (GIT) in
the field of natural disaster management, with an
emphasis on developing countries where natural
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
hazard vulnerability is high and disaster impacts
can be particularly devastating. The ability of GIT
to acquire, interpret, analyze, map and disseminate
information, are essential during all phases of the
natural disaster management cycle – the process
of mitigation, preparation, response and recovery.
Since disasters are spatial phenomenon there is a
strong relationship between disaster management
requirements and the spatial information and
decision support capabilities offered by GIT. In
this context, GIT includes geographic informa-
tion systems, remote sensing, global positioning
systems and Internet GIS. Among numerous other
tasks, GIT provides a basis for planning mitigation
strategies, hazard and risk assessment/mapping,
vulnerability assessment, vehicle dispatch and
supply routing, damage assessment, and response
resource mobilization. Consequently, by using
GIT, it is possible to identify and mitigate risk, be
better prepared for and respond to disasters, and
recover from disasters. However, we have also
identified and described a number of GIT imple-
mentation barriers that are especially relevant
to the application of GIT at local administrative
levels, where a strong disaster management ini-
tiative is required. These barriers include, but are
not limited to: (1) a lack of financial resources,
(2) a lack of spatial data, (3) political/institutional
instability and (4) a lack of local GIT knowledge/
expertise. Thus, until at least some of these bar-
riers are overcome the level of ability for GIT to
improve overall disaster management capacity at
the local level will remain low.
Many researchers have highlighted the par-
ticular opportunity that free and open source
software (FOSS) now provides for developing
countries that were previously unavailable. At-
tractive characteristics of FOSS from a develop-
ing country perspective include: cost, freedom,
accessibility, customizability, compatibility, and
software/technical capacity development oppor-
tunities. The cost aspect is especially significant
from a GIT perspective, as proprietary/commer-
cial GIT software are expensive and a lack of
financial resources is currently a very significant
GIT implementation barrier. Recent growth and
development in the FOSS-based GIT domain has
resulted in the emergence of many mature, very
capable and user-friendly software products that
in some cases offer functionality that is compa-
rable with commercial GIT software. As a result,
we propose that FOSS-based GIT products can
be used to improve the use of GIT in developing
countries, especially at the local level. Specifi-
cally, FOSS-based GIT provide opportunities to
improve local GIT knowledge and related skill
sets required for their effective application in the
field of natural disaster management. Additionally,
the current capabilities of FOSS-based GIT allow
for the development of spatial data infrastructures
(SDI) that are required by the disaster manage-
ment practitioner community to successfully
implement GIT, for purposes such as hazard and
risk assessment. Considering the lack of available
spatial data in developing countries, which is a
GIT implementation barrier, the development of
SDIs will represent a major step forward. How-
ever, future research is required that identifies
specific FOSS-based GIT and functionality that
can be effectively implemented at the local level
in developing countries to improve overall natural
disaster management capacity.
ACKNOWLEDGMENT
During the course of this research, the authors
appreciate the discussions with Mr. Chandeep
Corea, head of the GIS department of the Sri Lanka
Wildlife conservation Society (SLWCS). The
critical input of Dr. Barry Wellar was essential in
developing this paper. The authors also acknowl-
edge financial support to Sam Herold in the form
of a postgraduate scholarship from the Natural
Sciences and Engineering Research Council of
Canada. Further financial support was provided by
the Canadian Foundation for Innovation, Ontario
Innovation Trust and the University of Ottawa.
207
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
REFERENCES
Abbot, J., Chambers, R., Dunn, C., Harris, T.,
de Merode, E., & Porter, G. (1998). Participa-
tory GIS: opportunity or oxymoron? PLA Notes
International Institute for Environment and De-
velopment, 33, 24–34.
Alexander, D. E. (1995). A survey of the field of
natural hazards and disaster studies. In Carrara, A.,
& Guzzetti, F. (Eds.), Geographical information
systems in assessing natural hazards (pp. 1–19).
Amsterdam, Netherlands: Kluwer Academic.
Amdahl, G. (2001). Disaster response: GIS for
public safety. Redlands, CA: ESRI.
Amendola, A., Linnerooth-Bayer, J., Okada, N.,
& Shi, P. (2008). Towards integrated disaster risk
management: case studies and trends in Asia.
Natural Hazards, 44(2), 163–168. doi:10.1007/
s11069-007-9152-z
Anderson, G., & Moreno-Sanchez, R.., (2003).
Building web-based spatial information solu-
tions around open specifications and open source
software. Transactions in GIS, 7 (4), 447–466.
Becking, I. (2004, September 6-10). How space
can better support emergency management and
critical infrastructure protection. In Proceedings of
the Envisat & ERS Symposium, Salzburg, Austria.
Brewer, E., Demmer, M., Du, B., Ho, M., Kam, M.,
& Nedevschi, S. (2005). The case for technology
in developing regions. IEEE Computer Society,
38(6), 25–38. doi:10.1109/MC.2005.204
Briceño, S. (2004). Building disaster-resilient
communities: The road to the second World
Conference on Disaster Reduction, Kobe, Hyogo,
Japan. Natural Resources Forum, 28, 234–236.
doi:10.1111/j.1477-8947.2004.00093.x
208
Britton, J. M. R. (2000). GIS Capacity Build-
ing in the Pacific Island Countries: Facing the
realities of technology, resources, geography and
cultural difference. Cartographica, 37(4), 7–19.
doi:10.3138/B267-8663-4071-2375
Brodnig, G., & Mayer-Schönberger, V. (2000).
Bridging the gap: The role of spatial informa-
tion technologies in the integration of traditional
environmental knowledge and western science.
Electronic Journal on Information Systems in
Developing Countries, 1(1), 1–15.
Bui, T., Cho, S., Sankaran, S., & Sovereign,
M. (2000). A framework for designing a
global information network for multinational
humanitarian assistance/disaster relief. In-
formation Systems Frontiers, 1(4), 427–442.
doi:10.1023/A:1010074210709
Burton, I., Kates, R. W., & White, G. F. (1993).
The environment as hazard (2nd ed.). New York,
NY: Guilford Press.
Caldeweyher, D., Zhang, J., & Pham, B. (2006).
OpenCIS—Open Source GIS-based web commu-
nity information system. International Journal
of Geographical Information Systems, 20(8),
885–898. doi:10.1080/13658810600711378
Câmara, G., & Onsrud, H. (2004). Open-source
geographic information systems software: Myths
and realities. In Proceedings of the International
Symposium on Open Access and the Public Do-
main in Digital Data and Information for Science
(pp. 127-133).
Chen, P., Liew, S. C., & Kwoh, L. K. (2005, July
25-29). Tsunami damage assessment using high
resolution satellite imagery: A case study of Aceh,
Indonesia. In Proceedings of the IEEE Interna-
tional Symposium on Geoscience and Remote
Sensing (Vol. 2, pp. 1405-1408).
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
Cook, I., & Horobin, G. (2006). Implementing
eGovernment without promoting dependence:
Open source software in developing countries
in Southeast Asia. Public Administration and
Development, 26, 279–289. doi:10.1002/pad.403
Coppock, T. J. (1995). GIS and natural hazards: an
overview from a GIS perspective. In Carrara, A.,
& Guzzetti, F. (Eds.), Geographical information
systems in assessing natural hazards (pp. 21–34).
Amsterdam, The Netherlands: Kluwer Academic.
Cova, T. J. (1999). GIS in emergency management.
In Longley, M. A., Goodchild, M. F., Maguire, D.
J., & Rhind, D. W. (Eds.), Geographical informa-
tion systems: Principles, techniques, applications,
and management (pp. 845–858). New York, NY:
John Wiley & Sons.
Currion, P., De Silva, C., & Van De Walle, B.
(2007). Open source software for disaster man-
agement. Communications of the ACM, 50(3),
61–65. doi:10.1145/1226736.1226768
Cutter, S. L. (1996). Vulnerability to environ-
mental hazards. Progress in Human Geography,
20, 529–539. doi:10.1177/030913259602000407
Cutter, S. L. (2003). GI science, disasters, and
emergency management. Transactions in GIS,
7(4), 439–445. doi:10.1111/1467-9671.00157
Cutter, S. L., Boruff, B. J., & Shirley, W. L.
(2003). Social vulnerability to environmental haz-
ards. Social Science Quarterly, 84(2), 242–261.
doi:10.1111/1540-6237.8402002
Cutter, S. L., Mitchell, J. T., & Scott, M. S. (2000).
Revealing the vulnerability of people and places: A
case study of Georgetown County, South Carolina.
Annals of the Association of American Geogra-
phers. Association of American Geographers, 90,
713–737. doi:10.1111/0004-5608.00219
De La Ville, N., Diaz, A. C., & Ramirez, D.
(2002). Remote sensing and GIS technologies
as tools to support sustainable management of
areas devastated by landslides. Environment,
Development and Sustainability, 4, 221–229.
doi:10.1023/A:1020835932757
De Paratesi, S. R. (1989). Hazards and disasters:
concepts and challenges. In Barrett, E. C., &
Brown, K. A. (Eds.), Remote sensing for hazard
monitoring and disaster assessment: marine and
coastal applications in the Mediterranean region
(pp. 1–17). Philadelphia, PA: Gordon and Breach.
Dewan, A. M., Islam, M. M., Kumamoto, T., &
Nishigaki, M. (2007). Evaluating flood hazard
for land-use planning in greater Dhaka of Ban-
gladesh using remote sensing and GIS techniques.
Water Resources Management, 21, 1601–1612.
doi:10.1007/s11269-006-9116-1
Dewan, A. M., Kabir, M. D. H., Islam, M. M., Ku-
mamoto, T., & Nishigaki, M. (2007). Delineating
flood risk areas in Greater Dhaka of Bangladesh
using geoinformatics. Georisk, 1(4), 190–201.
Dong, P. (2005). Development of a GIS/GPS-
based emergency response system. Geomatica,
59(4), 427–433.
El-Masri, S., & Tipple, G. (2002). Natural disaster,
mitigation and sustainability: The case of devel-
oping countries. International Planning Studies,
7(2), 157–175. doi:10.1080/13563470220132236
Engler, N. J., & Hall, G. B. (2007). The Internet,
spatial data globalization and data use: The case
of Tibet. The Information Society, 23, 345–359.
doi:10.1080/01972240701572822
Environmental System Research Institute (ESRI).
(1993). Understanding GIS: The ARC/INFO
Method. Redlands, CA: Author.
209
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries

Environmental Systems Research Institute (ESRI). Hoe, N. S. (2006). Breaking barriers: The potential
(2006). GIS and emergency management in Indian of free and open source software for sustainable
Ocean earthquake/tsunami disaster. Retrieved human development. A Compilation of Case
from http://www.esri.com/library/whitepapers/ Studies from Across the World. Retrieved from
pdfs/gis-and-emergency-mgmt.pdf http://www.apdip.net/publications/ict4d/Break-
ingBarriers.pdf
Fitzgerald, B. (2004). A critical look at open source.
IEEE Computer, 37(7), 92–94. doi:10.1109/ Hussain, M., Arsalan, M. H., Siddiqi, K., Nas-
MC.2004.38 eem, B., & Rabab, U. (2005, June 9-11). Emerg-
ing geo-information technologies for natural
Freeman, P. K., & Pflug, G. C. (2003). Infrastruc-
disaster management in Pakistan: An overview.
ture in developing and transitional countries: Risk
In Proceedings of the 2nd International Confer-
and protection. Risk Analysis, 23(3), 601–609.
ence on Recent Advances in Space Technolo-
doi:10.1111/1539-6924.00340
gies. Retrieved from http://ieeexplore.ieee.org/
Goodchild, M. F. (2006). GIS and disasters: Plan- iel5/10107/32390/01512618.pdf?tp=&arnumbe
ning for catastrophe. Computers, Environment and r=1512618&isnumber=32390
Urban Systems, 3(30), 227–229. doi:10.1016/j.
International Federation of Red Cross and Red
compenvurbsys.2005.10.004
Crescent Societies (IFRC). (2001). World disasters
Guinau, M., Pallàs, R., & Vilaplana, J. M. (2005). report. Geneva, Switzerland: Author.
A feasible methodology for landslide susceptibility
International Panel on Climate Change (IPCC).
assessment in developing countries: A case-study
(2007). Climate change 2007: The Physical Sci-
of NW Nicaragua after Hurricane Mitch. Engi-
ence Basis, Contribution of Working Group I to the
neering Geology, 80, 316–327. doi:10.1016/j.
Fourth Assessment Report of the IPCC. Retrieved
enggeo.2005.07.001
from http://www.ipcc.ch/ipccreports/ar4-wg1.htm
Gunes, A. E., & Koval, J. P. (2000). Using
Jayaraman, V., Chandrasekhar, M. G., & Rao, U.
GIS in emergency management operations.
R. (1997). Managing the natural disasters from
Journal of Urban Planning and Development,
space technology inputs. Acta Astronautica, 40(2),
126(3), 136–149. doi:10.1061/(ASCE)0733-
291–325. doi:10.1016/S0094-5765(97)00101-X
9488(2000)126:3(136)
Johnson, R. (2000). GIS technology for disaster
Heeks, R. (2002). Information systems and de-
and emergency management. Redlands, CA:
veloping countries: Failure, success, and local
ESRI.
improvisations. The Information Society, 18,
101–112. doi:10.1080/01972240290075039 Karimi, H. A., & Blais, J. A. R. (1996). Cur-
rent and future directions in GISs. Computers,
Henderson, L. J. (2004). Emergency and disaster:
Environment and Urban Systems, 20(2), 85–97.
Pervasive risk and public bureaucracy in developing
doi:10.1016/S0198-9715(96)00002-6
nations. Public Organization Review, 4(2), 103–
119. doi:10.1023/B:PORJ.0000031624.46153.b2 Kerle, N., & Oppenheimer, C. (2002). Satel-
lite remote sensing as a tool in Lahar disaster
Hewitt, K. (1995). Excluded perspectives in the
management. Disasters, 26(2), 140–160.
social construction of disaster. International
doi:10.1111/1467-7717.00197
Journal of Mass Emergencies and Disasters,
13(3), 317–339.

210
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
Kienberger, S., & Steinbruch, F. (2005, September
7-10). P-GIS and disaster risk management: As-
sessing vulnerability with P-GIS methods – Ex-
perience from Buzi, Mozambiqu. In Proceedings
of the International Conference on Participatory
Spatial Information Management and Commu-
nication: Mapping for Change, Kenya, Nairobi.
Kondratyev, K. Y., Grigoryev, A. A., & Varotsos,
C. A. (2002). Environmental disasters: Anthro-
pogenic and natural. UK: Praxis.
Kraak, M.-J. (2004). The role of the map in a Web-
GIS environment. Journal of Geographical Sys-
tems, 6, 83–93. doi:10.1007/s10109-004-0127-2
Kumar, V., Bugacov, A., Coutinho, M., & Neches,
R. (1999, November 5-6). Integrating geographic
information systems, spatial digital libraries and
information spaces for conducting humanitarian
assistance and disaster relief operations in urban
environments. In Proceedings of the 7th ACM
Symposium on Advances in Geographic Informa-
tion Systems, Kansas City, MO.
Laben, C. (2002). Integration of remote sensing
data and geographic information system technol-
ogy for emergency managers and their applications
at the Pacific Disaster Center. Optical Engineer-
ing (Redondo Beach, Calif.), 41(9), 2129–2136.
doi:10.1117/1.1501137
Lay, T., Kanamori, H., Ammon, C. J., Nettles,
M., Ward, S. N., & Aster, R. C. (2005). The Great
Sumatra-Andaman Earthquake of 26 December
2004. Science, 308, 1127–1133. doi:10.1126/
science.1112250
Levy, J.K. (2005). Advances in decision support
systems for flood disaster management: challenges
and opportunities. International Journal of Water
Resources Development, 21(4), 593-612.
Magsud, M., Sann Oo, K., & Rajapaksha, J. (2005,
November 7-11). Tsunami disaster damage detec-
tion and assessment using high resolution satellite
data, GIS and GPS – Case study in Sri Lanka. In
Proceedings of the Asian Conference on Remote
Sensing, Hanoi, Vietnam.
Mansor, S., Shariah, M. A., Billa, L., Se-
tiawan, I., & Jabar, F. (2004). Spatial technol-
ogy for natural risk management. Disaster
Prevention and Management, 13(5), 364–373.
doi:10.1108/09653560410568480
Mansourian, A., Rajabifard, A., Valadan Zoej, M.
J., & Williamson, I. (2004). Facilitating disaster
management using SDI. Journal of Geospatial
Engineering, 6(1), 29–44.
McEntire, D. A. (1999). Issues in disaster relief:
progress, perpetual problems and prospective
solutions. Disaster Prevention and Management,
8(5), 351–361. doi:10.1108/09653569910298279
McEntire, D. A. (2001). Triggering agents, vulner-
abilities and disaster reduction: towards a holistic
paradigm. Disaster Prevention and Management,
10(3), 189–196. doi:10.1108/09653560110395359
Mercer, J., Dominey-Howes, D., Kelman, I.,
& Lloyd, K. (2007). The potential for combing
indigenous and western knowledge in reducing
vulnerability to environmental hazards in small
island developing states. Environmental Hazards,
7, 245–256. doi:10.1016/j.envhaz.2006.11.001
Mohamed, M., & Plante, R. (2002). Remote
sensing and Geographic Information Systems
(GIS) for developing countries. In Proceed-
ings of the IEEE International Geoscience and
Remote Sensing Symposium (Vol. 4, pp. 2285-
2287). Retrieved from http://ieeexplore.ieee.org/
iel5/7969/22039/01026520.pdf
Montoya, L. (2002, April 25-27). GIS and remote
sensing in urban disaster management. In Proceed-
ings of the 5th AGILE Conference on Geographic
Information Science, Balearic Islands, Spain.
211
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries

Montoya, L., & Masser, I. (2005). Management
of natural hazard risk in Cartago, Costa Rica.
Habitat International, 29, 493–509. doi:10.1016/j.
habitatint.2004.04.003
Morrow, B. H. (1999). Identifying and mapping
community vulnerability. Disasters, 23(1), 1–18.
doi:10.1111/1467-7717.00102
Murgia, J., Amemiya, N., & Turkstra, J. (2002,
May 15-18). Local spatial data infrastructure,
Trujillo, Peru. In Proceedings of the 6th Seminar
on GIS and Developing Countries. Retrieved from
http://www.gisdevelopment.net
Musinguzi, M., Bax, G., & Tickodri-Togboa, S.
S. (2004, June 7-9). Opportunities and challenges
for SDI development in developing countries – A
case study of Uganda. In Proceedings of the 12th
International Conference on Geoinformatics –
Geospatial Information Research: Bridging the
Pacific and Atlantic, Gävle, Sweden.
Nebert, D. (2004). Developing spatial data in-
frastructures: The SDI cookbook v.2.0. Global
spatial data infrastructure. Retrieved July 2005,
from http://www.gsdi.org
Nonogaki, S., Van Anh, T., Masumoto, S.,
Raghavan, V., Nemoto, T., & Mori, T. (2004).
Development of Vietnamese Version of GRASS
GIS. Geoinformatics, 15, 106–107.
Oloruntoba, R. (2005). A wave of destruction and
the waves of relief: issues, challenges and strate-
gies. Disaster Prevention and Management, 14(4),
506–521. doi:10.1108/09653560510618348
OSGeo. (2008). About the Open Source Geospatial
Foundation. Retrieved from http://www.osgeo.
org/content/foundation/about.html
Osti, R., Tanaka, S., & Tokioka, T. (2008).
Flood hazard mapping in developing coun-
tries: problems and perspectives. Disaster
Prevention and Management, 17(1), 104–113.
doi:10.1108/09653560810855919
Pande, R. K. (2006). Participation in practice and
disaster management: experience of Uttaranchal
(India). Disaster Prevention and Management,
15(3), 425–428. doi:10.1108/09653560610669909
Peng, Z. R., & Tsou, M. H. (2003). Internet GIS.
Hoboken, NJ: John Wiley & Sons.
Plewe, B. (1997). GIS Online: Information Re-
trieval, Mapping, and the Internet. Albany, NY:
OnWord Press.
Quarantelli, E. L. (1998). What is disaster?
Perspectives on the question. New York, NY:
Routledge.
Rabindra, O., Tanaka, S., & Tokioka, T. (2008).
Flood hazard mapping in developing coun-
tries: problems and perspectives. Disaster
Prevention and Management, 17(1), 104–113.
doi:10.1108/09653560810855919
Radke, J., Cova, T., Sheridan, M. F., Troy, A., Lan,
M., & Johnson, R. (2000). Application challenges
for geographic information science: Implications
for research, education, and policy for emergency
preparedness and response. URISA Journal,
12(2), 15–30.
Raghavan, V., Masumoto, S., & Hastings, D.
(2006, November 9-11). Free and open source
software for geoinformatics: Present status and
future prospects. In Proceedings of the Interna-
tional Symposium on Geoinformatics for Spatial
Infrastructure Development in Earth and Allied
Sciences, Ho Chi Minh, Vietnam. Retrieved from
http://wgrass.media.osaka-cu.ac.jp/gisideas06/
viewpaper.php?id=166
Rajani, N., Rekola, J., & Mielonen, T. (2003).
Free as in education: Significance of Free/Libre
Open Source Software for developing countries.
Retrieved from http://www.itu.int/wsis/docs/
background/themes/access/free_as_in_educa-
tion_niranjan.pdf

212
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
Ramasubramanian, L. (1999). GIS imple-
mentation in developing countries: Learn-
ing from organisational theory and reflective
practice. Transactions in GIS, 3(4), 359–380.
doi:10.1111/1467-9671.00028
Rambaldi, G., Kwaku Kyem, P. A., McCall,
M., & Weiner, D. (2006). Participatory spatial
information management and communication in
developing countries. The Electronic Journal on
Information Systems in Developing Countries,
25(1), 1–9.
Ramli, M. F., Sulaiman, W. N. A., Yusoff, M. K.,
& Low, Y. Y. (2005). Open source geographical
resources analysis support system (GRASS)
for landslide hazard assessment. Disaster
Prevention and Management, 11(4), 522–532.
doi:10.1108/09653560510618357
Ramsey, P. (2007). The state of open source GIS.
In Proceedings of the Global Conference on
Free and Open Source Software for Geospatial,
Vancouver, BC, Canada. Retrieved from http://
www.refractions.net/expertise/whitepapers/open-
sourcesurvey/ survey-open-source-2007-12.pdf
Rautela, P., & Pande, R. K. (2005). Implications of
ignoring the old disaster management plans: Les-
sons learnt from the Amparav tragety of 23 Septem-
ber 2004 in the Nainital district of Uttaranchal (In-
dia). Disaster Prevention and Management, 14(3),
388–394. doi:10.1108/09653560510605045
Rego, A. J. (2001, March 21-23). National disaster
management information systems & networks:
An Asian overview. In Proceedings of the Global
Disaster Information Network Conference, Can-
berra, Australia. Retrieved from http://www.adpc.
net/infores/adpc-documents/paperatgdin01.pdf
Renyi, L., & Nan, L. (2002). Flood area and dam-
age estimation in Zhejiang, China. Journal of En-
vironmental Management, 66, 1–8. doi:10.1006/
jema.2002.0544
Rivereau, J. C. (1995). Spot data applied to di-
saster prevention and damage assessment. Acta
Astronautica, 35(7), 467–470. doi:10.1016/0094-
5765(94)00281-P
Roets, R., Minnaar, M., & Wright, K. (2007). Open
Source: Towards successful systems development
projects in developing countries. In Proceedings
of the 9th International Conference on Social Im-
plications of Computers in Developing Countries,
São Paulo, Brazil. Retrieved from http://www.
ifipwg94.org.br/fullpapers/R0061-1.pdf
Rüdenauer, H., & Schmitz, M. (2005, March
14-16). Use of Landsat 7 ETM+ data as basic
information for infrastructure planning. In Pro-
ceedings of the ISPRS Joint Conference of the
3rd International Symposium on Remote Sensing
and Data Fusion Over Urban Areas, and the 5th
International Symposium on Remote Sensing of
Urban Areas, Tempe, AZ. Retrieved from http://
www.isprs.org/commission8/workshop_urban/
ruedenauer.pdf
San Miguel-Ayanz, J., Vogt, J., De Roo, A., &
Schmuck, G. (2000). Natural hazards monitoring:
Forest fires, droughts and floods – the examples
of European Pilot Projects. Surveys in Geophys-
ics, 21, 191–305. doi:10.1023/A:1006750412500
Sanyal, J., & Lu, X. X. (2003, October 13-15).
Application of GIS in flood hazard mapping: a
case study of Gangetic West Bengal, India. In
Proceedings of the Map Asia Conference, Kuala
Lumpur, Malaysia. Retrieved from http://www.
gisdevelopment.net/application/natural_hazards/
floods/pdf/ma03138.pdf
Schutzberg, A. (2005). MapServer Community,
Autodesk Announce MapServer Foundation. Di-
rections Magazine. Retrieved from http://www.
directionsmag.com/article.php?article_id=2037
Showalter, P. S. (1999). Remote sensing’s
use in disaster research: A review. Disaster
Prevention and Management, 10(1), 21–29.
doi:10.1108/09653560110381796
213
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries
Shrestha, S. (1994). Environmental issues in
Asia and the pacific - a perspective towards the
information in decision making. In Proceed-
ings of the International Conference on Expert
Systems for Development, Bangkok, Thailand.
Retrieved from http://ieeexplore.ieee.org/
iel2/973/7458/00302314.pdf?tp=&isnumber=&
arnumber=302314
Simonovic, S. P. (2002). Role of remote sensing
in disaster management (Paper Series No. 21).
London, ON, Canada: University of Western On-
tario, Institute for Catastrophic Loss Reduction.
Simpson, D. M., & Human, R. J. (2008). Large-
scale vulnerability assessments for natural hazards.
Natural Hazards, 47(2), 143–155. doi:10.1007/
s11069-007-9202-6
Sliuzas, R. (1999, March 11-12). Research is-
sues for the adoption of geographic information
technology for urban planning and management
in developing countries. In Proceedings of the
International Workshop on Concepts and Para-
digms of Urban Management in the Context of
Developing Countries, Venice, Italy. Retrieved
from http://www.naerus.net/sat/workshops/1999/
paper_sliuzas.html
Smith, K. (2004). Environmental hazards: As-
sessing risk and reducing disaster (4th ed.). New
York, NY: Routledge.
Space Agencies. (2005). International charter on
space and major disasters. Retrieved from http://
www.disasterscharter.org/main_e.html
Steiniger, S., & Bocher, E. (2009). An overview
on current free and open source desktop GIS de-
velopments. International Journal of Geographi-
cal Information Science, 23(10), 1345–1370.
doi:10.1080/13658810802634956
214
Thomson, C. N., & Hardin, P. (2000). Remote
sensing/GIS integration to identify potential
low-income housing sites. Cities (London,
England), 17(2), 97–109. doi:10.1016/S0264-
2751(00)00005-6
United Nations. (UN). (2006). The millennium
development goals report. Retrieved from http://
mdgs.un.org/unsd/mdg/Resources/Static/Prod-
ucts/Progress2006/MDGReport2006.pdf
Van Westen, C. J., & Hofstee, P. (2001). The role
of remote sensing and GIS in risk mapping and
damage assessment for disasters in urban areas.
Enschede, The Netherlands: International Institute
for Aerospace Survey and Earth Sciences (ITC).
Retrieved from http://www.dkkv.org/forum2001/
Datei61.pdf
Venkatachary, K.V., Manikiam, B. and Srivas-
tava, S.K. 2001. Harnessing information and
technology for disaster management. Retrieved
from http://www.aiaa.org/indiaus2004/Disaster-
management.pdf
Weber, S. (2004). The success of open source.
Cambridge, MA: Harvard University Press.
Weerawarana, S., & Weeratunge, J. (2004). Open
source in developing countries. Stockholm,
Swedish International Development Cooperation
(Sida) Agency. Retrieved from http://www.sida.
se/publications
Weichselgartner, J. (2001). Disaster mitigation:
The concept of vulnerability revisited. Disaster
Prevention and Management, 10(2), 85–94.
doi:10.1108/09653560110388609
Wheeler, D. A. (2007). Why Open Source Soft-
ware/Free Software (OSS/FS, FLOSS, or FOSS)?
Look at the Numbers! Retrieved from http://www.
dwheeler.com/oss_fs_why.html
A Review of Geospatial Information Technology for Natural Disaster Management in Developing Countries

Wu, M., & Lin, Y. (2001). Open source software Zerger, A., & Smith, D. I. (2003). Impediments to
development: An overview. IEEE Computer, using GIS for real-time disaster decision support.
33-38. Computers, Environment and Urban Systems, 27,
123–141. doi:10.1016/S0198-9715(01)00021-7

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 2,
edited by Donald Patrick Albert, pp. 24-62, copyright 2012 by IGI Publishing (an imprint of IGI Global).

215
216

Chapter 15
Virtual Environments for
Geospatial Applications
Magesh Chandramouli
Purdue University, USA

Bo Huang
Chinese University of Hong Kong, China

ABSTRACT
This article explores the application of virtual environments to 3D geospatial visualization and explora-
tion. VR worlds provide powerful functionalities for model generation and animation and are indeed a
valuable tool for geospatial visualization. Subsequently, related issues such as the constraints in pro-
gressive terrain rendering, geographic data modeling, photo-realism in virtual worlds, and the system
performance with relatively larger files are discussed. Nevertheless, to accomplish the desired results
and to attain a higher level of functionality, a good level of experience in VR programming and the ju-
risprudence to choose the appropriate tool are necessary. Although a standalone VR application is not
capable of a higher level of interaction, using the SCRIPT nodes and the External Authoring Interface
additional functionalities can be integrated. Intended for use over the internet with a VR browser, such
virtual environments serve not only as a visualization tool, but also a powerful medium for geospatial
data exploration.

INTRODUCTION animation. This chapter discusses related issues

This chapter explores the application of virtual
environments to 3D geospatial visualization, ani-
mation, and interaction. The authors describe the
design and implementation of some 3D models,
which offer a good level of user-interaction and
DOI: 10.4018/978-1-4666-2038-4.ch015
such as the constraints in progressive terrain ren-
dering, geographic data modeling, photo-realism
in virtual worlds, and the system performance
with relatively larger files. VR worlds provide
powerful functionalities for model generation
and animation and are indeed a valuable tool
for geospatial visualization. Nevertheless, to ac-
complish the desired results and to attain a higher

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Virtual Environments for Geospatial Applications
level of functionality, a good level of experience
in VR programming practices is mandatory. Even
though a standalone VR application is not capable
of a higher level of interaction, using the SCRIPT
nodes, JavaScript can be embedded in the program
to provide additional functionalities.
GEO-VIRTUAL ENVIRONMENTS:
EVOLUTION OVER THE YEARS
Since the 1960s and 70s, the past several decades
have witnessed the ‘information revolution’,
particularly in the domain of spatial information
technology, propelled by the advancements in data
acquisition techniques. The evolution of diverse
digital processing and image generation tech-
niques over the decades along with the parallel de-
velopments in Geographical Information Systems
GIS and remote sensing have resulted in colossal
volumes of digital spatial data. In order to make
the utmost use of this collected data, they must be
presented in the form of comprehensible informa-
tion. Geospatial data is increasingly being used
for addressing issues involving environmental
and urban planning, design, and decision-making
within a wide range of disciplines ranging from
urban landscape management to various other
applications. As geospatial data complexity in-
creases, besides the standard rasters, Triangulated
Irregular Networks (TINs) and Digital Elevation
Models (DEMs), which are used for data explora-
tion, additional tools such as photo-realistic models
are needed to provide advanced visualization and
query functionalities. Three-dimensional visual-
ization is a proven technique for exploring geo-
spatial data (Bonham-Carter, 1994). In the work
on urban modeling, Shiode (2001) explains the
development of 3D models and their role within
the domain of spatial information database and
remote sensing technology. The origins of concept
of spatial immersion can be dated back to 1965
when Ivan Sutherland (1965) made known the
ideas of immersion in virtual space in his influential
work, “The Ultimate Display”. Such immersive
virtual environments can serve as handy tools for
scientists and researchers that handle enormous
data volumes. By and large, visualization enables
large quantities of information to be presented in
a form that is comprehensible to a wide range of
users (Colin Ware, 2000).
3D Geospatial Data Visualization:
Tools and Techniques
Geospatial analysis and research require that the
data be in the 3D form. Geospatial data is inher-
ently three dimensional in nature since every
spatial element has its own position or location in
space (latitude, longitude, and altitude). A gamut
of applications involving geospatial analysis such
as environmental process simulation, infrastruc-
ture applications, landscape design, geological
applications, etc. necessitates three-dimensional
exploration and visualization. Traditionally, op-
erations such as town or country planning relied
heavily on drawings and these were eventu-
ally supplemented with Computer Aided Design
(CAD) drawings. However, one major handicap
with these forms of data is that they try to sym-
bolize 3D entities in 2D format. Albeit these may
offer a bird’s eye view of the place being studied,
such representations depicting 3D data using two
dimensions are incomplete and cannot replace
a 3D view. For instance, landscape and urban
modeling architecture applications of today are far
more complex and advanced tools are inevitable
to provide the required level of sophistication.
Several techniques have been tried and imple-
mented for visualizing 3D geospatial data. This
paper delineates some of the notable tools and
techniques that are employed in 3D geospatial
data visualization and briefly elaborates the basic
principles underlying the generation of static and
dynamic virtual environments.
A plethora of commercial software is available
for a wide range of purposes such as terrain or
building generation, photogrammetric measure-
217

ments, fly-through simulations, etc. Typically,
commercial software are designed for specific
application requirements with the objective of
saving time and costs. Even though such software
have been reasonably successful in accomplish-
ing desired tasks, they are not capable of being
extended to applications outside their intended
domain and therefore, are inherently limited. In
order to overcome such limitations and to attain
greater flexibility and efficiency, some research-
ers prefer the 3D virtual environments using a
variety of powerful programming languages.
Even though, initially, building the 3D worlds
from scratch might be time-consuming process,
with time and experience, using programming
languages to model becomes easier. Moreover,
the tremendous functionalities and flexibility of-
fered by different programming languages make
the final 3D virtual environments worth the ef-
fort. Some popular languages for writing code
to generate advanced 3D environments include
VRML, GeoVRML, X3D and OpenGL. VRML
is a specification designed to create interactive
3D worlds that can be used on the World Wide
Web. GeoVRML can be considered a derivative
of VRML because it is used to create geo-specific
applications. Of late, research involving the de-
velopment of 3D geospatial applications has been
gaining increasing significance, and the inherent
complexity in the efforts to visualize geographical
phenomena has necessitated the amalgamation
of the aforementioned tools with conventional
programming languages owing to the powerful
functionalities offered by the latter.
APPLICATIONS OF GEO-
VIRTUAL ENVIRONMENTS
Visualization facilitates not only presenting infor-
mation, but also enables seeing and understanding
hidden information among datasets. As mentioned
in the previous section, huge volumes of data are
available today and it is practically impossible
218
Virtual Environments for Geospatial Applications
to manually sift through these huge amounts of
data. Using visualization techniques, data can
be presented in a much more organized manner
that facilitates understanding the information that
may not otherwise be apparent. By proper use of
visualization tools, the same area can be viewed
at different scales, i.e., a small area in detail or
a bird’s eye view of a larger area. In order to see
the overall landscape of a whole country we need
to view the entire country at a glance. However,
the advantage of modern visualization is that
such visualizations are not mere depictions of
scenes, but also interactive environments capable
of animating the scenes, and simulating phenom-
ena. Another kind of information that needs to be
discussed in this context is associated informa-
tion. In order to understand these links among the
various components of a system, tools that can
reveal the various concurrent processes among the
various sub-systems are vital. Solutions to many
complex problems can be found by understanding
the relationships among the system components.
Urban planning authorities and town plan-
ners face several problems such as managing
water shortages, transportation problems, urban
housing and land use problems, natural and man-
made disasters, etc. Several of these problems
are mutually dependant and trying to solve them
in isolation will never lead to a permanent or
long-lasting solution. One of the foremost steps
in solving these problems is to get a bird’s eye
view of the problem scenario as a whole, while
simultaneously concentrating on the minutiae
(Figure 1). This kind of visualization is of im-
mense value to town and country planners and
urban infrastructure management in understanding
the link among the various components. Also, the
influences on the ambient environment as a result
of the aforementioned project can be studied by
means of the virtual settings. The visual impact
of new buildings and surroundings on each other
can be vividly seen on the screen.
Two main classes of models are discussed here,
namely static and dynamic models. The static 3D
Virtual Environments for Geospatial Applications

Figure 1. Geo-virtual environments for visualizing interlinked components of urban infrastructure.

Clockwise from top-left a. A Sample urban landscape, b. Land parcel, c. Road network, d. parcel with
proposed building layout

Table 1. A summary of applications of Geo-virtual Static Geo-Virtual Environments

environments
Static virtual environments are those that are
• Geospatial visualization and animation
composed of objects that lack animation capabili-
• Urban planning and Infrastructure applications
ties. Simply stated, the depiction of a 3D model
• Environmental process simulations
of a building is a static representation, while 3D
• Resource management and conservation
illustrations of phenomena such as debris flow
• Hosting cyber cities online
or volcanic activity involves dynamic represen-
• Imparting training and demonstration purposes
tation. Researchers use different approaches to
• Simulating applications such as mining, seismic processes,
etc.
model or create the 3D virtual worlds and depict
the constituents of those worlds. One very effi-
cient method of describing real-world scenarios
is the use of a hierarchical scene tree structure.
models are used for planning and design pur-
Especially, several 3D modeling languages (X3D,
poses, while the dynamic models are used for
VRML, and GeoVRML) model real-world ob-
simulation and training purposes. Such immersive
jects as shapes with geometry and appearance.
models are of immense value in the planning and
All features such as buildings, roads, trees, and
decision-making processes involving terrain and
rivers can be designed and modeled as shapes
building databases. Such models serve as valuable
which can be grouped together and transformed
tools in solving the previously mentioned urban
(translated or rotated).
infrastructure problems. The emphasis here is not
Objects in the real-world have positional
only on the utilization of virtual worlds for visu-
attributes and these are represented in the form
alization but also for simulation and animation.
of the x, y, and z coordinate points within the
Three-dimensional virtual environments are being
virtual world. The topological relationships
increasingly used for disaster mitigation and
among the constituent elements of a scene can
managing events such as debris flow, volcanic
be represented and the virtual representations can
eruption, seismic activities, etc.
be built to scale. The virtual world scene is built
within a virtual coordinate system, in which the
x and y axes are along the length and breadth of

219
 Virtual Environments for Geospatial Applications

the computer monitor and the z axes extends in
a direction perpendicular to the monitor. Hence,
theoretically, an object that is at a distance of 10
units along positive z-axis is in front of an object
at 5 units along positive z-axis. Also, the notion
of the parent-child relationship (Figure 2) implies
that smaller objects can be grouped to form larger
ones or objects higher in the hierarchy and these
are, in turn, grouped to form objects still higher
in the hierarchy.
Typically, various modeling approaches use
the notion of point, line, and polygon, as in GIS,
to build virtual worlds. All constituent elements
of a 3D scene can be built using fundamental
entities such as point, line, and polygon or faces.
Simple objects such as cuboids might be composed
of just 6 faces, while complex objects might ne-
cessitate a greater number of faces.
Photo-realistic environments can be built by
the judicious use of textures and by scaling them
to accurately match the faces. A vast number of
environmental and landscape applications gen-
erated using Virtual Reality (VR) or VRML are
available on the Web. Several researchers have
generated virtual environments for geospatial ap-
plications (e.g., Huang et al., 1999; Chandramouli
et al., 2004), resource management (e.g., ref:
SGILICGF, UW), 3D urban applications (e.g.,
Shiode, 2001), etc. For some applications a mere
representative environment depicting the relative

Figure 2. a) Scene tree structure; b) Parent-child relationship

Figure 3. a) A point in 3D space; b) Vertices of a polygon; c) Polygon with 4 sides

220
Virtual Environments for Geospatial Applications
positioning of various elements would suffice, but
others might necessitate photo-realistic rendering
of the 3D scene. Figure 4a (left) shows a simple 3D
environment showing a general layout of houses
within a community without much photo-realism.
On the other hand, Figure 4b (right) shows an
advanced infrastructure application wherein the
finished product, or final environment, is required
for viewing by planners in advance and hence the
virtual world is depicted with a higher degree of
photo-realism.
One advantage of the hierarchical mode of
representation, in which the scene elements are
modeled as individual objects that are grouped
using parent-child relationships, is that objects
can be built once and used any number of times
either in the same virtual scene or in other vir-
tual environments. This is known as the concept
Figure 4. a) Community planning with less photo-realism; b) Sample urban landscape with higher
degree of photorealism
Figure 5. Reusable software objects – can be generated once and used multiple times
of ‘reusable software objects’. For instance, a
lamp object or a tree object that has been created
for a particular project can be reused in another
application. Just like using a math function from
a C library, objects modeled within virtual envi-
ronments (Figure 5) as part of one project can be
referenced and included as part of another ap-
plication, thereby saving considerable program-
ming time and effort.
Dynamic Virtual Environments
As mentioned in the previous section, dynamic
virtual representations refer to those 3D worlds in
which the objects are incorporated with animation
capabilities. Quite frequently, research involving
the study of real-world geospatial processes and
phenomena, for instance seismic activity or hydro-
221

logical processes, necessitates that the virtual envi-
ronments possess dynamic animation capabilities.
The inherent nature of such activities makes on
site evaluation of such phenomena an extremely
hazardous task. Moreover, such disastrous events
might be triggered by a variety of factors, both
known and unknown, and hence, predicting their
occurrence accurately still remains a challenging
task. Hence, 3D geospatial environments offer an
efficient means of modeling such phenomena and
studying them at the comfort of an office desktop.
Virtual reality serves as an extremely potent and
flexible tool for applications involving simulation
and animation since it has functionalities capable
of being extended to achieve desired results.
Modeling languages such as GeoVRML, VRML,
and X3D provide additional functionalities known
as Scripts that can be programmed and executed
to create simple animations. However, these 3D
modeling languages lack advanced program-
ming functions to achieve curvilinear motion or
more sophisticated forms of animation. In order
to incorporate dynamic behavior in the virtual
world objects and replicate complex animation
patterns, an existing programming language with
advanced functionalities, for instance Java or
JavaScript, can be used in conjunction with the
3D modeling language. Typically, the objects in a
parent-child relationship within a virtual scene are
grouped into structures called nodes. In order to
achieve animation, these nodes must be made to
Figure 6. Flow of events within a virtual world wherein objects are grouped I the form of nodes
222
Virtual Environments for Geospatial Applications
sense user actions or events (Figure 6). In case of
simple actions such as the movement of an object
upon impact, there is a ‘trigger’ or a stimulation,
upon which something acts. In programming, the
‘triggering action’ is referred to as an event. Upon
the occurrence of a particular action, an object
or group of objects (node) behaves accordingly.
Events may not always be explicit or need not
only be ‘physical movement’. In programming
context, even the reaching of an instant of time,
say 11:59 pm, is considered an event. Hence, an
object or a group of objects (Parent node) can be
programmed to ‘sense’ the particular instant of
time and behave in a particular fashion upon
reaching that moment. The word ‘sense’ has been
highlighted as it is inevitable that an object or a
group of object recognize a change of condition
(state). Typically, there are three kinds of sensors
that are used in order to sense time changes, cur-
sor movement over the screen or objects, or the
field of view from a particular position within the
virtual world. These are respectively called time
sensors, touch sensors, and visibility sensors in
the context of virtual worlds (Figure 7). Upon
reaching a particular instant of time, or when the
mouse cursor is pointed on a particular object, or
when the viewer is within the virtual world is at
a specific position, the objects in the scene can
be programmed to behave in a desired fashion.
This behavior or response of objects might be a
Virtual Environments for Geospatial Applications

Figure 7. Diagram illustrating the sensing of actions of events by the objects and the corresponding
movement based on the flow of events (From Chandramouli et al., 2004, 2006)

translation along the x, y, or z axis or a rotatory
motion, or a combination of both.
Based on the above principles of animation,
dynamic geo-virtual environments can be gener-
ated to simulate various processes or geospatial
phenomena. Environmental processes and natural
disasters can be studied in detail by generating
dynamic virtual worlds. By programming the
properties and behavior of the constituent objects,
the 3D virtual environments can be used to study
various occurrences which are not easy to be
observed in real time, e.g., Figure 8 describes
flow events, seismic activities and the responses
of structures to such events.
Critical Issues in the Design and
Implementation of Geospatial Virtual
Environments
This section concisely discusses some limitations
and constraints typically experienced in several
virtual world generations as well. One noteworthy
issue is that in visualizing real-world scenarios,
there is an inevitable trade-off amid performance

Figure 8. a) 3D virtual environment depicting geospatial processes (1 picture of a series) such as land-
scape change over time etc.; b) 3D virtual environment depicting water flow in a reservoir

223

and resolution. Exploiting the complete capabili-
ties of virtual environments over the Web continues
to pose problems. As the number of objects in
a virtual environment increases, online hosting
becomes an issue as spontaneous rendering of
numerous objects is no easy task. Scenes with a
greater number of polygons decelerate the system
and make the interactivity poor. Several factors
need to be considered during visualization such
as the type and volume of data to be visualized,
memory constraints, and system performance.
Figure 9 presents a summary of the significant
issues concerning geo-virtual environments. In
their work on information visualization, Robertson
et al. (1993) have presented a terse compilation
of the important issues.
In his work on dynamic and interactive web-
based visualizations, Huang and Lin (1999, 2001,
and 2002) discuss in detail some of these concerns
and also address some critical issues concerning
online hosting of interactive visualizations. The
Java-3D based hybrid method that Huang and Lin
(1999, 2001) propose offers a standard framework
for visualizing dynamic environmental processes.
Figure 10 illustrates a 3-tier configuration that
Huang and Lin (1999) proposed in GeoVR. The
visualization server that is interlinked to the spa-
tial database accesses the geospatial information
Figure 9. Table showing a summary of critical issues in designing and implementing 3D virtual worlds
224
Virtual Environments for Geospatial Applications
from the data repository and the web server ac-
cesses the visualization server for 3D information.
This framework efficiently handles requests for
visualizing dynamic processes and based on the
client requests, the web server provides the ap-
propriate information in the conventional HTML
or 3D VRML format.
DISCUSSION AND CONCLUSION
Over the past several decades, information presen-
tation has inspired the development of several new
tools and techniques. The information revolution
has resulted in vast amounts of data that are far
too complex, both in quality and quantity, to be
handled by conventional tools and techniques. Re-
cent technological advances in the realm of remote
sensing have dramatically increased the amount
of geospatial data available. Virtual environments
are an efficient means of visualizing voluminous
geospatial data and are efficient in elucidating the
intricate patterns as well as hidden and associated
information. Such virtual environments facilitate
understanding of the complex relationships among
the various components of a multi-level scenario.
This paper discussed the design and imple-
mentation of virtual worlds that can be used to
Virtual Environments for Geospatial Applications
Figure 10. Online hosting of interactive visualization (From Huang et al.,1999)
generate both static representations depicting
real-world settings and dynamic representa-
tions that can simulate geospatial processes and
environmental phenomena. The paper discussed
the generation of such geo-virtual environments
with examples and provided explanations as to
how such geo-visualization applications facilitate
understanding of various geospatial phenomena
and environmental processes. The fundamental
principles underlying the generation of virtual
worlds, both static and dynamic, were elaborated
and the common issues involved in the generation
of such 3D virtual worlds were discussed. Further-
more, the issues related to the online hosting of
such virtual environments were tersely delineated
and possible solutions to frequently encountered
problems were provided.
REFERENCES
Ames, L. A., Nadeau, D. R., & Moreland, J. L.
(1996). VRML 2.0 Sourcebook.
Bonham-Carter, G. F. (1994). Geographic Infor-
mation Systems for Geoscientists: Modeling with
GIS. Pergemon: Oxford (p. 398).
Boyd, D. S., Lansdown, J., & Huxor, A. (1996).
The Design of Virtual Environments. SIMA.
Chandramouli, M., Huang, B., Yin Chou, T., Kun
Chung, L., & Wu, Q. (2006). Design and Imple-
mentation of Virtual Environments for planning
and Building Sustainable Railway Transit Systems,
COMPRAIL July 2006, Prague.
Chandramouli, M., Lan-Kun, C., Tien-Yin, C., &
vChing-Yi, K. (2004). Design and Implementation
of Virtual Environments for Visualizing 3D Geo-
spatial Data. TGIS Conference, Oct 28-29 2004.
Colin, W. (2000). Information Visualization:
Perception for Design. Morgan Kaufmann Series
in Interactive Technologies GeoVRML, (www.
geovrml.org)
Huang, B., Jiang, B., & Lin, H. (2001). An inte-
gration of GIS, virtual reality and the Internet for
spatial data exploration. International Journal of
GIS, 15(5), 439–456.
Huang, B., & Lin, H. (1999). GeoVR: A Web-
based tool for virtual reality presentation from
2D GIS data. Computers & Geosciences, 25(10),
1167–1175. doi:10.1016/S0098-3004(99)00073-4
225

Huang, B., & Lin, H. (2002). A Java/CGI approach
to developing a geographic virtual reality toolkit
on the Internet. Computers & Geosciences, 28(1),
13–19. doi:10.1016/S0098-3004(01)00008-5
Karel, C., & Jiri, Z. (n/d). Using VRML for creat-
ing interactive demonstrations of physical models.
Department of computer science and Engineering.
Czech Technical University.
Robertson, G., Card, S., & Mackinlay, J. D. (1993).
Information visualization using 3d interactive
animation. Communications of the ACM, 36,
57–71. doi:10.1145/255950.153577
SGILICGF. UW Sea Grant Institute and Land
Information and Computer Graphics Facility
http://coastal.lic.wisc.edu/visualization/Visual-
ization.htm
Shiode, N. (2001). 3D urban models: recent
developments in the digital modeling of urban
environments in three-dimensions. GeoJournal,
52(3), 263–269. doi:10.1023/A:1014276309416
Sutherland, I. E. (1965). The ultimate display. In
the proceedings of the IFIPS Congress, 2, 506-
508. New York City, NY.
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 320-331,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
226
Virtual Environments for Geospatial Applications
KEY TERMS AND DEFINITIONS
Immersion: A Sense of being present within
the virtual world and a ‘sense’ being able to visu-
alize objects by being amidst their surroundings
and navigating through the world.
Node: An entity within the hierarchical scene
structure that represents a group of objects.
OpenSource: Source code or computer soft-
ware that is freely offered and is available to the
public for building software applications.
Scene-Hierarchy: The organization of the
elements of a 3D virtual scene into successive
levels, in such a way that the object under which
other objects are grouped is called the parent and
the grouped objects are called its children. When
a parent object is transformed, the children are
also transformed.
SCRIPT: Program scripts that are used to
perform calculations and return values to the
calling programs.
Transformation: Operations such as transla-
tion, rotation, or scaling involving objects in a
virtual environment.
Virtual Reality: A three-dimensional visual
immersive setting that facilitates user to navigate
within the scene and perform operations in real
time.
 Section 2
Development and Design
Methodologies

This section provides in-depth coverage of conceptual architecture frameworks to provide the reader with a com-
prehensive understanding of the emerging developments within the field of Geographic Information Systems.
Research fundamentals imperative to the understanding of developmental processes within Geographic Informa-
tion Systems are offered. From broad examinations to specific discussions on methodology, the research found
within this section spans the discipline while offering detailed, specific discussions. From basic designs to abstract
development, these chapters serve to expand the reaches of development and design technologies within the Geo-
graphic Information Systems community. This section includes 15 contributions from researchers throughout the
world on the topic of Geographic Information Systems.
228

Chapter 16
GIS Technique for
Territorial Analysis:
Spatial MultiCriteria Decision Analysis

Francesco Riccioli
University of Florence, Italy

Toufic El Asmar
University of Florence, Italy

ABSTRACT
Territorial analysis is related to the complexity of an area where human activities and environmental
characteristics are the main development factors. Human activities have a direct (e.g. farm activities)
and indirect (e.g. urban development/sprawl) influence on a territory and its environment, which gives an
important function to such activities when strategies for territorial planning have to be defined. Several
actors are involved in the assessment of a territory, which is a complicated feature of decision making
to be achieved with the use of complex data-analysis process. The evolution of personal computer and
specific software for the analysis of spatial data is giving important contribution to “Territorial Planning
Strategies.” Spatial Decision Support System is an example of how problems occurred during a deci-
sion making process can be solved using methodologies that combine “Multicriteria Approach” and
“Geographic Information System.” One of these methodologies is represented by “Spatial Multicriteria
Decision Analysis.”

DECISION SUPPORT FOR
TERRITORIAL ANALYSIS
It is possible to explain the complexity of a terri-
tory as a mosaic of interactions between human
activities and its surrounding environment. Dur-
DOI: 10.4018/978-1-4666-2038-4.ch016
ing the last century anthropic expansion and the
consequent urban sprawl produced a great loss of
natural resources therefore “Territorial Planning
Strategies” became an important tool to attain
decision makers targets. Many variables are
concerned in land-use changes, and many factors
(human, political, economic) are involved in the
urbanization growth, thus through the application

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
GIS Technique for Territorial Analysis
of a “decision support analysis” is an important
and complicated strategy for a correct territorial
planning.
Territorial planning is based on decision: a
decision is a choice between alternatives. The
alternatives may represent different courses of
action, different hypotheses about the character
of a feature, different classifications, and so on.
With rapid increases in population and con-
tinuing expectations of growth in the standard
of living, pressures on natural resource use have
become intense. For the resource manager, the task
of effective resource allocation has thus become
especially difficult. Clear choices are few and the
increasing use of more marginal lands puts one
face-to-face with a broad range of uncertainties.
Add to this a very dynamic environment subject
to substantial and complex impacts from human
intervention, and one has the ingredients for a
decision making process that is dominated by
uncertainty and consequent risk for the decision
maker.
Uncertainty is inevitable in the decision mak-
ing process. Uncertainty includes any known or
unknown error, ambiguity or variation in both the
database and the decision rule. Thus, uncertainty
may arise from such elements as measurement
error, inherent variability, instability, conceptual
ambiguity, over-abstraction, or simple ignorance
of important model parameters.
Managing the environment implies dealing
with dynamic systems that are only partly under-
stood, with multiple interests and multiple actors,
with long-term implications that range from the
local to the global scale. “It has become more
and more difficult to see the world around us in
a unidimensional way and to use only a single
criterion when judging what we see. We always
compare, rank, and order the objects of our choice
with respect to criteria of choice. But only in a
very simple, straightforward, or routine situation
can we assume that a single criterion of choice
will be fully satisfactory” (Zeleny, 1982). Suc-
cessful decision making depends on the quality
and quantity of information available to decision
makers. In general, two types of information
are used in the spatial decision-making process:
geographical information and information related
to decision maker’s preferences. The amount
and type of information required depend on the
nature of a particulate decision problem. Several
characteristics of spatial decision problems can be
singled out. Spatial problems typically involve a
large number of alternatives that are evaluated on
the basis of multiple criteria. Some of the criteria
may be qualitative while others may be quantita-
tive. There may be more than one decision maker
and they may have different preferences with
respect to the relative importance of the evaluation
criteria and the decision consequences. The spatial
problem may involve uncertainty and imprecision
that make it difficult to predict the outcome of
particular decisions. Given these characteristics,
many real-world spatial decision problems give
rise to Geographical Information System (GIS)
based multicriteria decision making.
Decision Theory is concerned with the logic by
which one arrives at a choice between alternatives.
What those alternatives are varies from problem
to problem. They might be alternative actions,
alternative hypotheses about a phenomenon, alter-
native objects to include in a set, and so on. In the
context of GIS it is useful to distinguish between
policy decisions and resource allocation decisions.
The latter involves decisions that directly affect
the utilization of resources (e.g., land) while the
former is only intended to influence the decision
behavior of others who will in turn make resource
commitments. “Resource allocation decisions are
also prime candidates for analysis with a GIS.
Indeed, land evaluation and allocation is one
of the most fundamental activities of resource
development” (FAO, 1976).
One of several GIS application tools is based
on territorial analysis with the integration of a
decision support methodology that is the “Spatial
Decision Support System” (SDSS).
229

SDSS provides complex data elaborations
that are useful for planning process, in fact as
referred by Malczewski (2004): “An DSS aims
to improve the effectiveness of decision making
by incorporating judgments and computer-based
programs within the decision making process.
The system should support a variety of decision
making styles that may be present in a particular
problem solving process. The purpose of such
a system is to support a user in achieving better
decision.”
Spatial Decision Support Systems are com-
monly based on 4 utilities that are: data input,
data storage, data analysis and data output. All of
these utilities are focused on the support of deci-
sion making procedures. Evidently GIS play an
important role in an SDSS process with the aim
to store and elaborate quantitative and qualitative
data or information. GIS help users to resolve their
decision making problems through a participa-
tory approach in order to incorporate decision
making analytical techniques into it. GIS-based
approaches to territorial analysis are mainly rep-
resented by Artificial Intelligence methods like
neural network models or cellular automata and
Spatial Multicriteria Decision analysis. Previous
applications.”..can help in modeling and describ-
ing complex systems for inference and decision
making” nevertheless “the major limitation of
these methods is their “black box” style of analyz-
ing spatial problems” (Malczewski, 2004).
Spatial Multicriteria Decision analysis com-
bine Multicriteria Analysis and Geographical
Information System with the aim of converting
input data (spatial and aspatial data) in output
data (resultant decision). Conventional MCDM
techniques have largely been aspatial in the sense
that they assume a spatial homogeneity within
the study area. This assumption is unrealistic in
many decision situations because the evaluation
criteria vary across space. Consequently, there is a
need for an explicit representation of geographical
dimension in MCDM. As Ascough et al. (2002)
referred, Spatial Multicriteria Decision Analysis
230
GIS Technique for Territorial Analysis
are “far from maturity” and “continued changes
in computer hardware and software technology
are fundamentally altering the way that decision
makers, stakeholders, policy makers, and analysts
interact with computers.”
This chapter is organised in two parts or under
– chapters. In the first part a general overview
of Multicriteria methodology and Geographi-
cal Information System is described, whilst the
second part will consider “Spatial Multicriteria
Decision Analysis” with the presentation of a
specific case study.
MULTICRITERIA ANALYSIS AND
GEOGRAPHICAL INFORMATION
SYSTEM: A BRIEF INTRODUCTION
Multicriteria Decision Analysis (MCDM) is a
methodology commonly used to support decision
makers that have to face numerous and conflict-
ing evaluations. MCDM aims at highlighting
these conflicts and deriving a way to come to a
compromise in a transparent process. MCDM is
based on two processes: Multi-Objective Decision
Making (MODM) and Multi-Attribute Decision
Making (MADM).
MODM studies decision-making problems
in which decisional space is continuous, while
MADM focus on problems with discrete decisional
spaces. Both MODM and MADM problems are
assessed by a single decision-maker or group of
several decision-makers, that are distinguished in
the number of involved persons within a decision
analysis process (goals, preferences and beliefs).
The decisions are classified in deterministic,
probabilistic and fuzzy decisions. In the case of
deterministic decision process, required data and
information are well known. During probabilistic
analysis decisional problems made under uncer-
tainty are deal randomly. Fuzzy decision analysis
consider the inherent imprecision of information
involved in decision making.
GIS Technique for Territorial Analysis
Broadly speaking MCDM is one of the most
well known branch of decision-making and can
be classified as a decision support method which
aim at seeking the best solution1.
In general, MCDM problems involves six
components:
1. A goal or a set of goals the decision maker
attempts to achieve;
2. The decision maker or group of decision
makers involved in the decision-making
process along with their preferences with
respect to evaluation criteria (consist of a
single person or a group of people, such as
government or corporate organizations) ;
3. A set of evaluation criteria (called also
objectives and/or attributes) on the basis of
which the decision makers evaluate alterna-
tive courses of action;
4. The set of decision alternatives (the decision
or action variables);
5. The set of uncontrollable variables or states
of nature;
6. The set of outcomes or consequences as-
sociated with each alternative-attribute pair.
Figure 1. MCDM matrix
In literature we find several definitions of
Multicriteria Analysis, Figure 1 shows Pitz and
McKillip, (1984) scheme (one goal, one decision
maker, three objectives).
The columns consist of a level representing
the decision makers, their preferences and evalu-
ation criteria. These elements are organized in a
hierarchical structure. The most general level is
represented by goal and at this level a desired end
state resulting from decision-making activity is
specified. The decision require analysis of the
values of person affected by the decision, who
are often characterized by unique preferences with
respect to the relative importance of criteria on
the basis of which alternative decisions are
evaluated. The preferences are typically opera-
tionalized in terms of weights assigned to the
evaluation criteria.
The rows of MCDM matrix represent decision
alternatives. All decision are made in some kind of
environmental context and therefore involve many
factors beyond the control of the decision maker.
The uncontrollable factors reflect the degree of
uncertainty about decision-making processes
and outcomes (consequences). Therefore, for
each decision alternative there is a set of possible
outcomes: the decision outcomes depend on the
231

set of attributes for evaluating alternatives, so a
particular intersection of a row and column is a
specific (decision) outcome associated with a
particular alternative and attribute. The decision
problem requires that the set of outcomes are or-
dered so that the best alternative can be identified.
MCDM techniques commonly apply numerical
analysis to a performance matrix in two stages:
scoring and weighting.
• Scoring: the expected consequences of
each alternative are assigned a numerical
score on a strength of preference scale for
each alternative for each criterion. More
preferred alternatives score higher on the
scale, and less preferred alternatives score
lower. In practice, scales extending from 0
to 100 are often used, where 0 represents a
real or hypothetical least preferred alterna-
tive, and 100 is associated with a real or
hypothetical most preferred alternative.
All alternatives considered in the MCDM
would then fall between 0 and 1002.
• Weighting: numerical weights are as-
signed to define, for each criterion, the
relative valuations of a shift between the
top and bottom of the chosen scale. A key
feature of MCDM is its emphasis on the
judgement of the decision making team, in
establishing objectives and criteria, esti-
mating relative importance weights and, to
some extent, in judging the contribution of
each alternative to each performance cri-
terion. The subjectivity that pervades this
can be a matter of concern. Its foundation,
in principle, is the decision makers’ own
choices of objectives, criteria, weights and
assessments of achieving the objectives,
although ‘objective’ data such as observed
prices can also be included.
Mathematical routines, which may be written
into computer programmes, then combine these
two components (score and weight) to give an
232
GIS Technique for Territorial Analysis
overall assessment of each alternative being ap-
praised.
Reference literature (Nijkamp 1976, Narisim-
han 1980, Romero and Amador 1987) indicates
several methods for matrix solution that are based
on Multi-Objective Programming, Goal Program-
ming and Compromise Programming.
These approaches are often referred as “com-
pensatory concept”: since low scores of one
criterion may be compensated by high scores
of another. The most common way to combine
scores on criteria, and relevant weights between
criteria, is to calculate a simple weighted average
of scores. Use of such weighted averages depends
on the assumption of mutual independence of
preferences. This means that the judged strength
of preference for an alternative on one criterion
will be independent of its judged strength of
preference on another.
An initial step can be to see if any of the al-
ternatives are dominated by others. Dominance
occurs when one alternative performs at least as
well as another on all criteria and strictly better
than the other on at least one criterion. In principle,
one alternative might dominate all others, but in
practice this is unlikely. When it does occur, it is
helpful to ask if there is some advantage of the
dominated alternative that is not represented by
the criteria; this may reveal new criteria that have
been overlooked. Dominance is more likely just
to enable the decision-making team to eliminate
dominated alternatives from further consideration.
Once any dominance analysis has been concluded,
the next stage is for the decision-making team to
determine whether trade-offs between different
criteria are acceptable, so that good performance
on one criterion can in principle compensate for
weaker performance on another. Most public deci-
sions admit such trade-offs, but there may be some
circumstances, perhaps where ethical issues are
central, where trade-offs of this type are not accept-
able. If it is not acceptable to consider trade-offs
between criteria, then there are a limited number of
non-compensatory MCDM techniques available.
GIS Technique for Territorial Analysis
Where compensation is acceptable, most MCDM
methods involve implicit or explicit aggregation
of each alternative’s performance across all the
criteria to form an overall assessment of each
alternative, on the basis of which the set of alter-
natives can be compared. The principal difference
between the main families of MCDM methods is
the way in which this aggregation is done.
In general the aim of matrix solution is fo-
cused on the ideal solution choice, which does
not coincide with the maximum values that can
be obtained by Mathematical programming
viewpoint. The ideal solution definition is based
on Pareto efficiency concept “Given a set of al-
ternative allocations of goods or outcomes for a
set of individuals, a change from one allocation
to another that can make at least one individual
better off without making any other individual
worse off” (Pareto, 1906).
The matrix solution implies a compromise
solution that is represented by the alternative with
minor distance to the ideal point (solution). So,
given a goal and n objectives, the alternative with
minor distance to ideal solution will be considered
the best alternative. It is possible to quantify the
distance of n alternative from ideal solution using
Cartesian representation of each alternative: the
distance is based on p parameter3.
A parallel can be found in the study of geom-
etry. Simple principles like the shortest distance
between two points is a straight line (Formula 1)
are combined using the rules of logic to prove
theorems that are not obvious, like the Pythagorean
principle, that the square of the hypotenuse equals
the sum of the squares of the other two sides.
Another two way for calculating the distance
from alternative and ideal point, are showed in
Formula 2 and 3.
1
 n 22
L2 =  ∑ x i 1 − x i 2  (1)
 i =1 
 
where: L2 = Euclidean distance of alternative from
ideal solution (p = 2); xi1 xi2= Cartesian coordinates
of alternative i
 n 
L1 =  ∑ x i 1 − x i 2  (2)
 i =1 
 
where: L1 = Manhattan distance of alternative
from ideal solution (p = 1); xi1 xi2= Cartesian
coordinates of alternative i
 n 
1
p p
L∞ = lim  ∑ x i 1 − x i 2  (3)
 i =1 
 
with lim p → ∞where: L∞ = Infinite metrics
distance of alternative from ideal solution (p =
∞); xi1 xi2= Cartesian coordinates of alternative i
Formula 1 and 2 are considered compensatory
methods because low scores on one criterion may
be compensated by high scores on another. For-
mula 3 is considered no compensatory method that
does not permit trade-offs between criteria (i.e. low
values for an alternative on one criterion cannot
be offset by better values on another criterion).
Using different solution methods can produce
different results. “At this point one may reasonably
claim that there may never be a single MCDM
method that can ensure that the derived ranking
of the alternatives is the correct one. The prob-
lem of determining the best alternative is an ill
defined one by its own nature. It is hard to know
the best solution even under perfect knowledge
of input data. What can be done, however, is to
analyze the behaviour of numerical methods un-
der certain evaluative criteria that are motivated
by widely accepted notions of logical stability”
(Triantaphyllou 2000).
233

GEOGRAPHICAL
INFORMATION SYSTEM
Geographic Information System (GIS)4 is based
on a set of hardware, software an human skills
that have the aim of processing, analysing and
creating graphical representation of a database
that is associated to a territory.
Since 1960 GIS were performed in Canada and
United States for military purposes, later (since
1980), thanks to personal computer diffusion, they
become available for private users.
It is possible to divide a Geographic Informa-
tion System in three processing methods: data
acquisition, data analysis and data results.
Data acquisition is performed by different type5
of photo or cartographic and scanned maps, land
use rendering. Each acquired item is linked to a
numeric database. This database correspond to a
numeric representation of different spatial items
like land use, slope degree, road and cities location,
etc. Database can be composed by categorical
data (nominal data with no inherent ordering like
land use types, country names or ordinal data with
inherent order like road class or stream class) or
numerical data (like temperature degree or rainfall
measurement). Spatial items can have different
representations like unbounded areas (land use),
bounded areas (cities, country boundaries), net-
works (roads or transmission lines) etc.
All spatial items are related to a specific pro-
jection, scale and resolution. Specific projection
is the related to Earth representation through X,Y
coordinates on a 2D map surface; the scale is
the ratio of distance on a 2D map respect to the
equivalent distance in real world, while resolu-
tion is the dimension of the smallest spatial item
which can be recognized.
There are two different layer representations:
the vector layer and the raster layer. Vector layer
is based on lines, points and polygons for items
representation, and raster layer are pixel grid that
represent items in a map.
234
GIS Technique for Territorial Analysis
Each layer is created by alphanumeric data
and it is possible to elaborate it using several
operations: measurement, reclassification, buffer-
ing operations that are the fundamental utilities
of a GIS.
Measurement utility enables the calculation
of points, lines, polygons or pixels to estimate
areas, volumes, distance or other. Reclassification
function transform the attribute data associated
with a specific “geographic primitives” or pixels.
Buffering operation gives specific values to an area
surrounding a geographical primitives or pixels.
To make map making easy, a GIS system
should include a variety of standard map compo-
nents that can be readily added to a layout. These
include map grids, scale bars, legends, annotation
text, and a means of mixing georeferenced and
ungeoreferenced groups (north arrows, company
logos) to complete the map. Each of these map
components should be easily customizable; for
example, with map grids you should be able to
control the size and color of the text and lines, the
grid spacing, the components of the grid, and so
on. Theme mapping is a powerful means to look
at quantitative trends and classes of attributes as-
sociated with vector and raster elements. Theme
maps provide a ready means for visualization
of this information without the need for you to
design the scripts that identify the classes and
assign meaningful styles. Theme mapping treats
the values in a designated field statistically and
assigns styles to the classes identified according
to the parameters you specify.
One very important feature of an advanced GIS
system is that it continues to evolve and incorpo-
rate new technologies. In territorial planning and
decision making strategies previous techniques are
increased by introducing decision support tools
for instance statistical, mathematical modeling
and map algebra overlay tools.
Statistical tools are mainly performed by Mul-
tivariate analysis and time series analysis. Former
is based on Cluster analysis and discriminant
analysis that “are used to explore the similarities
GIS Technique for Territorial Analysis
between objects and to define groups of objects
by considering simultaneously all variables that
have been measured” (Griffith and Amrhein 1991;
Mather, 1991). Time series analysis “is a set of
observations measured at successive points in time
or over successive periods of time” (Malcewski,
1999). Mathematical modeling tools are mainly
performed by the optimization approach and the
simulation approach. The first approach is based
on finding the best solution of a well defined
management problem. There are several type of
optimization process, like linear programming,
integer programming, dynamic programming
etc. Simulation approach is “a methodology for
performing experiments using a model of the
real world system” (Mather, 1991). Map Algebra
operations perform overlay process of different
layers (each geographical primitives or pixels in
a specific layer is expressed by numbers): the
final layer is the aggregation of some layers using
different type of operators. A list of most typical
operators is showed below.
Algebraic operators:
• Addiction: territorial entity value of lay-
er A is added to territorial entity value of
layer B;
• Subtraction: territorial entity value of
layer A is subtracted from territorial entity
value of layer B;
• Multiplication: territorial entity value of
layer A is multiplied by territorial entity
value of layer B;
• Division: territorial entity value of layer
A is divided by territorial entity value of
layer B;
• Average: the average value of territorial
entity value of layer A and correspond-
ing territorial entity value of layer B is
calculated;
• Maximum: the maximum value between
territorial entity value of layer A and cor-
responding territorial entity value of layer
B is calculated;
• Minimum: the result is the minimum val-
ue between territorial entity value of layer
A and corresponding territorial entity value
of layer B is calculated.
Logical operators:
The logical operators are based on Boolean
algebra and performed on binary variables (0,1
or true, false):
• Intersection: logical AND (algebraic
multiplication);
• Union: logical OR (algebraic addiction);
• Complement logical NOT (inversion).
Fuzzy operators6:
• Intersection: membership value of x to
both territorial entity values of layer A and
B:
μA (x) AND μB (x) (MIN[μA (x), μB (x)]);
• Union: membership value of x to territo-
rial entity value of layer A or territorial en-
tity value of layer B:
μA (x) OR μB (x) = (MAX[μA (x), μB (x)]);
In conclusion, GIS has had an enormous impact
on virtually every field that manages and analyzes
spatially distributed data. Investing in GIS requires
more than an investment in hardware and software.
Indeed, in many instances this is the least issue
of concern. Most would also recognize that a
substantial investment needs to be placed in the
development of the database. However, one of the
least recognized yet most important investments
is in the analysts who will use the system. The
system and the analyst cannot be separated, one
is simply an extension of the other. In addition,
235

the process of incorporating GIS capabilities into
an institution requires an investment in long-term
and organization-wide education and training.
In many ways, learning GIS involves learning
to think, learning to think about patterns, about
space and about processes that act in space. Most
procedures have many more general applications
and can be used in many novel and innovative ways
SPATIAL MULTICRITERIA
DECISION ANALYSIS
Decisions may be characterized as single- or
multi-objective in nature, based on either a
single criterion or multiple criteria. While one
is occasionally concerned with single criterion
problems, most problems approached with a GIS
are multicriteria in nature. For example, we might
wish to identify areas of concern for soil erosion
on the basis of slope, land-use, soil type and the
like. In these instances, our concern lies with how
to combine these criteria to arrive at a composite
decision. As a consequence, the first major area
of concern in GIS with regard to Decision Theory
is MultiCriteria Evaluation.
“Earlier, the multi-criteria land suitability was
assessed more non-spatially, assuming the spatial
homogeneity over the area under consideration.
This, however, is unrealistic in cases like land suit-
ability studies, where decisions are made using
criteria which vary across in space” (Malczewski
1999). Spatial MultiCriteria Decision Support can
be used to overcome previous limit. It is an SDSS
methodology that combines Multicriteria analysis
and Geographical Information System. Starting
from available resources (input data), this meth-
odology uses classic overlay operations through
specific decisional process (decision rules) with
the aim of achieving Decision Makers goals (out-
put data). “GIS-based-MCDM can be thought of
as a process that combines and transforms spatial
and aspatial data (input) into a resultant decision
(output). The MCDM procedures (or decision
236
GIS Technique for Territorial Analysis
rules) define a relationship between the input
maps and the output map” (Malczewski, 2004).
Multiobjective and multiattribute decision
making represent two mainly Spatial Multicriteria
Decision Analysis methods. First one is mathemat-
ical programming model oriented method “where
the alternatives must be generated” (Keeney R.L.,
Raiffa H., 1976). Second one is mathematical
programming data oriented method where the
alternatives is given explicitly: this method is
used with the aim of solving our case study that
will be illustrated in six boxes during this section.
The structure of Spatial MultiCriteria Decision
Analysis is divided into the 5 following defini-
tion points:
1. decisional problem;
2. evaluation criteria;
3. constrains and alternatives;
4. decisional rules;
5. territorial analysis results.
It is possible to combine previous points and
MultiCriteria matrix: the result is showed in
Figure 2.
Decisional Problem Definition
Decisional problem definition is prevalently
related to territorial planning objectives that are
represented by the difference between the ideal
and the existing situation: this difference is the
core of decision makers targets.
Frequently (in literature) territorial analysis is
based on the evaluation of three main aspects of
a generic territory that represent the interaction
between human activities and environment:
• economic aspects;
• social aspects;
• natural aspects.
Territorial planning can be focus on suitable
areas for urban development or suitable land use
GIS Technique for Territorial Analysis
Figure 2. Spatial Multicriteria Decision Analysis structure
for some agricultural or forest activities. “Land-
use planning is sometimes misunderstood as
being a process where planners tell people what
to do. Land-use planning means the systematic
assessment of physical, social and economic
factors in such a way as to encourage and assist
land users in selecting options that increase their
productivity, are sustainable and meet the needs
of society” (FAO, 1993).
Generally numeric data are used for decisional
problem quantification and some objectives are
used for the solution of a decisional problem (Box
1). “An objective is a statement about the desired
state of geographical system under consideration.
It indicates the directions of improvement of one
or more criteria. In the context of decision analy-
sis, the criteria can be thought of as indicators of
future outcomes (outcome measures) rather than
past achievements” (Starr and Zeleny, 1977).
Box 1. Decisional problem definition (Goal)
Definition of suitable areas for forest activities (harvesting).
Goal is related to following objectives:
1. to minimize the costs of forest harvesting
2. to maximize the forest profit.
Evaluation Criteria Definition
The second decisional process phase is based
on definition of evaluation criteria (Box 2). In
GIS analysis Evaluation criteria are qualitative
or quantitative factors used to define the area
characteristics. They are represented by criterion
maps (raster or vectorial layers) like slope degree
layer, basin roads layer, etc. Each criterion map
can be obtained through fundamental GIS func-
tionalities (i.e. measurement, reclassification,
buffering operations).
Frequently criteria are represented by in-
creasing or decreasing values in a qualitative or
quantitative scales. Former is used in order to
identify numbers values with indirect mathemati-
cal correlation. For example they can be used for
land use classification (i.e number 1 is related to
conifer forest, number 2 is related to croplands
and number 3 is related to vineyards. In this case
all numbers have the same importance value).
Latter is used in order to identify numbers values
with direct mathematical correlation: for example
they can be used for forest production calculation
(i.e. Canada coniferous forest produce n m3 per
237

hectare per year, while Italian coniferous forest
produce n-1 m3 per hectare per year. In this case
n m3 value is higher than n-1 m3 value).
During this phase is important to compare
different evaluation criteria: the problem is rep-
resented by different unit of measure (i.e. it is
impossible to related m3 and kilometres per hour
or hectares and litres).
A standard measure scales must to be used
with the aim of confrontation of evaluation crite-
ria. A normalization process can be used for this
purpose. Following Formula (4-7) show different
normalization methodologies.
v jmax − v j
vn = (4)
v jmax − v jmin
where: vnj = normalized value of criteria j; vjmax =
maximum value of criteria j; vj = value of criteria
j; vjmin = minimum value of criteria j
v j − v jmin
v nj = (5)
v jmax − v jmin
where: vnj = normalized value of criteria j; vj =
max
maximum value of criteria j; vj = value of criteria
j; vjmin = minimum value of criteria j
vj
v nj = (6)
v jmax
where: vnj = normalized value of criteria j; vjmax =
maximum value of criteria j; vj = value of criteria j
v jmin
v nj = 1 − (7)
vj
where: vnj = normalized value of criteria j; vjmin =
minimum value of criteria j; vj = value of criteria j
238
GIS Technique for Territorial Analysis
Another normalization methodology fre-
quently used under uncertainty scenarios is based
on Fuzzy logic (Zadeh, 1965). “The inability
of the normal decision making methods to ad-
dress the imprecision and uncertainty paved the
path for the fuzzy decision making techniques.
Goals, constraints and consequences are known
imprecisely in much of the real world decision-
making processes and in such a situation fuzzy
set theory becomes functional” (Bellman R. E.,
Zadeh L. A., 1970).
In Fuzzy logic theory an element x (criteria)
has membership value in real unit interval [0,1]
related to an y set. Membership values in vaguely
defined sets are an extension of classic number
binary range {0,1}. An element x value close to
1 is considered like an high membership value
in defined set, while an element x value close to
0 is considered like a low membership value in
defined set. Formula 8 shows a membership value
function in a generic set. The function derived set
is called fuzzy set.
F = {(x, μ(x))|x X} (8)
where: F = fuzzy set; X = generic set; x = criteria
value; μ(x) = membership value of criteria
Most common fuzzy normalizations are related
to linear, j-shaped and sigmoidal normalization
functions. The y axes represent membership value
of criteria μ(x) real unit interval [0,1] while a, b,
c and d represent inflection points (called also
control points) of the curve for increasing function
normalization, decreasing function normalization
and symmetrical function normalization.
Figure 3 represent an example of member-
ship value of criteria μ(x) to high slope soils set.
The high slope soils set is between 50% (slope
percentage maximum membership value) and 5%
(slope percentage minimum membership value).
μ(x) =1: it identifies high slope percentage soils
(50% slope percentage)
GIS Technique for Territorial Analysis
Figure 3. Membership value of criteria (linear function)
μ(x) = 0,7: it identifies medium-high percentage
slope soils (35% slope degree)
μ(x) = 0: it identifies almost flat soils (5% slope
degree)
Box 2. Evaluation criteria definition
Suitable areas for forest harvesting are identified by areas with
low labour costs (minimize costs of forest harvesting objec-
tive) and areas with high forest production (maximize forest
profit). So evaluation criteria are based on following layers:
1. basin roads layer
2. forest production layer
First layer is used with the aim of minimize costs objective,
while second layer is used with purpose of maximize the profit
objective. A Fuzzy linear normalization process is used using
following parameters
1. basin road layer-linear decreasing- control point 0, 7 where
7 is the maximum distance from roads expressed in kilometres
2. forest production layer-linear increasing-control point 0, 324
where 324 is the maximum forest production expressed in m3/
ha
Unit of measures of each criteria (kilometres and cubic metres
per hectare) are transformed in 0-2557 range values. Areas
close to roads will have normalized values close to 255 (0
Kilometres), and good forest production areas will have nor-
malized values close to 255 (324 m3/ha).
Constrains and Alternatives
Definition
“Constrains are limitations imposed by nature or
by human beings that do not permit certain actions
to be taken” (Keeney, 1980). The nature constrains
are usually related to orographic characteristics
while human constraints are related to planning
strategies, budget constraints, etc. Constrains are
generally valued only by binary values (0 or 1)
of their outcomes. They are often represented by
Boolean maps (0,1).
Alternatives represent the choices of stakehold-
ers and they are valued through numeric values
of outcomes that compose themselves. In GIS
structure each alternative is represented by pixels
(in raster models) or geographical primitives (in
vector models). Geographical primitives can be
grouped into two types of alternatives: simple and
complex. Former are represented by single type of
object like points, lines and polygons, while latter
are represented by interaction between objects like
239

Box 3. Constrains and alternatives definition
Constrains (Figure 4): suitable areas must to be covered by
forest and they must to be included into case study boundar-
ies (dark coloured pixels show forest areas into case study
boundaries).
Alternatives are represented by pixel values of criterion raster
maps where dark coloured pixels show best areas location for
forest production criteria (Figure 5) and dark coloured pixels
show areas close to roads (Figure 6).
line-line, line-polygon, point-line etc. Given objec-
tive n, the stakeholders decision focus on the best
location finding (alternative). Alternatives values
are related to outcomes value that can be binary
values (0 or 1), discrete values (finite values) or
continuous values (infinite values). Thematic
maps often represent them (Box 3).
Decision Rules
The procedure by which criteria are selected and
combined to arrive at a particular evaluation, and
by which evaluations are compared and acted
upon, is known as a decision rule. Decision rules
typically contain procedures for combining criteria
into a single composite index and a statement of
Figure 4. Forest areas map
240
GIS Technique for Territorial Analysis
how alternatives are to be compared using this
index (Box 4).
Broadly speaking evaluation criteria are ag-
gregated with multicriteria operators for choose
the best alternative with the aim of the objectives
definition.
In GIS models it is possible to do it using
overlay process with compensatory, no compensa-
tory operators. During recent years also partially
compensatory operators are used in territorial
analysis.
Weighted Linear Combination (WLC) is typi-
cal compensatory operator through which alterna-
tive value is defined as the average of each criteria
value multiplied by the weights value (Formula 9).
It is important the role of criteria weights:
they are assigned to given criteria to indicate its
relative importance according to the preferences
of decision makers. They are called importance
weights.
n
∑c ij
⋅ pi
j =1
Vj = (9)
n
GIS Technique for Territorial Analysis

Figure 5. Forest production map

where: Vj = alternative j value; cij = criteria i
value relating to alternative j; pi = weight value
of criteria i (∑pi = 1); n = total criteria number
This is a compensatory method because an
high alternative value hides (compensates) a low
alternative value and a low alternative value hides
an high alternative value back.
Minimum operator (Min) is a no compensatory
method through which alternative value is defined

Figure 6. Distance from roads map

241

by minimum value of each criteria multiplied by
the weights value (Formula 10).
Vj = Min (cij) ∙ pi (10)
where: Vj = alternative j value; cij = criteria i
value relating to alternative j; pi = weight value
of criteria i (∑pi = 1)
Maximum operator (MAX) is also a no com-
pensatory method through which alternative value
is defined by maximum value of each criteria
multiplied by the weights value (Formula 11).
Vj = MAX (cij) ∙ pi (11)
where: Vj = alternative j value; cij = criteria i
value relating to alternative j; pi = weight value
of criteria i (∑pi = 1)
Formula 10 and 11 show no compensatory
methods because the final criteria value is based
on highest alternative value (MAX operator) or
lowest alternative value (Min operator).
Ordered Weighted Average (OWA) (Eastman
and Jiang, 1996). This method offers a complete
spectrum of decision strategies along the primary
dimensions of degree of tradeoff involved and
degree of risk in the solution.
Ordered Weighted Average is a partially com-
pensatory operator and it can be considerate the
intermediate solution between previous opera-
tors (Formula 12). This is a weighted sum with
ordered evaluation criteria. Thus in addition to
the criterion weights (importance weights), order
weights are used. The order weights are associated
with criteria values on location by location basis.
They are assigned to a location’s criteria values
in decreasing order without considering which
criteria value comes from.
n
∑L ij
Vj = i =1
∙ pi (12)
n The AND operation may be usefully described
242
GIS Technique for Territorial Analysis
where: Vj = alternative j; n = total criteria number;
Lij = order weights of criteria i and alternative j (L
≤ n, if L = 1 MIN operator, L close to n = compen-
satory method, L close to 0 = no compensatory
method); pi = weight value of criteria i (∑pi = 1)
OWA is almost identical to that of WLC, with
the exception that a second set of weights appears.
This second set of weights, the order weights,
controls the manner in which the weighted factors
are aggregated (Eastman and Jiang, 1996; Yager,
1988). Indeed, WLC turns out to be just one vari-
ant of the OWA technique. To introduce the OWA
technique, let’s first review WLC in terms of two
new concepts: tradeoff and risk.
About tradeoff: Factor weights are weights
that apply to specific factors, i.e., all the pixels
of a particular factor image receive the same fac-
tor weight. They indicate the relative degree of
importance each factor plays in determining the
suitability for an objective. In the case of WLC
the weight given to each factor also determines
how it will tradeoff relative to other factors. For
example, a factor with a high factor weight can
tradeoff or compensate for poor scores on other
factors, even if the unweighted suitability score
for that highly-weighted factor is not particularly
good. In contrast, a factor with a high suit- ability
score but a small factor weight can only weakly
compensate for poor scores on other factors. The
factor weights determine how factors tradeoff but,
as described below, order weights determine the
overall level of tradeoff allowed.
About risk: Boolean approaches are extreme
functions that result either in very risk-averse
solutions when the AND operator is used or in risk-
taking solutions when the OR operator is used.53
In the former, a high aggregate suitability score
for a given location (pixel) is only possible if all
factors have high scores. In the latter, a high score
in any factor will yield a high aggregate score,
even if all the other factors have very low scores.
as the minimum, since the minimum score for
any pixel determines the final aggregate score.
GIS Technique for Territorial Analysis
Similarly, the OR operation may be called the
maximum, since the maximum score for any pixel
determines the final aggregate score. The AND
solution is risk-averse because we can be sure
that the score for every factor is at least as good
as the final aggregate score. The OR solution is
risk-taking because the final aggregate score only
tells us about the suitability score for the single
most suitable factor.
The WLC approach is an averaging technique
that softens the hard decisions of the Boolean ap-
proach and avoids the extremes. In fact, given a
continuum of risk from minimum to maximum,
WLC falls exactly in the middle; it is neither risk-
averse nor risk-taking.
About order weights, tradeoff and risk: The
use of order weights allows for aggregation solu-
tions that fall anywhere along the risk continuum
between AND and OR. Order weights are quite
different from factor weights. They do not apply
to any specific factor. Rather, they are applied
on a pixel-by-pixel basis to factor scores as de-
termined by their rank ordering across factors at
each location (pixel). Order weight 1 is assigned
to the lowest-ranked factor for that pixel (i.e., the
factor with the lowest score), order weight 2 to
the next higher-ranked factor for that pixel, and so
forth. Thus, it is possible that a single order weight
could be applied to pixels from any of the various
factors depending upon their relative rank order.
Using different operators the process can
produce different values. From operative point of
view compensatory operators are used for envi-
ronmental improvement projects. In this case all
Figure 7. OWA scheme
criteria and alternatives values can have the same
importance. No compensatory operators are fre-
quently used for valuation of projects possibly
damaging to the environment. The lowest alterna-
tive value can be represent a dangerous factor that
is more important than others. In this case it is
significant to underline the value of dangerous
alternative in order to avoid the wrong choice. An
example is showed in Box 6.
Territorial Analysis Results
Each Criterion map is aggregated by decisional
rules (operators): the result is a final map that
represents an easily understanding method for all
public or private stakeholders that can choose the
correct territorial planning strategies (Box 5, Box
6). The end of a decision making process is not
only focus on the resolution of a specific decisional
problem, but it is represent a recommendation
for future actions. A large number of alternatives
are evaluated on the basis of multiple criteria and
then the best solution (best alternatives or ranking
of alternatives) is defined. Spatial Multicriteria
Decision Analysis is planned with the aim to help
the decision maker during planning process: it
performs complex multiple criteria evaluation
and at same time it includes the preferences of
decision maker.
Box 4. Decision rules
A compensatory operator (WLC) is used for criterion maps
aggregation (Figure 8).
243

Figure 8. Flowchart of WLC aggregation
Box 5. Territorial analysis results (cartographic
layout)
The territorial analysis results are showed by final map (Figure
9) where dark coloured pixels represent more suitable areas for
forest activities (areas with high forest production and close to
roads).
FUTURE RESEARCH DIRECTIONS
The main goal is to realise a correct territorial
planning strategy through the prediction of future
territorial scenarios. Decision processes become
Figure 9. Suitable areas for forest activities
244
GIS Technique for Territorial Analysis
an important phase that need support tools in
order to considerate many decisional variables.
The decisions can be categorized in decision
involving stochastic information and imprecise
information. So it is possible to divide decisions
in probabilistic and fuzzy group. Recently both
stochastic (probabilistic) and imprecise (fuzzy)
methodologies was included in GIS software in
order to resolve decision problems under uncer-
tainty, such as territorial characteristics. The future
research directions move to a full integration of
previous statistical methodologies and automatic
techniques aiming to the extraction of the involved
GIS Technique for Territorial Analysis

Box 6. Territorial analysis results (numeric layout)

According to Formula 9 (Weighted Linear Combination) the following examples provide 2 different suitable areas definitions.
Case study area 1
1. Criteria values
Forest production 324 m3/ha (maximum production is 324 m3/ha)
Distance from roads = 6 kilometres (maximum distance is 7 kilometres)
2. Criteria values after normalization
Forest production = 255
Distance from roads = 15
3. Suitable area value for forest activities using WLC aggregation
WLC = (255*0,58+15*0,5)/2 = 67,5
Case study area 2
1. Criteria values
Forest production 122 m3/ha (maximum production is 324 m3/ha)
Distance from roads = 3 kilometres (maximum distance is 7 kilometres)
2. Criteria values after normalization
Forest production = 120
Distance from roads = 100
3. Suitable area value for forest activities using WLC aggregation
WLC = (120*0,5+100*0,5)/2 = 55
The first example area is far from roads but has the highest forest production, while the second area has medium values for all criteria.
Using WLC method the lowest alternative value (case study area 1 distance from road) is hidden by high production values, so the first
area is more suitable for forest activities than second one.
Same territorial analysis can have different results using different operators. The previous example is also performed using Min operator.
Case study area 1
1. Criteria values
Forest production 324 m3/ha (maximum production is 324 m3/ha)
Distance from roads = 6 kilometres (maximum distance is 7 kilometres)
2. Criteria values after normalization
Forest production = 255
Distance from roads = 15
3. Suitable area value for forest activities using Min aggregation
Min = Min (255*0,5; 15*0,5) = 7.5
Case study area 2
1. Criteria values
Forest production 122 m3/ha (maximum production is 324 m3/ha)
Distance from roads = 3 kilometres (maximum distance is 7 kilometres)
2. Criteria values after normalization
Forest production = 120
Distance from roads = 100
3. Suitable area value for forest activities using Min aggregation
Min = Min (120*0,5; 100*0,5) = 50
No compensatory operators focus on the lowest alternatives value. In this case the lowest alternative value (case study area 1 distance
from road) is not hidden by high values of other alternatives, so the second area is more suitable for forest activities than first one.

decisional rules in territorial planning strategies
considering temporal dimension. “GIS have not
fully realized their potential as systems to sup-
port and facilitate spatial modeling processes.
They continue to handle the temporal dimension
very poorly” (Wagner, 1997). For such purposes
sophisticated techniques are performed: decisional
rules are extracted using Discrete Choice models,
Neural Network System or other non linear models
and are implemented in automatic learning tools.
Markov Chains, Bayesian networks, Cellular
Automata are used in order to predict future ter-
ritorial scenarios. They use decisional rules with
the aim of create transitional rules for modelling
spatial territorial scenarios.
So the integration of GIS and statistical
techniques will be the new frontier of territorial
analysis. “Model integration is a deep scientific

245

problem but even so a pragmatic multi- criteria
multi-objective problem. Model integration re-
quires dealing with end-use aspects: what is ap-
propriate to be integrated in view of the intended
use, the scientific aspect: what can and cannot be
integrated on scientific grounds; and the techni-
cal aspects: how will the integrated model be
assembled and run. Yet, and despite the fact that
the terms ‘integral model’ and ‘integrated model’
are used all over, there are very few recipes or
procedures for model integration available from
the scientific literature. Thus, model integration
seems more an art than a science at this moment”
(Engelen et all, 2002).
CONCLUSION
A correct territorial planning must consider the
increasing of anthropic pressure and the protection
of territorial resources: the protection of a territory
is an important key issue, where it is significant
the respect of rules for a correct exploitation of
natural resources.
Spatial Decision Support System carry out
several procedures when analysing complex de-
cisional problems. “The basic strategy is to divide
the decision problem into small, understandable
parts; analyze each part; and integrate the parts in a
logical manner to produce a meaningful solution”
(Malczewski, 1999). One of SDSS procedures is
based on Spatial Multicriteria Decision Analysis
that combines Multicriteria analysis and Geo-
graphical Information System. Given a decisional
problem, some objectives are defined and are ana-
lyzed through criteria, alternatives and constrains
definition. Thus a spatial analysis is provided by
using multicriteria operators in order to choose
best alternatives. This phase is characterized by
two main problems related to the complexity of
data analysis and to the difficulty of working under
uncertain scenarios: GIS techniques executed by
246
GIS Technique for Territorial Analysis
personal computer are used to resolve the first
problem, while Multicriteria rules applied to Fuzzy
logic are used to represent imprecise, incomplete
and vague information. In conclusion this chapter
focus on a decision support methodology where
the final results are thematic maps that represent
friendly-understanding method for territorial
analysis and an important tool open to public or
private stakeholders who can choose the proper
territorial planning strategy.
REFERENCES
Amrhein, C., Griffith, D., & Desloges, J. R. (1991).
Statistical analysis for geographers. Prentice Hall
publisher.
Arrow, K. J. (1950). A difficulty in the concept of
social welfare. The Journal of Political Economy,
58(4), 328–346. doi:10.1086/256963
Ascough., et al. (2002). Multicriteria spatial deci-
sion support systems: Overview, applications, and
future research directions. In A. E. Rizzoli & A. J.
Jakeman (Eds.), Proceedings of the 1st Biennial
Meeting of the iEMSs.
Bellman, R. E., & Zadeh, L. A. (1970). Decision-
making in a fuzzy environment. Management
Science, 17(4), B141–B164. doi:10.1287/
mnsc.17.4.B141
Burrough, P. A. (1989). Fuzzy mathemati-
cal methods for soil survey and land evalua-
tion. Journal of Soil Science, 40, 477–492.
doi:10.1111/j.1365-2389.1989.tb01290.x
Eastman, J. R., & Jiang, H. (1996). Fuzzy measures
in multi-criteria evaluation. Proceedings, Second
International Symposium on Spatial Accuracy
Assessment in Natural Resources and Environ-
mental Studies, (pp. 527-534). May 21-23, Fort
Collins, Colorado.
GIS Technique for Territorial Analysis
Engelen, G., White, R., van der Meulen, M.,
& Hahn, B. (2002). Sustainable developments
of islands: A policy support framework for the
integrated assessment of socio-economic and
environmental development. In Hsiao, H.-H. M.,
Liu, C.-H., & Tsai, H.-M. (Eds.), Sustainable
development for island societies: Taiwan and the
world (pp. 251–287). Taipei, Taiwan: Asia-Pacific
Research Program, Academia Sinica and SARCS
Secretariat.
FAO. (1976). A framework for land evaluation.
Soils Bulletin, 32. Rome, Italy: Food and Agricul-
tural Organization of the United Nations.
FAO. (1993). Guidelines for land-use planning.
FAO Development Series No. 1. ISBN 92-5-
103282-3, ISSN 1020-0819
Keeney, R. L. (1980). Siting energy facilities. New
York, NY: Academic Press.
Keeney, R. L., & Raiffa, H. (1976). Decision
with multiples objectives: Preferences and value
tradeoffs. New York, NY: Wiley.
Malczewski, J. (1999). GIS and multicriteria deci-
sion analysis. John Winley & Sons, Inc.
Malczewski, J. (2004). GIS–based land-use suit-
ability analysis: A critical overview. Elsevier.
Mather, P. M. (1991). Computer applications in
geography. Chichester, UK: John Wiley & Sons.
Narisimhan, R. (1980). Goal programming
in a fuzzy environment. Decision Sciences,
11(2), 325–336. doi:10.1111/j.1540-5915.1980.
tb01142.x
Nijkamp, P. (1976). Multi-objective programming
models: new ways in regional decision making.
Regional Science and Urban Economics, 6,
253–274. doi:10.1016/0166-0462(76)90002-8
Pareto, V. (1906). Manuale di economia politica.
Romero, C., & Amador, F. (1987). Multiple ob-
jectives in agricultural planning: A compromise
programming application. American Journal
of Agricultural Economics, 69(1), 78–86.
doi:10.2307/1241308
Starr, M. K., & Zeleny, M. (1977). Multiple cri-
teria decision making. North-Holland Pub. Co.
Triantaphyllou, E. (2000). Multi-criteria decision
making methods: A comparative study. Kluwer
Academic Publishers.
Wagner, D. F. (1997). Cellular automata and geo-
graphic information systems. Environment and
Planning B, 24, 219–234. doi:10.1068/b240219
Watson, S. R., & Buende, D. M. (1987). Deci-
sion synthesis. The principles and practice of
decision analysis. Cambridge, UK: Cambridge
University Press.
Yager, R. (1988). On ordered weighted averaging
aggregation operators in multicriteria decision
making. IEEE Transactions on Systems, Man, and
Cybernetics, 8(1), 183–190. doi:10.1109/21.87068
Yager, R. R. (1987). Fuzzy sets and application:
Selected papers by L.A. Zadeh. New York, NY:
John Wileys & Sons.
Yu, P. L. (1973). A class of solutions for group
decision problems. Management Science, 19(8).
doi:10.1287/mnsc.19.8.936
Zadeh, L. (1965). Fuzzy sets and systems. In Fox,
J. (Ed.), System theory. Brooklyn, NY.
ADDITIONAL READING
Bernetti I., Fagarazzi C., (2002). L’impiego dei
modelli multicriteriali geografici nella pianificazi-
one territoriale. Aestimum dicembre 41, pp 1-26.
Chen, S. J., & Hwang, C. (1992). Fuzzy multiple
attribute decision making. Springer-Verlang,
Berlino.
247

Chirici G., (1998). I GIS nella pianificazione e
gestione forestale: l’esempio del Comune di M.te
Armelio. Genio Rurale n°7/8, pp 19-25.
Ferrucci D., (1999). Applicazione di un GIS per
il confronto e l’integrazione di cartografie di uso
del suolo. Genio Rurale n°5, pp 55-63.
Funtowicz, S. O., & Ravetz, J. R. (1990). Uncer-
tainty and quality in science for policy. Dordrecht:
Kluwer.
Gisotti G.,Bruschi S., (1990). Valutare l’ambiente.
La nuova Italia Scientifica editore, cap 1,7,9.
Nijkamp, P., & Rietveldt, P. (1996). Multiple Ob-
jective Decision Analysis in regional economic.
Amsterdam, North Holland: Handbook of regional
and urban economics.
Nuti F., (2001). La valutazione economica delle
decisioni pubbliche. Giappichelli editore.
Riccioli, F. (2009). Una proposta metodologica
per lo studio della multifunzionalità: l’analisi
MultiCriteriale Geografica. Economia e diritto
agroambientale n.3 (pp. 99–118). Firenze Uni-
versity Press.
Sugeno, M. E., & Yasukawa, M. (1993). A fuzzy-
logic-based approach to qualitative modeling.
IEEE Transactions on Fuzzy Systems, 1, 7–31.
doi:10.1109/TFUZZ.1993.390281
Yu, P. L. (1973). A class of solutions for group
decision problems. Management Science, 19(8).
doi:10.1287/mnsc.19.8.936
Zeleny, M. (1974). A concept of Compromise
solutions and method of the displaced ideal -.
Computers & Operations Research, 1(4).
Zeleny, M. (1982). Multiple Criteria Decision
Making. New York: McGraw Hill.
Zimmermann, H. J. (1987). Fuzzy sets, decision
making and expert system. Boston: Kluwer Ac-
cademic Publisher.
248
GIS Technique for Territorial Analysis
KEY TERMS AND DEFINITIONS
Decision Rules: Territorial planning choices
driven by Stakeholders decisions that shape a
specific territory.
Geographical Information System (GIS):
Software for acquisition, storage and analysis of
spatial data in order to represent real world items
with vector o raster structure.
Multicriteria analysis: (MCDM): Decision
support method based on objectives, alternatives,
criteria definition with the aim of solving a specific
decisional problem (goal).
Raster Structure: Representation of real
world items with grid formed by pixels (minimum
grind unit).
Spatial Decision Support System: (SDSS):
Computer-based systems with the aim of solving
spatial decision problems.
Spatial Multicriteria Decision Analysis:
Decision support tool used in territorial analysis
based on integration of Geographical Information
System and Multicriteria analysis.
Territorial Planning: Assessment process
driven by Decision Makers objectives with the
aim of economics, social and environmental
characteristics analysis of a specific territory.
Vector Structure: Representation of real
world items with geographical primitives (points,
lines, polygons).
ENDNOTES
1
Best solution meaning is based on Arrow’s
impossibility theorem (Arrow, 1950)
2
The measures used in performance matrices
of this kind are also qualitative descriptions,
or natural units.
3
p is the metrical parameter used for distance
from ideal measurement
4
The word GIS is considered both plural and
singular during text
5
Aerial photo, satellite photo, etc.
GIS Technique for Territorial Analysis

6
Based on fuzzy theory (see Spatial Multi- 7
The range scale 0-255 allow an higher cal-
Criteria Decision Analysis section for more culation speed than 0-1 range scale from
details) computer performance point of view
8
Same weights values for each criteria are
considered

This work was previously published in ural and Environmental Informatics, Governance and Management: Emerging Research
Applications, edited by Zacharoula Andreopoulou, Basil Manos, Nico Polman and Davide Viaggi, pp. 425-445, copyright 2011
by Information Science Reference (an imprint of IGI Global).

249
250

Chapter 17
Cognitively Ergonomic
Route Directions
Alexander Klippel
University of Melbourne, Australia

Kai-Florian Richter
Universität Bremen, Germany

Stefan Hansen
Spatial/Information Systems Ltd./LISAsoft, Australia

ABSTRACT
This contribution provides an overview of elements of cognitively ergonomic route directions. Cognitive
ergonomics, in general, seeks to identify characteristics of cognitive information processing and to formal-
ize these characteristics such that they can be used to improve information systems. For route directions,
an increasing number of behavioral studies have, for example, pointed to the following characteristics:
the use of landmarks, changing levels of granularity, the qualitative description of spatial relations.
The authors detail these aspects and additionally introduce formal approaches that incorporate them
to automatically provide route directions that adhere to principles of cognitive ergonomics.

COGNITIVE ASPECTS OF ROUTE aspects of route directions has increased. Next to

DIRECTIONS
Route directions fascinate researchers in several
fields. Since the 70s linguists and cognitive scien-
tists have used verbal route directions as a window
to cognition to learn about cognitive processes
that reflect structuring principles of environmen-
tal knowledge (e.g., Klein, 1978). Over the last
decade, the number of publications on various
DOI: 10.4018/978-1-4666-2038-4.ch017
the general aspects of how to provide route direc-
tions and how to identify principles that allow us
to define what makes route directions cognitively
ergonomic, technical aspects of navigation support
systems have become an additional focus. The
question required from the latter perspective is
part of a broader approach that aims to formally
characterize the meaning (semantics) of spatial
relations. In other words, if we want to bridge the
gap between information systems and behavioral

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Cognitively Ergonomic Route Directions
analysis we have to answer how we perform the
transition from data to knowledge.
Several key elements can be identified based
on psychological and linguistic literature on route
directions that are pertinent for cognitively ergo-
nomic route directions (Denis, 1997; Lovelace,
Hegarty, & Montello, 1999; Tversky & Lee,
1999). These comprise the conceptualization of
directions at decision points, the spatial chunking
of route direction elements to obtain hierarchies
and to change the level of granularity, the role
of landmarks, the communication in different
modalities, the traveling in different modes, and
aspects of personalization (see Table 1). Most
research on routes and route directions deals with
navigation in urban structures such as street net-
works. The results discussed in this article focus
on this domain.
APPROACHES TO REPRESENTING
ROUTE KNOWLEDGE
Behavioral studies have substantiated key ele-
ments of cognitively ergonomic route directions.
To implement these aspects in information sys-
Table 1. Cognitive ergonomics of route directions
• are qualitative, not quantitative,
• allow for different levels of granularity and organize spatial
knowledge hierarchically,
• reflect cognitive conceptualizations of directions at decision
points,
• chunk route direction elements into larger units to reduce
cognitive load,
• use landmarks to:
° disambiguate spatial situations,
° anchor turning actions,
° and to confirm that the right actions have been taken,
• present information in multimodal communication systems
allowing for an interplay of language and graphics, but re-
specting for the underlying conceptual structure,
    • allow for an adaptation to the user’s familiarity with an en-
vironment, as well as personal styles and different languages.
tems detailed formal characterizations of route
knowledge are required. The approaches discussed
below are a representative vocabulary that allows
for the characterization of mental conceptualiza-
tion processes reflecting the results from behav-
ioral studies (see Table 1). In this sense we can
refer to them as Ontologies of Route Knowledge
(Chandrasekaran, Josephson, & Benjamins, 1999;
Gruber, 1993). In Guarino’s terminology these
approaches would most likely be called domain
ontologies (Guarino, 1998).
One of the earliest approaches is the TOUR
model by Kuipers (Kuipers, 1978) that later devel-
oped into the Spatial Semantic Hierarchy (SSH)
(Kuipers, 2000). Kuipers and his collaborators
developed this approach to add the qualitative-
ness that can be found in the organization of a
cognitive agent’s spatial knowledge to approaches
in robotics. The latter classically relied more on
quantitative spatial descriptions. The SSH al-
lows for modeling cognitive representations of
space as well as for building a framework for
robot navigation, i.e. qualitative and quantita-
tive aspects are combined. The SSH especially
reflects the aspect of hierarchical organization of
spatial knowledge by providing different levels of
information representation: the sensory, control,
causal, topological, and metrical level. Ontological
characterizations are developed for each level to
match human cognitive processes.
The Route Graph model (Werner, Krieg-
Brückner, & Herrmann, 2000) describes key
elements for route based navigation. Similar to
the SSH, it allows representing knowledge on
different levels of granularity. However, it is
much more abstract and does not provide any
processes for acquiring this knowledge. It is
intended to provide a formalism expressing key
notions of route knowledge independent of a
particular implementation, agent, or domain. Its
focus is on a sound formal specification of basic
elements and operations, like the transition from
route knowledge to survey knowledge by merging
routes into a graph-like structure.
251

A linguistically grounded approach with the
aim to generate verbal route directions is the
CORAL project by Dale and coworkers (e.g.,
Dale, Geldof, & Prost, 2005). One of the central
aspects of their approach is the organization of
parts of a route into meaningful units, a process
they call segmentation. Instead of providing
turn-by-turn directions, this approach allows for
a small number of instructions that capture the
most important aspects of a route. The employed
modeling language is called Route Planning
Markup Language (RPML).
Formalisms that model route knowledge on
the conceptual level can be found in the theory of
wayfinding choremes (Klippel, Tappe, Kulik, &
Lee, 2005) and context-specific route directions
(Richter & Klippel, 2005). These approaches
model route knowledge modality-independent
on the conceptual level. The wayfinding choreme
theory employs conceptual primitives—as the
result of conceptualization processes of a cogni-
tive agent incorporating functional as well as
geometrical environmental aspects—to define
basic as well as super-ordinate valid expressions
on different levels of granularity. The approach
to context-specific route directions builds on
this theory. A systematics of route direction ele-
ments determines which, and how, entities may
be referred to in route directions. Accordingly,
abstract relational specifications are inferred by
optimization processes that adapt route directions
to environmental characteristics and inherent
route properties.
Human wayfinding, however, may not be
restricted to a single mode of transportation. A
typical example is public transport, where travelers
frequently switch between pedestrian movement
and passive transportation (trains, buses, etc.).
Timpf (2002) analyzed route directions for multi-
modal wayfinding and developed two different
ontologies of route knowledge: one representing
knowledge from the perspective of the traveler and
one taking the perspective of the transportation
system. The former focuses on movement along
252
Cognitively Ergonomic Route Directions
a single route, i.e., actions to perform to reach
the destination, while the latter provides concepts
referring to the complete transportation network.
An industry approach for formalizing route
knowledge can be found in Part 6: Navigation
Service of the OpenLS specification. The Open-
GIS Location Services (OpenLS) Implementation
Specification (Mabrouk, 2005) describes an open
platform for location-based application services,
the so called GeoMobility Server (GMS) proposed
by the Open Geospatial Consortium (OGC). It
offers a framework for the interoperable use of
mobile devices, services and location-related data.
The Navigation Service described in Part 6 of
the OpenLS specification provides the accessing
client, amongst other services, with preprocessed
data that is required for the generation of route
directions. Based on XML specifications, it defines
a data structure that allows clients to generate their
own route directions which may accord more to a
user’s preferences. The used data model structures
the route in maneuvers (descriptions combining
a turn at a decision point and proceeding on the
following route segment) and enhances them with
additional information about route elements.
CORE ASPECTS OF COGNITIVELY
ERGONOMIC ROUTE DIRECTIONS
In the following, three aspects that are at the core
of cognitively ergonomic route directions will be
discussed in greater detail: cognitively adequate
direction concepts, the use of landmarks, and
spatial chunking to obtain hierarchies and change
the level of granularity.
Conceptualization of Directions
at Decision Points
The specification of direction changes is the
most pertinent information in route directions.
While current route information systems heavily
rely on street names to identify the proper direc-
Cognitively Ergonomic Route Directions

tion to take, behavioral research (Tom & Denis,
2003) has shown that from a cognitive perspec-
tive, street names are not the preferred means to
reorient oneself. People rather rely on landmarks
(as discussed in the next section) and appropri-
ate direction concepts. On the most basic level
we have to specify the correspondence between
a direction change (in terms of the angle) and a
direction concept. For example, which sector is
applicable to a concept like “turn right”? On a
more elaborate level, we have to specify alter-
native direction concepts and detail their scope
of application. Figure 1 shows some examples
of how the same direction change can result in
different direction concepts (and corresponding
verbalizations) depending, among other things, on
the spatial structure in which the change occurs.
We need this level of specificity for two reasons.
First, a qualitative but precise direction model
allows for verbally instantiating a situation model
(Zwaan & Radvansky, 1998) of the encountered
intersections. Second, intersections can function
as landmarks. Just like classical examples of
landmarks, such as the Eiffel Tower, in the context
of a specific route, a salient intersection can be
used to organize spatial knowledge. This aspect
has not yet gained much attention.
Enriching Route Directions with
Landmarks
Analyzing human route directions shows how
prominently landmarks are used to structure
the respective spatial knowledge, to give the
instructed the possibility to assure that they are
still following the correct route, and to anchor
required turning actions. Since landmarks seem
to be such an important part of human-generated
route directions their integration is pertinent for
automatically generating cognitively ergonomic
instructions.
Several classifications of landmarks and their
characteristics have been discussed in the litera-
ture. One of the first assessments is presented by
Lynch (1960) who distinguishes Landmarks as one
of five elements that structure urban knowledge:
path, edges, districts, nodes, and landmarks. It
is commonly agreed that the landmark account

Figure 1. A change of a direction is associated with different conceptualizations according to the inter-
section at which it takes place. The ‘pure’ change may be linguistically characterized as take the second
exit at the roundabout (a). At intersection (b) it might change to the second right; at intersection (c) it
may change to fork right, and at (d) it becomes veer right.

253

should comprise all five elements, as according
to Presson and Montello (1988) everything that
stands out of the background may serve as a
landmark. That is, given the right spatial context
different features of an environment may serve
as landmarks. Sorrows and Hirtle (1999) distin-
guish three characteristics important for making
an object a landmark: its visual, semantic, and
structural characteristics. Additionally, landmarks
can be categorized according to their cognitive
function within route directions, their geometry,
and their spatial relation to the route. Humans
conceptualize landmarks either as point-like,
linear, or area-like entities. However, these con-
ceptualizations do not necessarily correspond to
the geometric characteristics of objects but reflect
the schematization processes cognitive agents
apply (Herskovits, 1986). A detailed description
of the different roles of landmarks is necessary
to allow for their integration in an automatic gen-
eration process. For example, a simple, yet as of
today unexplored way to enrich route directions
with landmarks is to include references to salient
intersections, like T-intersections or roundabouts,
which are easy to identify automatically. This also
reflects the direction concepts humans employ
with such structures (see also Figure 1).
Spatial Chunking: Hierarchies
and Levels of Granularity
The hierarchical organization of spatial infor-
mation and flexibly changing between levels
of granularity are omnipresent in the cognitive
organization of spatial knowledge (Hobbs, 1985;
Kuipers, 2000). Chunking elementary wayfinding
actions (such as turns at intersections) in order
to impose a hierarchical structure and to change
the level of granularity reflects not only cogni-
tive conceptualization processes but organizes
route knowledge in a cognitively ergonomic
way. Especially users who are familiar with an
environment can profit from such an approach.
In general, providing a user with too much detail
254
Cognitively Ergonomic Route Directions
violates findings of cognitive science, as for ex-
ample formulated in Clark’s 007 Principle: “In
general, evolved creatures will neither store nor
process information in costly ways when they
can use the structure of the environment and their
operations upon it as a convenient stand-in for the
information-processing operations concerned.
That is, know only as much as you need to know
to get the job done.” (Clark, 1989, p. 64)
Structuring route descriptions by subsuming
instructions gives users a coarse overview over a
route, which is easier to perceive and quite often
sufficient for successful wayfinding, especially
if the user is familiar with the environment. Of
course, the subsumed information still has to be
accessible in case the user needs it (or, as discus-
sions on positioning technologies in this volume
show, the user may simply re-query a new route
from his new position). This may either be pos-
sible by zoom-in operations, i.e., by accessing the
next, more detailed level of the hierarchy, or by
(mental) inference processes. Such inferences, for
example, extract from an instruction like “turn left
at the dead end” information on which action to
perform at all intersections before the dead end,
namely to continue straight (e.g., Duckham &
Kulik, 2003). The following cognitive strategies
for spatial chunking are discussed in the litera-
ture (Dale et al., 2005; Klippel, Tappe, & Habel,
2003): numerical chunking, structure chunking,
landmark chunking, and chunking using the street
level hierarchy.
THE MULTIMODAL PRESENTATION
OF ROUTE KNOWLEDGE
The multimodal communication of spatial in-
formation is a core aspect of human cognition:
linguistic expressions, graphical representations
such as sketch maps, and gestures are channels
along which humans naturally communicate (Ovi-
att, 2003). Each representational medium—each
channel—has advantages in specific contexts but
Cognitively Ergonomic Route Directions
may fail in other situations (Kray, Laakso, Elting,
& Coors, 2003). For example, natural language
expressions are inherently underspecified: a term
like turn right is applicable to a range of different
turning angles at an intersection and therefore
may be sufficient in many situations. Figure 2,
however, shows a situation that requires a complex
explanation if a description is provided in linguis-
tic terms. In this case, a graphic representation
is more suitable to communicate the situation at
hand. Communication channels also differ with
respect to their suitability in the identification of
landmarks. A salient object at an intersection might
be visually easily identifiable and recognisable,
but hard to describe linguistically. An expression
like follow the road to the dead end on the other
hand, may chunk a large part within a route lin-
guistically and therefore, communicate the spatial
situation more efficiently if the dead end is a long
way away and hard to depict on a small screen.
The communication of route information,
whether visually, linguistically, or in any other
modality, has to follow the same guidelines as
established for the structuring of route knowledge.
Cluttering any communication process has shown
to violate cognitive ergonomics and to slow down
information processing. This confinement to
sparseness has been shown for visual route direc-
tions, for example, by Agrawala and Stolte(2000),
who based their route direction tool on results
obtained from sketch maps (Tversky & Lee, 1999).
Figure 2. Complex intersection
SUMMARY
In the last decades, research on route directions in
linguistics and cognitive science revealed many
underlying principles and processes of human
route direction production and comprehension,
and, thus, provides us with an understanding of
what constitutes cognitively ergonomic route
directions. However, this understanding has to
be formally specified to be implemented in in-
formation systems for wayfinding assistance, like
internet route-planners. In essence, three cognitive
principles need to be implemented in wayfind-
ing assistance systems to generate cognitively
ergonomic route directions: adequate direction
concepts, the enrichment of route directions with
landmarks, and spatial chunking which allows for
a hierarchical structuring of route knowledge and
representations on different levels of granularity.
To this end, we need a thorough understanding of
which direction concept humans apply in which
situation, a detailed ontology of the different kinds
of landmarks and the role they may take in route
directions, as well as formal characterizations
that model hierarchical structures and guide the
changes of granularity.
REFERENCES
Agrawala, M., & Stolte, C. (2000). A design and
implementation for effective computer-generated
route maps. In AAAI Symposium on Smart Graph-
ics, March 2000. Stanford.
Chandrasekaran, B., Josephson, J. R., & Ben-
jamins, V. R. (1999). What are ontologies, and
why do we need them? IEEE Intelligent Systems
and Their Applications, 14(1), 20–26.
Clark, A. (1989). Microcognition: Philosophy,
cognitive science, and parallel distributed pro-
cessing. Cambridge, MA: MIT Press.
255
 Cognitively Ergonomic Route Directions

Dale, R., Geldof, S., & Prost, J.-P. (2005). Using
natural language generation in automatic route
description. Journal of Research and practice in
Information Technology, 37(1), 89-105.
Denis, M. (1997). The description of routes: A
cognitive approach to the production of spatial
discourse. Cahiers de Psychologie Cognitive,
16, 409–458.
Duckham, M., & Kulik, L. (2003). “Simples”
paths: Automated route selection for navigation.
In W. Kuhn, M. Worboys, & S. Timpf (Eds.),
Spatial information theory: Foundations of
geographic information science. Conference on
Spatial Information Theory (COSIT) 2003. (pp.
182-199). Berlin: Springer.
Gruber, T. R. (1993). A translation approach to
portable ontologies. Knowledge Acquisition, 5(2),
199–220. doi:10.1006/knac.1993.1008
Guarino, N. (1998). Formal ontology and infor-
mation systems. In N. Guarino (Ed.), Formal
Ontology in Information Systems. Proceedings of
FOIS’98, Trento, Italy, 6-8 June 1998. (pp. 3-15).
Amsterdam: IOS Press.
Herskovits, A. (1986). Language and Spatial
Cognition: An Interdisciplinary Study of the
Representation of the Prepositions in English.
UK: Cambridge University Press.
Hobbs, J. R. (1985). Granularity. In A. Joshi (Ed.),
Proceedings of the 9th International Joint Confer-
ence on Artificial Intelligence. Los Angeles, CA
(pp. 432-435). San Francisco: Morgan Kaufmann.
Klein, W. (1978). Wegauskuenfte. Zeitschrift für
Literaturwissenschaft und Linguistik, 33, 9–57.
Klippel, A., Tappe, T., & Habel, C. (2003). Pic-
torial representations of routes: Chunking route
segments during comprehension. In C. Freksa, W.
Brauer, C. Habel & K. F. Wender (Eds.), Spatial
Cognition III. Routes and navigation, human
memory and learning, spatial representation and
spatial learning. (pp. 11-33). Berlin: Springer.
Klippel, A., Tappe, T., Kulik, L., & Lee, P. U.
(2005). Wayfinding choremes - A language for
modeling conceptual route knowledge. Journal of
Visual Languages and Computing, 16(4), 311–329.
doi:10.1016/j.jvlc.2004.11.004
Kray, C., Laakso, K., Elting, C., & Coors, V.
(2003). Presenting route instructions on mobile
devices. Paper presented at the IUI’03, January
12-15, 2003, Miami, Florida, USA.
Kuipers, B. (1978). Modelling spatial knowledge.
Cognitive Science, 2(2), 129–154.
Kuipers, B. (2000). The spatial semantic hierarchy.
Artificial Intelligence, 119, 191–233. doi:10.1016/
S0004-3702(00)00017-5
Lovelace, K., Hegarty, M., & Montello, D. R.
(1999). Elements of good route directions in
familiar and unfamiliar environments. In C.
Freksa & D. M. Mark (Eds.), Spatial information
theory. Cognitive and computational foundations
of geographic information science (pp. 65-82).
Belin: Springer.
Lynch, K. (1960). The image of the city. Cam-
bridge, MA: MIT Press.
Mabrouk, M. (2005). OpenGis Location Services
(OpenLS): Core Services. OGC Implementa-
tion Specification 05-016 Version 1.1 Open Gis
Consortium Inc.
Oviatt, S. L. (2003). Multimodal interfaces. In J.
Jacko & A. Sears (Eds.), The Human-Computer
Interaction Handbook: Fundamentals, Evolving
Technologies and Emerging Applications. (pp.
286-304). Mahwah, NJ: Lawrence Erlbaum.
Presson, C. C., & Montello, D. R. (1988). Points of
reference in spatial cognition: Stalking the elusive
landmark. The British Journal of Developmental
Psychology, 6, 378–381.

256
Cognitively Ergonomic Route Directions
Richter, K.-F., & Klippel, A. (2005). A model for
context-specific route directions. In C. Freksa,
M. Knauff & B. Krieg-Brueckner (Eds.), Spatial
Cognition IV. Reasoning, Action, and Interaction:
International Conference Spatial Cognition 2004,
Frauenchiemsee, Germany, October 11-13, 2004,
Revised Selected Papers (pp. 58-78). Berlin:
Springer.
Sorrows, M., & Hirtle, S. C. (1999). The nature
of landmarks for real and electronic spaces. In C.
Freksa & D. M. Mark (Eds.), Spatial information
theory. Cognitive and computational foundations
of geographic information science (pp. 37-50).
Berlin: Springer.
Timpf, S. (2002). Ontologies of wayfinding: A
traveler’s perspective. Networks and Spatial
Environments, 2, 3–33.
Tom, A., & Denis, M. (2003). Referring to land-
mark or street iniformation in route directions:
What difference does it make? In W. Kuhn, M.
Worboys & S. Timpf (Eds.), Spatial information
theory. Foundations of geogrpahic information
science. International conference, COSIT 2003,
Kartause Ittingen, Switzerland, September 2003.
(pp. 362-374). Berlin: Springer.
Tversky, B., & Lee, P. U. (1999). Pictorial and
verbal tools for conveying routes. In C. Freksa &
D. M. Mark (Eds.), Spatial information theory.
Cognitive and computational foundations of geo-
graphic information science (pp. 51-64). Berlin:
Springer.
Werner, S., Krieg-Brückner, B., & Herrmann,
T. (2000). Modeling navigational knowledge by
route graphs. In C. Freksa, W. Brauer, C. Habel
& K. F. Wender (Eds.), Spatial cognition II.
Integrating abstract theories, empirical studies,
formal methods, and practical applications. (pp.
295-316). Berlin: Springer.
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 230-238,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
Zwaan, R. A., & Radvansky, G. A. (1998).
Situation models in language comprehension
and memory. Psychological Bulletin, 123(2),
162–185. doi:10.1037/0033-2909.123.2.162
KEY TERMS AND DEFINITIONS
Granularity: Here, it refers to the detail in
route directions; from coarse levels for general
planning to finer levels to provide context-specific
information, for example at decision points.
Landmark: Any entity in the environment
that sticks out from the background.
OpenLS: Specification of an open platform
for location-based services defining their core
functionality (directory service, gateway service,
location utility service, presentation service, route
service).
Personalization: Adaptation of information
presentation and interaction with a device / soft-
ware to the needs and preferences of a specific,
individual user.
Route Directions: A set of instructions that
allow a wayfinder in known or unknown envi-
ronments to follow a route from a start point to
a destination.
Spatial Semantic Hierarchy (SSH): A com-
putational model defining acquisition and repre-
sentation of spatial knowledge on different levels
of abstraction ranging from sensory information
to topological knowledge.
Wayfinding: The cognitive conceptual activity
of planning and finding ones way.
Wayfinding Choremes: Mental conceptu-
alizations of functional wayfinding and route
direction elements.
257
258

Chapter 18
Design and Implementation
Approaches for Location-Based,
Tourism-Related Services
George Kakaletris
University of Athens, Greece

Dimitris Varoutas
University of Athens, Greece

Dimitris Katsianis
University of Athens, Greece

Thomas Sphicopoulos
University of Athens, Greece

ABSTRACT
The globally observed recession of mobile services market has pushed mobile network operators into
looking for opportunities to provide value added services on top of their high cost infrastructures. Recent
advances in mobile positioning technologies enable services that make use of the mobile user location
information, offering intuitive, attractive applications to the potential customer. Mobile tourism services
are among the primary options to be considered by service providers for this new market. This chapter
presents the key concepts, capabilities, and considerations of infrastructures and applications targeted
to the mobile tourist, covering data and content delivery, positioning, systems’ interactions, platforms,
protocols, security, and privacy as well as business modelling aspects.

INTRODUCTION traditionally considered as being only suited to

During the last decade of the 20th century, wire-
less data networks have invaded everyday life
and have gradually started taking over areas
DOI: 10.4018/978-1-4666-2038-4.ch018
wired applications. Due to their versatility, wire-
less telecommunications systems have become a
widespread standard, leading to hardware price
drops and radical quality increases. Today there ex-
ist a bunch of technologies that allow the delivery
of information to mobile or wireless devices and

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Design and Implementation Approaches for Location-Based, Tourism-Related Services
their users, all presenting different characteristics
in performance/ quality, autonomy and cost. These
technological advances accompanied by the reach
of the saturation level (Ellinger, Barras, & Jackel,
2002; Gruber, 2005; Gruber & Verboven, 2001)
in the mobile telephony market pushed hardware
vendors and network and service providers into
looking for new business opportunities. The needs
of tourism-related information provision and
services were amongst the first to be considered
for new applications in the field of communica-
tion devices.
In traditional fixed systems, the location of
a terminal and its user was a part of its identity
and remained constant for a long period during its
lifetime. In this new mobility era, this observation
no longer holds: the physical position of the user
might be highly variable, introducing a whole
new range of issues and opportunities to be taken
into account. The use of intelligent systems that
exploit the positional information of the client,
accompanied by the ability to provide feedback
over a wireless medium, can lead to the provision
of innovative highly intuitive services that were not
available in the near past (Grajski & Kirk, 2003;
Kakaletris, Varoutas, Katsianis, Sphicopoulos, &
Kouvas, 2004; Rao & Minakakis, 2003; Staab &
Werthner, 2002; Yilin, 2000).
But, although mobile telephony networks offer
maximum mobility, they are not the only means
for providing location-based services (LBS) for
tourism. Local fixed wireless networks in their
various forms are another of the modern and
popular technologies facilitating relevant services.
In addition to telecommunication systems and
from a technological perspective, there are a wide
range of other systems such as global positioning
system (GPS) (Dana, 1994; ETSI, 2006; GAR-
MIN, n.d.), or ID tags (Bohn & Mattern, 2004;
Tarumi, Morishita, & Kambayashi, 2000) which
might have a significant role in the development
and deployment of e-tourism applications based
on location information.
This chapter presents the technological con-
cepts associated with the provision of location-
aware tourism-related services under a service-
oriented approach capable of supporting open
value chains and to lead financially viable open
and powerful communication systems. The rest
of the chapter is organised as follows: The “Back-
ground” section presents the technological and
business background of location-based services;
the “Technology Overview” section gets into
details of the technological aspects and issues
raised in the domains of positioning and data/
content delivery, which are fundamental elements
of the examined class of services; the section on
“Mobile Tourism Services” captures the specific
needs and opportunities in the specific application
area and presents issues and considerations with
respect to integrating the various parts into an open
system capable of delivering such services. In the
“Conclusion,” technology and market conclusions
and trends are presented. Finally, due to the large
number of acronyms and the frequency of their
appearance, a table of acronyms is provided at
the end of the chapter in order to ease reading
through it (see Appendix).
BACKGROUND
The application of the above-mentioned technolo-
gies and concepts in tourism gave birth to the
ubiquitous tourism1 concept (OTC, 2003), which
refers to the existence and access of tourism related
services at any place, any time. Although tourism-
related services are mostly related to content
provision, more applications can be identified.
In its entirety, content provision for e-tourism
covers a large number of thematic areas: culture,
urgencies, transportation, events, and so on. Thus,
content might be both temporally and spatially
labelled (LoVEUS, 2002; M-Guide, 2002). In
addition, information seeking and avalanche-like
content provision might guide the user to areas
quite outside her/his initial focus areas.
259
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
The information technology (IT), the Internet
and the mobile telecommunications revolutions
of the last decades of the 20th century made it
possible for enterprises to enable massive access
to their applications and data. Users are able to
access applications and information through a
variety of integrated “channels” including the In-
ternet, mobile telephony, and voice interfaces and
thus bring forward the concept of multi-channel
architectures. Consequently, multi-channel con-
tent delivery and media-independent publishing
have emerged in order to address the demand for
personalised content that can adapt to the end-user
device capabilities. Devices, such as PDAs, cel-
lular phones, smartphones, and television set-top
boxes, introduced the need for additional channels
for publishing content. The approach of maintain-
ing independent content sets per channel proved
to be highly inefficient in terms of maintenance,
until the wide adoption of eXtensible Markup
Language (XML) and related technologies,
such as eXtensible Stylesheet Language / XSL
Transformation (XSL/XSLT), offered a standard
solution to this challenge.
Technology is not the sole reason behind the
emergence of the ubiquitous tourism concept.
The existing 2/2.5G2 mobile market has reached
saturation as analysts have predicted, but its ef-
fects have only been acknowledged lately, due to
the high expectations of the emerging 3G3 mar-
kets. The costs of licensing (Andersson, Hulten,
& Valiente, 2005; Katsianis, Welling, Ylonen,
Varoutas, Sphicopoulos, Elnegaard, et al.,2001;
Yan 2004) and deployment of 3G networks led
mobile network operators4 (MNOs) into a global
recession era and a global pessimism for their
adoption which actually reflects user attitudes
towards the new standard. In order to confront
that, business opportunities based on existing
mobile and wireless networks have been further
investigated (Katsianis et al., 2001; Varoutas,
Katsianis, Sphicopoulos, Loizillon, Kalhagen,
& Stordahl, et al., 2003). The provision of value
added services over 2.5/3G networks not only al-
260
lows providers and users to make the most out of
the existing infrastructures, but also encourages
usage and drives expectations for the next gen-
eration of mobile networks (Varoutas, Katsianis,
Sphicopoulos, Stordahl, & Welling, 2006). To
provide such services, the integration of various
components and base-services is required, which
breaks the current status of most MNOs that have
traditionally been formed as almost monolithic
self-contained service(s) providers.
This need for integration of various market
stakeholders in complex business models aim-
ing for the provision of high quality services has
been indicated not only by mobile market analysts
but also by information systems architects. The
service-oriented approach (Brown, Johnston,
& Kelly 2003; Colan, 2004), a whole new IT
perspective which is rushing into the industry,
underlies the concepts and offers the guidelines
that render possible such complex collaboration
schemes. In the LBS domain, mobile positioning
protocol and mobile location protocol (Ericsson;
OMA, 2002) already exercise concepts in-line
with current service-oriented architectures (SOA)
common practices. Nevertheless, the design of
services, such as location-based ones, will always
have to face domain specific challenges concern-
ing technical, economical or even ethical and
social factors of the service application (Daoud
& Mohan 2002).
Nowadays it is possible and desirable to build
open systems that can support the delivery of
tourism-related location-dependent content to an
end-user on top of the technological and business
background already described, allowing:
• Seamless interoperability of systems and
content provided by several market stake-
holders towards providing a large range of
high-quality location-based content deliv-
ery services, through standards and loosely
coupled elements
Design and Implementation Approaches for Location-Based, Tourism-Related Services
• Exploitation of state-of-the-art and future
technology in positioning, mobile devices,
and network infrastructures
• Compliance with requirements and stan-
dards for personalisation and quality of
service (QoS)
• Low-cost implementation and upgrade
roadmap from 2/2.5G to 3G and other
current and future mobile and wireless
networks
• Guarantees of privacy
As already mentioned, provision of tourism-
related content can be shown that covers a large
portion of the information that is usually delivered
through location-based services. A number of stud-
ies already exist that focus on various aspects of
technologies, architectures and business models
of this area (Devine & Holmquist, 2001; EMILY,
2002; M-Guide, 2002). This chapter presents the
design aspects of such services in a generic way,
capturing the needs of many location-dependent
services since it assumes a highly heterogeneous
network infrastructure leveraged by the Internet
protocol (IP) layer. In this way, dealing with
the details of mobile or other wireless network
infrastructures is avoided yet interoperability
and integration issues are been identified and
investigated.
TECHNOLOGY OVERVIEW
In the following sections the technologies involved
in the provision of mobile tourism services are
introduced. Connectivity, which essentially allows
delivering data to a device, and positioning, which
is the ability to locate a device and consequently
its user in space, are the fundamental enabling
technologies for the provision of location-based
services. Assuming these, tourism related infor-
mation could be delivered to devices capable
of presenting it (e.g., mobile phones), with a
multitude of options (quality, depth, size, etc.),
derived from exactly the same content that would
drive traditional applications (Web sites, printed
elements, etc.). The driving force behind these
is modern software platforms and system archi-
tectures that facilitate the creation of the various
nodes of a complex structure of collaborating
service elements.
Wireless and Mobile Data Services
Since the last decade of the 20th century and the
beginning of the 21st, the mobile user has come
to enjoy the provision of many technologies and
services that were hard to even imagine several
years before (Lin & Chlamtac, 2001). Besides
voice, some of the most common ones are:
• Information services (News, Directories,
Weather, Athletics, Financial, etc.)
• Entertainment (Chat & Flirt Services,
Guess who, Alerts, Horoscope, ringtones,
etc.)
• Communication tools (SMS, MMS, e-
mail, instant messaging, etc.)
Apart from these common services, a series
of other, more complex ones are being offered to
the user like navigation, local news, SMS vote,
microbilling, and so forth. Enabling these services
is achieved through various means the most im-
portant of which being the Web, SMS, and MMS.
These higher-level information exchange media
are based on lower level communication channels
offered by an infrastructure provider. The most
important ones are briefly described below:
• Global system for mobile telecommu-
nications (GSM) refers to 2nd generation
mobile telephony networks (Mouly &
Pautet, 1995), which although digital, was
designed with voice communications in
mind, thus giving data rates of 9.6kbis/s,
which is rather slow for multimedia ap-
plications. Additionally data transfer is
261
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
not packet switched thus not optimised for
computer type communications requiring
circuits to be allocated even if no data are
exchanged. Since its initial appearance,
several enhancements where proposed as
side-by-side technologies that enable high-
er performance data transfers (Korhonen,
Aalto, Gurtov, & Lamanen, 2001).
• General packet radio services (GPRS) is
a wireless communication protocol based
on the same modulation as GSM, designed
to be provided as a complementary medium
to facilitate data transfers over GSM net-
works. It is packet-based and delivers data
rates of approximately 40kbps5 (Korhonen
et al., 2001; Pahlavan & Krishnamurthy,
2002; Patil, 2003; Tisal 2001). It supports
continuous connection to the Internet for
mobile equipment users. Since GPRS ra-
dio resources are utilised only when devic-
es have data to exchange, its end-user cost
is lower in both terms of money and power
consumption. Packet switching allows
more users to be simultaneously connected
to the Internet, yet performance drops on
high load and no strict guarantees can be
given.
• Enhanced data rates for global evolu-
tion (EDGE) facilitates high-speed mobile
data transfer over which can reach a peak
rate of 384kbps and is aimed to mobile net-
work operators that might not be able to
obtain UMTS (further information is pro-
vided below) spectrum but would not like
to be left out the modern high speed data
services (Halonen, Romero, & Melero,
2003; Rysavi 2005). Even higher speeds
may be available in good radio conditions.
EDGE provides the same benefits of GPRS
(e.g., packet switching, always connected)
however by using a different modulation
schematic achieves much higher speeds.
• High speed circuit switched data
(HSCSD) overcomes the limitation of
262
GSM circuit switched data, which supports
the allocation of one user per channel per
time slot and allows multiple channels to
be virtually merged thus offering higher
data rates (Halonen et al., 2003; Korhonen
et al., 2001). However the allocation of
multiple channels raises the connection
cost of the end-user, rendering the service
rather inappropriate when compared to
other modern techniques.
• Universal mobile telecommunications
system (UMTS) utilises WCDMA (wide-
band CDMA) over a 5MHz bandwidth
thus allows speeds the increase of mobile
network speed in order to allow high-
speed transfers. UMTS is one of the five
types of 3G radio interfaces specified in
the ITU6’s IMT-2000 recommendation. It
allows various classes of service, ranging
from more than 100kbps for a fast mov-
ing user up to a 2Mbps for a fixed client
“lab” speed (3GPP, 2002; Lin & Chlamtac,
2001; UMTS, n.d.).
• High-speed downlink packet access
(HSDPA) is deployed as an upgrade to
UMTS networks and captures the obser-
vation that most end-user high-bandwidth
demanding applications require one-way
high-speed communications, downstream
(i.e., towards the end-user). On-demand
video, TV and data downloading are some
applications that expose such a transfer
pattern thus can benefit quite significantly
from the speed offered by HSDPA which is
up to 3.5 times faster compared to the max-
imum rate of 14Mbps of today’s 3G UMTS
(Holma & Toskala, 2004; Kaaranen, 2005;
Rysavy, 2005).
• Wireless fidelity (WiFi) is a term that
in general refers to the 802.11 family of
wireless network protocols (Muller, 2003;
Smith, 2003). A variety of protocols (e.g.,
802.11b, 802.11g) that operate in 2.4GHz
and 5GHz bands, being the most popular
Design and Implementation Approaches for Location-Based, Tourism-Related Services

ones, belong to this family, offering nomi-
nal speeds of up to 108MBps (802.11.
Super-g). WiFi networks are fixed, local-
area, wireless networks thus do not of-
fer the mobility capabilities provided by
mobile networks. Additionally, although
they can operate without an infrastructure
(e.g., on a computer-to-computer way),
yet another (usually fixed) connection to
the Internet is required in order to obtain
worldwide access.
• Bluetooth (IEEE 802.15) is a short-range,
relatively low performance communica-
tions protocol. It is designed so that it al-
lows low power consumption and it is very
simple to implement so that it can be easily
adopted by “dummy” devices such as head-
phones, computer mice, keyboards,and
so forth. (Bluetooth.org, 2001; Miller &
Bisdikian, 2002; Morrow, 2002; Muller,
2001; Xiao & Pan, 2005).
Location-Based Services
Location-based services are an entire class of
mobile services that utilise positional (mostly geo-
spatial) information of the mobile user in order to
provide intuitive, easy access to content and tools.
The term covers a quite large group of services
since it can range from global scale services to
highly spatially constrained ones (within a room
or a building, such as a museum or a conference
center). See Table 1.
One of the main aspects of location-based
services is positioning, that is, the estimation of
the user position, a topic to be covered in detailed
sub-sequent sections. Depending on the service
class, the mobile equipment and the infrastructure,
different approaches and accuracy levels can be
applied (Dao, Rizos, & Wang, 2002; Northstream,
2001). In this chapter, the focus is on global scale
services, but the elements and concepts presented
are not restricted to this particular class of ser-
vices (Kakaletris et al., 2004). Most location-based
services can be categorised into four main busi-
ness applications:
• Tracking services (with two sub catego-
ries): emergency services and fleet man-
agement services. In the case of emergency
services (such as the E-911), the network
has the ability to locate persons who are
in danger or missing (with special inter-
ests for kids) and give them the necessary
protection (Reed, Krizman, Woerner, &
Rappaport, 1998). Emergency services are
first priority for USA (FCC, 2001) and US
companies focus on these class services.
Road assistance and tracking of stolen
equipment (cars, boats, etc.) are other sim-
ilar services (Fritsch & Scherner, 2005).
Fleet management services cover scenar-

Table 1. Typical wireless and mobile data networks

Technology Typical Performance7 End-user relative cost Scope Compatibility

GSM 9.6Kbps High Global Very high
GPRS 40Kbps Medium Global High
HSCSD 64Kbps Very High Global Low
EDGE 115Kbps Medium Global Low
UMTS 220Kbps Medium Global Medium (in deployment)
HSDPA 750kbps Medium Global Low
WiFi 11Mbps 8
Low Local/indoors High
Bluetooth 500Kbps Low Indoors High

263
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
ios such as radio-taxi coordination, trans-
portations, delivery, and so on, and in the
general case, require high accuracy (Feng,
Zhu, Mukai, & Watanabe, 2005; Iwasaki,
Kawaguchi, & Inagaki, 2005).
• Information services: In this category of
services content relative to the location of
the user is provided to him/her. However in
global scale services the focus is not on the
accuracy of user’s position acquisition but
rather on the content and the way it is pre-
sented (Taylor & Ryan, 1995). Local news,
cultural information, events highlighting,
or even advertising are some of the appli-
cations of this category. Such services may
be provided near sightseeing or within mu-
seums (Zimmermann, Lorenz, & Specht,
2003). In high granularity services (e.g.,
within museums) positioning has to be
accurate and in many cases highly sophis-
ticated since it might even need 3-dimen-
sional location of the user and directional
clues (Pateli, Giaglis, & Spinellis, 2005).
• Fun and entertainment: Player position-
aware games are a new opportunity for
service providers and the first flavors are
already out. Despite the criticism, chat and
flirt is another very popular type of ser-
vice. In this area, location-aware systems
will have the opportunity to refine partner
matches within a certain distance (Gratsias,
Frentzos, Delis, & Theodoridis, 2005;
Karagiozidis, Markoulidakis, Velentzas,
& Kauranne, 2003; Lee, Prabhu, & Park,
2002).
• Billing: Billing also can adopt location-
aware schemes. Creating attractive options
such as allowing users to exercise cheaper
communications when in certain hot spots
(such as home, etc.) is a possible scenar-
io of location-based billing (Gratsias et
al., 2005; Koutsopoulou, Panagiotakis, &
Alonistioti, 2005).
264
Positioning
Location-based tourist information requires posi-
tioning of the mobile user with a variable accuracy
(Kakaletris et al., 2004; Yilin, 2000). Yet this does
not imply that automated acquisition of positional
information is always required or desired in order
to consume such a service, as when checking
available information by forehand, not being
physically present in some area. It is obvious that
locating the user requires that one is equipped
with a module that has some kind of connection
to a fixed infrastructure (Figure 1). However the
elements of this composition might vary:
• Equipment can be a PDA, a 2G or 3G mo-
bile phone, a personal computer, an ID tag,
an IR Scanner, a GPS receiver, and so on.
• Connection to the infrastructure can
be unidirectional or bidirectional utiliz-
ing systems such as WiFi, GSM, GPRS,
UMTS, satellite antenna/receiver, IR re-
ceiver/transmitter, and so on.
• Infrastructure can be mobile or fixed net-
work such as a satellite system, a mobile
telephony/data network (GSM, GPRS,
UMTS, etc.), a set of WiFi or Bluetooth
access points, installed RF tags, fixed IR
scanners, and so on.
Positioning can be categorised under two
main classes:
• Active: The client is the only one respon-
sible for the computational load of the po-
sition estimation.
• Passive: The client (user/mobile equip-
ment) is being located without adding any
logic to positioning by itself. Infrastructure
obtains simple to extract information from
client and calculates the position.
Hybrid methods also are very popular, allow-
ing for high accuracy and availability systems.
Design and Implementation Approaches for Location-Based, Tourism-Related Services
Figure 1. GPS operation
Exploitation of the positional information also
can vary quite significantly:
• In self-contained systems, such as GPS
enabled PDAs, the content might be al-
ready present on the mobile equipment
and a local piece of software acts on them,
thus no further connection to the outside
world is required. The range of location-
based services to be supported in this case
is limited, navigation being the most popu-
lar one.
• In always-connected systems a medium
for exchanging information with the in-
frastructure and usually the world (the
Internet) is required. This can be done in
order to acquire the position, or exchange
information that will allow position esti-
mation, or access the core service/content.
The rest of this section presents the details of
some of the most important positioning methods,
their requirements, and their characteristics. There
are a number of ways for obtaining user’s position
and the following sections describe some of the
current applied automated methods:
The GPS
GPS positioning (ETSI, 2006; GARMIN, n.d.) is
based on a network of 24 earth orbiting satellites.
It was originally designed and implemented to
cover the needs of the US military forces, however
since the 1980’s it has been used for wide spectrum
of applications ranging from civil engineering
to recreational systems. Communication is one-
way; consequently clients only receive signals,
thus guaranteeing privacy. A GPS client/device
receives signals from several satellites with at
least three satellites needed for 2-dimensional
coordinate estimation (latitude, longitude) while
four are required for three-dimensional position-
ing (latitude, longitude, and altitude). Signals
are time-tagged, enabling the client to calculate
distance from each satellite using the send/receive
time difference. Accurate position estimation
requires combining the aforementioned distances
for multiple satellites. Bookkeeping of coordinates
is a way to calculate the speed and direction of
moving GPS devices, a facility provided by almost
all modern receivers. In such systems end-user
devices are quite small and due to low power
consumption tend to be highly autonomous.
265
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
Civilian GPS satellite signal is low power
(i.e., less than 50 watts in 1575.42 MHz). It pen-
etrates glass and plastic, but cannot go through
metal, ground or concrete, effectively preventing
indoors GPS usage. Dense urban use also can be
problematic in some cases. Additionally, the GPS
signal is vulnerable to signal travel time errors
that lead to false distance estimation. Reflection
of signal on solid objects as well as other orbital/
stratospheric phenomena can result to wrong time
estimations. Utilizing more satellites is a way to
deal with such problems.
One of the strongest points of GPS however is
accuracy, which can range from 1 to 15 meters for
civilian systems. This fact, accompanied by the
simplicity of the relevant devices and the avail-
ability of the service which can be utilised at no
cost (apart from equipment) makes it capable of
driving major successful commercial applications,
such as navigation for tourists. As already men-
tioned, GPS is controlled by the U.S. Department
of Defence, nevertheless it is expected that the
European Galileo system will be competing with
it by 2008 (Di Fazio, Mocci, Rossini, D’Angelo,
Lorelli, & Jarosh, 2004; El-Rabbany, 2002; Prasad
& Ruggieri, 2005). GPS has to be accompanied
by a supplementary network in order to drive
interactive on line systems where satellite cover-
age is unavailable. Pseudo-GPS systems emulate
the existence of satellites for indoors use without
requiring any additional equipment, yet they are
not widely adopted solutions (Schmid, Neubauer,
Ehm, Weigel, Lemke, Heinrichs, 2005).
GSM-Positioning
GSM positioning (ETSI, 2006; Mao & Douligeris,
2000; Spirito, 2001; Zhao, 2002) is a facility
potentially provided by GSM mobile network
operators. Its operation is based on the fact that
there is always some type of raw information on the
location of a certain GSM mobile device in order
for the network to be able to deliver information
to the user. Since it can be originated from the
266
network, it is raising serious privacy and security
concerns, which can be overlooked for emergency
purposes but not without risk of misuse.
GSM positioning is always available in some
form, as long as network coverage is provided.
However depending on the network infrastruc-
ture and method utilised its accuracy might vary
quite significantly ranging from 100m to more
than 500m (even several kilometers) (Caffery &
Stuber, 1998b; Drane, Macnaughtan, & Scott,
1998). Although due to this low level of accuracy
GSM positioning is of little use for high accuracy
demanding application, it is accompanied by a
bidirectional communication channel (voice or
data) thus enabling interactive applications. Its
network-side activation, whenever applicable,
makes it ideal for some special emergency and
tracking cases.
• Cell ID is a location technology that
utilises the well-known location of fixed
network elements, that is, the base station
transceivers (BTS), to identify the mobile
equipment location (Figure 2). It can be
easily combined with timing advance (TA)
in GSM networks and round trip time9
(RTT) information in WCDMA networks
in order to improve accuracy. TA is a tech-
nique that utilises the timing advance in-
formation applied by the GSM network to
determine the approximate distance a MS
is from a base station.
• Enhanced-observed time difference
(E-OTD) is a more complex method for
calculating device position (Caffery &
Stuber 1998a, 1998b). It requires loca-
tion measurement units (LMUs), a type
of device used to provide precise timing
information for asynchronous networks.
Although this method can increase accu-
racy, it also increases infrastructure cost
and it still suffers from issues with network
density in rural areas. E-OTD and time of
arrival (TOA) methods are very similar to
Design and Implementation Approaches for Location-Based, Tourism-Related Services
Figure 2. Cell ID/Cell ID + TA
GPS positioning already described. The
location of mobile devices is calculated us-
ing the signaling time from two or more
stations (see Figure 3).
Typically GSM-positioning would not always
be enabled, since this requires significant re-
sources from the operator’s side. In the typical
usage scenario, the user has to ask for locating
him/her, either through an SMS or through a re-
quest over the standard Internet protocol (IP). Due
to its privacy risk, GSM positioning is usually
surrounded by rules of activation and bureau-
cratic procedures/agreements, which can be even
harder to overcome in cross-MNO cases unless
a standard user-driven mechanism is utilised.
Unfortunately despite the existence of specifica-
tions, in mid 2001 decade there is still no standard
for GSM positioning that is widely adopted by
operators. A final restriction on its use is its cost,
which can vary quite vastly depending on the
level of consumption of the service.
WLAN
Wireless local area network (WLAN) position-
ing (Wang, Jia, Lee, & Li, 2003) is a local type
of positioning usable within range of WLAN hot
spots, which is mostly restricted indoors or in quite
Figure 3. E-OTD operation
constrained areas (e.g., building blocks, etc.).10 It is
the primary candidate technology for big buildings
like airports, museums, market places and other
sites where radio interference is not an issue and
it should generally be accurate enough to guide a
visitor through the coverage area. Positioning is
very similar to GPS and GSM positioning, thus it
uses the time difference between the signals from
some known positions. This information can be
corrected and improved with site calibration, a
method which also can be utilised in some GSM
positioning cases but not without significant dif-
ficulties. In the WLAN positioning case, informa-
tion about the physical structure of the target area
(walls, obstacles, etc.) is of crucial information
in order to obtain reliable results.
According to systems’ vendors, WLAN posi-
tioning can be quite accurate (one meter indoors
according to Ekahau, 2006) yet site calibration is
definitely required for such accuracy. Its cost is
quite reasonable since infrastructure and equip-
ment are quite cheap and it is well suited for in-
teractive on-line applications since data network
coverage is implied.
Other Methods
Despite its peculiarity considered in the context
of the e-services described above, observation ac-
companied by a traditional map (i.e., paper-print-
ed) also could be a useful method of information
267
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
dissemination for tourists even in the electronic
age. This does not suffer from service unavail-
ability, privacy, mobility, and autonomy issues,
and its price is usually low. Using an electronic
map instead of a printed one would add in detail
and in some cases reduce in size while posing
some restrictions on autonomy, however if not
accompanied by a positioning system it would
still require observation and manual operation in
order to consolidate other information.
Positioning through mapping addresses to
geographical coordinates is very similar to using
maps since it requires observation and manual
utilisation of information. However the location is
obtained through lookups in appropriate databases,
such as geographical information systems11 (GIS).
It is quite usable in urban areas and roadsides
of rural areas and it can be both electronic and
traditional (maps with indices).
There are several other methods, like local sen-
sor networks, ID tags based on infrared or radio
frequencies (IR or RF), gyroscopes, and statistical
models (Hightower & Borriello 2001; Nellen,
Bronnimann, Held, & Sennhauser, 2004; Spratt
2003). Out of these one could expect low-cost
gyroscopes to be the most attractive mechanisms
for mobile travelers in the future (whenever it
becomes possible to build lightweight, low-cost
gyroscopes) since they could drive the most
fascinating classes of applications. For example
taking into account the direction of sight of the
user could even further enhance the simplicity of
interaction with a service, especially in complex
(with respect to location and quantity of content)
environments such as museums or exhibitions.
Further issues and opportunities can be found in
the section on “Mobile Tourism Services.”
Summary
Each method presented has advantages and
shortcomings. It seems that GPS is currently the
primary means of reliable positioning when global
positioning comes into play, however hybrid po-
268
sitioning (i.e., by blending of different location
technologies) utilizing GSM and WLAN meth-
ods could be optimal. In the rest of this chapter
“hybrid” will refer to combining GPS with other
location techniques.
There are several whitepapers, guidelines,
studies, and surveys showing the details of each
positioning approach (Caffery & Stuber, 1998b;
Hightower & Borriello, 2001), however a higher
level evaluation of the ones applicable to tourism
as it appeals to mobile network operators will be
introduced below. Other types of systems like
personnel/patient tracking in hospitals (Dao et
al., 2002; Douglas, 2004), information delivery
in museums (Sparacino, 2002) or emergency
related ones (FCC, 2001; GSM, 2002) generally
require more sophisticated means for delivering
their services. See Table 2.
When designing an open infrastructure then,
positioning should not have any strong relation
to the implementation of services for the mobile
traveler. Realisations of different services should
leave methods of getting positioning information
open, allowing the end-user to choose among
them: GPS, GSM and map outdoors, WLAN and
GSM inside. Services should be prepared to re-
ceive coordinates as bunches of numbers in the
three-dimentional + direction space accompanied
by definition of coordinate-standard and accu-
racy.
Such an approach would leave all possibilities
open for the user. Also this method would keep
service providers’ monetary costs down, because
service provider can leave all costs of positioning
to be decided by the traveler.
When building services for the mobile traveler,
designers should ideally therefore forget the actual
positioning method and provide a service in a way
that it would be able to cope with any degree of
accuracy or availability of positioning information.
Obtaining the position would then be a task to be
carried out by external parties such as equipment
manufacturers or infrastructure providers, both of
Design and Implementation Approaches for Location-Based, Tourism-Related Services

Table 2. LBS summary of performance, implementation, and cost trends (Sources: EMILY, 2002; High-
tower & Borriello, 2001; Moureu, 2000; SnapTrack, 2001)

Metric CELL-ID + TA E-OTD GPS A-GPS Hybrids

100m 100m 10m 10m 1m
Accuracy
10km 500m undefined 100m 100m
Kick-off time Low Low High Very high Very high
Dimensions 2 2 3 2/3 2/3 (+direction)
Mobile Equipment Require-
None Low High High High
ments
Applicability Good Poor High Medium Good
Infrastructure Cost None High None None-to-high High
Urban / Dense
Best area applicability Dense urban Suburban Suburban, urban Outdoors, Indoors
urban
Coverage High High Partial High High
Overall Quality Poor Medium High High Excellent
Compatibility Medium Poor Full Medium Poor

them working on low details of the software that
manages the “networking” hardware.
Positioning Standards
An attempt to standardise positioning has been
carried out during the past few years. Solutions
and best practices have been proposed by vendors,
associations and individual researchers (Adams,
Ashwell, & Baxter, 2003; GSM, 2002; Rao &
Minakakis, 2003). As a consequence the terms
mobile positioning center (MPC) and gateway
mobile location center (GMLC) systems emerged
in order to label the entities of the infrastructure
that where used in order to extract positioning
information and estimate end-user locations. The
mobile location protocol (MLP) is an application
protocol developed by the location interoperabil-
ity forum (LIF) that specifies the messages to be
exchanged over standard Internet technologies
in order to obtain coordinates from a positioning
center. It is focusing on network based position-
ing and has its origin in proprietary commercial
protocols. OpenLS (OpenGIS location services

Figure 4. Indicative LBS technologies quality per area (accuracy, availability, calculation time)12

269
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
[OGS]) defines the interfaces to an open location-
based services platform, such as the GeoMobility
server. Utilizing XML for location services (XSL)
defines the data exchanged in the interactions
with the geographic information system (GIS) for
various classes of services that capture facilities
such as routing (navigation), position extraction,
lookup (directory), map drawing (presentation),
and so on.
Popular, yet proprietary protocols such mobile
positioning protocol (MPP) (Ericsson) also can
be an option since many existing systems sup-
port these.
Content Technologies
One of the major aspects of mobile tourism ser-
vices is the delivery of content to the end-user.
This in the general case involves identifying the
content to be displayed, (optionally) moving it
to the device14 and presenting it. In the following
paragraphs focus is given on the case that content
is passed to the device upon request, that is, does
not reside pre-installed on the device.
Figure 5. LBS technologies services and accuracy13
270
Delivering content of location-based services
to the end-user is not as simple as it seems at first
sight. There is a multitude of devices that content
can be delivered to and a number of media through
which it can be transferred, all of them presenting
different capabilities and shortcomings. Typical
examples are SMS, MMS, and WAP:
• SMS market penetration is significantly
higher than the penetration of WAP. The
majority of the location-based systems im-
plemented by MNOs, use the SMS for ex-
changing data. However presentation and
interaction facilities are severely limited
over this transport medium; furthermore
SMS can be extremely expensive per byte
of data exchanged
• MMS significantly increases the quality
of presentation and reduces the price per
byte of exchanged data. Yet its penetration
is significantly lower than SMS, it suffers
the same interaction restrictions and finally
its cost can be significant over SMS for
simple text information exchange
Design and Implementation Approaches for Location-Based, Tourism-Related Services
• WAP adds to presentation and interactivity
quite significantly. However its penetration
is quite limited and it is not expected to be
widely adopted unless significant conver-
gence to Web technologies is achieved
Research on service types show that the SMS/
MMS solution might be more appropriate, while
WAP/Web is the only way for others (Bennett,
2001; Heijden & Taylor 2000). Taking into ac-
count that current end-user mobile terminals have
limited displays and end-users are not accustomed
to consuming interactive services through the
mobile phone interface, SMS/MMS may be a
commercially viable solution for a wide range of
location-based services. Special sub-sections of
tourism related services such as directory lookups
can be ideally served through simple SMS ex-
changes.
Positioning also is influenced by device ca-
pabilities. The majority of mobile phones are not
equipped with GPS devices, which could have a
significant impact on their autonomy, size, and
weight since the cost of a GPS receiver is insignifi-
cant compared to the cost of the rest of the device.
Even PDAs or laptops do not usually come out of
the box accompanied by GPS receivers, making
it a requirement to purchase one separately. The
primary option is a Bluetooth GPS device that
can be easily connected to portable computing
devices such as latest mobile phones, PDAs, or
laptops; however despite the small size of these
devices this scenario requires that the user carry
one extra component.
Mobile Terminals That Support LBS
The development and full deployment of high
quality location-based services for tourists as-
sumes the existence of terminals with extended
capabilities in battery capacity, processing power,
main memory, application capabilities, size, and
weight. Such devices may be categorised accord-
ing to an incremental rating of the applications they
can support. Today, four categories of terminals
can be found in market:
• Standard GSM mobile terminals
• GPRS/UMTS phones: There are two cat-
egories of GPRS/UMTS phones. The WAP
based and the HTML based ones. The for-
mer category offers facilities for consum-
ing WAP information sources (WAP sites)
while the latter one exchanges information
in Web standard formats such as HTML,
JPEG, etc. The require a Web server, which
will host the mapping applications and will
be responsible for routing requests to ex-
ternal systems.
• PDAs/Smartphones: Smartphones are
latest generation mobile phones equipped
with enhanced displays, more powerful
processors and capabilities of carrying out
tasks that were formerly responsibility of
computer systems. PDAs are devices in
some extent similar to the smartphones,
stressing on presentation and application
capabilities rather than communications.
Latest generation devices are usually
equipped with an HTML browser able to
interface with standard Web sites. PDAs
offer built in or expandability options for
GPS, GSM, WiFi, and so on.
• Laptops with communication facilities:
These offer all the advanced options of
computers in terms of computational and
storage capacity, yet they have limited mo-
bility features.
Multi-Channel Content Delivery
Multi-channel content delivery is a very important
aspect of the content provision system. Infrastruc-
ture and content should be suitably constructed so
that they allow delivery of information through
various media and means, be it SMS message
exchanges, GPRS, or fixed networks connections,
computers/PDAs/phones, and so on. Nowadays
271
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
there is a whole suite of technologies and a large
number of systems that enable multi-channel
content delivery, under the author once—publish
many approach, where the same source content
is used for several targets ranging from printed
media to mobile devices.
Nowadays the most widely applied solutions
for this approach are based on a set of standards and
tools build around XML, a simple text language
developed by the World Wide Web Consortium
(W3C) that has its roots on SGML. XML can be
used with a wide range of applications, such as au-
thoring tools, content display engines, translation
tools, and database applications (Anderson 2000;
Bosak 1998; Bosak & Bray 1999). However, its
major use in the context of mobile tourism services
is for structuring the tourism related content in a
form that it becomes meaningful for a series of
software elements and operations. More precisely,
the various parts of a document are labeled in a
special manner (tagged) so that can be extracted
by various applications to fulfill their own needs
(Anderson, 2000). A relevant technology is the
well-known HTML which uses a somewhat similar
tagging mechanism yet it neither enforces structure
nor allows for application specific labeling that
would facilitate a customised use of the document.
HTML uses a fixed set of tags and describes only
formatting instructions for the Web. On the other
hand XML is completely extensible since it does
not have predefined tags, unless it explicitly con-
forms to a schema. Various schemas exist in order
to capture industrial and domain specific needs
and as long as two systems comprehend the same
schema they can use XML documents to exchange
information. News Industry Text Format (NITF)
and News Markup Language (NewsML) are two
examples of markup languages that could be of
some use to tourism related services, since they
enable the sharing of syndicated news articles
from alternative news providers.
In addition to XML, the combination of XSL/
XSLT is one of the fundamental parts of the
described content delivery chain. A stylesheet
272
is a declaration that describes how to display a
document and XSL is a language for expressing
stylesheets for XML documents. By keeping
content and its presentation information separate
in the XML document and the XSL stylesheet
respectively, display of the same original docu-
ment on different media and preferences sets is
achieved by simply modifying the stylesheet.15
However modifying presentation might not be
enough in several cases; transformation of the
source document might be required and this is
where XSLT comes into play. Although XSLT
transformation was originally intended to perform
complex styling operations, like the generation
of tables of contents and indexes, currently it is
often used as a complete general-purpose XML
transformation language. In that manner it is the
standard means for transforming the source XML
document (e.g., tourism content) into presentable
forms such as HTML or WML16 pages, or other
XML schemas or even into other more complex
formats.
Apart the content structure, a number of fea-
tures are exploitable in the context of establishing
advanced, reliable multi-channel content delivery
services, such as mobile tourism related ones:
• Security, in order to authenticate users and
authorise access to sections of the system
and potentially charge usage, required in
order to safeguard mobile user’s privacy
and content/service providers resources
• Session management, in order to drive
stateful applications driven by a series of
user actions, a feature required for almost
every type of modern interactive Web
application
• Automated device/channel detection, in
order to enable adaptation to device ca-
pabilities, with minimal user intervention
that drive better application experience
• Client and/or server form processing
and validation that allows meaningful in-
Design and Implementation Approaches for Location-Based, Tourism-Related Services
teraction with the system while minimiz-
ing message exchanges
• Content rendering that adapts to the de-
vice capabilities and user preferences
(personalisation) required for delivering
content to a multitude of heterogeneous
devices
• Support for off-line operation of devic-
es, able to drive different application and
connectivity scenarios, especially useful in
several forms of application in the mobile
tourism area17
Unfortunately, it can be shown that with today’s
applied technologies, depending on the device/
channel, almost all requested features have to be
achieved by a different set of technologies and
interactions each time. Approaching this problem
through the Model-View-Controller design pattern
(Gamma, 1995) offers a potential design solution
to the challenge, by breaking an application into
modules so the logic can be dealt separately from
the interface. This allows developers to maintain
a single code base for all channels and face each
device/channel as a different view for a given
transaction. The Model-View-Controller pattern
can have as many views as desired and isolates the
request handler (aka the Controller) and enterprise
system (the Model) from any knowledge of the
view. Achieving the desired features (e.g., security,
session management, form validation, and content
rendering across all devices and channels) is case/
device/channel specific.
In the proposed multi-channel approach con-
sidering the Web as the primary channel allows
an all-in-one solution for already established Web
content provision services. The value of a multi-
channel content delivery enabling platform is
that it provides to the hosting enterprise a single
interaction model for all external systems or users
and leverages security, session management, and
content rendering.
Commercial, freeware and open source soft-
ware markets offer a wide range of tools that relate
to the multi-channel content delivery process.
One can easily locate a series of software tools
that transform rich content into tailored formats
mainly for adopting Web content to mobile devices
or vice versa, or systems that offer some degree
of support for multi-channel content delivery.
These systems mostly belong to one of the fol-
lowing categories:
• Content management systems (CMS)
• Authoring and publishing tools
• Wireless application gateways (WAG)
Platforms such as enterprise portals and e-
commerce suites also might be offering some
degree of support for multi-channel content de-
livery, however they are considered of rather little
use in the case of rich mobile tourism services by
themselves, since they are capturing this aspect
under the perspective of their specific application
domain. Such facilities also can be offered by ap-
plication servers, however at a much lower level.
Currently there is no product that provides a full
solution that can take raw content and transform it
automatically to multiple formats for any request-
ing device taking into account all the aspects of
multi-channel content delivery referenced in the
previous section.
Content Types
There are various types of content that can be
delivered and displayed by end-user devices. Text
is the easiest to handle, however it is relatively
less attractive compared to the others due to its
nature, which only uses vision through the chan-
nel reading. On the opposite side, rich multimedia
content such as Flash18 (Against the Clock [Firm],
2002; Kozak, 2002) is highly attractive, however
its current support by mobile devices is not global,
yet it is expected to increase in the near future,
as convergence of technologies dictates. Direct
execution of “binary” code on the devices through
industry standards also can be referenced, however
273
 Design and Implementation Approaches for Location-Based, Tourism-Related Services

support is relatively limited and minimal or no
cross device compatibility is provided.
The Semantic Web
The Semantic Web (W3C, 2001) comes into the
scene in order to cover, among others, many of the
issues met in the multi-channel content delivery
area. It is led by W3C and provides a whole frame-
work so that data obtain a well defined meaning
and can be reused under different cases potentially
leading to quite different end-user experiences and
content compositions. The key-enabling concept is
the resource description framework that is heavily
dependent on XML contracts. The Semantic Web
can be considered as the next step of the current
World Wide Web (Berners-Lee & Hendler, 2001;
Berners-Lee et al., 2001).
Software in Control
All of the above mentioned technologies are pro-
vided through software elements which might be
either proprietary, bound to a specific device and/
or case or generic enough to fit several scenarios
or even totally irrelevant services. Commercial or
open source platforms do exist that can support
almost all of the stages of mobile tourist content
delivery, with little or even no need for integra-
tion development.20 Nevertheless, provisioning
of custom services will most probably demand
explicit development of custom software elements.
In scenarios that several market players are
involved for the provision of a rich service, such
as the case of mobile tourism ones, the traditional
component oriented architecture, which modular-
ises software in order to be reused among various
cases, needs to be enriched by concepts that allow
independence among the various stakeholders.
The concept of service-oriented architecture
(SOA) captures architectural style that aims to
achieve loose coupling among interacting software
entities (Erl, 2004, 2005; Newcomer & Lomow,
2005). In this context a service is a unit of logic
and data provided by the service provider that is
assigned to provide the appropriate results for
its consumer. In this framework the provider and
consumer are both software agents that act behalf
of their owners. As a concept SOA is quite abstract
and can be utilised both for B2B and B2C interac-
tions. Typically lots of protocols and operations
fall under its meaning, however in mid-2001, more
concrete forms of service oriented paradigms are
being applied, accompanied by relevant protocols
and enabling platforms.
The model of publisher/subscribers, assisted
by a directory service over a transport consisting
mostly of XML and Internet based protocols is
the current roadmap for applying the SOA:

Table 3. LBS content types (Source: M-Guide, 2002)

Power
Content class Compatibility Features Size Sample technologies
requirements
Text Very high Poor Very Low Very low HTML, Plain text
Audio High Medium Low Medium MP3, MIDI
Image High Medium Low Medium JPEG, WBMP, GIF
Video Medium High High Very high MPEG, 3GP19 etc
Integrated interac-
Poor Very high Medium Very high Macromedia Flash
tive multimedia
Applets / ActiveX Compo-
Executable entities Very poor Very high Medium Very high
nents

274
Design and Implementation Approaches for Location-Based, Tourism-Related Services
• Service publishers (might) announce their
characteristics in a directory service (e.g.,
UDDI, 2004).
• Service consumers can either look in the
directory service and locate the service
that meets their requirements and obtain a
reference to the access point of the service
or they directly obtain this through other
media.
• Consequently consumers use the access
point reference to directly communicate
with the server.
• SOAP over XML over HTTP (i.e., TCP/
IP) are the most common protocols for
consumer/provider interaction, yet mes-
sage details are service dependent.
Recent extensions of the initial concept of Web
services for the SOA approach add state manage-
ment, thus allowing a standardised way to access
resources hosted by suppliers (Czajkowski, Foster,
Ferguson, Frey, Graham, & Snelling, 2004; Foster,
Frey, Graham, Tuecke, Czajkowski, & Ferguson,
2004; Huber & Huber 2002; Snell, Tidwell, &
Kulchenko, 2002). Typically consumers have no
internal knowledge of the operation of the service
provider and a particular service in question. The
provider might revise implementation, as well as
other characteristics of a service without explicitly
notifying clients.21
One level below the overall software archi-
tecture is the software technology, which in the
case of location-based services has two facets:
infrastructure and client. Developing on the
infrastructure side is facilitated by numerous
technologies and platforms and is of little interest
from the mobility perspective. All Web service
enabling technologies, assisted by technologies
to deliver reach content are quite applicable for
the purposes of tourism location-based services.
Dozens of open-source free or fully commercial
platforms exist in the today’s software market,
making it available for anyone to adapt its in-
frastructure to their own needs and capabilities.
However, due to its relatively recent appearance,
development on the device has somewhat limited
options. For the time being, the most common
platforms for mobile device development are the
two described bellow and come from commercial
software vendors, however they are both freely
available to the development communities:
• The micro edition of the Java 2 platform
is the port of Java to portable devices that
have minimal capabilities of processing
and storage. It is a runtime environment
that provides on one hand a subset of the
facilities offered by the typical desktop
java engines, yet it provides microdevice-
specific capabilities access, potentially as-
sisted by vendor specific libraries. J2ME is
very common in latest 2G and 3G mobile
phones as well as non PocketPC PDAs
(SUN).
• The Microsoft.NET compact framework
is a subset of the Microsoft.NET frame-
work, the Microsoft technology for desk-
top and Web application on the Windows
platform. Although meant to be platform
independent through standardisation
(ECMA22), currently.NET compact frame-
work is only available for PocketPC and
other flavours of the Windows platform.
Due to its requirements it is not suitable
for current low-end devices, yet its perfor-
mance is very promising in addressed de-
vices (Microsoft).
MOBILE TOURISM SERVICES
“Tourism-related” content refers to any content
that might be of interest to the visitor of a location
(Beatty 2002; Kakaletris et al 2004; Karagiozidis
et al 2003; Kim, Kim, Gautam, & Lee, 2005). This
content, be it static (e.g., maps, routes), of low
refresh rate (e.g., scheduled or periodic events),
highly dynamic (e.g., traffic), cultural (e.g., muse-
275
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
ums, exhibitions, local historical info), informative
(e.g., weather, local news) or commercial (e.g.
restaurants, commercial fun parks, etc.), has an
increased degree of interest to the “consumer”,
especially when he/she is in particular locations.
The term “location” is a varying-“size” descriptor
which might range from the actual spot where one
is standing in a room, to just a rough approxima-
tion of a country. In this context of course, the
term “tourist” is very loosely related to the typical
“tourist” and mostly fits the definition of “mobile
equipment user”. It is obvious that a very large
number of services related to content delivery fit
this definition, such as map delivery, archaeologi-
cal information delivery, events announcements,
emergency, or health services, transportation
information acquisition, even in-doors museum
exhibition presentations. These are all forms of
this type of location-based tourism-related content
delivery service.
When designing an infrastructure able to sup-
port such services, requirements that have to be
met come from various sources:
• Technical issues that have to be ex-
ploited and addressed in order for it to be
applicable
• Regulatory and social/ethical restric-
tions that have to be met in order for the
supplied services to be publishable
• End-user expectations and require-
ments that have to be considered in order
to achieve successful commercialisation,
and so on.
The requirements presented here arise from a
careful study of the 2/2.5G and 3G mobile net-
work technical specifications, location-based ser-
vices related whitepapers and applied paradigms
(Laitinen, 2001; Ludden, 2000; Northstream,
2001; Searle, 2001), modern IT system architec-
ture concepts (Varoutas, Katsianis, Sphicopoulos,
Cerboni, Kalhagen, Stordahl, et al., 2002; Varou-
tas et al., 2006), regulations & standards (FCC,
276
2001; GSM, 2002; IST-REPOSIT, 2001; LIF-IOT,
2002), and finally, market and end-user surveys
(Beatty, 2002; Dao et al., 2002; Katsianis et al.,
2001; LoVEUS, 2002; M-Guide, 2002). This
comprehensive study led to the identification of
the following key points that require careful con-
sideration when implementing a location-aware
service that aims to provide tourist information
content:
• QoS in quantifiable technical terms (re-
sponse time, throughput, availability, scal-
ability, coverage, accuracy, etc.)
• QoS in non-quantifiable terms (qual-
ity of content e.g coverage, depth, media,
multilinguality/multiculturality, etc.)
• Integration capabilities (relevant services
support and integration capabilities)
• Security (authentication, authorisation,
privacy, etc.)
• Service related procedures (e.g., activa-
tion / de-activation, billing, pricing, per-
sonalisation, stakeholders interoperability,
etc.)
• Service specific features (e.g., notifica-
tion, positioning triggering)
• Content related issues (e.g., ontology)
• Present and emerging technology capa-
bilities (positioning, presentation, mobile
networks, etc.)
Having carefully examined alternatives and the
potential impact of decisions on the above-men-
tioned key points, several interesting conclusions
can be identified as referenced below. Although
technology enthusiasts would think otherwise,
end-users are not generally amazed by high ac-
curacy positioning mechanisms or very high data
rate demanding services. The already referenced
user surveys show that they rather prefer low cost
intuitive services that will satisfy their expecta-
tions, within reasonable quality limits. In the
same context, widely used equipment lacks high
interactivity capabilities (i.e., means of passing
Design and Implementation Approaches for Location-Based, Tourism-Related Services
information to the service and presenting content
to end-users). Although emerging devices are
expected to vastly improve this particular aspect,
average users are not enticed by such capabilities.
High accuracy in positioning is not always a
real concern either. For example delivering some
sort of transportation information or providing a
local directory service, requires only a rough ap-
proximation of the position. On the other hand,
presenting information in an actual exhibit, which
the visitor is looking at, requires not only precise
position as to where one is located but also infor-
mation about the direction of sight, and even that
detail of information may not be adequate. At this
point, the conclusion is that “less accuracy” does
not render LBS useless but rather restricts the
range of services that can be provided. Therefore,
although almost identical content might be avail-
able for presentation to each end-user visiting a
museum or an open archaeological site, equipment
differentiation (e.g., display capabilities, channel
usage, positioning mechanism, and capabilities),
can drive a completely different degree of content
exploitation.
Another issue is that some requirements,
for example accurate positioning and enhanced
content, come with some extra “cost” to the end
user. Cost is not only in financial terms, which
are mostly obvious, positioning accuracy (e.g.,
through a GPS module) and advanced display
capabilities for example, might have a severe
impact on device autonomy and size. Enhanced
content (video, virtual reality representations,
etc.) will require more bandwidth and even more
powerful equipment (having indirect impact on
autonomy and size).
When to locate a user is yet another confusing
issue. User-requested positioning requires an extra
step to be taken by the end user, who actually has
to request to be located or explicitly state his/her
location at the moment. Alternatively, the user can
be continuously located (tracked) by the system.
Continuously positioning the mobile user is not
usually required, even in the case when the un-
doubted demand it implies in terms of infrastruc-
ture and equipment resources is not a problem.
A typical example is a highly mobile user who is
accessing some location-based directory of con-
tent, related to a specific position he/she has been
some time earlier. While the user may be attracted
to these particular sites of interest, mobility might
cause the delivery of quite different content in
subsequent requests for information, which could
come to be quite irritating. Thus the user should
have the possibility to allow either continuous or
on-request triggered positioning. A good practice
would be to totally avoid automatic positioning
instead of temporarily ignoring new position
information, however certain applications could
require two positional indicators (“live” position
and “querying” position). The issue of delay also
needs to be taken into consideration when locating
the user. In on-demand positioning schemes, the
system might introduce quite significant delays
due to high load or computational issues which
might render some applications rather unattractive.
Finally, a very important observation that
guides the proposed approach is network con-
vergence; xxML/HTTP/TCP/IP “stacks” tend to
provide a uniform roadmap for offering a very
large range of services to each-and-every consumer
oriented device type.
System Concepts
In the rest of this chapter various concepts of an
open architecture, capable of providing rich mobile
tourism services, are pointed out. Although an
architectural approach is introduced, the intention
is not to propose a single solution that could be
identified as the best in one way or the other for
the tourist; the purpose of the analysis to follow
is rather to position the enabling technologies
and identify the issues, shortcomings and strong
points of various approaches. Depending on the
particular application one has to provide, differ-
ent roadmaps can be followed. An hypothetical
structure of an LBS provision system targeting
277
 Design and Implementation Approaches for Location-Based, Tourism-Related Services

tourism content delivery is shown in Figure 6. In
this scenario, the elements are separated in three
conceptually different clusters:
• End-user devices, be it mobile or fixed
ones. For simplicity, it will be considered
that mobile ones have some sort of posi-
tioning capabilities either on their own
(e.g., GPS) or network assisted/provided.
• Network infrastructure, which is either
a mobile network or a fixed wireless local
area network, or a combination of both. In
case of mobile equipment that lack posi-
tioning capabilities, it is necessary that the
network provide the ability to locate the
user through connection information, spa-
tial information and potentially the user’s
previous behavioural data. The network
might provide various means of submit-
ting information to the user, such as MMS/
SMS (GSM, 2003).
• Service infrastructure, which relates to
the systems that delivers information and
facilities to the user utilizing as one of its
implied inputs, the positional information
of the user.
The mobile/wireless network details such as
access points, MSCs, VLRs, HLRs and so forth
are almost totally hidden and only interactions of

Figure 6. Example of layout of LBS system (Based on Kakaletris et al., 2004)

278
Design and Implementation Approaches for Location-Based, Tourism-Related Services
higher-level components are being referenced. Ad-
ditionally SOA constructs which obey to specific
patterns are omitted (Brown et al., 2003; Colan,
2004; UDDI, 2004). The following sections high-
light important structural and operational aspects
of such a system.
Service Interactions
In the presented architecture, the main element is
the “service” (i.e., to tourists). Under the modern
understanding of realizing a service-oriented
architecture, services are implemented under the
WebServices paradigm, utilizing in almost 100
percent of the cases the SOAP/XML - HTTP - TCP/
IP route. Thus messages are exchanged in a very
standardised manner which can be found to be
fully supported by vendors’ major server systems
(content management systems, relational database
management systems, runtime environments,
etc.), and mobile/client systems as stated by the
leading technologies of the area (provided by Mi-
crosoft, SUN, IBM, etc.). Under this architecture,
the system elements such as services are loosely
coupled and can be even dynamically composed
upon interaction/request by using SOA enabling
systems such as the UDDI registry.23 Composi-
tion can be entirely left to the servicing side, thus
the service access point (SAP) and its delegates,
however exploring mobile equipment capabilities
allows that this task also can be performed at the
client side through the same standard mechanisms.
Service Access Point
In the proposed scheme the concept of a service
access point (SAP) is introduced, which is respon-
sible for accepting customer (i.e., tourist) requests
for a bunch of services. This is not required to
be part of an MNO infrastructure, but in the case
that the full Webservice approach is followed this
element is a Web service hosting server, accessed
by the client devices through typical WS address-
ing constructs. In case typical World Wide Web
(WWW) interactions are performed through a
browser then this server would be a normal Web/
application server. A unifying approach can be that
both servers exist (even in the same system) in
the provider’s infrastructure and the Web server
redirects incoming calls to the same SAP that the
Webservice consuming capable clients access.
This point can be responsible for a series of ac-
tions, which among others include:
• Authenticating end-user even beyond typi-
cal username/password challenges by util-
ising sophisticated network provided au-
thentication mechanisms
• Orchestrating service interactions in an
application specific scenario when an end-
user delivered service requires invocation
of a series of services
• Providing facilities for dynamic service
composition allowing the end-users, de-
vices or services to seek and locate suitable
content and service provided that match
their needs with regards to quality, cost,
availability, and so on; delegating lookups
to registries such as UDDI is performed at
this level
• Providing a repository for open-schema
persistent end-user profiles that can be se-
lectively partially offered to subsequent
request handlers. Alternatively a distrib-
uted profile schema could be used however
this would raise interoperability issues and
would increase the risk of compromising
user privacy. Updating profiles could be
triggered by users and other services to the
extent allowed by their role in the system
• Providing a “service-side triggered” posi-
tioning acquisition point, in case end-user
equipment is incapable of autonomously
handling positioning which is the main
the case when SMS/MMS channels are
utilised. In this case requesting services
would address this element in order to ob-
279
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
tain user position upon and special security
concerns have to be handled
• Providing a mechanism for content lookup
in registered content providers acting as a
content broker, so that content that meets
specific minimal structural requirements
can be delivered to end users without ad-
ditional processing nodes
• Maintaining session state on top of state-
less protocols
• Acting as a front-end for non-IP network
connected channels, such as SMS/MMS
ones, to coordinate requests and replies be-
tween channel gateways and content/sub-
service providers
• Providing presentation transformation lay-
er to adapt content to client device/channel
needs
Optimisation features as a caching of aggre-
gated content are application specific ones that
can be considered as part of the SAP. Advanced
value added services on top of positioning such
as movement prediction also can be considered
however they should preferably be implemented
as a typical Web service to be accessed on demand.
Homogenisation of Positioning
Examining the details of user positioning, be it
network driven, client driven, or even manual,
shows a significant heterogeneity not only in the
area of the actual position estimation, but also in
the process of obtaining this information which
is a rather quite important aspect of the system
design. Even the relatively simple case of manual
positioning might have various implementations:
manually supplying the coordinates of the loca-
tion or indirectly acquiring them out of location
information (e.g., road intersections, etc.). The
following graph shows various potential interac-
tions among equipment, software, and data that
could be utilised to drive user positioning.
280
It is desirable to hide as much as possible of
this heterogeneity from the ASP point of view. A
potential solution would be as to separate position-
ing into three distinct layers as follows:
• One layer is concerned with the actual cal-
culation of the position. This layer is quite
different in each and every implementation,
not only due to the various alternatives of
positioning mechanisms but also due to
differentiation to details of the actual algo-
rithms and infrastructures. Very little can
be done in order to unify this layer other
than defining the appropriate interfaces
• The second layer is the SAP positioning
layer, which has to provide the actual end-
user position to the application layer. This
layer is separated in a server-side and a
client-side part (components to extract de-
vice data where applicable) and its purpose
is to determine the actual position supplier
and extract information data, or provide a
means for manually identifying position
• The top-level layer is the application lay-
er, which has to determine whether new
positioning is required, reuse of previous
information is adequate or actual position
is known (e.g., within query), forward a
request to SAP layer and ultimately return
location information back to the end-user
(within response)
The aforementioned layers might make use
of components (e.g., coordinates translators) or
services (e.g., GIS). A subset of the previous in-
teractions, through eliminating the direct message
exchange of the positioning server and SAP could
potentially minimize the risk of building privacy
holes in the system.
Push and Pull Services
There are two ways one can consider delivery of
content to the end user. In both cases, it assumed
Design and Implementation Approaches for Location-Based, Tourism-Related Services
Figure 7. Positioning interactions (Based on Kakaletris et al., 2004)
that the user has posed a request for content and
they are not entirely network driven, since this
would raise several regulatory issues. In each case
the user interaction pattern is somewhat different:
• A pull service is the one that delivers re-
quest to the user upon his/her explicit re-
quest within a reasonable interaction time;
• A push service is one that submits infor-
mation to the user when special criteria
are met and there seems to be no explicit
binding of a request to a response in the
form of interaction. These criteria could
be location, time or more advanced ones
like matched lookups for information and
so on.
Over the typical HTTP/TCP/IP stack found
in advanced mobile equipment one can consume
such services by various mechanisms. The sim-
plest approach would be by periodically polling
through a proprietary client for any kind of pushed
information to be presented to the user. Trigger-
ing polling by identifying significant positional
change also could be considered for appropriate
devices. This solution introduces network usage
even when no information is to be delivered to the
user, and is actually a mixture of pushing and con-
cepts services. Traditional mobile devices could
be contacted through typical gateways (SMS/
MMS) whenever network-monitoring procedures
decide that interesting information is available.
Push services on location criteria can however
become a potential scalability and performance
threat for an LBS infrastructure, because of the
281
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
computational and bandwidth demands required
for scanning vast collections of data on each and
every end-user move.
User Authentication and Identification
A very important part of the service is user identifi-
cation, authentication, and authorisation (Gajparia,
Mitchell, & Yeun, 2005; Soppera & Burbridge,
2005). The SAP is proposed for taking over the
responsibility of authorizing end-users. Authen-
tication is actually optional since it is considered
that just identifying the user can be quite enough
in several application scenarios, such as free-of-
charge information delivery. Simple identification
might be utilised for some personalisation facili-
ties. A basic scenario for authentication would be
based on a username/password challenge scheme,
which will be verified against a user directory. Cli-
ent certificates are a more advanced option. Using
secure channels for performing authentication is
required to avoid common security penetration
cases, yet not all devices might capable of all
common types of secure information exchange.
SAP is the responsible for maintaining the user
directory where users (via their credentials) are
mapped to unique identifiers and authorisation
information.
Authentication is of crucial importance when
positioning takes place under a network provided
positioning mechanism. Acquiring end-user’s
location relying on simple username/password
challenge opens a wide opportunity for service
misuse. Taking advantage of the “spying effect,”24
an unauthorised user can attach a typical mobile
device related to a specific user account on a
person to be “watched,” and request positioning
information through a secondary device (e.g., fixed
PC or other mobile equipment [ME]) on behalf
of the above-mentioned account. If no network
session data are being used, location information
will potentially be delivered to the “spy”. This
might be a well-accepted feature of some imple-
mentations but the typical perspective of privacy
282
forces counter measures for avoidance of such
situations. In the SMS/MMS channels case this
information is being incorporated in the state of
the gateway/SAP session that is being established
to fulfil a client request. On the TCP/IP typical
channel the situation is a bit more complicated.
Since the client is requesting position over an
HTTP request, the MNO typically receives the
request for locating the user by the SAP. Thus
some user identifier must be presented (other than
the username). Fortunately, the SAP already has
the TCP/IP session information (source address/
port) that is being used in order for the user to be
serviced. Typically end-user mobile equipment is
behind a NAT gateway and this address is not a
real one, thus the port number must also be used
in order for the related MNO to locate the correct
user. The SAP can select the correct MNO either
by the internal directory service (the source ad-
dress of the request directly or indirectly identifies
the MNO) or by the end-user information record
being used. The MNO can therefore identify the
actual user through NAT layer information and
GPRS IP address assignment records.
Nevertheless a highly secure approach could
be restricting positioning information exchange
internally to the MNO infrastructure, by having
positioning servers to replying only to requests
raised internally by MNO subscribers, and more
specifically only by the mobile device being
located. In this scenario, the client device acts as
an autonomously positioned piece of equipment,
at least as this can be understood from the SAP
point of view.
Content Structure
Limiting delivered content to “tourism” related
categories might not be a significant restriction
for the location-related content to be provided, but
certainly poses some requirements on the structure
this content must meet. There exist rich schemas
for describing “tourism” content (TourML 2003),
which attempt to fully capture the correspond-
Design and Implementation Approaches for Location-Based, Tourism-Related Services
ing ontology. However from our point of view
forcing “over” structuring of content is usually a
source of problems for the content producers and
providers. It is a strategic decision that content
providers expose content compliant with a minimal
set of requirements. These requirements could
include position information (in various forms,
e.g., rectangles, series of points, etc.), language
identifiers, labels, multiple classifications in a two
level hierarchy (though not a hard constraint) and
a content rating scheme. Content classifiers are
open to extension by CPs and are being aggregated
by the SAP. The first level is mainly proposed to
be a means to define end-user content selection
preferences, but a more sophisticated scheme can
make use of a free hierarchy. Content rating is being
required in order for free of charge, simple or rich
content to be differentiated by the SAP, in order
for it to be delivered to various clients, although
the actual content manipulation is to be performed
by the content provider (CP). Internally CPs might
realise a more complex structure, but the SAP is
concerned only about the above-mentioned tags.
Usage of extra tags can be handled by additional
user preferences, for which an extensible mecha-
nism should be provided.
Presentation
A final, but quite important issue is that eventually
data must be presented to the end-user. Presenta-
tion can be realised by a wide series of equipment,
which might range from high quality multimedia
laptops to simple SMS capable mobile phones.
As already mentioned the XML/XSL/XSLT set
of standards and related tools provide a powerful
mechanism for implementing an author-once-
publish-many times scheme, capable of covering
a wide range of end user preferences and equip-
ment. At this point, it can be identified that these
are two level transformations in order to achieve
maximum content reusability:
• The first level is internal CP transformation
and renders original data into a form that
will comply with the SAP required mini-
mal structure;
• The second level of transformations is the
one that will transform the content in a way
that it will fully exploit the capabilities of
the end user device and user preferences.
Such rendering can result into simple text
fragments to be posted over SMS, MMS
fragments, WML, HTML, DHTML pages,
and even advanced SMIL25 presentations.
Presentation transformation should not be
considered as a Web-only requirement; even if
data are accessed through proprietary Web service
clients such transformations would be of use.
Application Gateways
The assumed infrastructure of these models of ser-
vice delivery is heavily based on the assumption of
a homogeneous underlying network—on HTTP/
TCP/IP layers. It is obvious that this is not a real-
istic assumption especially when mobile networks
come to play. In order to hide the diversity of the
underlying infrastructures, gateways are required
that allow for the translation of non-SAP native
requests to be fulfilled. These gateways receive
client requests and deliver them to the SAP in a
form they can be understood. At this point it is
obvious that in order for uniform interaction to be
achievable by all CPs, translation of requests is
being handled by a specific reception point of the
gateway that analyses the content of the incoming
message and forms the appropriate request to the
SAP. When requests are being replied, the data
are being appropriately packaged and delivered
to the actual network operator infrastructure com-
ponents that will push it back to the requesting
equipment through the appropriate “channel.”
Typically channels that require gateways are the
SMS, EMS, MMS ones, while the functionality of
283
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
such gateways also is significant for implementa-
tion of alerting services.
Pricing/Billing
Authorizing users is not only required for enabling
access to the services and content authentication
but also for charging and billing. When requests
and responses are being exchanged among the
various parties, the SAP user identification is
being utilised in order for the CP or application
service provider (ASP) to authorise and charge the
appropriate subscriber. Alternatively ASPs or CPs
perform no user identification and deliver data on
behalf of the SAP, which is the actually charged
partner. This would cover the CP/ASP/SAP part of
the service usage. On the other hand the network
operator will charge network usage, potentially
posing a premium on top of the traffic which is
being delivered to specific ASP/CPs/SAPs, and
so on. Another scenario is that ASPs/CPs might
vary the charges depending on user location or
user preferences (such as content quality, type, or
location-based advertising allowance), thus al-
lowing for quite complex pricing/billing schemes.
CONCLUSION
Making use of standards and protocols (e.g., LIF,
XML, and HTTP), avoiding proprietary solutions
and limiting interoperability requirements to a
minimum renders architecture capable of integrat-
ing various stakeholders with limited cost and
implications. The technologies being proposed
allow interoperability and redirection among
various sub-service providers and the integration
of services in order to provide a homogenous ap-
plication to the end-user.
Although the proposed solution for tourists has
been tested26 on a single network operator else-
where, support for location-based services is not
restricted to a single network provider, even when
positioning is performed by the mobile network
284
operator. Despite its “tourism content delivery”
orientation, the infrastructure in this testbed can be
made capable of providing more complex services
with minimum impact on service providers, by
making use of state-of-the-art client equipment
features, such as Java, DHTML, and so on. In this
regard, utilizing a directory service under a stan-
dard protocol, within the SAP, is mostly adequate
for creating a totally uniform platform for in-doors
and outdoors LBS and quite heterogeneous po-
sitioning methods and capabilities (e.g., RFID,
wireless access point identification, etc.). Since
the motivation behind location-based services is
its potential beneficiary commercial exploitation,
a preliminary technoeconomic evaluation based
on hypothetical deployment and commercialisa-
tion scenarios has been attempted by the authors
and can be found in M-Guide (2002). The study
presents basic financial indicators that showed
the viability of the proposed solution as well as
a full description of the evaluation scenarios and
results. The details of the methodology used in
technoeconomic evaluation can be found in IST-
TONIC (2000).
There are various aspects of the presented
approach where potential improvements will
lead to further benefits for the end-user and the
content/network providers. These largely involve
standards identification and adoption, as well
as refinements of current implementations. An
essential improvement would be to fully fit the
design discussed here to the open service-oriented
architecture initiative, thus providing dynamic
service discovery. Introducing the attributes and
rules that will enable the creation of a universal
directory service to support each and every rel-
evant information and service lookup is a step
that should be made.
Although business-to-customer (B2C) infor-
mation interchange is sufficiently faced under
our perspective, business-to-business (B2B) is
partially dealt with. B2B data exchange is limited
to areas where information is essential for servicing
a particular user request, for example, position-
Design and Implementation Approaches for Location-Based, Tourism-Related Services
ing, authentication, and so on. However B2B
collaboration can be further exploited in subjects
such as content modifications due to relocation,
caching, restructuring service organisation, and
so on. Further work on this area would potentially
enhance scalability of “push services,” another
area that has not been fully exploited.
Work also needs to be done on roaming-user
scenarios in order for the infrastructure to fully
comply with market requirements, since han-
dling and network positioning of roaming users
is currently performed inefficiently. Moreover, a
more detailed technoeconomic study is needed,
based on specific deployment scenarios, taking
into consideration up-to-date technological and
financial data in order for a more precise profit/
loss prediction of relevant service deployment to
be made possible.
ACKNOWLEDGMENT
A part of this work has been performed with the
support of the E-Content M-Guide Project, which
was partially funded by the European Union. The
authors would like to acknowledge the support and
contributions of their colleagues from M-Guide
project, namely Exodus SA, Vodafone-Panafon
SA, University of Athens, Municipality of Athens,
Adaptia and Iternet. The authors would also like
to thank the two anonymous reviewers for the
fruitful comments.
REFERENCES
W3C. (2001). The Semantic Web. Retrieved from
http://www.w3.org/2001/sw/
Adams, P. M., Ashwell, G. W. B., & Baxter, R.
(2003). Location-based services: An overview
of the standards. BT Technology Journal, 21(1),
34–43. doi:10.1023/A:1022572210026
Against the Clock (Firm). (2002). Macromedia
flash 5: Animating for the Web. Upper Saddle
River, NJ: Prentice Hall.
Anderson, R. (2000). Professional XML. Birming-
ham, UK; Chicago: Wrox Press.
Andersson, P., Hulten, S., & Valiente, P. (2005).
Beauty contest licensing lessons from the 3G
process in sweden. Telecommunications Policy,
29(8), 577–593. doi:10.1016/j.telpol.2005.06.005
Beatty, C. (2002). Location-based services:
Navigation for the masses, at last! Journal
of Navigation, 55(2), 241–248. doi:10.1017/
S0373463302001790
Bennett, C. (2001). Practical WAP: Developing
applications for the wireless Web. Cambridge, UK;
New York: Cambridge University Press.
Berners-Lee, T., & Hendler, J. (2001). Publish-
ing on the semantic Web: The coming Internet
revolution will profoundly affect scientific
information. Nature, 410(6832), 1023–1024.
doi:10.1038/35074206
Berners-Lee, T., Hendler, J., & Lassila, O. (2001).
The Semantic Web: A new form of Web content
that is meaningful to computers will unleash a
revolution of new possibilities. Scientific Ameri-
can, 284(5), 3.
Bluetooth.org. (2001). The bluetooth membership
site. Retrieved from http://www.bluetooth.org/
Bohn, J., & Mattern, F. (2004). Super-distributed
RFID tag infrastructures. Ambient Intelligence.
Proceedings, 3295, 1–12.
Bosak, J. (1998). Media-independent publish-
ing: Four myths about XML. Computer, 31(10),
120–122. doi:10.1109/2.722303
Bosak, J., & Bray, T. (1999). XML and the second-
generation. Scientific American, 280(5), 89–93.
285
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
Brown, A., Johnston, S., & Kelly, K. (2003). Using
service-oriented architecture and component-
based development to build Web service applica-
tions: IBM Corporation.
Caffery, J., & Stuber, G. L. (1998a). Subscriber
location in CDMA cellular networks. IEEE
Transactions on Vehicular Technology, 47(2),
406–416. doi:10.1109/25.669079
Caffery, J. J., & Stuber, G. L. (1998b). Overview
of radiolocation in CDMA cellular systems.
IEEE Communications Magazine, 36(4), 38–45.
doi:10.1109/35.667411
Colan, M. (2004). Service-oriented architecture
expands the vision of Web services: IBM Cor-
poration.
Czajkowski, K., Foster, I., Ferguson, D. F., Frey,
J., Graham, S., Snelling, D., et al. (2004). The
ws-resource framework. Retrieved from http://
www.globus.org/wsrf/specs/ws-wsrf.pdf
Dana, P. H. (1994). GPS overview. Department of
Geography, University of Texas at Austin.
Dao, D., Rizos, C., & Wang, J. (2002). Location-
based services: Technical and business issues.
GPS Solutions, 6(3), 169. doi:10.1007/s10291-
002-0031-5
Daoud, F., & Mohan, S. (2002). Strategies for
provisioning and operating VHE services in
multi-access networks. IEEE Communications
Magazine, 40(1), 78. doi:10.1109/35.978053
Devine, A., & Holmquist, S. (2001) Mobile inter-
net content providers and their business models.
The Royal Institute of Technology, Sweden.
Di Fazio, A., Mocci, G., Rossini, E., D’Angelo,
G., Lorelli, A., & Jarosh, A. (2004). Gauss satellite
solution for location-based services. International
Journal of Satellite Communications and Network-
ing, 22(1), 55–85. doi:10.1002/sat.775
286
Douglas, E. (2004). Indoor positioning system can
track equipment and people. Anesthesiology News.
Drane, C., Macnaughtan, M., & Scott, C. (1998).
Positioning GSM telephones. IEEE Communica-
tions Magazine, 36(4), 46. doi:10.1109/35.667413
Ekahau. (2006). Ekahau Positioning Engine
3.1—Ekahau Site Calibration. Retrieved from
http://www.ekahau.com/
El-Rabbany, A. (2002). Introduction to GPS: The
global positioning system. Boston: Artech House.
Ellinger, F., Barras, D., & Jackel, H. (2002).
Mobile phone market study. TRLabs/IEEE Wire-
less 2002. Retrieved from http://www.ife.ee.ethz.
ch/~ellinger/Homepage/mp_rev.pdf
EMILY. (2002). Business models: EMILY (IST-
2000-26040) Project.
Ericsson. Mobile positioning system / mobile
positioning protocol. Ericsson.
Erl, T. (2004). Service-oriented architecture: A
field guide to integrating XML and Web services.
Upper Saddle River, NJ: Prentice Hall Professional
Technical Reference.
Erl, T. (2005). Service-oriented architecture:
Concepts, technology, and design. Upper Saddle
River, NJ: Prentice Hall Professional Technical
Reference.
ETSI. (2006). ETSI telecom standards. Retrieved
from http://www.etsi.org/
Extensible markup language (XML). (n.d.). Re-
trieved from http://www.w3.org/XML/
FCC. (2001). FCC wireless 911 requirements.
FCC.
Feng, J., Zhu, Y. L., Mukai, N., & Watanabe,
T. (2005). Search on transportation network for
location-based service. Innovations in Applied
Artificial Intelligence, 3533, 657–666.
Design and Implementation Approaches for Location-Based, Tourism-Related Services

Foster, I., Frey, J., Graham, S., Tuecke, S., Cza- GSM. (2002). Location based services. Retrieved
jkowski, K., & Ferguson, D. F. (2004). Modeling from http://www.gsmworld.com/documents/lbs/
stateful resources with Web services. Retrieved se23.pdf
from http://devresource.hp.com/drc/specifica-
GSM. (2003). MMS interworking guidelines.
tions/wsrf/ModelingState-1-1.pdf
Retrieved from http://www.gsmworld.com/docu-
Fritsch, L., & Scherner, T. (2005). A multilaterally ments/ireg/ir52310.pdf
secure, privacy-friendly location-based service
Halonen, T., Romero, J., & Melero, J. (2003).
for disaster management and civil protection.
GSM, GPRS, and EDGE performance: Evolution
Networking ICN, 2(3421), 1130–1137.
towards 3G/UMTS (2nd ed.). Chichester, West
Gajparia, A. S., Mitchell, C. J., & Yeun, C. Y. Sussex, UK; Hoboken, NJ: J. Wiley.
(2005). Supporting user privacy in location based
Heijden, M. d., & Taylor, M. (2000). Understand-
services. IEICE Transactions on Communications.
ing wap: Wireless applications, devices, and
E (Norwalk, Conn.), 88B(7), 2837–2847.
services. Boston: Artech House.
Gamma, E. (1995). Design patterns: Elements of
Hightower, J., & Borriello, G. (2001). Location
reusable object-oriented software. Reading, MA:
systems for ubiquitous computing. Computer,
Addison-Wesley.
34(8), 57. doi:10.1109/2.940014
GARMIN. (n.d.). GPS guide for beginners. Re-
Holma, H., & Toskala, A. (2004). Wcdma for
trieved from http://www.garmin.com/aboutGPS/
UMTS: Radio access for third generation mobile
manual.html
communications (3rd ed.). Hoboken, NJ: J. Wiley
3GPP. (2002). 3rd Generation Partnership Proj- & Sons.
ect—3GPP. Retrieved from http://www.3gpp.org/
Huber, A. J., & Huber, J. F. (2002). UMTS and
Grajski, K. A., & Kirk, E. (2003). Towards a mo- mobile computing. Boston: Artech House.
bile multimedia age—location-based services: A
IST-REPOSIT. (2001). Real time dynamic band-
case study. Wireless Personal Communications,
width optimisation in satellite networks. IST. EU.
26(2-3), 105–116. doi:10.1023/A:1025518301575
IST-TONIC. (2000). Technoeconomics of IP
Gratsias, K., Frentzos, E., Delis, V., & Theodo-
optimised networks and services. IST. EU.
ridis, Y. (2005). Towards a taxonomy of location
based services. Web and Wireless Geographical Iwasaki, Y., Kawaguchi, N., & Inagaki, Y. (2005).
Information Systems. Proceedings, 3833, 19–30. Azim: Direction-based service system for both
indoors and outdoors. IEICE Transactions on
Gruber, H. (2005). The economics of mobile
Communications. E (Norwalk, Conn.), 88B(3),
telecommunications. New York: Cambridge
1034–1044.
University Press.
Kaaranen, H. (2005). UMTS networks: Architec-
Gruber, H., & Verboven, F. (2001). The diffusion
ture, mobility, and services (2nd ed.). Chichester,
of mobile telecommunications services in the
West Sussex, UK; Hoboken, NJ: J. Wiley & Sons.
european union. European Economic Review,
45(3), 588. doi:10.1016/S0014-2921(00)00068-4

287
 Design and Implementation Approaches for Location-Based, Tourism-Related Services

Kakaletris, G., Varoutas, D., Katsianis, D., Sphi- LIF-IOT. (2002). Location services (LCS) inter-
copoulos, T., & Kouvas, G. (2004). Designing and operability test (IOT) specification in GSM. Loca-
implementing an open infrastructure for location- tion Interoperability Forum (LIF), Interoperability
based, tourism-related content delivery. Wireless Testing Group.
Personal Communications, 30(2-4), 153–165.
Lin, J. Y.-B., & Chlamtac, I. (2001). Wireless and
doi:10.1023/B:WIRE.0000049396.99006.0e
mobile network architectures. New York: Wiley.
Karagiozidis, M., Markoulidakis, Y., Velentzas, S.,
LoVEUS. (2002). User requirements of the
& Kauranne, T. (2003). Commercial use of mobile,
LoVEUS system (LoVEUS IST-2000-30155).
personalised location-based services. Journal of
the Communications Network, 2, 15–20. Ludden, B. (2000). Location technology. Voda-
fone.
Katsianis, D., Welling, I., Ylonen, M., Varoutas,
D., Sphicopoulos, T., & Elnegaard, N. K. (2001). M-Guide. (2002). User needs—market analysis
The financial perspective of the mobile networks (Deliverable D2.1).
in europe. [see also IEEE Wireless Communica-
Mao, Z., & Douligeris, C. (2000). A location-
tions]. IEEE Personal Communications, 8(6),
based mobility tracking scheme for pcs networks.
58–64. doi:10.1109/98.972169
Computer Communications, 23(18), 1729–1739.
Kim, J. W., Kim, C. S., Gautam, A., & Lee, Y. doi:10.1016/S0140-3664(00)00202-4
(2005). Location-based tour guide system using
Microsoft. (n.d.). Microsoft.Net compact frame-
mobile gis and Web crawling. Web and Wireless
work 2.0, Retrieved from http://msdn.microsoft.
Geographical Information Systems, 3428, 51–63.
com/mobility/netcf/
Korhonen, J., Aalto, O., Gurtov, A., & Lamanen, H.
Miller, B. A., & Bisdikian, C. (2002). Bluetooth
(2001). Measured performance of GSM, HSCSD
revealed (2nd ed.). Upper Saddle River, NJ: Pren-
and GPRS. Paper presented at the IEEE Interna-
tice Hall PTR.
tional Conference on Communications (ICC’01).
Morrow, R. (2002). Bluetooth operation and use.
Koutsopoulou, M., Panagiotakis, S., Alonistioti,
New York: McGraw-Hill.
A., & Kaloxylos, A. (2005). Customised billing
for location-based services. Distributed Applica- Mouly, M., & Pautet, M. B. (1995). Current evolu-
tions and Interoperable Systems, 3543, 27–37. tion of the GSM systems. IEEE Personal Com-
munications, 2(5), 9–19. doi:10.1109/98.468359
Kozak, B. (2002). Macromedia flash MX freehand
10: Advanced training from the source. Berkeley, Moureu L. (2000), Impact of location-based
CA: Macromedia Press. services on the mobility market. Alcatel Telecom-
munications Review 2nd Quarter 2000.
Laitinen, H. (2001). Cellular location techniques.
Paper presented at the MLW. Muller, N. J. (2001). Bluetooth demystified. New
York: McGraw-Hill.
Lee, Y., Prabhu, N., & Park, E. K. (2002). Middle-
ware for location based information services in mo- Muller, N. J. (2003). WiFi for the enterprise. New
bile environment. Developing and Infrastructure York: McGraw-Hill.
for Mobile and Wireless Systems, 2538, 173–182.
doi:10.1007/3-540-36257-6_16

288
Design and Implementation Approaches for Location-Based, Tourism-Related Services
Nellen, P. M., Bronnimann, R., Held, M., &
Sennhauser, U. (2004). Long-term monitoring
of polarization-mode dispersion of aerial optical
cables with respect to line availability. Journal
of Lightwave Technology, 22(8), 1848–1855.
doi:10.1109/JLT.2004.832423
Newcomer, E., & Lomow, G. (2005). Understand-
ing soa with Web services. Upper Saddle River,
NJ: Addison-Wesley.
Northstream. (2001). Location based services:
Considerations and challenges. Retrieved from
http://www.northstream.se
OGS. (n.d.). Opengis location services, open
geospatial consortium. Retrieved from http://
www.opengeospatial.org/functional/?page=ols
OMA. (2002). Mobile location protocol, location
working group: Open Mobile Alliance. OTC.
(n.d.). Open tourism consortium. Retrieved from
http://www.opentourism.org/
Pahlavan, K., & Krishnamurthy, P. (2002). Prin-
ciples of wireless networks: A unified approach.
Upper Saddle River, NJ: Prentice Hall PTR.
Pateli, A. G., Giaglis, G. M., & Spinellis, D. D.
(2005). Trial evaluation of wireless info-com-
munication and indoor location-based services
in exhibition shows. Advances in Informatics.
Proceedings, 3746, 199–210.
Patil, B. (2003). IP in wireless networks. Upper
Saddle River, NJ: Prentice Hall Professional
Technical Reference.
Prasad, R., & Ruggieri, M. (2005). Applied satellite
navigation using GPS, Galileo, and augmentation
systems. Boston: Artech House.
Rao, B., & Minakakis, L. (2003). Evolution
of mobile location-based services. Com-
munications of the ACM, 46(12), 61–65.
doi:10.1145/953460.953490
Reed, J. H., Krizman, K. J., Woerner, B. D., &
Rappaport, T. S. (1998). An overview of the chal-
lenges and progress in meeting the e-911 require-
ment for location service. IEEE Communications
Magazine, 36(4), 30–37. doi:10.1109/35.667410
Rysavy, P. (2005). Data capabilities: GPRS
to HSDPA and beyond. Retrieved from http://
www.3gamericas.org/pdfs/rysavy_data_
sept2005.pdf
Schmid, A., Neubauer, A., Ehm, H., Weigel, R.,
Lemke, N., & Heinrichs, G. (2005). Combined
Galileo/GPS architecture for enhanced sensitivity
reception. AEÜ. International Journal of Elec-
tronics and Communications, 59(5), 297–306.
doi:10.1016/j.aeue.2005.05.013
Searle, M. (2001). Location base services. Paper
presented at the InSig.
Smith, R. J. (2003). WiFi home networking. New
York: McGraw-Hill.
SnapTrack. (2001). Location technologies for
GSM, GPRS and WCDMA networks. White paper.
Snell, J., Tidwell, D., & Kulchenko, P. (2002).
Programming Web services with soap (1st ed.).
Sebastopol, CA: O’Reilly & Associates.
Soppera, A., & Burbridge, T. (2005). Wireless
identification: Privacy and security. BT Technol-
ogy Journal, 23(4), 54–64. doi:10.1007/s10550-
006-0007-z
Sparacino, F. (2002). The museum wearable:
Real-time sensor-driven understanding of visitors’
interests for personalized visually-augmented
museum experiences. Paper presented at the
Museums and the Web.
Spirito, M. A. (2001). On the accuracy of cellular
mobile station location estimation. IEEE Transac-
tions on Vehicular Technology, 50(3), 674–685.
doi:10.1109/25.933304
289
 Design and Implementation Approaches for Location-Based, Tourism-Related Services
Spratt, M. (2003). An overview of positioning
by diffusion. Wireless Networks, 9(6), 565–574.
doi:10.1023/A:1025942800145
Staab, S., & Werthner, H. (2002). Intelligent
systems for tourism. IEEE Intelligent Systems,
17(6), 53–55. doi:10.1109/MIS.2002.1134362
SUN. (n.d.). Java 2 platform, micro edition (j2me).
Retrieved from http://java.sun.com/j2me/
Tarumi, H., Morishita, K., & Kambayashi, Y.
(2000). Public applications of spacetag and
their impacts. Digital Cities, 1765, 350–363.
doi:10.1007/3-540-46422-0_28
Taylor, J. H., & Ryan, J. (1995). Museums and
galleries on the internet. Internet Research-
Electronic Networking Applications and Policy,
5(1), 80. doi:10.1108/10662249510084499
Tisal, J. (2001). The GSM network: GPRS evolu-
tion: One step towards UMTS (2nd ed.). Chichester;
New York: J. Wiley.
Tour, M. L. (2003). TourML. Retrieved from http://
www.tourml.org/
UDDI. (2004). Enabling service-oriented archi-
tecture, Retrieved from http://www.uddi.org/
UMTS. (n.d.). The UMTS forum. Retrieved from
http://www.umts-forum.org/
Varoutas, D., Katsianis, D., Sphicopoulos, T.,
Cerboni, A., Kalhagen, K. O., Stordahl, K., et
al. (2002, June 16-19). 3G MVNOs financial
perspectives. Paper presented at the IST Mobile
and Wireless Telecommunications Summit, Thes-
saloniki, Greece.
Varoutas, D., Katsianis, D., Sphicopoulos, T.,
Loizillon, F., Kalhagen, K. O., & Stordahl, K.
(2003). Business opportunities through UMTS—
WLAN networks. Annales Des Telecommuni-
cations-Annals of Telecommunications, 58(3-4),
553–575.
290
Varoutas, D., Katsianis, D., Sphicopoulos, T.,
Stordahl, K., & Welling, I. (2006). On the eco-
nomics of 3G mobile virtual network operators
(MVNOs). Wireless Personal Communications,
36(2), 129–142. doi:10.1007/s11277-006-0027-5
Wang, Y., Jia, X., Lee, H. K., & Li, G. Y. (2003).
An indoors wireless positioning system based on
wireless local area network infrastructure. Paper
presented at the SatNav 2003, the 6th International
Symposium on Satellite Navigation Technology
Including Mobile Positioning & Location Ser-
vices, Australia.
Xiao, Y., & Pan, Y. (2005). Wireless lans and
bluetooth. New York: Nova Science Publishers.
Yan, X. (2004). 3G licensing in Hong Kong:
The debate. Telecommunications Policy, 28(2),
213–226. doi:10.1016/j.telpol.2003.12.006
Yilin, Z. (2000). Mobile phone location determina-
tion and its impact on intelligent transportation sys-
tems. IEEE Transactions on Intelligent Transpor-
tation Systems, 1(1), 55. doi:10.1109/6979.869021
Zhao, Y. L. (2002). Standardization of mobile
phone positioning for 3G systems. IEEE Commu-
nications Magazine, 40(7), 108–116. doi:10.1109/
MCOM.2002.1018015
Zimmermann, A., Lorenz, A., & Specht, M.
(2003). User modeling in adaptive audio-aug-
mented museum environments. User Modeling
2003. Proceedings, 2702, 403–407.
ENDNOTES
1
Being redefined for the electronic era.
2
2nd generation (2G) of mobile telephony is
widely known as GSM, while the term 2.5G
is used for the 2G networks with data and
Internet access capabilities mainly based on
GPRS.
3
The 3rd generation of mobile telephony is
mainly known as UMTS.
Design and Implementation Approaches for Location-Based, Tourism-Related Services
4
Also labelled as “operator” when no ambigu-
ity is introduced.
5
Speed might vary upon device capabilities,
reception quality, network load, and so forth.
6
International Telecommunication Union
whose standardisation sector is ITU-T,
known as CCITT up to 1993 (CCITT origi-
nates back to 1865).
7
Performance highly varies upon network
load, infrastructure specifications, reception
and mobile equipment capabilities.
8
Nominal speed varies from 11 (802.11b) to
108MBps (Super-G).
9
The time required for a signal (or minimal
data element) to be send by the source,
transmitted back by the target and received
back at the source.
10
Outdoors WLAN access is becoming more
and more popular since year 2005.
11
Basic geographical information systems
are data repositories that offer storage and
retrieval facilities suited to the type of infor-
mation in focus (geospatial) accompanied by
appropriate management and visualisation
tools.
12
Source: Derived from Searle (2001)
13
Source: Derived from Searle (2001)
14
There are cases that content might be pre-
installed on a device.
15
The XSL Formatting Objects (XSL-FO)
vocabulary is designed in order to facilitate
content display on a wide variety of media
such as screen, paper, or even voice.
16
Mark-up language technologically similar
with HTML for use with WAP and 2G/2.5G
Web-enabled mobile phones.
17
In advance downloading of the content to the
mobile device and consequently displaying
it to the tourist by utilizing combined user-
interaction and equipment (such as a GPS
device) provided information.
18
Flash technology (introduced by MacroMe-
dia Inc. in 1996) is mainly used for creating
interactive multimedia presentations on
the Web.
19
3GP is a format defined by 3GPP as a
simplified version of MPEG-4 Part 14 for
multimedia objects exchange and playback
on 3G mobile phones.
20
Development effort required to achieve
systems’ inter-working when no sufficient
configuration/customisation capabilities
exist in the involved software entities.
21
Special concerns apply to this assumption.
22
An initially European standardisation
organisation, currently named as ECMA
International.
23
The UDDI registry is essentially a “yellow
pages” like service for locating Web services
capable of serving a particular request.
24
The term has been proposed by Antti Dam-
ski, Adaptia Ltd. in M-GUIDE Project (M-
Guide, 2002)
25
A markup language of similar technology
to HTML that allows display and playback
of interactive multimedia content adding
special features for synchronisation/presen-
tation, partially supported by some of the
well-known Web browsers.
26
The solution has been developed and tested
as part of the M-Guide Project, which has
been partially funded project by EU under
the eContent initiative, during 2002-2004.
291
 Design and Implementation Approaches for Location-Based, Tourism-Related Services

APPENDIX: ACRONYMS

Acronym Description
2G 2nd generation mobile telephony (e.g., GSM)
2.5 G 2nd generation mobile telephony (e.g., GSM + GPRS)
3GPP 3G Partnership Project
A-GPS assisted GPS
AOA angle of arrival
ASP application service provider
B2B business to business
B2C business to client
BS base station
BTS base transceiver station
CCITT International Telegraph and Telephone Consultative Committee
CDMA code division multiple access
CMS content management system
COO cell of origin
CSS cascading stylesheet
DHTML Dynamic HTML
ECMA European Computer Manufacturers Association
EDGE enhanced data GSM environment
E-OTD enhanced observed time difference
GIS geographical information system
GMLC gateway mobile location center
GPRS general packer radio service
GPS global positioning system
GSM global system for mobile communication
HSCSD high speed circuit switched data
HSDPA high-speed downlink packet access
HTML Hypertext Markup Language
HTTP hypertext transfer protocol
ICT information and telecommunication technology
IEEE Institute of Electrical and Electronics Engineers
IP Internet protocol
IR infrared
IS information systems
IT information technology
ITU International Telecommunication Union
JPEG Joint Photographic Experts Group’
LAN local area network
LBS location-based services

continued on following page

292
Design and Implementation Approaches for Location-Based, Tourism-Related Services

Appendix: Acronyms continued

Acronym Description
LIF location interoperability forum
LMU location measurement unit
ME mobile equipment
MLP mobile location protocol
MMS multimedia messaging service
MNO mobile network operator
MPC mobile positioning center
MPEG Motion Picture Expert Group
MPP mobile positioning protocol
MSID mobile subscriber ID
NAT network address translation
OMA Open Mobile Alliance
OpenLS OpenGIS location services
PDA personal data assistant
QoS quality of service
RDF resource description framework
RF radio frequency
RTT round trip time
SAP service access point
SMS short message system
SOA service oriented architecture
SOAP simple object access protocol
TA timing advance
TCP transport control protocol
TDMA time division multiple access
TOA time of arrival
UDDI universal description, discovery, and integration
UMTS universal mobile telecommunications system
URI uniform resource identifiers
URL uniform resource locator
W3C World Wide Web Consortium
WAG wireless application gateway
WAN wide area network
WAP wireless access protocol
WCDMA wideband CDMA
WiFi wireless fidelity, any type of 802.11xx network
WLAN wireless LAN
WML Wireless Markup Language
WSRF Web services resource framework

continued on following page

293
 Design and Implementation Approaches for Location-Based, Tourism-Related Services

Appendix: Acronyms continued

Acronym Description
XLS XML for location services
XML eXtensible Markup Language
XSL eXtensible Stylesheet Language
XSL-FO XSL formatting objects
XSLT XSL transformation

This work was previously published in Information and Communication Technologies in Support of the Tourism Industry, edited
by W. Pease, M. Rowe, and M. Cooper, pp. 114-161, copyright 2007 by IGI Publishing (an imprint of IGI Global).

294
 295

Chapter 19
Methodical Spatial Database
Design with Topological
Polygon Structures
Jean Damascène Mazimpaka
National University of Rwanda, Rwanda

ABSTRACT
Spatial databases form the foundation for a Spatial Data Infrastructure (SDI). For this, a spatial database
should be methodically developed to accommodate its role in SDI. It is desirable to have an approach
to spatial database development that considers maintenance from the early stage of database design
and in a flexible way. Moreover, there is a lack of a mechanism to capture topological relations of spa-
tial objects during the design process. This paper presents an approach that integrates maintenance of
topological integrity constraints into the whole spatial database development cycle. The approach is
based on the concept of Abstract Data Types. A number of topological classes have been identified and
modelling primitives developed for them. Topological integrity constraints are embedded into mainte-
nance functions associated with the topological classes. A semi-automatic transformation process has
been developed following the principles of Model Driven Architecture to simplify the design process.

INTRODUCTION tem (DBMS) is a software system used to create

With the current advances in Information and Com-
munication Technology, there has been a problem
shift from data availability to data maintenance.
A database is a large, structured and integrated
collection of data. A Database Management Sys-
DOI: 10.4018/978-1-4666-2038-4.ch019
and manage databases. Today’s best approach to
managing huge amounts of data is to use a DBMS.
The field of Geographic Information Systems
(GIS) makes no exception. However, in GIS the
problem becomes much more complex due to the
nature of spatial data that make a spatial database
(Tveite, 2001). A spatial database is a special type
of database which, in addition to containing con-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Methodical Spatial Database Design with Topological Polygon Structures
ventional attribute values of the objects, contains
data about their geographic location and shape.
In the field of spatial database development, a lot
of work has been done to capture the semantics
of spatial objects such as their geographic loca-
tion and shape (Shekhar et al., 1999). However,
less importance has been given to capturing their
spatial relations. For instance, spatial objects
such as provinces of a country can be modelled
as polygons and their geographic coordinates are
recorded to indicate their geographic location.
However, their spatial relations that they cannot
overlap and they form a contiguous area which is
the country, are not modelled or it is done in an
inflexible way if done. These spatial relations will
later form database integrity constraints that need
to be maintained. The spatial integrity constraints
of the objects are among the most important prob-
lems in modelling spatial databases (Shekhal et
al., 1999). These considerations highlight the need
for a mechanism to capture spatial relations, and
an improved methodology for spatial database
design which considers the maintenance of spa-
tial integrity constraints throughout the database
development cycle.
Egenhofer and Franzosa (1991) distinguish
three categories of spatial relations; topological
relations, metric relations and relations concern-
ing the partial and total order of spatial objects.
Topological relations are spatial relations that
are not affected by elastic transformations such
as rotation and scaling (Egenhofer et al., 1994).
This paper presents an improvement to existing
spatial database design methodology by including
also topological relations in the conceptual model
of the database. Through a semi-automatic trans-
formation process, topological relationships are
carried at all design steps till the implementation
where they become database integrity constraints
that are enforced using a set of maintenance func-
tions. The work presented in this paper focuses on
collections of area features that display topological
dependencies. The paper is structured as follows;
after this introduction, we present an overview of
296
currently existing approaches of data modelling in
spatial databases. Next, we describe the proposed
spatial database design methodology. To show the
applicability of the proposed approach, we then
present an example of implementing the proposed
transformational design. Finally, we give some
conclusions and the direction for future work.
Related Work
The practice of data modeling has been adopted
in the field of spatial database design because it
simplifies the task of database design (Worboys et
al., 1990). The data modeling approaches that have
been widely used are Entity Relationship (ER) or
Extended Entity Relationship (EER) and the Uni-
fied Modeling Language (UML). The extensibility
of UML has made it the best choice as a modeling
language for sophisticated applications, such as
GIS dealing with complex objects and complex
relationships (Marcos et al., 2003). In the domain
of GIS, Tveite (2001) suggests a technique of
extending the EER and the UML approaches with
geographical icons. In his work, Tveite proposed
different geographical icons for different object
types (e.g., points, lines, regions, volumes). The
geometry of a spatial object is modeled by adding
the appropriate geographical icon to a standard
modelling primitive. In this approach different
object types are identified and their geometry is
modeled but different topological relationships
are not modeled. The consequence of this is that
if the translation of the model into the database
model is ever done, the spatial relationships are not
included and the design does not provide any sup-
port for maintaining spatial integrity constraints.
Wang and Reinhardt (2007) present a method to
extend geographic data modeling also taking into
account topological relations. This method uses
a Constraint Decision Table (CDT) to store the
topological integrity constraints. The problems of
this approach are linked to the separation of the
constraints from the UML conceptual schema and
the mastery required for defining the constraints.
Methodical Spatial Database Design with Topological Polygon Structures
The two elements (UML model and the Constraint
Decision Table) do not make a compact structure
and the propagation of updates on one element to
the other element needs to be handled explicitly
and carefully. The model reader does not find
easily the connection between the components of
the UML model and the constraints in the Con-
straint Decision Table. Another approach that is
sometimes used to express spatial relationships in
UML models of spatial data is to use OCL (Object
Constraint Language) constraints as presented in
Belussi et al. (2004) and Pinet et al. (2007). OCL
has a high expressive power (Warmer & Kleppe,
2003) but we argue that the textual description
of OCL constraints makes the model less under-
standable by the model readers and writing OCL
constraints requires the database designer to master
OCL syntax. It is desirable to restrict the use of
OCL constraints to the cases where they are the
only possible option.
One of the issues treated in this research is the
implementation of topology related to the work by
Hoel et al. (2003). Their work presents a physical
topology implementation which is built into the
ArcGIS Geodatabase (ESRI, 2002). This topol-
ogy implementation has an interesting feature of
flexibility and it is supported in ArcGIS products
which are very popular. While the ArcGIS Geo-
database is an ESRI (Environmental Systems
Research Institute) proprietary spatial database
format, the current trend in the GIS domain is to
use a full-fledged DBMS and a GIS in conjunction
(De By, 2004). In this situation, popular DBMSs
such as Oracle (Loney, 2004) and PostgreSQL
(PostgreSQL Global Development Group, 2008)
are good candidates for spatial data management
while popular GIS packages such as ArcGIS are
good candidates for spatial data manipulation.
These DBMSs have a well known standard data
management technology and some of them have
an extension to support spatial data. For instance
Murray (2003) describes the spatial data support
in Oracle while Ramsey (2005) and Refractions
Research (2005) describe the spatial data support in
PostgreSQL. Our work does not intend to replace
the topology modeling approach used in ArcGIS
products but to enhance the spatial data support
in those popular DBMSs that are increasingly be-
ing adopted. The enhancement concerned here is
on the maintenance of topological relationships.
Moreover, we want to provide an approach which
is flexible throughout the whole spatial database
development cycle. The approach is such that
from the early stage of design the user designs the
database visually with high clarity and less effort.
Then the design is transformed semi-automatically
to implementation enabling choice among avail-
able implementation options but in such a way
that irrespective of the chosen implementation
option, the implementation contains functions
that will help the user to maintain topological
integrity constraints at the time of manipulating
the database developed.
TRANSFORMATIONAL DESIGN
OF SPATIAL DATABASES
The spatial database design methodology pre-
sented in this paper is a transformational design
which follows the principles of Model Driven
Architecture (MDA). The MDA approach has been
chosen for its feature of simplifying the design
task by separating concerns at different design
steps as explained in the MDA guide (Miller
& Mukerji, 2003). A conceptual schema of the
database is built using a set of modeling primi-
tives. The conceptual schema is then transformed
into a logical schema which is then transformed
into a physical schema for the implementation.
A detailed discussion of database schema and its
transformation can be found in (Halpin & Proper,
1995). While the concepts of database design and
transformational design are not new, in this paper
a focus is on handling spatial data peculiarities in
those existing concepts. Figure 1 illustrates this
transformational design process. Simple concepts
are modeled using standard UML modeling primi-
297
 Methodical Spatial Database Design with Topological Polygon Structures

Figure 1. The transformational design process

For each of the topological classes, a modeling
primitive has been designed. Each topological
class has properties which are the attributes of the
object being modeled and methods which are the
functions that will be used to perform database
modification operations such as deleting the object
without violating topological integrity constraints.
An area collection can be defined as a spatial
object made of one or more related area objects.
It can be represented by a polygon if it is made of
only one element or by a set of polygons (a mul-
tipolygon) if it is made of more than one element.
The inherent characteristic of area collections is
that a collection is made of areas that have mutu-
ally disjoint interiors. The consideration of other
specific characteristics allows us to define a number
of categories of area collections. Specific charac-
teristics of area collections are such as whether
the objects making all the related area collections
form a contiguous area, whether the elements in
one collection are disjoint, touch in points or in line
segments and whether there are area collections
with a semantic spatial aggregation connection. A
contiguous area is an area in which from any point
there exists a path fully contained in the area to any
other point in the area. A semantic spatial aggrega-
tion connection means a relationship in which an
object at one level is made of the geometric union
tives. For modeling area features with topological
of some objects at another level.
dependencies, we have designed new modeling
Based on these characteristics, we have defined
primitives by extending UML.
the following categories of area collections which
are illustrated in Figure 2:
Area Collection Modeling Primitives
• Collections made of polygons that make
The spatial objects of main interest in this work are
up a contiguous area (Figure 2a)
collections of area features that display topological
• Collections made of multipolygons that
dependencies referred to as area collections. We
make up a contiguous area (Figure 2b)
have designed the modeling primitives for area
• Collections made of disjoint areas (Figure
collections based on the concept of Abstract Data
2c)
Types (ADT). A detailed discussion of the concepts
• Collections of discretely touching areas
of abstract data types can be found in Dale and
(Figure 2d)
Walker (1996). A number of classes of area col-
• Collections of continuously touching areas
lections have been identified based on the types
(Figure 2e)
of topological dependencies that they display.
• Hierarchy of areas (Figure 2f)

298
Methodical Spatial Database Design with Topological Polygon Structures
Figure 2. Example of categories of area collections
This is not an exhaustive list of categories of
area collections. A different categorisation can
be made based on other specific characteristics
for a specific application. For instance if a land
parcels management application is considered a
more focused categorisation can be developed
The categorisation given here is a possible general
example that considers mostly encountered real
world objects. It is up to the user to decide which
modeling primitive most suits the application
at hand. For each of these classes, a modeling
primitive has been developed. The new modeling
primitives are made available as a UML profile
containing six stereotypes, one for each topological
class. The names of the six stereotypes in the order
of their corresponding classes are; contiguous-
PolygonalArea, contiguousMultipolygonalArea,
disjointArea, pointTouchArea, lineTouchArea and
areaHierarchyAssociation.
One of the unique features of the spatial data-
base design methodology presented in this paper
is the consideration of maintenance of topologi-
cal integrity constraints from the early stages of
the design. Topological integrity constraints are
maintained through the maintenance functions
carried by the modeling primitives. For instance,
the following functions are associated with the
modeling primitive contiguousPolygonalArea:
299
 Methodical Spatial Database Design with Topological Polygon Structures
• AddToContiguousArea used to add a new
contiguousPolygonalArea object to the
set of existing contiguousPolygonalArea
objects,
• DeleteFromContiguousArea used to delete
a contiguousPolygonalArea object,
• MergePolygonalAreas used to merge two
contiguousPolygonalArea objects,
• ModifyPolygonalAreaBoundary used to
modify a part of the boundary of a con-
tiguousPolygonalArea object shrinking or
enlarging it.
The specification of the function AddToCon-
tiguousArea is as follows:
Input: The name of the table that will contain the
objects, an array of names of non geometry
columns into which values are to be inserted,
an array of values to insert into non geom-
etry columns and the geometry value of the
object to be added.
Output: the text Success if the operation was
successful or the cause of failure otherwise.
Algorithm: AddToContiguousArea (T: text, u:
text[] v: text[], w: geometry): text
Figure 3. Example of a conceptual function
300
Step 1: Check if the new object has an intersec-
tion relationship other than “touches” with
any existing object. If it does, set the error
message then go to step 4
Step 2: Check if the geometric union of existing
and new objects would have gaps or holes.
If it would, set the error message then go
to step 4.
Step 3: Insert the new object. Set the message
to Success.
Step 4: Return the message
This function is conceptually expressed using
the pseudocode and the relational calculus nota-
tions as shown in Figure 3.
A Semi-Automatic Transformation
Process
The transformational design presented in this
paper separates three design levels: the concep-
tual design, the logical design and the physical
design. The conceptual design uses UML to build
a conceptual schema of the database. A detailed
discussion of the object-oriented concepts used in
UML can be found in (Ambler, 2004). The logi-
cal design uses the concepts of the relational data
Methodical Spatial Database Design with Topological Polygon Structures
model to produce the structure of the database. A
detailed discussion of the relational data model
can be found in (Elmasri & Navathe, 2007). The
physical design specifies the schema obtained
from logical design in the SQL (Structured Query
Language) of the DBMS chosen for implementa-
tion. The physical design also fine-tunes the logi-
cal schema to make efficient use of the physical
storage space and enhance the performance of
the database according to the requirements of the
application (Marcos et al., 2003). In particular at
this design step, spatial indexes are considered.
The major advantage of the MDA approach is
the ability to transform one type of model into
another type with a certain level of automation.
In the context of MDA, the conceptual schema is
a Platform Independent Model (PIM) whereas the
logical schema and the physical schema are Plat-
form Specific Models (PSM). The design approach
presented here contains two levels of transforma-
tion; from conceptual schema to logical schema
and from logical schema to physical schema. The
transformation process is semi-automatic. The
semi-automation simplifies the design task as the
database designer has to concentrate on specify-
ing the data content of the database (conceptual
schema) and not on its structure and performance.
The semi-automation also provides flexibility by
allowing the database designer to influence the
transformation; during the transformation, the
designer can choose among available options
such as the implementation model.
The transformation from conceptual schema
to logical schema takes a UML class diagram,
which is the conceptual model of the database
and produces another UML class diagram that
represents the relational model of the database.
The transformation is done following a set of
MDA transformation rules. The transformation
from conceptual schema to logical schema is done
using the functionalities offered by the modeling
tool. These functionalities include the standard
transformation definitions built in the tool and
the transformation definitions that can be devel-
oped using the programming environment of the
modeling tool.
The transformation from logical schema to
physical schema produces from the final logical
schema a set of statements written in the SQL of
the DBMS chosen for implementation. This is a
PSM to PSM transformation. These SQL state-
ments include Data Definition Language (DDL)
statements used to create the database structure
and Data Manipulation Language (DML) state-
ments used to work on the data in the database.
The DML statements include the maintenance
functions to manipulate area collections. Some
modeling tools provide the functionality to gen-
erate DDL and DML statements from the logical
schema. However, this built-in functionality is
only suitable in a few simple cases. For instance,
these modeling tools do not come with support for
modeling spatial database and therefore their DDL
statement generation functionality does not handle
spatial data peculiarities in DBMS such as specific
treatment for geometry columns. For this reason
a custom tool for performing the logical schema
to physical schema transformation is preferred.
IMPLEMENTATION OF THE
TRANSFORMATIONAL SPATIAL
DATABASE DESIGN
The implementation of the proposed transfor-
mational spatial database design consists of the
creation of the UML profile containing the mod-
eling primitives for area collections followed by
the development of tools to perform the transfor-
mation operation. In the current implementation
example, Enterprise Architect has been chosen
as the modeling tool and PostgreSQL/PostGIS
as the target DBMS mainly because of the wide
use of these tools and the support of concepts
required in this work. The UML profile has been
created using the UML extensibility mechanisms
(Jacobson et al., 1998) available in Enterprise
Architect. This UML profile is shown in Figure
301
 Methodical Spatial Database Design with Topological Polygon Structures
4. The transformation from conceptual schema
to logical schema has been implemented using
Enterprise Architect’s transformation definitions
(Sparx Systems, 2008). Existing transformation
definitions have been customised by modifying
and adding transformation templates to handle
specific needs especially the transformation of
concepts specific to spatial data such as spatial data
types and Spatial Reference System information.
The transformation from logical schema to physi-
cal schema has been implemented by developing
a custom tool. The approach adopted here is to
export the final logical schema to XMI (XML
Metadata Interchange) file, an XML (eXtensible
Markup Language)-based file format, which is
then taken as input to the custom transforma-
tion tool for the second transformation step. The
custom transformation tool parses the XMI file
and generates SQL statements to create database
Figure 4. UML profile for modelling area collections
302
objects based on the information extracted from
the data structures that have been built during
XMI file parsing. In order to test the implemen-
tation, The design of the spatial database of the
provinces of Rwanda has been carried out using
the implemented transformational design. The
database was designed using the new modeling
primitives and implemented though the transfor-
mation process. Finally the database was populated
and the administrative units were manipulated
using the implemented functions that contributed
to the maintenance of topological relationships.
CONCLUSION AND FUTURE WORK
In this paper we have presented a spatial database
design methodology that integrates the mainte-
nance of topological integrity constraints into the
Methodical Spatial Database Design with Topological Polygon Structures
whole database development cycle. In this respect
a set of modeling primitives have been developed
to model area features that display topological
dependencies, and a semi-automatic transforma-
tion process has been developed to simplify the
design task by transforming semi-automatically
the conceptual schema to logical schema and then
the logical schema to physical schema. The meth-
odology is flexible; the new modeling primitives
are easy to use and the database designer has the
opportunity to influence the transformation for
specific needs for instance by choosing among the
spaghetti and topological implementation models.
The methodology provides a user-friendly way
of maintaining topological integrity constraints
in a spatial database. The future work will focus
on developing more maintenance functions with
editing capabilities such that the user will get
more support on solving topological integrity
constraint violations in case they would occur. The
future work will also be on improving the area
feature topology classification by considering a
specific application area and on the application
of the methodology on the class of hierarchy of
area features. A work is also needed on the imple-
mentation of the methodology by targeting more
Database Management Systems and on testing
maintenance functions for their performance on
very large datasets.
REFERENCES
Ambler, S. (2004). The object primer (3rd ed.).
Cambridge, UK: Cambridge University Press.
doi:10.1017/CBO9780511584077
Belussi, A., Negri, M., & Pelagatti, G. (2004).
GEOUML: A geographic conceptual model
defined through specialization of ISO TC211
standards. Paper presented at the 10th EC GI &
GIS Workshop on ESDI State of the Art, Warsaw,
Poland.
Dale, N., & Walker, H. M. (1996). Abstract data
types: Specifications, implementations, and ap-
plications. Lexington, MA: Heath and Company.
De By, R. (Ed.). (2004). Principles of geographic
information systems (3rd ed.). Enschede, The
Netherlands: ITC Education Textbook Series.
Egenhofer, M., Clementini, E., & Felice, P.
(1994). Topological relations between regions
with holes. International Journal of Geo-
graphical Information Systems, 1, 129–142.
doi:10.1080/02693799408901990
Egenhofer, M., & Franzosa, D. (1991). Point-
set topological relations. International Journal
for Geographical Information Systems, 1(2),
161–174. doi:10.1080/02693799108927841
Elmasri, R., & Navathe, S. (2007). Fundamentals
of database systems (5th ed.). Reading, MA:
Addison-Wesley.
ESRI. (2002). Building a geodatabase. Redlands,
CA: ESRI Press.
Halpin, T. A., & Proper, H. A. (1995). Database
schema transformation & optimization. In Pro-
ceedings of the 14th International Conference on
Object-Oriented and Entity-Relationship Model-
ing (pp. 191-203).
Hoel, E., Menon, S., & Morehouse, S. (2003).
Building a robust relational implementation of
topology. In T. Hadzilacos, Y. Manolopoulos, J.
Roddick, & Y. Theodoridis (Eds.), Proceedings
of the 8th International Symposium on Advances
in Spatial and Temporal Databases (LNCS 2750,
pp. 508-524).
Jacobson, I., Booch, G., & Rumbaugh, J. (1998).
The unified modeling language reference manual.
Reading, MA: Addison-Wesley.
Loney, K. (2004). Oracle database 10g: The
complete reference. New York, NY: McGraw-Hill.
303
 Methodical Spatial Database Design with Topological Polygon Structures

Marcos, E., Vela, B., & Cavero, J. (2003). A meth- Shekhar, S., Vatsavai, R. R., Chawla, S., & Burk,
odological approach for object-relational database T. E. (1999). Spatial pictogram enhanced concep-
design using UML. Software and Systems Model- tual data models and their translation to logical
ing, 2(1), 59–72. doi:10.1007/s10270-002-0001-y data models. In P. Agouris & A. Stefanidis (Eds.),
Proceedings of the International Workshop on
Miller, J., & Mukerji, J. (2003). MDA guide ver-
Integrated Spatial Databases (LNCS 1737, pp.
sion 1.0.1. Needham, MA: Object Management
77-104).
Group (OMG).
Sparx Systems. (2008). Enterprise architect.
Murray, C. (2003). Oracle spatial topology and
Retrieved from http://www.sparxsystems.com.
network data models, 10g release 1 (10.1). Nashua,
au/products/ea/index.html
NH: Oracle Corporation.
Tveite, H. (2001). Structured modelling for
Pinet, F., Duboisset, M., & Soulignac, V. (2007).
geographical information systems using icons.
Using UML and OCL to maintain the consistency
In Proceedings of the Scandinavian Research
of spatial data in environmental information sys-
Conference on Geographical Information Science
tems. Environmental Modelling & Software, 22(8),
(pp. 185-201).
1217–1220. doi:10.1016/j.envsoft.2006.10.003
Wang, F., & Reinhardt, W. (2007). Extending geo-
PostgreSQL Global Development Group. (2008).
graphic data modeling by adopting constraint de-
PostgreSQL 8.3.0 documentation. Retrieved from
cision table to specify spatial integrity constraints
http://www.postgresql.org/files/documentation/
(pp. 435–454). Berlin, Germany: Springer-Verlag.
pdf/8.3/postgresql-8.3-A4.pdf
Warmer, J., & Kleppe, A. (2003). The object
Ramsey, P. (2005). PostGIS manual. Victoria, BC,
constraint language: Getting your models ready
Canada: Refractions Research Inc.
for MDA (2 ed.). Reading, MA: Addison-Wesley.
Refractions Research. (2005). Postgis topology.
Worboys, M. H., Hearnshaw, H. M., & Maguire,
Retrieved from http://postgis.refractions.net/sup-
D. J. (1990). Object-oriented data modelling for
port/wiki/index.php?PostgisTopology
spatial databases. International Journal of Geo-
Shekhar, S., Chawla, S., Ravada, S., Fetterer, graphical Information Systems, 4(4), 369–383.
A., Liu, X., & Lu, C. (1999). Spatial databases doi:10.1080/02693799008941553
– Accomplishments and research needs. IEEE
Transactions on Knowledge and Data Engineer-
ing, 11(1), 45–55. doi:10.1109/69.755614

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 1,
edited by Donald Patrick Albert, pp. 21-30, copyright 2012 by IGI Publishing (an imprint of IGI Global).

304
 305

Chapter 20
Methodologies for
Augmented Discovery of
Geospatial Resources
Mattia Santoro Cristiano Fugazza
National Research Council, Italy Institute of Environment and Sustainability, Italy

Paolo Mazzetti Carlos Granell

National Research Council, Italy Universitat Jaume I de Castellón, Spain

Stefano Nativi Laura Díaz

National Research Council, Italy Universitat Jaume I de Castellón, Spain

ABSTRACT
Presently, solutions for geo-information sharing are mainly based on Web technologies, implementing
service-oriented frameworks, and applying open (international or community) standards and interoper-
ability arrangements. Such frameworks take the name of Spatial Data Infrastructures (SDIs). The recent
evolution of the World Wide Web (WWW), with the introduction of the Semantic Web and the Web 2.0
concepts and solutions, made available new applications, architectures, and technologies for sharing and
linking resources published on the Web. Such new technologies can be conveniently applied to enhance
capabilities of present SDIs—in particular, discovery functionality.

Different strategies can be adopted in order to enable new ways of searching geospatial resources, le-
veraging the Semantic Web and Web 2.0 technologies. The authors propose a Discovery Augmentation
Methodology which is essentially driven by the idea of enriching the searchable information that is
associated with geospatial resources. They describe and discuss three different high-level approaches
for discovery augmentation: Provider-based, User-based, and Third-party based. From the analysis of
these approaches, the authors suggest that, due to their flexibility and extensibility, the user-based and
the third-party based approaches result more appropriate for heterogeneous and changing environments
such as the SDI one. For the user-based approach, they describe a conceptual architecture and the main
components centered on the integration of user-generated content in SDIs. For the third-party approach,
the authors describe an architecture enabling semantics-based searches in SDIs.

DOI: 10.4018/978-1-4666-2038-4.ch020

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Methodologies for Augmented Discovery of Geospatial Resources
INTRODUCTION
In recent years, the World Wide Web (WWW) has
undergone several important changes in terms of
available applications, architecture, and related
technologies. The need for a more effective re-
source sharing through the Web raised awareness
on efforts aiming to enable machine-to-machine
applications on top of the Web architecture by mak-
ing semantics explicit. These efforts are currently
coordinated in the W3C Semantic Web Activity
which “provides a common framework that allows
data to be shared and reused across application,
enterprise, and community boundaries […]. It is
based on the Resource Description Framework
(RDF)” (W3C, 2011). At the same time, new use
cases, new available applications and technologies
have made possible the WWW revolution which
is known as Web 2.0 (O’Reilly, 2005). This term
actually refers to an entirely new paradigm in
the use of the Web as a platform for applications
characterized by features like: delivery of services
instead of packaged software, with cost-effective
scalability; control over unique, hard-to-recreate
data sources that get richer as more people use
them; trusting users as co-developers; harnessing
collective intelligence; leveraging the long tail
through customer self-service; design of software
above the level of a single device; lightweight
user interfaces, development models, and busi-
ness models (O’Reilly, 2005; O’Reilly & Bat-
telle, 2009).
These two main changes jointly make new
resources available, and new technologies to
discover them through semantic relationships.
Unavoidably, these changes would and should
affect the geo-information sharing domain that
is mostly based on web paradigms and technolo-
gies. Recently many efforts aim to provide more
powerful tools for the discovery of geospatial
information that is made available through tradi-
tional or Web 2.0 services, basing on explicit or
implicit semantics (Klien, et al., 2004; Smits &
306
Friis-Christensen, 2007; Lemmens, et al., 2006;
NASA-JPL, 2011).
Information technology and geo-science are
worlds in continuous change. Semantics and Web
2.0 are the present challenges, but new ones will
emerge in the future. This raises the conceptual is-
sue of enhancing geospatial information discovery
capabilities in order to accommodate present and,
possibly, future needs. This chapter describes two
approaches based on the methodology of augment-
ing semantically the discovery process to enhance
the search and retrieval of geospatial resources.
Data Discovery in the Geospatial
Information Domain
It is estimated that more than 80% of data that
human beings have collected so far are geospatial
data in a wide sense (Frankling & Hane, 1992;
MacEachren & Kraak, 2001), i.e. data with an
explicit or implicit spatial/temporal reference.
Moreover, this spatial/temporal reference is rel-
evant, and even fundamental, for many applica-
tions. Therefore, it is of major importance to be
capable of discovering geospatial data according
to their content and geospatial characteristics
(i.e., spatial coverage and temporal extent), and
to effectively describe them.
Presently, solutions for sharing geo-infor-
mation implement service-oriented frameworks
applying open (international or community) stan-
dards and interoperability arrangements (Nativi,
2010). Such frameworks take the name of Spatial
Data Infrastructures (SDIs). Typically, in service-
oriented frameworks such as SDIs, discovery
functionalities are provided by catalog compo-
nents. A formal definition of these components
is given by the ISO/TC211-Geographic Informa-
tion/Geomatics, stating that a catalog service is a
“service that provides discovery and management
services on a store of metadata about instances.
The metadata may be for dataset instances, e.g.,
dataset catalogue, or may contain service metadata,
e.g., service catalogue” (ISO, 2003b). At present,
Methodologies for Augmented Discovery of Geospatial Resources
the Open Geospatial Consortium (OGC1) Catalog
Service Specification (OGC, 2007) defines a com-
mon core specification for geospatial resources
discovery providing a consolidated framework.
Services implementing this standard can be que-
ried using common geospatial constraints (i.e.,
basing on what, where, when, who, etc.).
In addition to standard solutions based on
ISO metadata and data models and OGC Web
Services (OWS), many other de-facto standards
are extensively used, defined in the context of
specific Communities-of-Practice (CoP), such
as THREDDS Data Server and OPeNDAP in
the Meteo-Ocean domain, Global Biodiversity
Resource Discovery Service (GBRDS) in the
biodiversity domain, and so on. The advent of
Web 2.0 further widens the geo-information world
with services providing geospatial resources (Geo-
Names2, GeoCommons3, OpenStreetMap4, etc.).
Statement of the Problem
In state-of-the-art catalog implementations, while
the information about geospatial reference is
usually precise (through the spatial and temporal
extent expressed with well defined Spatial and
Temporal Reference Systems), the information
about content is often inadequate for some rel-
evant use-cases. Indeed, it is often expressed us-
ing free text in multiple metadata fields or using
controlled vocabularies for some specific fields
(e.g. keywords), as it happens in the INSPIRE
profile (INSPIRE, 2010). This limits the effective-
ness of geospatial data discovery because of the
limitations of searching words in free text instead
of matching concepts. Therefore, the fact that
the same word may have different meanings in
different domains, languages, or contexts makes
the search error-prone.
Moreover, existing geospatial catalog solu-
tions do not address the issue of searching and
retrieving resources from the emerging world of
Web 2.0 services. In such a context, new content
and resources are continuously created and made
publicly available by thousands of users. Such an
amount of information cannot be ignored when
designing and developing a modern SDI.
The conceptual challenge behind the above
issues concerns the enablement of new ways of
searching in present SDIs. To address this, we
define a Discovery Augmentation Methodology.
This methodology is essentially driven by the
idea of enriching the searchable information that
is associated with geospatial resources. In fact,
whether we consider semantic information or
user-generated content, the current capabilities of
geospatial catalog solutions need to be enhanced
for enabling searches on additional content.
Next section introduces three high-level
approaches for discovery augmentation taking
into account not only semantics and Web 2.0,
but with a more general perspective in order to
address possible future needs. In the following
two sections, we analyze in depth the User-based
Approach and the Third-Party Approach. For the
former, we describe a conceptual architecture and
the main components centered on the integration
of user-generated content in SDIs. For the latter,
we describe an architecture based on a third-party
approach to enable semantics-based searches in
SDIs. The chapter is finished with some conclud-
ing remarks.
DISCOVERY AUGMENTATION
METHODOLOGY: HIGH-
LEVEL APPROACHES
Enhancing geospatial resource discovery capabili-
ties can be achieved by augmenting the searchable
descriptions of resources. Examples of additional
descriptions (that is, something not searchable
with typical geospatial discovery services) are:
semantic information and user-generated anno-
tations. Searching for geospatial resources that
are constrained by one or more of the previous
descriptions implies to characterize such resources
properly. Several approaches can be followed in
307
 Methodologies for Augmented Discovery of Geospatial Resources
order to cope with the required additional descrip-
tions. Roughly, these can be classified in:
• Provider-based approach
• User-based approach
• Third-party approach
Each of the above approaches has advantages
and drawbacks. To analyze them in the next sec-
tions it is important to underline which are the
main issues to be addressed by a discovery aug-
mentation solution, in particular: interoperability,
accuracy, and extensibility.
Interoperability is critical for the geo-informa-
tion domain, where large and distributed infra-
structures must manage resources from different
and heterogeneous scientific domains. A good
solution for discovery augmentation must then
be able to address the different interoperability
issues related to interconnecting resources in such
a heterogeneous environment. Secondly, Accuracy
of search results is central to all discovery sys-
tems; it should be as high as possible also when
discovery capabilities are augmented. An impor-
tant aspect that impacts on accuracy is metadata
quality because poorly documented resources
(e.g. a wrong translation of a term indicating its
semantics) may lead to lower precision-recall
values. Finally, extensibility, which is more related
to the augmentation approach. As stated in the
introduction, a general approach should be able
to address possible future needs and not only the
present need raised by semantics interoperability
and Web 2.0 resources. A general augmentation
approach that is relatively easy to extend for
satisfying new requirements is highly desirable.
Provider-Based Approach
This approach represents the straightforward
solution. In this case, data providers enrich
resource metadata by adding related semantic
information based on controlled vocabularies
or even on full ontologies. In fact, this approach
308
consists of making explicit as much as possible
additional metadata description for each resource.
Clearly, this approach is provider-based since the
characterization of resources with new searchable
content is completely entrusted to the provider.
The main advantage of this approach consists of
the high accuracy in the description of resources.
In fact, they are directly supplied by the resource
providers and then allow agents to execute que-
ries against the additional description. Thus, an
authoritative, quality description can be ensured.
On the other hand, it is not always possible
to apply this approach. Resources, especially for
global systems, are maintained by several provid-
ers at different stages. Two providers, one the role
of creator and the other as a custodian, should
generate additional descriptions in a consistent
way. This implies extra synchronization tasks
among diverse providers involved in the life cycle
of resources, which is not always possible. This
might be a quite complex and expensive operation,
since the continuous changes in geo-information
domain would require constant updates of large
repositories. In conclusion, this approach does
not result an extensible solution.
User-Based Approach
This approach moves the task of augmenting
the searchable resources from data providers to
data users. From this perspective, we address the
discovery augmentation in two different ways:
a) augmenting information sources by extending
current systems in order to access Web 2.0 ser-
vices, which are typically based on user-generated
content; and b) by delegating to users the descrip-
tion of resources by adding the so-called resource
annotation capability to resource sharing systems
or providing mechanisms that allow users to an-
notate resources once they are found.
This approach distributes the task of enriching
large repositories of metadata to a wide range of
users, scaling and making potentially use of a
much higher amount of knowledge. Allowing users
Methodologies for Augmented Discovery of Geospatial Resources
to annotate resource with new meta-information
highly increases the available knowledge; how-
ever, this raises the challenge of metadata qual-
ity control. This is still an open issue not only
in the geospatial domain (Stvilia, et al., 2008),
but it assumes great importance in this domain
as scientific data must be described in a proper
way to support scientific and decision-making
applications (Craglia, et al., 2008).
Third-Party Approach
The main principle of the third-party approach
is to build on existing systems, a widely applied
concept in creating SDIs based on the System of
Systems approach (Global Earth Observation Sys-
tem of Systems5). The idea is that existing systems
continue to operate within their own mandates,
because additional capabilities are provided by
new components that interconnect with existing
systems generating added value, for instance,
additional meta-information to perform searches.
In our case, existing systems are: a) currently
available discovery services and b) repositories
of additional meta-information such as controlled
vocabularies, ontologies, user-generated annota-
tions, etc. Following this approach, a third-party
component is in charge of classifying existing re-
sources according to available meta-information.
Clearly, this approach is not as accurate as
the provider-based one. Indeed metadata quality
is still ensured by the data provider but the auto-
matic classification required to elicit additional
information (e.g. semantic information) may be
inaccurate in some cases. However, this approach
allows to characterize resources with proper ad-
ditional meta-information (provided with a basic
set of metadata) stored in existing archives and
repositories, without any modification. Moreover,
it does not prevent from adding explicit new meta-
information on existing resources.
Besides, another advantage of this approach
consists of being able to accommodate future
needs in a relatively easy way, i.e., it is extensible.
In fact, the business logic necessary to classify
resources is concentrated in a separate component
(provided by a third party) that can be adapted to
satisfy new requirements without affecting the
other existing systems.
AN ARCHITECTURE TO INTEGRATE
AND ANNOTATE USER RESOURCES
With the emergence of the Web 2.0, ordinary citi-
zens have begun to produce and share geospatial
information on the Internet. These Web 2.0-based
geospatial activities show that users are willing
to engage more actively in the production and
provision of content. The aim of this section is to
describe how web services offering volunteered
geographic content can be considered to augment
the number of data sources in Geospatial Infor-
mation Infrastructures (GII) (Díaz, et al., 2011).
Data contained in GIIs are usually produced
and maintained by the official providers, like sci-
entific or government institutions that guarantee
data quality and completeness. These providers
also register metadata descriptions of data and
resources in standard catalog services. In this
way, common tasks of GII users (like data search,
discovery, and evaluation for a particular purpose)
are performed against these catalog services, since
they contain metadata descriptions that should
point to the resource itself or the data service
serving it.
Web 2.0 services are meant to be easily ac-
cessible, via specific web sites or APIs, in order
to integrate them in different applications. The
contents of Web 2.0 services are mainly user-
generated and are being continuously updated by
non experts. There is then a lack of authoritative
indicators regarding completeness, accuracy, or
even veracity of data, but on the other hand, it
can be easily rated, improved, and updated by
users. Due to the easy deployment and publica-
tion mechanisms, the rate of participation is high
and then resources are also quite up-to-date. In
309
 Methodologies for Augmented Discovery of Geospatial Resources
contrast, GII publication mechanisms are still
complex and users do not have the knowledge to
publish easily new content.
GII is built on services implementing OGC
standard interfaces. Service interoperability is
reached by applying the same standard interfaces
to the different components deployed in the in-
frastructure to be used in as many use cases as
possible, at the cost of certain abstraction level
and format complexity. The use of OGC standards
is then beneficial in terms of integration and in-
teroperability, though, these can be rather complex
when compared to Web 2.0 services, which are
built upon simple application-level protocols (the
so-called APIs) and lightweight data formats re-
lying on simplicity, ease of implementation, and
fast adaptation to user’s needs. Open standards are
used where they serve the keep-it-simple principle.
Each service offers different functionalities, so
each one provides its own public API. Some com-
mon operations (such as searching and accessing
content) could benefit from simple standards as
GeoRSS6, GeoJSON7, or KML (OGC, 2008), thus
increasing interoperability to some extent.
An application of the user-based approach for
discovery augmentation is the implementation of
user annotation or tagging, which covers a descrip-
tive perspective in terms of discovering resources
through user-generated tags close to the content
of resources. There are user-centered techniques
that may improve the discovery experience from
the user perspective. A recent study (Strohmaier,
et al., 2010) differentiates between users who use
tags for categorization and those who use tags for
description purposes. The first group of users is
motivated toward tagging because they want to
construct and maintain a navigational aid for the
resources being tagged. This implies a limited,
stable set of tags that represent categories. As the
tags are very close to the hierarchical, structured
representations of certain models (e.g. forestry,
environment, floods, etc.) they can act as suitable
facilitators for navigation and browsing.
310
In the user-based approach, we propose a cou-
ple of techniques around the concept of tagging:
• Augmenting the discoverable sources:
Users define tag-based search queries that
expand over a great range of heterogeneous
sources, augmenting the range of potential
discoverable sources.
• Annotating the discovered resources: After
discovery, when a user annotates and ag-
gregates one or more search results into a
collection to improve future searches.
Augmenting the
Discoverable Sources
The first architecture consists of exposing a simple
common query interface and common response
formats for several discoverable sources. These
backend sources represent Web 2.0 services, which
in principle contain distinct types of resources,
both in nature and format. In order to provide
interoperability across these services, increased
data accessibility, and make client implementa-
tion much easier, we proposed the use of a Web
2.0 Broker (Nuñez, et al., 2011) that enables a
simple common search interface to query the set
of Web 2.0 services. In this context, tags play
an essential role in this architecture by allowing
users to discover heterogeneous resources from
different sources.
Figure 1 illustrates the proposed architecture
for tag-based discovery and the augmentation
of discoverable sources via the Web 2.0 Broker
integrated with the EuroGEOSS Discovery Bro-
ker (EuroGEOSS, 2010). This architecture is
supported, at least, by the following components:
• Recommendation Module, which seeks
recommended tags from previous user’s
queries.
• Web 2.0 Broker, which forwards the user’s
query over a large set of Web 2.0 services.
Methodologies for Augmented Discovery of Geospatial Resources
Figure 1. Tag-based discovery and sources augmentation via the web 2.0 broker
In a normal discovery scenario, the user starts
typing the tags that she thinks best describe the
information she is looking for. The Qtag in Figure
1 denotes the initial user query in terms of typed
tags in an unrestricted manner. In the following,
we describe the main components involved in a
tag-based query (Qtag) augmented over discover-
able resources.
Recommendation Module
The functionality of this module is to help users
choosing the tags to improve search accuracy, on
the basis of historical tags used in previous queries.
The Recommender component takes a tag (Qtag in
Figure 1) and provides a list of tags (Q’tag) related
to the input tag. Its aim is to propose related tags
in function of the actual tag by analyzing histori-
cal searches. The idea is to find correspondences
between tags that, although syntactically different,
could be semantically related. For instance, if a
user types the tag “fire,” this component should
recommend other tags that were used in previous
queries and have been related somehow with the
actual tag, the strategy could consider synonyms
that have been previously highly ranked, highly
used, etc. In this sense, we benefit from the previ-
ous tags used by users to perform similar queries,
that is, exploiting the social knowledge in terms
of tag clouds built by the users. Recent works
(Barragáns-Martínez, et al., 2010) are exploring
with success the use of social tags to improve rec-
ommendations to users, other works like (Vilches-
Blázquez & Corcho, 2009) disambiguate between
syntactic terms to find semantic correspondences
in known taxonomies. On the other hand, the
Query Log component will store and keep track
of the different tag-based queries performed by
users. This allows us to improve the quality and
accuracy of further tag-based queries.
The Recommendation Module will augment
the original query Qtag into an extended query
Q’tag ready to be consumed by the main Discovery
Broker or directly by the Web 2.0 Broker. We
311
 Methodologies for Augmented Discovery of Geospatial Resources
also identify some challenges in implementing
this module. One for example is to deal with
multilingual tags. Another challenge is to identify
the target fields in the ISO metadata records used
to match the tags. Current metadata records are
not annotated with user tags as other Web 2.0
services are.
Web 2.0 Broker Component:
OpenSearch as Integration Protocol
As mentioned, Web 2.0 services offer different
contents and functionalities and provide their
own public API. In our approach, we look for
a mechanism to access common functionalities
(search interface, geographic content data type) by
offering a common entry point and experimenting
with simple open standards.
OpenSearch (Clinton, 2009) and its geo exten-
sion (Turner, 2010; Gonçalves, 2010) is proposed
as the minimal query interface that can be used to
access geospatial content, actually for both Web
2.0 services and SDI services. In this sense, Open-
Search becomes the “common query interface”
to query the set of Web 2.0 services (Figure 1).
This would allow for easy client implementations
that could search and integrate all data sources
regardless of their origin. In addition, the use of
the OpenSearch as a common interface to access
to all Web 2.0 services greatly alleviates the inte-
gration with the Discovery Broker. This approach
implies that the Web 2.0 Broker is designed as a
set of adapters that transform and propagates the
original OpenSearch query to the different Web
2.0 APIs. The Web 2.0 broker encompasses all the
adapters for the selected services for the integration
scenario, which for an initial prototype would be:
Twitter, Flickr, OpenStreetMap, and Geonames.
Annotating Discovered Resources
Complementary to the previous approach, where
the Web 2.0 Broker component augmented the set
of discoverable sources, the following approach is
312
based on the use of annotation techniques around
the concept of collections. This technique makes
use of the OAI-ORE8 (Open Archive Initiative—
Object Reuse & Exchange) abstract model to
compose and annotate collections of resources that
are of interest to the users, along with the use of a
RDF-based repository to persist such collections.
In this context, tags play an essential role in this
architecture in letting users annotate collections of
heterogeneous resources from different sources.
Figure 2 illustrates the proposed architecture
for the annotation of collections of search results
(discovered resources) through the set of brokers.
Note that this architecture may be an extension
of the previous, to provide added-value func-
tionalities over the previous tag-based discovery
approach. For this reason, some components are
shared by both approaches, like the Recommenda-
tion and Web 2.0 Broker components:
• Recommendation Module and Web 2.0
Broker.
• Annotation Module, which let users anno-
tate collections of resources and then pub-
lish them in a repository.
• Repository Module, which persists re-
source collections as RDF triples.
Before going into the description of the An-
notation and Repository modules, we describe in
the following the basic notions and entities behind
the OAI-ORE specification.
OAI-ORE Specification
The Open Archive Initiative - Object Reuse and
Exchange (OAI-ORE) protocol (OAI, 2008)
defines an abstract data model (OAI, 2008b) for
describing, reusing, and exchanging collections of
Web resources. The OAI-ORE protocol is initially
conceived of in the context of digital libraries and
e-print resources, in order to expose rich content
(text, images, data, video, etc.) into aggregations
to be then reused by client applications.
Methodologies for Augmented Discovery of Geospatial Resources
Figure 2. Tag-based discovery and annotation of collections with the broker
Conceptually, the OAI-ORE’s abstract data
model builds strongly on the notion of “address-
able resources” to indicate that any resource (file,
image, text document, metadata, process, etc.) is
identified using HTTP URI (Uniform Resource
Identifiers9). The simplified diagram in Figure 3
shows the main entities that form part of the OAI-
ORE’s abstract model. The entity Aggregation
plays a central role as it represents a collection
of addressable resources that in turn are called
Aggregated Resources (AR). The ore:aggregates
relation denotes the inclusion of related resources
in the same collection.. In addition, both Aggrega-
tions and AR entities are addressable resources in
the sense that both use HTTP URI as referencing
method, i.e., which just means looking up a URI
on the Web in order to get either the resource itself
or its representation.
Aggregation and AR are abstract entities that
must still refer to concrete resources, which can
be discovered resources of any type and even a
chain of geospatial services (Abargues, et al.,
2010). The OAI-ORE specification makes use of
the Resource Map entity to provide a concrete
representation for the whole aggregation, mostly
derived from RDF. Some suggested formats in
the specification are the Atom syndication for-
mat10, RDF/XML11, and RDFa12 (a microformat
for extending XHTML to support RDF).
Finally, OAI-ORE also defines a useful abstract
entity called Proxy (P) by which it is possible
to express the role an aggregated resource has
explicitly in the context of an aggregation. The
intended meaning of the role is expressed via
ore:proxyFor relationships. For example, two
resources may have a temporal relationship that
connects to each other and this is only meaning-
313
 Methodologies for Augmented Discovery of Geospatial Resources
Figure 3. Simplified OAI-ORE’s abstract data model
ful within the aggregation context in which they
are defined.
Annotation Module
Annotating search results is a means to refine future
searches. The use of collections (aggregations)
of resources combined with user knowledge can
improve the discovery of these concepts and oth-
ers related to them. Currently, there exist different
techniques to model collections or aggregations
of heterogeneous resources. After some analysis,
we selected the OAI-ORE paradigm because it
defines a straightforward abstract data model for
the description and exchange of aggregations/
collections of Web resources. The resources that
compose an aggregation are identified by their
URIs. Although OAI-ORE specification does not
specify a concrete serialization format, RDF, and
Atom are the preferred ones.
The aims of the Annotation Module is to
allow the user to group different search results
314
(resources) into OAI-ORE-based collections,
annotate them, and finally publish all this infor-
mation in a repository. Several components are
required to carry out these tasks. The Collection
Annotator (Figure 3) component allows users to
create a collection from a set of aggregated re-
sources. Such resources are the result of a search
made by the Discovery Broker. An interesting
point here is that, by using the OAI-ORE abstract
model, the collection is no further a plain list
but a graph or map of resources with typed links
and established relationships among them and
even with entities outside of the collection (i.e.,
Geonames13 or DBpedia14 entries). Users should
interact with the Annotation Module to discard
and select only those resources that better fits their
requirements to form a collection. At this point,
the user can add extra metadata (tags, terms) to the
aggregated resources but also and most important
to the aggregation itself. Tags at collection-level
may describe common features of the aggregated
Methodologies for Augmented Discovery of Geospatial Resources
resources and meaningful information that makes
only sense as a whole.
Building a collection amounts to grouping
a list of resources with regard to some links or
relationships. In a first moment, maybe only a
basic set of relationships can be set, such as those
specifying the internal structure of the aggrega-
tion, and some others regarding the resources’
metadata obtained in the search. These types of
relationships could be incremented by using the
Relationship Generator component (Figure 2)
that generates new ones based on the geometric
and temporal topology that the different resources
exhibit among them. These new relationships
could be generated automatically by gathering
the resource’s information collected by the broker,
and by calculating the different relationships using
for instance ontologies or vocabularies that could
formally specify them, similarly to the ones used
currently by the Ordnance Survey15. Terms from
controlled vocabularies and taxonomies might
be also used at this stage to identify aggregated
resources and typed links.
As the OAI-ORE defines an abstract model,
the newly-created collections need to be serialized
in a concrete format. The Collection Publisher
component serializes an abstract collection into
a RDF or Atom representation to be added in
suitable repositories.
Repository Module
The Repository Module, in right side of Figure2,
will make available all the information, struc-
tured and annotated as collections, to the user.
This module will be composed by a specialized
RDF storage system commonly known as “triple
store”16. These systems usually offer a SPARQL
endpoint as a way to access and query the under-
lying repository.
As the Discovery Broker supports OpenSearch-
interfaced sources (accessors), the OS2SPARQL
component will address the translation of Open-
Search-based queries into SPARQL syntax. The
OpenSearch interface ensures the communication
between the broker and the RDF-aware reposi-
tory. As the collections of resources can appear
in the search results, the contained resources may
point to other related resources, enriching then
the search experience. Nevertheless, a dedicated
SPARQL-interfaced accessor would be preferred
to grant access to a great amount of public RDF
triples available elsewhere.
SEMANTIC AUGMENTATION
Grounding Semantics-
Aware Discovery
Organizing the concepts that web contents (and,
among them, geospatial resources) are referring
to is essentially what the Semantic Web (SW)
is all about. Despite the immense work done in
this research field in the last decade, most of the
design principles driving the development of the
SW can be found in the enlightening book, Weav-
ing the Web: The Original Design and Ultimate
Destiny of the World Wide Web by its Inventor
(Berners-Lee & Fischetti, 1999). Since contents
on the Web are constituted by (or at least anno-
tated by) text, the only means for indexing them
has traditionally been text-matching criteria. Of
course, hyperlinking of web resources has been
playing a predominant role in the ranking of search
results (Brin & Page, 1998); however, hyperlink-
ing functionalities are not available in all formats
currently delivered by the Web and, particularly,
not in the spatial data sets and services we are
addressing, nor in the metadata annotating them.
Moreover, text-matching techniques are charac-
terised by a number of shortcomings that make
them inefficient; among these, the more relevant
are related to homonymy, synonymy (biasing,
respectively, precision, and recall in the discovery
process) and of course the difficulty of carrying
out searches in a multilingual context.
315
 Methodologies for Augmented Discovery of Geospatial Resources
In a nutshell, the SW consists of statements
that relate resources to each other or, alternatively,
relate resources to literal values (e.g., the string
“foo”); statements are defined by specifying a
subject, a predicate (or property) and an object
according to the RDF data model (W3C, 2004),
a triple in RDF jargon. Resources are singled out
by using URIs, which provide a straightforward
means to provide unique identifiers to complex
entities. As an example, consider the sentence
“chapter ‘Methodologies for Augmented Dis-
covery of Geospatial Resources’ is included in
the book Discovery of Geospatial Resources:
Methodologies, Technologies, and Emergent Ap-
plications.” Apparently, the sentence is defining
a single statement relating a chapter to a book.
Instead, if we want to unambiguously define the
resources involved, we need to specify the URIs
representing them, increasing the number of neces-
sary statements. A set of sentences for expressing
this may be the following:
• chapter <http://example.com/chapter123>
is entitled ‘Methodologies for Augmented
Discovery of Geospatial Resources
• book <http://example.com/book456>
is entitled Discovery of Geospatial
Resources: Methodologies, Technologies,
and Emergent Applications
• resource <http://example.com/chap-
ter123> is included in resource <http://ex-
ample.com/book456>
As customary in RDF triple representation,
URIs are distinguished by enclosing them in angle
brackets. Even in the extended set of sentences,
some of the semantics conveyed by the original
one is still not formalized. Specifically, the infor-
mation that the two resources are, respectively, a
book chapter and a book is missing; also, prop-
erties “is entitled” and “is included in” do not
explicitly refer to entities that an automated agent
may understand. The former can be expressed by
using the type predicate of RDF for referencing
316
elements from some widely acknowledged data
schema for structuring publications (here, we
consider the chapter and book elements of the
DocBook format17). The titles associated with
the two resources can be rendered by specifying
the title Dublin Core metadata term18; instead, for
expressing compositions of chapters in a book
we can leverage on predicate aggregates from
the OAI-ORE standard (OAI-ORE, 2008), that
has been introduced earlier in this chapter. The
resulting triples are shown in Listing 1.
Lists of triples are quite verbose and not apt
to human consumption; a more convenient rep-
resentation format is that of directed labelled
graphs: In this formalism, ellipses represent re-
sources (note that, in the SW, everything that is
not a literal value is a resource) whose identifiers
are often shortened by substituting a long
namespace (e.g., “http://example.com/”) with
prefixes (e.g., “ex:”). Rectangles represent literal
values, such as the string ‘Discovery of Geospa-
tial Resources: Methodologies, Technologies, and
Emergent Applications.’ Finally, arcs represent
predicates. Figure 4 is showing the graph repre-
sentation of the triples in Listing 1. Despite its
simplicity, the RDF model is capable of represent-
ing very complex data structures, such as n-ary
relations and statements about statements (that is,
considering a whole triple as subject or object of
a statement) (see Figure 4).
The graph representation of RDF makes ap-
parent one of the main strengths of the data
model, that is, its unstructured nature that makes
it possible to seamlessly aggregate information
(statements) from heterogeneous sources. How-
ever, RDF is just the medium for expressing in-
formation on the SW; for structuring it as spe-
cific data models, more standards were layered
on top of RDF, with different degrees of expres-
sive power and computational capabilities. These
will be briefly introduced in the following section.
Methodologies for Augmented Discovery of Geospatial Resources

Listing 1. Example of RDF triples

<http://example.com/chapter123>
<http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
<http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd#chapter>

<http://example.com/book456>
<http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
<http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd#book>

<http://example.com/chapter123>
<http://purl.org/dc/terms/title>
‘Methodologies for Augmented Discovery of Geospatial Resources’

<http://example.com/book456>
<http://purl.org/dc/terms/title>
‘Discovery of Geospatial Resources: Methodologies, Technologies, and Emer-
gent Applications’

<http://example.com/book456>
<http://www.openarchives.org/ore/terms/aggregates>
<http://example.com/chapter123>

Representing Domain Knowledge may need to characterise book456 as a scientific

book (that is, ascribe the resource to a specific
An aspect of data modelling that the basic RDF data class of resources) and not, for instance, as a novel,
model cannot handle is characterizing the entities as a cooking manual, etc. Still we may want to
that are referred to by triples. As an example, we retain some of the characteristics that all books are

Figure 4. RDF graph induced by the statements in Listing 1

317
 Methodologies for Augmented Discovery of Geospatial Resources
sharing, such as the editor, the number of pages,
the selling price, etc. In a word, we would like to
structure data in a more object-oriented fashion.
The schema languages that were defined for the
SW, such as the RDF Schema (RDFS) (W3C,
2004b) and the Web Ontology Language (OWL)
(W3C, 2004c), are addressing this requirement;
schemata defined through the language primitives
defined by RDFS and OWL are typically referred
to as ontologies. Another desirable feature is the
capability of structuring predicates and defining
relations among them. As an example, OAI-ORE
defines predicate isAggregatedBy for expressing
part-whole relations (not the whole-part relations
expressed by predicate aggregates) but the relation
between these (one being the inverse of the other)
is not made explicit. The RDF/XML serialisation
of an OAI-ORE resource map containing the last
triple in Listing 1 may, or may not, feature also
the following triple:
• <http://example.com/chapter123>
• <http://www.openarchives.org/ore/terms/
isAggregatedBy>
• <http://example.com/book456>
In fact, one of the main advantages of SW data
formats is that an RDF graph may convey more
information than that explicitly stated by the graph
itself. This is due to the semantics underlying the
schema languages, which defines the entailments
(that is, the logical implications) that shall hold.
These may be given at different formalization
levels; for example, RDFS provides a set of
rules indicating exactly which inferences shall
be supported by RDFS reasoners (the automated
agents deriving implications). Instead, most OWL
sub-languages (e.g., OWL-Lite, OWL-DL) reflect
the expressive power of some (description) logics
(Baader, et al., 2003) and then do not define entail-
ments as a finite set of production rules. RDFS
and the different flavors of OWL have a well-
defined expressiveness. As an example, RDFS is
expressive enough to arrange classes and predi-
318
cates in a hierarchal way (e.g., for defining class
ScientificBook as a specialization of class Book)
but OWL expressiveness is required to express
that predicates aggregates and isAggregatedBy
are one the inverse of the other. Because of the
associated inference capabilities, data sources that
are defined by SW schema languages are typically
referred to as Knowledge Bases (KB) (Russell &
Norvig, 2005).
A prominent example of OWL ontology in the
geospatial domain is the schema grounding the
GeoNames geographical database19. Queries to the
GeoNames web service can specify the parameter
type=rdf for obtaining results encoded as RDF
according to the schema defined by the service.
As an example, Listing 2 is showing a fragment
taken from the RDF response retrieved from the
service for the search pattern “france.” On the
basis of the associated schema, a reasoner can
derive that France (represented by the URI http://
sws.geonames.org/3017382) is a parentFeature
of Île-de-France because predicate parentCountry
is a sub-property of the former. Another interest-
ing feature of GeoNames is that it also provides
data as LinkedData: this means that, by accessing
any of the URIs in a query response (such as the
one corresponding to France) a user agent may
retrieve the RDF data fragment corresponding to
the resource. This feature allows clients to selec-
tively navigate the huge RDF graph represented by
GeoNames data without requiring the download
of the whole resource.
Other than creating data schemas that can be
instantiated by individuals, such as ex:chapter123,
SW schema languages can also be used for the
sole purpose of structuring domain knowledge
(that is, defining classes and predicates) with only
few or no individuals (members of a class) in-
stantiating them. This approach allows for a high
expressiveness when defining controlled vo-
cabularies (i.e. thesauri) for modelling a given
domain. In the context of SDI-related thesauri,
this approach is exemplified by the Semantic Web
for Earth and Environmental Terminology
Methodologies for Augmented Discovery of Geospatial Resources

Listing 2. Fragment of the RDF output of the GeoNames web service

<gn:Feature rdf:about=”http://sws.geonames.org/3012874/”>
<rdfs:isDefinedBy>http://sws.geonames.org/3012874/about.rdf</
rdfs:isDefinedBy>
<gn:name>Île-de-France</gn:name>
<gn:alternateName xml:lang=”fr”>Île-de-France</gn:alternateName>
<gn:alternateName xml:lang=”en”>Île-de-France</gn:alternateName>
<gn:alternateName xml:lang=”es”>Isla de Francia</gn:alternateName>
<gn:alternateName xml:lang=”de”>Île-de-France</gn:alternateName>
<gn:featureClass rdf:resource=”http://www.geonames.org/ontology#A”/>
<gn:featureCode rdf:resource=”http://www.geonames.org/ontology#A.ADM1”/>
<gn:countryCode>FR</gn:countryCode>
<gn:population>11598866</gn:population>
<wgs84_pos:lat>48.5</wgs84_pos:lat>
<wgs84_pos:long>2.5</wgs84_pos:long>
<gn:parentCountry rdf:resource=”http://sws.geonames.org/3017382/”/>
<gn:childrenFeatures rdf:resource=”http://sws.geonames.org/3012874/con-
tains.rdf”/>
<gn:locationMap rdf:resource=”http://www.geonames.org/3012874/region-ile-
de-france.html”/>
</gn:Feature>

(SWEET) (NASA-JPL, 2011), a set of 201 Developing Thesauri for SDIs

highly modularised OWL ontologies whose terms
are expressed as Class instances and arranged
hierarchically by using predicate subClassOf from
RDFS. The lexical representation of terms is
constituted by the fragment identifiers of URIs
(e.g., “#ClimateChange”). The exploitation of
constructs from the meta-language (OWL) allows
for defining custom predicates that class in-
stances may have and constrain ontology con-
structs according to them. As an example, Listing
3 is portraying the definition of class #Mesoclimate
whose members are defined as either instances
of super-class #Climate and also instances whose
property #hasSpatialScale takes values from class
#Mesoscale.
The discovery of geospatial data and services on
the Internet is a topic that, despite the hype nour-
ished by the novel notion of SDI (Masser, 2005),
still lacks efficient technologies and techniques for
enabling an effective retrieval of resources. The
non-textual nature of this category of resources,
the scarce set of metadata that is often annotat-
ing them, the multilingual gap (and in general
the linguistic issues hindering the indexing of all
categories of resources) motivates the recourse to
specialized applications, geoportals (Maguire &
Longley, 2005), for retrieval of spatial data and
services. Even when an exhaustive set of metadata
is provided, the distance between the terminology
adopted by the domain expert during annotation
and that of the end user (which can be herself a
thematic user, but coming from a different do-
main) makes it difficult to reconcile the metadata

319
 Methodologies for Augmented Discovery of Geospatial Resources

Listing 3. Definition of class #Mesoclimate in the SWEET ontology

<owl:Class rdf:about=”#Mesoclimate”>
<rdfs:subClassOf rdf:resource=”#Climate”/>
<rdfs:subClassOf>
<owl:Restriction>
<owl:onProperty rdf:resource=”&scale;#hasSpatialScale”/>
<owl:allValuesFrom rdf:resource=”&scale;#Mesoscale”/>
</owl:Restriction>
</rdfs:subClassOf>
<rdfs:comment xml:lang=”en”>The climate of a natural region of small ex-
tent, for example, valley, forest, plantation, and park. Because of subtle
differences in elevation and exposure, the climate may not be representative
of the general climate of the region.</rdfs:comment>
</owl:Class>

descriptions on the one hand and the search pat-
terns on the other. Also, specifying metadata in
a language that is different from that of the end
user may utterly prevent retrieval of a resource.
Albeit easing access to geospatial information,
geoportals often fall short of providing advanced
search functionalities that allow the end user for
bridging this gap.
Acknowledging these shortcomings, the SDI
community is increasingly considering recourse
to thesauri as one of the most promising means to
annotate resources in a consistent way and to ease
retrieval of this category of resources in a multi-
lingual and cross-thematic context. In particular,
thesauri based on SW formats provide language-
neutral identifiers (URIs) that can be related with
distinct syntactic forms in order to be selected for
annotation through multiple, language-dependant
textual representations. Also, the internal structure
of thesauri (relating terms with each other ac-
cording to specificity and relatedness) allows for
advanced functionalities such as query refinement
on the basis of semantic properties among terms.
Finally, by aligning independent thesauri produced
in distinct thematic areas, it is possible to easily
bridge between different domains. Using SW
schema languages for articulating thesauri, such
as for the SWEET ontologies, may be awkward
to domain experts because of the inherent tech-
nicalities. Moreover, applying inference to large
schema definitions may be particularly expensive
on the computational side.
Instead, a different approach that is gaining
wide acceptance is constituted by relying on a
fixed OWL schema and articulate thesauri by
instantiating the classes and predicates defined by
the former. The clear advantage with respect to
the previously described approach is the capabil-
ity of bounding the complexity of data structures
on the basis of the underlying schema. As for the
expressiveness, it is depending on the specific
data schema that is taken into consideration. The
research on Knowledge Organisation Systems
(KOS) (Bechhofer & Goble, 2001) is focusing on
the development of thesauri organizing terms in
a coherent way. The intuition this research elabo-
rated on is that it is possible to take advantage of
the Semantic Web infrastructure in a lightweight,
easily implementable way. The Simple Knowledge
Organisation System (SKOS20) constitutes the
main output of this research thread: it is a low
complexity ontology that allows to easily structure
terms (also providing translations into multiple
languages), to group them into collections of terms

320
Methodologies for Augmented Discovery of Geospatial Resources

Figure 5. The SKOS ontology: a) class hierarchy

and to relate terms from independent thesauri to
and b) property hierarchy
one another.

The Simple Knowledge

Organisation System

SKOS represents a breakthrough in knowledge

organisation inasmuch it provides an exhaustive
set of constructors for creating thesauri without
involving the technicalities of more expressive
ontology schema languages. Making a comparison
with XML technologies, it is like exploiting an
XML custom vocabulary for expressing a given
category of data structures versus dealing with
XML Schema constructs for expressing them.
Figure 5 is showing a general view of the entities
defined by SKOS, as rendered in the Protégé on-
tology editor21 (the names in bold building on the
entities of OWL and RDFS). Class ConceptScheme
is used for creating a whole thesaurus; Collection
and OrderedCollection allow to group terms;
finally, class Concept represent individual terms
in the thesaurus (see Figure 5).
More complex is the organisation of properties.
A set of properties can be used to associate labels
(that is, human-readable text representations) to
terms that are, by themselves, represented by
URIs: prefLabel, altLabel, and hiddenLabel rep-
but are typically not introduced explicitly because
resent, respectively, preferred labels, alternative
it is up to the underlying reasoning facilities to
labels and labels that are possible misspellings of
deduce them. As we will see, all categorisations
term labels. A second set of properties allow the
of SDI-related terms considered in this Chapter
grouping of entities in a thesaurus: properties
are based on SKOS or on an extension to it.
inScheme and topConceptOf allow to associate
terms (Concepts) with the thesaurus (Con-
SKOS Thesauri in the SDI-Domain
ceptScheme) they belong to and allow to specify,
among terms, which are the more general. Fi-
The development of thesauri is a key factor for
nally, a set of properties can be used to categorise
geographic information retrieval and led to the
terms according to specificity (broader, nar-
inclusion in the GEO Work Plan 2009-2011 (GEO,
rower) and relatedness (related); their variants
2010) of a sub-task of the GEOSS Common In-
terminating with “Match,” together with proper-
frastructure devoted to ontology and taxonomy
ties exactMatch and closeMatch, are used for
development. Early adopters of SKOS as either
relating terms from distinct thesauri while those
the native or an alternative format for represent-
terminating with “Transitive” are introduced to
ing thesauri include the United Nations Food and
support the transitive closure of sub-properties

321
 Methodologies for Augmented Discovery of Geospatial Resources
Agriculture Organization which is maintaining
AGROVOC (FAO, 2011), a multilingual resource
comprising over 28.000 terms covering all sub-
ject fields in agriculture, forestry, fisheries, food
and related domains. The United States National
Agricultural Library is maintaining a bilingual ag-
ricultural thesaurus made up by more than 80.000
terms (USDA). In Europe, the GEMET Thesaurus
(EIONET, 2011) is providing a narrower set of
terms (around 6000 terms) but is covering all
languages spoken in the European Union.
Also in Europe, the EU Publications Office is
providing the SKOS version of EuroVoc (2011), a
multilingual and multidisciplinary thesaurus cov-
ering all areas of activity of the EU. Recently, the
categorization of Societal Benefit Areas (SBAs)
underlying GEOSS has been encoded as SKOS
and translated into four more languages in the
context of the EuroGEOSS FP7 project22. Even if
this resource can barely be considered a thesaurus,
since it is featuring only 68 terms, it constitutes
a good example of how resources not natively
encoded as SKOS can be easily translated and
extended in order to support multiple languages.
Also, a large amount of data sets and services are
expected to be categorized according to SBAs
and, consequently, the availability of this small
thesaurus, when properly interconnected with
other reference thesauri, may prove very useful
for resource discovery.
Finally, several other institutions and proj-
ects are similarly digitising SDI-related domain
knowledge by using SKOS as the encoding
format. Starting from the CUAHSI Hydrologic
Information System23, a thematic thesaurus for
water has been developed in GEOSS AIP-3
and is currently being translated into SKOS for
inclusion in the knowledge base considered in
this Chapter. The Australian agency CSIRO is
developing vocabulary services24 for thematic
thesauri that are made accessible through an API
akin to the one implemented by GEMET; among
these, the extensive water thesaurus derived from
the WDTF schema of the Australian Bureau of
322
Meteorology. In general, since knowledge man-
agement has a direct application in the internal
functioning of companies and institutions, more
and more domain-specific thesauri are expected
to be developed in the future.
With regard to the thesauri that were exploited
as reference thesauri during development of the
semantics-aware component that is described in
this Section, the source that was taken into con-
sideration has been the RDF repository developed
in the context of the GENESIS FP7 project25. The
repository is an instance of the Sesame frame-
work26, an open source Java web application for
storing and querying RDF data, that is currently
hosting the following thesauri in the SKOS format:
• The General Multilingual Environmental
Thesaurus (GEMET27): 28 of the 31 lan-
guages currently provided by the EIONET
portal.
• The INSPIRE Feature Concept Dictionary
and Glossary: 21 of the 23 EU official lan-
guages for INSPIRE Themes, monolingual
the other terms.
• The ISO 19119 categorisation of spatial
data services: 21 of the 23 EU official
languages.
• The GEOSS Societal Benefit Areas: 5
languages.
However, as explained in the following
Section, in principle any SPARQL-compliant
repository can be seamlessly integrated and also
sources based on a different protocol can be easily
integrated. Another important precondition for
applying the semantic augmentation paradigm
to the retrieval of geospatial resources was the
harmonization of the distinct thesauri by align-
ing corresponding terms among the resources
listed above.
Methodologies for Augmented Discovery of Geospatial Resources
An Architecture for Transparent
Semantic Augmentation
One possible concrete application of the third-
party discovery augmentation approach is to en-
able semantic discovery of geospatial resources.
The objective is to develop a flexible framework
that aims to provide users with semantics-aware
query capabilities to discover geospatial resources
from one (or more) traditional/standard discovery
services. The semantic augmentation mechanism
is transparent to resource providers because no
additional meta-information needs to be explicitly
added to existing metadata repositories.
According to the third party approach, the
framework we designed makes use of a specific
Discovery Augmentation Component (DAC)
that implements the business logic required for
semantic augmentation. The DAC implements
a query expansion strategy according to which
multiple, traditional geospatial queries are derived
from a single semantic query. Existing semantic
services are used to expand queries and combine
the related results in a meaningful way.
In fact, the DAC is able to accept a query, ex-
pand it as multiple, semantically-related queries by
accessing a customizable set of external semantic
services (thesauri, ontologies, gazetteers, etc.),
and finally issue them to a geospatial discovery
service. The DAC combines query results in a
meaningful way and sends them back to the client.
This framework realizes the Information
Expert design pattern (Larman, 2005) assigning
specific tasks (responsibilities) to components
that have the information needed to carry out the
task, making the architecture flexible and scalable.
Moreover, it does not affect existing geospatial
services interfaces by implementing a loosely-
coupled solution. However, in order to enable
this architecture, specific solutions are required in
order to define the exposed interface, to address
resources heterogeneity and performances issues.
This framework was implemented and tested
in the context of the FP7 EuroGEOSS (A Euro-
pean approach to GEOSS) project28 to provide
semantics-aware capabilities to the EuroGEOSS
discovery broker (EuroGEOSS, 2010; Nativi &
Bigagli, 2009).
System Architecture Overview
The transparent semantic augmentation frame-
work was designed applying the following general
principles:
• Layered Architecture;
• Separation of Concerns;
• System of Systems (SoS) Approach.
The layered architecture is a well-known
approach for designing complex systems (ISO,
1994), where functionalities are grouped and lay-
ered according to their abstraction level. Figure 6
shows the three layers of the proposed architecture,
following the SOA taxonomy:
• Service Consumer Layer: This layer
implements presentation functionalities, in
fact it is comprised of components which
implement Graphical User Interfaces
(GUIs);
• Semantic Augmentation Service Layer:
This layer is composed of service compo-
nents which implement the business logic
necessary to integrate semantic and geo-
spatial services;
• Query Service Provider Layer: This lay-
er contains services providing both seman-
tic and geospatial query functionalities.
The DAC clearly falls into the semantic aug-
mentation service layer, which makes use of the
query service provider layer.
By separating the concerns, we obtain a
set of “dedicated” components: each of them
implementing a precise and well-defined set of
functionalities. The catalog service component
is dedicated to execute “traditional” geospatial
323
 Methodologies for Augmented Discovery of Geospatial Resources
Figure 6. Architecture of the transparent semantic augmentation framework
queries. The semantic service provides access to
semantic repositories (e.g. thesauri, ontologies,
etc.). The DAC is in charge of executing the query
expansion (accessing the semantic services) and
distributes the resulting set of “traditional” queries
to the federated catalog service. Finally, the user
agent provides results visualization.
The System of Systems approach implies the
capability of connecting with existing systems
without any modification to their interfaces and/
or functionalities. This is achieved by providing
interoperability arrangements to mediate from a
geospatial semantic query to a set of semantic
queries and traditional geospatial queries.
The choice of the service interfaces was mainly
driven by the need of being compliant with widely
adopted catalog service specifications and to en-
324
sure interoperability with existing systems. For
the interaction between the DAC and the catalog
service, the OGC CSW/ISO AP (OGC, 2007b)
(Application Profile) interface is used. Among
the present OGC CSW APs, ISO AP is presently
one of the most widely implemented. Besides, it is
the INSPIRE-compliant catalog service interface.
The query syntax to access the semantic service
follows the W3C standard SPARQL (W3C, 2008).
In keeping with the SoS approach, the DAC was
conceived to be flexible enough to interoperate
with other query languages published by different
types of semantic services.
The DAC publishes an interface for user agent
bindings. This interface allows clients to submit
queries with a possible combination of semantic,
geospatial, and free text constraints. At the time
Methodologies for Augmented Discovery of Geospatial Resources
being, there is no standard interface or syntax
allowing such combined queries. Hence, we de-
cided to introduce an extension of the OpenSearch
interface (Clinton, 2009). The OpenSearch is a
lightweight interface that allows agents to query
catalogs with a simple free text search. There
exist several extensions of the basic OpenSearch
syntax; two widely used extensions for geospatial
queries are:
• The Geo extension (Turner, 2010;
Gonçalves, 2010): this allows agents to
specify a spatial extent/location as con-
straint in a query;
• The Time extension (Gonçalves, 2010):
this allows agents to build queries based
on time and time span constraints.
We introduced a new extension called: “Con-
cept-driven extension.” This allows the discovery
of well-defined concepts and their relations from
semantic services. The DAC query interface
implements all these three extensions.
OpenSearch Concept-Driven
Extension and Query Expansion
The introduced extension is based on the concept
of RDF triple (see see paragraph ‘Grounding
semantics-aware discovery’) (W3C, 2004). Con-
Table 1. Semantic query parameters, example 1
searchTerm Subject
drought http://www.eionet.europa.eu/gemet/concept/9221
This query selects all concepts more specific than the concept “water resource” and whose description contain the keyword “drought”
Table 2. Semantic query parameters, example 2
searchTerm Subject
- http://www.eionet.europa.eu/gemet/concept/9221
This query selects all concepts more specific than the concept “water resource”
sidering the basic OpenSearch interface, three
additional semantic parameters can be specified
in the query:
A. Subject
B. Predicate (semantic relation)
C. Object
All parameters are optional and can be com-
bined with the free text (searchTerm) parameter
in many different ways. The three parameters
are used to represent an RDF statement that the
selected concepts must satisfy. When the search-
Term (free text) parameter is also specified, the
selected concepts are filtered to match the given
searchTerm—that is, concepts in whose descrip-
tions the free text (searchTerm) appears.
The three semantic parameters represent se-
mantic concepts; thus, they can be represented in
different ways by semantic services. When the
user agent and the semantic service make use
of different representations, the DAC is able to
implement the necessary mediation functionality.
Examples of possible query parameters com-
binations are given in Tables 1, 2, 3, 4. They use
the URI “http://www.eionet.europa.eu/gemet/
concept/9221” from the GEMET Thesaurus
for representing the semantic concept “water
resource.”
Predicate Object
skos:narrower -
Predicate Object
skos:narrower -
325
 Methodologies for Augmented Discovery of Geospatial Resources
Table 3. Semantic query parameters, example 3
searchTerm Subject
drought http://www.eionet.europa.eu/gemet/concept/9221
This query selects all concepts which are related to the concept “water resource” according to the DAC default relation (customizable)
and whose description contain the keyword “drought”
Table 4. Semantic query parameters, example 4
searchTerm Subject
drought - -
This query selects all concepts whose description contain the keyword “drought”
The implementations of the “Concept-driven
extension” must support at least the GeoRSS and
X-Suggestion29 return types. GeoRSS is a recent
extension of the RSS30 feed format; it enriches
traditional feeds with geographic and temporal
information. When the query asks for a GeoRSS
return type, the DAC builds a set of traditional
geospatial queries using the multilingual text
representation of the selected concepts in order
to retrieve metadata records form the catalog
service.
In order to create requests such as those in
Table 1, 2, 3 the client software must be able to
retrieve the URIs (or any other kind of represen-
tation of semantic concepts) from the semantic
service. To achieve this, the X-Suggestion return
type is used. When this return type is requested,
the DAC selects the concepts without submit-
ting queries to the catalog service. The selected
concepts and their textual description are returned
to the components of the service consumer layer
(user agents) in order to be viewed by the user
and, eventually, used in subsequent queries to the
DAC. This feature enables the use of the DAC
for browsing the content of the semantic service.
More return types can be defined and imple-
mented for addressing specific needs. Examples
of suitable return types are: XML documents for
RDF graphs, ISO 19115 (ISO, 2003), metadata
encodings for geospatial resources.
326
Predicate Object
- -
Predicate Object
-
Using the concept-driven extension, it was
possible to implement two different types of
query expansion:
1. Automatic query expansion
2. User-assisted query expansion
For the first type, the main discovery steps are
depicted in Figure 7.
A. The query keywords (the what constraint)
are “expanded” with the concepts retrieved
from the semantic services accessed by the
DAC;
B. The selected concepts are used to build a
set of “traditional” geospatial queries to be
submitted to the catalog service accessed
by the DAC;
C. The DAC performs a “smart” aggregation of
the queries results and provides them back
to the client (user agent).
The second type differs from the first one only
for the first step. In fact, in this case, the selection
of the concepts of interest is not automatic; the
user can freely navigate the content of the seman-
tic services and select which concepts will be used
to query the catalog service. Figure 8 depicts the
steps of this second type of query expansion
Methodologies for Augmented Discovery of Geospatial Resources

Figure 7. Sequence diagram of the automatic query expansion

Figure 8. Sequence diagram of the user-assisted query expansion

327
 Methodologies for Augmented Discovery of Geospatial Resources
Discovery Augmentation Component
In order to implement the architecture described
above, the DAC design is crucial. As for the
functional requirements, the main functionality of
the DAC is the implementation of the two query
expansion strategies; the DAC must be able to:
• Interpret a semantic geospatial query;
• Expand the semantic geospatial query into
one or more semantic queries;
• Build a set of geospatial queries and sub-
mit them to the catalog service;
• Combine the results of a set of geospatial
queries in a meaningful way;
• Transform the results to the requested re-
turn type.
Figure 9. DAC internal modules
328
The design of the DAC internal modules is
based on the Mediation pattern: a central module
(the mediator) orchestrates a set of modules that
operate independently from each other. This way,
it is possible to concentrate the business logic that
is required to expand the query into one central
module: the Orchestrator. The other modules
provide specific functionalities, such as generat-
ing and executing semantic queries, generating
geospatial queries, executing geospatial queries,
and transforming the results to the desired return
type. In order to achieve the desired flexibility,
communication between the Orchestrator and
the other modules is completely decoupled. The
Orchestrator module interprets the semantic geo-
spatial query and, accordingly, calls the necessary
modules in the right order to process the request.
Figure 9 depicts the main DAC internal modules.
Methodologies for Augmented Discovery of Geospatial Resources
The OpenSearch Profiler reads incoming re-
quests, checks for validity, and forwards them to
the Orchestrator in a more structured form. These
functionalities must not be implemented by the
Orchestrator to guarantee more flexibility; in fact,
the Orchestrator must be agnostic to the com-
munication protocols published by the DAC. Thus,
future useful protocols can be added just by imple-
menting new profilers.
The Semantic Service Manager is in charge of
generating and submitting requests to the semantic
service(s). This module makes of use of a set of
Adapters to manage semantic services heterogene-
ity. These adapting modules mediate between the
internal Semantic Service Manager interface and
the specific semantic service query language (e.g.
SPARQL). Thus, in order to add a new semantic
service type (e.g. a gazetteer service), it is suf-
ficient to implement the corresponding Adapter.
The OpenSearch2GetRecords Mapper module
is able to read the non-semantic parameters of the
request (i.e. spatial and temporal constraints) and
create an ISO GetRecords request—which can be
submitted to a standard OGC CSW/ISO catalog.
Given an ISO GetRecords request and a set of
concepts (represented as terms), the CSW Request
Builder module expands the original request gen-
erating a set of related ISO GetRecords requests.
Figure 10. DAC data model
The CSW Manager module communicates with
the catalog service. It should be noted that this
component does not make use of any adaptation
functionalities. In fact, solutions for mediating
different types of geospatial discovery services
already exist (EuroGEOSS, 2010). Thus, where
necessary, they should be coupled with the DAC.
Finally, the Result Converter module encodes
the results (which can be both metadata records
and semantic concepts) into the desired return type.
The DAC data model considers two main
information elements:
• Metadata Records
• RDF elements (concepts and semantic
relations).
Figure 10 depicts a simplified schema of the
DAC data model. This model connects ISO meta-
data records and concepts represented as RDF
nodes, allowing building an RDF graph between
metadata records (see Figure 11).
Referring to Figure 11, relations among meta-
data records are more “relaxed” than those ob-
tained by linking the metadata elements directly:
Figure 12a) depicts the case of direct links; while
Figure 12b) shows the case of metadata elements
linked through concepts.
329
 Methodologies for Augmented Discovery of Geospatial Resources
Figure 11. Graph connecting concepts and metadata records
In the design of the data model, an important
choice is the strategy adopted to link metadata
records to concepts. In fact, these two kinds of ele-
ments can be linked in several ways, determining
the reliability of the Relaxed Relation. Basically,
this can be decided case by case depending on
the overall system needs.
Metadata records are composed of several ele-
ments (e.g. title, abstract, keywords, etc). These
elements are used to link the metadata records to
one or more matching semantic concepts—see
Figure 12
One strategy is to match the “keyword” ele-
ment content to one or more semantic concepts.
This strategy generates a very reliable Relaxed
Relation, resulting quite restrictive however.
A more general but less reliable strategy
consists of taking into account also other meta-
data elements such as “title,” “abstract,” etc. We
adopted this strategy, defining the link between
metadata records and semantic concepts as it fol-
lows: a metadata record is linked to a semantic
330
concept if this appears in one of metadata record
“keyword” elements OR in any other textual ele-
ment of the metadata record.
CONCLUSION
In this chapter, we discussed the enhancement
of geospatial resources discovery services to
support semantic queries. After explaining the
general vision, we exposed specific issues related
to the semantic discovery of existing geospatial
resources provided through standards services and
Web2.0 services. Methodologies, architectures,
and current experimentations are also discussed.
A possible methodology to enhance discovery
capabilities of SDIs is to augment the searchable
content that is associated with geospatial resources.
We have defined three high-level approaches for
providing the required additional meta-informa-
tion: provider-based, user-based, and third party.
We have selected the more extensible and flex-
Methodologies for Augmented Discovery of Geospatial Resources
Figure 12. Relations between metadata records in the DAC data model
ible ones: the user-based approach, and the third
party approach.
Two architectures have been presented, each
implementing one of the selected approaches. In
the first one, regarding the user-based approach,
the architecture makes use of tags in two different
but complementary scenarios. The first scenario
considers tagging at discovery time, when users
launch a query based on a set of tags. These tags
are either introduced by the users or suggested by
the system given the pool of previous tags used
by others. In this case, the previous queries give
clues to non-expert users to successfully discovery
geospatial resources. The second scenario involves
the grouping of the resources that are discovered
into collections. Users are allowed to annotate
with tags the collection of discovered resources
on the basis of certain relationships. In this case,
not only individual resources are annotated but
whole collections that contain resources that are
related to the same topics.
The second solution we presented implements
the third party approach in order to extend tra-
ditional discovery functionalities with semantic
capabilities. Leveraging the Semantic Web
approach, we described a flexible framework
that transparently augments user queries with
concepts related to the free-text constraint in the
original query. This architecture is centered on a
third party component, the Discovery Augmenta-
tion Component (DAC), which implements the
business logic needed to integrate semantic and
geospatial capacities.
The two architectures have been prototyped in
the context of the EC-funded EuroGEOSS project
and are part of the EuroGEOSS multidisciplinary
interoperability infrastructure. This was success-
fully tested in several use scenarios of the GEOSS
AIP-3 (Architecture Implementation Pilot—Phase
3) in collaboration with the GENESIS project
(Nativi, et al., 2011; Fugazza, et al., 2011; Pozzi,
et al., 2011). Demonstration videos are available
at http://www.ogcnetwork.net/pub/ogcnetwork/
GEOSS/AIP3/pages/Demo.html.
With regard to the user-centered approach,
future research threads will be the inclusion of
new Web 2.0 services, exchange formats and
protocols, and the potential integration with of-
ficial SDI content, the recommender module is
still an open issue and existing and innovative
techniques must be tested.
With regard to the DAC, future work will
extend the set of semantic service interfaces that
are currently supported. Besides, new controlled
331
 Methodologies for Augmented Discovery of Geospatial Resources
vocabularies and ontologies will be integrated. In
our experimentation, the alignment of different
thesauri and ontologies was carried out manu-
ally by domain experts; an automatic alignment
approach for the DAC is another challenge that
will be considered in the future.
ACKNOWLEDGMENT
The work was partially funded by the European
Commission’s Seventh Framework Programme
under contracts 226487 and 223996.
REFERENCES
Abargues, C., Granell, C., Díaz, L., & Huerta, J.
(2010). Aggregating geoprocessing services using
the OAI-ORE data model. International Journal
on Advances in Intelligent Systems, 3(3-4).
Baader, F., Calvanese, D., McGuinness, D., Nardi,
D., & Patel-Schneider, P. (2003). The descrip-
tion logic handbook: Theory, implementation
and applications. Cambridge, UK: Cambridge
University Press.
Barragáns-Martínez, A. B., Rey-López, M.,
Costa-Montenegro, E., Mikic-Fonte, F. A., Bur-
guillo, J. C., & Peleteiro, A. (2010). Exploitation
of social tagging for outperforming traditional
recommender system techniques. IEEE Internet
Computing, 14(6), 23–30.
Berners-Lee, T., & Fischetti, M. (1999). Weaving
the web: The original design and ultimate destiny
of the world wide web by its inventor. San Fran-
cisco, CA: Harper.
Brin, S., & Page, L. (1998). The anatomy of a
large-scale hypertextual web search engine. Paper
presented at the Seventh International World-Wide
Web Conference. New York, NY.
332
Clinton, D. (2009). OpenSearch 1.1 specification
(draft 4). Retrieved June 8, 2011, from http://www.
opensearch.org /Specifications/OpenSearch/1.1/
Draft_4.
Craglia, M., Goodchhild, M. F., Annoni, A.,
Camara, G., Gould, M., & Kuhn, W. (2008).
Next-generation digital earth: A position paper
from the Vespucci initiative for the advancement
of geographic information science. International
Journal of Spatial Data Infrastructures Research,
3, 146–167.
Díaz, L., Granell, C., Gould, M., & Huerta,
J. (2011). Managing user generated informa-
tion in geospatial cyberinfrastructures. Future
Generation Computer Systems, 27(3), 304–314.
doi:10.1016/j.future.2010.09.002
EIONET. (2011). The general multilingual envi-
ronmental thesaurus (GEMET). Retrieved May 20,
2011, from http://www.eionet.europa.eu /gemet.
EuroGEOSS WP2. (2010). Deliverable 2.2.2
– Specification of the EuroGEOSS initial op-
erating capacity. Retrieved May 20, 2011,
from http://www.eurogeoss.eu /Documents/
EuroGEOSS_D_2_2_2.pdf.
EuroVoc. (2011). EuroVoc – The EU’s multilingual
thesaurus. Retrieved May 20, 2011, from http://
eurovoc.europa.eu/drupal/.
FAO. (2011). FAO – The AGROVOC thesaurus.
Retrieved May 20, 2011, from http://aims.fao.org
/website/ AGROVOC-Thesaurus/sub.
Franklin, C., & Hane, P. (1992). An introduction
to GIS: Linking maps to databases. Database,
15(2), 17–22.
Fugazza, C., Nagai, M., Nativi, S., Santoro, M.,
& Pozzi, W. (2011). GEOSS AIP-3 engineering
report – Vocabularies and semantics scenario.
Retrieved May 20, 2011, from http://www.ogc-
network.net /pub/ogcnetwork/GEOSS/AIP3/
documents/AIPscenario_Semantics_1.5.pdf.
Methodologies for Augmented Discovery of Geospatial Resources
Gonçalves, P. (2010). OpenGIS® OpenSearch
geospatial extensions draft implementation stan-
dard, ver 0.0.2. OGC 09-084r3. Retrieved from
http://www.ogcnetwork.net.
Group on Earth Observation. (2010). GEOSS
– 2009-2011 work plan. Retrieved May 20,
2011, from http://www.earthobservations.
org /documents/work%20plan/ geo_wp0911_
rev3_101208.pdf.
INSPIRE. (2010). INSPIRE metadata implement-
ing rules: Technical guidelines based on EN ISO
19115 and EN ISO 19119, version 1.2. Retrieved
May 20, 2011, from http://inspire.jrc.ec.europa.eu
/documents/Metadata/ INSPIRE_MD_IR_and_
ISO_v1_2_20100616.pdf.
International Organization for Standardization.
(1994). Information technology -- Open systems
interconnection -- Basic reference model: The
basic model. ISO/IEC 7498-1. Geneva, Switzer-
land: ISO.
International Organization for Standardization.
(2003a). Geographic information – Metadata. [E]
[. Geneva, Switzerland: ISO.]. ISO, 19115, 2003.
International Organization for Standardization.
(2003b). Geographic information – Services.
ISO/IS, 19119:2003. Geneva, Switzerland: ISO.
Klien, E., Einspanier, U., Lutz, M., & Hubner, S.
(2004). An architecture for ontology-based discov-
ery and retrieval of geographic information. In F.
Toppen & P. Prastacos (Eds.), Proceedings of the
Seventh Conference on Geographic Information
Science (AGILE 2004), (pp. 179-188). Heraklion,
Greece: Crete University Press.
Larman, C. (2005). Applying UML and patterns
- An introduction to object-oriented analysis and
design and iterative development (3rd ed.). Upper
Saddle River, NJ: Prentice Hall PTR.
Lemmens, R., Wytzisk, A., de By, R., Granell, C.,
Gould, M., & van Oosterom, P. (2006). Integrat-
ing semantic and syntactic descriptions to chain
geographic services. IEEE Internet Computing,
10(5), 42–52. doi:10.1109/MIC.2006.106
MacEachren, A. M., & Kraak, M. J. (2001).
Research challenges in geovisualization. Cartog-
raphy and Geographic Information Science, 28,
3–12. doi:10.1559/152304001782173970
Maguire, D. J., & Longley, P. A. (2005). The
emergence of geoportals and their role in spatial
data infrastructures. Computers, Environment
and Urban Systems, 29, 13–14.
Masser, I. (2005). GIS worlds: Creating spatial
data infrastructures. New York, NY: ESRI Press.
NASA-JPL. (2011). Semantic web for earth and
environmental terminology (SWEET). Retrieved
May 20, 2011, from http://sweet.jpl.nasa.gov/.
Nativi, S. (2010). The implementation of interna-
tional of geospatial standards for earth and space
sciences. International Journal of Digital Earth,
3(1), 2–13. doi:10.1080/17538941003764412
Nativi, S., & Bigagli, L. (2009). Discovery, me-
diation, and access services for earth observation
data. IEEE Journal of Selected Topics in Applied
Earth Observations and Remote Sensing, 2(4),
233–240. doi:10.1109/JSTARS.2009.2028584
Nativi, S., Santoro, M., Dubois, G., Skoien, J.
O., De Jesus, J., De Longueville, B., & Fugazza,
C. (2011). GEOSS AIP-3 engineering report –
eHabitat. Retrieved May 20, 2011, from http://
www.ogcnetwork.net /pub/ogcnetwork/GEOSS/
AIP3/documents/ CCBio-eHabitat-ER-v2.0-
FINAL.pdf.
Nuñez, M., Díaz, L., Granell, C., & Huerta, J.
(2011). Web 2.0 broker: A tool for massive col-
lection of user information. In Proceedings of the
European Geosciences Union (EGU) General As-
sembly 2011 (EGU 2011). Retrieved 8 June, 2011,
from http://meetings.copernicus.org /egu2011/.
333
 Methodologies for Augmented Discovery of Geospatial Resources
O’Reilly, T. (2005). What Is web 2.0 - Design
patterns and business models for the next genera-
tion of software. Retrieved May 20, 2011, from
http://oreilly.com /pub/a/web2/archive/what-is-
web-20.html.
O’Reilly, T., & Battelle, J. (2009). Web squared:
Web 2.0 five years on. Retrieved May 2, 2011,
from http://www.web2summit.com /web2009/
public/ schedule/detail/10194.
Open Archives Initiative. (2008a). Object reuse
and exchange specification. Retrieved May 20,
2011, from http://www.openarchives.org /ore/.
Open Archives Initiative. (2008b). ORE user
guide – Abstract data model. Retrieved May 20,
2011, from http://www.openarchives.org /ore/1.0/
datamodel.
Open Geospatial Consortium. (2007a). OpenGIS®
catalogue services specification, ver. 2.0.2. OGC
07-006r1. Retrieved from http://www.ogcnet-
work.net.
Open Geospatial Consortium. (2007b). OpenGIS®
catalogue services specification - ISO metadata
application profile, ver. 1.0.0. OGC 07-045. Re-
trieved from http://www.ogcnetwork.net.
Open Geospatial Consortium. (2008). OGC KML,
ver 2.2.0. OGC 07-147r2. Retrieved from http://
www.ogcnetwork.net.
Pozzi, W., Fugazza, C., Brewer, M. J., Santoro,
M., Nativi, S., Lee, B., & Enenkel, M. (2011).
GEOSS AIP-3 engineering report – Global
drought monitoring service through the GEOSS
architecture. Retrieved May 20, 2011, from http://
www.ogcnetwork.net /pub/ogcnetwork/GEOSS/
AIP3/documents/ GEOSS_AIP3_DroughtWa-
ter_ER.pdf.
Richards, D. (2007). A social software/Web 2.0
approach to collaborative knowledge engineer-
ing. Information Sciences, 179(15), 2515–2523.
doi:10.1016/j.ins.2009.01.031
334
Russell, S., & Norvig, P. (2005). Artificial intel-
ligence: A modern approach. Upper Saddle River,
NJ: Prentice Hall.
Smits, P., & Friis-Christensen, A. (2007). Resource
discovery in a European spatial data infrastruc-
ture. IEEE Transactions on Knowledge and
Data Engineering, 19(1), 85–95. doi:10.1109/
TKDE.2007.250587
Strohmaier, M., Koerner, C., & Kern, R. (2010).
Why do users tag? Detecting users’ motivation for
tagging in social tagging systems. In Proceedings
of the Fourth International AAAI Conference on
Weblogs and Social Media (ICWSM 2010). AAAI.
Stvilia, B., Twidale, M. B., Smith, L. C., & Gasser,
L. (2008). Information quality work organization
in Wikipedia. Journal of the American Society
for Information Science and Technology, 59,
983–1001. doi:10.1002/asi.20813
Turner, A. (2010). OpenSearch Geo extension 1.0
(draft 2). Retrieved June 8, 2011, from http://www.
opensearch.org /Specifications/OpenSearch/ Ex-
tensions/Geo/1.0/Draft_2.
USDA. (2011). National agricultural library
thesaurus and glossary. Retrieved May 20, 2011,
from http://agclass.nal.usda.gov /agt.shtml.
Vilches-Blazquez, L. M., & Corcho, O. (2009).
A heuristic approach to generate good quality
linked data about hydrography. In Proceedings of
the Database and Expert Systems Applications,
International Workshop, (pp. 99-103). IEEE Press.
World Wide Web Consortium. (2004a). Resource
description framework (RDF): Concepts and
abstract syntax. Retrieved May 20, 2011, from
http://www.w3.org /TR/rdf-concepts/.
World Wide Web Consortium. (2004b). RDF vo-
cabulary description language 1.0: RDF schema.
Retrieved May 20, 2011, from http://www.w3.org
/TR/2004/REC-rdf-schema-20040210/.
Methodologies for Augmented Discovery of Geospatial Resources

World Wide Web Consortium. (2004c). OWL web 11

http://www.w3.org/TR/rdf-syntax-gram-
ontology language reference. Retrieved May 20, mar/
2011, from http://www.w3.org /TR/2004/REC- 12
http://www.w3.org/TR/rdfa-syntax/
owl-ref-20040210/. 13
http://www.geonames.org
14
http://dbpedia.org
World Wide Web Consortium. (2008). SPARQL 15
http://www.ordnancesurvey.co.uk/ontology
query language for RDF. Retrieved May 20, 2011, 16
http://en.wikipedia.org/wiki/Triplestore
from http://www.w3.org /TR/rdf-sparql-query/ 17
http://www.docbook.org/specs/cs-docbook-
World Wide Web Consortium. (2011). Semantic docbook-4.2.html
web activity. Retrieved May 20, 2011, http://www. 18
http://www.dublincore.org/documents/
w3.org /2001/sw/. dcmi-terms/
19
http://www.geonames.org/ontology/ontol-
ogy_v2.2.1.rdf
ENDNOTES 20
http://www.w3.org/2004/02/skos/
21
http://protege.stanford.edu/
1
http://www.opengeospatial.org/ 22
http://www.eurogeoss.eu/
2
http://www.geonames.org/about.html 23
http://his.cuahsi.org/
3
http://geocommons.com/help/About 24
http://auscope-services.arrc.csiro.au/vocab-
4
http://wiki.openstreetmap.org/wiki/Main_ service/client/
Page 25
http://www.genesis-fp7.eu/
5
http://www.earthobservations.org/geoss. 26
http://www.openrdf.org
shtml 27
http://www.eionet.europa.eu/gemet
6
http://www.georss.org/Main_Page 28
http://www.eurogeoss.eu/
7
http://geojson.org/ 29
http://www.opensearch.org/Specifications/
8
http://www.openarchives.org/ore OpenSearch/Extensions/Suggestions/1.1
9
http://tools.ietf.org/html/rfc3986 30
http://www.rssboard.org
10
http://www.ietf.org/rfc/rfc4287.txt

This work was previously published in Discovery of Geospatial Resources: Methodologies, Technologies, and Emergent Ap-
plications, edited by Laura Díaz, Carlos Granell and Joaquín Huerta, pp. 172-203, copyright 2012 by Information Science
Reference (an imprint of IGI Global).

335
336

Chapter 21
Semantic Interoperability
of Geospatial Services
Iftikhar U. Sikder
Cleveland State University, USA

Santosh K. Misra
Cleveland State University, USA

ABSTRACT
This article proposes a multi-agent based framework that allows multiple data sources and models to be
semantically integrated for spatial modeling in business processing. The paper reviews the feasibility of
ontology-based spatial resource integration options to combine the core spatial reasoning with domain-
specific application models. We propose an ontology-based framework for semantic level communica-
tion of spatial objects and application models. We then introduce a multi-agent system, ontology-based
spatial information and resource integration services (OSIRIS), to semantically interoperate complex
spatial services and integrate them in a meaningful composition. The advantage of using multi-agent
collaboration in OSIRIS is that it obviates the need for end-user analysts to be able to decompose a
problem domain to subproblems or to map different models according to what they actually mean. We
also illustrate a multi-agent interaction scenario for collaborative modeling of spatial applications us-
ing the proposed custom feature of OSIRIS.

INTRODUCTION to manage spatial data, a trend often described as

Recent studies have revealed that over 80% of
local governments in the U.S. have locational
references in their data (Nedovic-Budic & Pinto,
1999) and a majority of local governments use
geographic information systems (GIS) technology
DOI: 10.4018/978-1-4666-2038-4.ch021
a “growth surge”(Warnecke, Beattie, Cheryl, &
Lyday, 1998). With the growth of Internet, there
is an increasing demand for location specific data
and analytical solutions requiring GIS to locate
and integrate multiple databases. This, in turn,
requires federal, state, and local government
agencies to develop capabilities so that their data
can interoperate. For example, a real estate entre-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Semantic Interoperability of Geospatial Services
preneur, looking for a suitable location for a new
business, would require data that combines GIS
data with that of the local government’s zoning
and tax incentive areas. Home owners and home
buyers, looking for information about environ-
mental hazards, can use e-maps that combine
data from several sources including the EPA’s
environmental data, and the HUD’s (Department
of Housing and Urban Development) housing
community programs (GAO, 2003). Similarly, a
water/sewer storm water utility company evaluat-
ing the feasibility of a new project to expand the
existing infrastructures in mountain areas may
need information about geomorphologic forma-
tions and associated potential landslide risk from
the local and federal government databases.
Agencies in various levels of government rarely
coordinate the development of their applications.
Consequently there is often redundancies and
duplications of data even within organizations
belonging to the same jurisdiction (Nyerges,
1989). Users often have to deal with proprietary
systems that require the understanding of the
systems’ native command language, input data
format, and output presentations. The problem
is further compounded when there is a need for
communicating with more than one modeling
paradigms or when spatial analysis and model-
ing techniques are used in application areas for
which they were not necessarily designed. In most
cases, users’ access, inference, and analytical
ability of spatial dataset and services are limited
by proprietary standards, platform dependence,
and incompatibility.
In an e-government environment, simple trans-
actions can require interactions among multiple
resources possibly from different entities within
the government, and meaningful understanding
of system architectures and the service composi-
tions. Interagency transactions become simple
if the agencies involved in a transaction have
homogeneous representation structures as well as
the same discourse domain (Malucelli, Palzer, &
Oliveira, 2006). A geospatial application can use
business services with relative ease if it can un-
derstand another application’s service descriptions
and representations of workflows and information
flows within and across organizations. However,
these representations become complicated when
one needs to embed complex data structures and
models into an application. For instance, suppose
we are interested in a mobile commerce applica-
tion that would provide geospatial information as
a prelude to completing a business transaction. The
transaction protocol for such an application would
require access to and representation of geographic
data and models. These models themselves may
require chaining of multiple services that depend
on service level description of geo-processing
models, spatial geometries, spatial analysis, and
implementation logic. Typical query such as “Find
the nearest Italian restaurant along the highway”
could possibly be answered by chaining multiple
services such as geocoding points of interest,
integrating transport networks, creating dynamic
segmentation of network, providing routing net-
work, rendering cartographic information, and
possibly converting text to voice. It is possible
to envision integration and chaining of services
to provide higher levels of functionality if such
services are distributed all over the enterprise
and are accessible in a uniform standard manner.
(Peng & Tsou, 2003).
Methodological artifacts, techniques for cor-
rect description, and interpretation of resources,
collectively known as the semantic layer (Vetere
& Lenzerini, 2005), are pre-requisites to high level
interoperability in a service-oriented environment.
High level or semantic interoperability is of vital
importance if collaborative business processes
are involved (Padmanabhuni, 2004). A complex
collaborative process is often needed to compose
complex services like dynamic visualization and
query processing of geo-spatial data in real time.
The representation of semantic content of space
and spatial knowledge has a special significance
in information exchange and integration. As
representation of spatial features are scale depen-
337

dent, multiple representations of spatial objects
require contextual information of the spatial
reference domain that allows spatial features to
be integrated into application models. Therefore,
typical business transactions involving spatial
data interoperability require processes beyond the
data translation or the conversion of geometric
primitives. In collaborative business processes
involving multi-agent interactions, the semantics
of the complex spatial application models needs
to be shared to infer or draw conclusions from
such annotations. However, understanding the
collaborative aspects of multi-agent interaction
to solve complex spatial problems is still in its
infancy (Sikder & Gangapadhayay, 2002).
This article proposes a framework for a
semantic-level communication between geo-
spatial services in business processes and appli-
cation models. The article presents an overview
of interoperability efforts with specific reference
to geo-spatial databases and application models
and reviews the feasibility of an ontology-based
spatial resource integration to combine the core
spatial reasoning with domain-specific applica-
tion models. Existing industry standards and
practices in geo-spatial interoperability are identi-
fied. This is followed by a discussion of the role
of ontology in explicating the implicit semantics
of spatial data models and the need for formal-
ism in descriptions of spatial categories. Use of
markup languages for spatial resource description
and the tagging of spatial ontology are illus-
trated. Finally, a multi-agent based architecture
(OSIRIS-Ontology-Based Spatial Information
and Resource Integration Services) for semantic
interoperability of spatial data sets and models is
proposed. The architecture is illustrated using an
application model that uses domain ontology of
urban environmental hydrology.
338
Semantic Interoperability of Geospatial Services
GIS SERVICES IN THE FRAMEWORK
OF E-GOVERNMENT
As stated earlier, government agencies within
the same jurisdiction rarely coordinate their ap-
plication development as well as standardization
in data representation. Making data available
electronically then becomes a problem whether the
potential user of the data is a government agency
or otherwise. A major initiative in the USA seeks
to address this problem. The National Spatial Data
Infrastructure (NSDI) seeks to build an organiza-
tional and virtual network to promote the sharing
of spatial data among federal, state, regional, and
local government agencies and the private sector.
The Federal Geographic Data Committee (FGDC)
(FGDC, 2006) is tasked to develop a spatial data
and metadata standard as well as to create data
clearinghouse. FGDC is also responsible for
coordinating the development of a “framework”
data (Wayne, 2005). Similar initiatives have been
taken by European communities for developing a
spatial data infrastructure (SDI) for e-government
(Craglla & Signoretta, 2000). As a part these initia-
tives to share data and services among public and
private agencies as well as to achieve interoper-
ability, various legal frameworks and standards
have been instituted. The E-government Act of
20021 requires federal agencies to coordinate the
development of standard protocols for sharing
geographic information to reduce redundant data
collection and promote collaboration and the use
of standards (GAO, 2003). As a result, the role
of NSDI and SDI is evolving from a simple data
access facilitator to integrated service provider
where complex geospatial services and function-
alities can be chained in a distributed environment
to meet user needs. The framework of SDI and
the role of e-government are viewed as a portal
that can be used to access e-government services.
The geospatial one-stop portal, implementing the
NSDI framework, seeks to coordinate and align
geospatial data collection and maintenance across
all levels of government. This portal envisions a
Semantic Interoperability of Geospatial Services
virtual repository of spatial data and Web services
and can provide support to local, state, and federal
programs as well as decision making at various
level of e-government operation. The portal can
be used by users to access the following services
(Cummens, 2003):
1. Searching the geospatial one-stop network
for image and feature Web services;
2. Registering for notification when new or
updated data, maps, activities, and references
are added to system repositories;
3. Viewing metadata to determine if a product
is suitable for the intended use; and
4. Publishing (registering) map services, im-
ages, geographic data sets, geoservices,
spatial solutions, geographic, and land
reference materials.
The geospatial one-stop portal can to be ap-
plied in the real world environment. For example,
in a real estate business, the portal can perform
the task of a ‘broker’ to discover and chain land
information or cadastre services from various
nodes according to the specifications set by ser-
vice requesters (Radwan, Bishir, Emara, Saleh,
& Sabrah, 2005).
The application of Web services in geospatial
service management offers an opportunity towards
the local autonomy of databases. Web services
can be used to dynamically query various GIS
layers while maintaining local agency level in-
dependence in a distributed environment. From
an organizational point of view, this may be very
appealing. Local governments, such as counties
can independently collect and manage data and
still integrate information and services using
Web services. A client of the local government
entity, for example, a transportation company,
can directly access the government’s base map
without maintaining its own dataset. With ap-
propriate permissions, a client can also update
the government’s dataset from the client’s own
record. An extended collaboration and partnership
between multiple agencies and clients using Web
services can provide opportunity to interoper-
ate through open interfaces and communication
protocols. Typical geo-processing services may
include data management tools like projection and
transformation, topology manipulation, indexing,
and spatial join.
The interoperability issue in the context of
e-government services is not limited to technical
issues such as linking computer networks. There
is a fundamental requirement to share and re-use
knowledge networks and reorganize adminis-
trative processes to better support the services
themselves. The key areas of interoperability
in need of consideration when implementing e-
government services include the organizational,
semantic, and technical issues (CEC, 2003, 2006).
At the semantic level, there is an imperative that
the meaning of the information exchanged is not
lost in the process of acquiring e-government
services from all levels of government. Thus,
semantic interoperability entails seamless integra-
tion of information and services from agencies at
different levels whether they are local, regional,
or national. This ensures organizational or local
autonomy as well as global integration.
However, there is growing skepticism whether
the vision of geospatial one-stop’s objectives will
be fulfilled (GAO, 2003) and interoperability
can be achieved. The main obstacles are lack of
standards and the requirement of a huge metadata.
While existing FGDC metadata provides content
standards, they lack a framework for semantic
annotation of the geospatial content and services.
The geospatial one-stop portal requires a semantic
layer that would contain domain level specifica-
tion of knowledge, mediator, and mapping agents
to coordinate among mapping components. The
semantic layer should be able to decompose
users’ declarative requests, compose complex
spatial services within the context of user-defined
constraints.
339

EPISTEMOLOGY OF
INTEROPERABLE AGENTS IN
BUSINESS PROCESSES
Epistemologically, the intrinsic state of an agent’s
reflection of a business process is partially char-
acterized by the internal schema (e.g., relational
or object relational) of the agent. In addition, the
agent also offers manipulation methods that can be
used to interface with other agents through specific
access protocol. Such autonomous agents can be
considered as a proxy who works collectively on
behalf of some higher entity having rich schema
or domain knowledge. For example, in a B2B
e-commerce system, a cooperative mechanism
agent may negotiate with multiple suppliers,
monitor auctions, and infer tactical business strate-
gies in a competitive environment. Agents may
not only involve in collaborative interactions to
achieve a common goal, they may also identify
best partner through negotiations (He, Jennings,
& Leung, 2003). Business applications can then
exploit these agent benefits. Specifically, a large
complex problem can be decomposed into subtasks
that can then be distributed among the agent com-
munities specializing in solving domain problems.
In a multi-agent environment, the internal state
of agent’s reflection needs to be translated to an
export schema to map conceptual correspondence.
Therefore, multi-agent communication paradigms
require semantic level agreement. Such semantic
interoperability eventually boils down to the prob-
lem of the identification of semantically similar
objects that belong to different components and
the resolution of their semantic differences (Yeap
& Handley, 1990). In this sense, the semantic
interoperability problem is somewhat similar to
the problem of schema integration in traditional
multi-databases.
Different levels of business process conceptu-
alization have given rise to many domain specific
heterogeneous systems, which are often very
difficult to interoperate in applications. In recent
years, there has been a significant level of inter-
340
Semantic Interoperability of Geospatial Services
est to study the role of an agent in system level
interoperability by offering interface with special-
ized systems that require agents to understand
the native command language, input data format,
and output presentations. In such heterogeneous
environments, agents require mapping one-to-one
objects on a syntactic basis or measuring the se-
mantic equivalence of objects. However, when the
representation is complex, for example, spatial da-
tabases, simple semantic translation is inadequate
as it can result in semantic conflicts. Three different
types of semantic conflicts have been identified
in spatial applications (Goh, 1997): confounding
conflicts, scaling conflicts, and naming conflicts.
Confounding conflicts occur when information
items have the same meaning but have different
realizations. Information items representing the
same value but with different reference systems
lead to scaling conflicts; and naming conflicts
are generated when same items have different
rules in different systems. A typical example of
spatial semantic heterogeneity can start with both
naming and confounding conflicts. For instance, a
“river” and a “body of water” have two different
realizations with respect to the scale and context
of user perceptions. A location-based query such
as “find the intersections of a river and a high-
way in Ohio,” would need a semantic translator
to identify the spatial object that corresponds to
the concepts referred by “river” and “highway”.
The spatial agent needs to be able to characterize
the semantic content of spatial data and model.
The representation of “river” or “body of water”
depends on the context of the requesting agent.
Depending on the context, the geometric repre-
sentations of “river” and “body of water” could
be also different; the resulting intersection of a
“river” and a “highway” could be a point feature
in one representation while it could be a polygon
feature in another.
An agent’s access to data layers of spatial data-
bases is dependent on the semantic interpretation
and corresponding topological definition where
spatial relationships such as adjacency relation-
Semantic Interoperability of Geospatial Services
ships can be reflected in a topological model. A
topological model of a highway can be a graph
theoretic model characterized by its “from” and
“to” nodes as well as by its left and right polygon
objects. An agent can interpret the semantics of a
textual query that includes “longest highway in
Ohio” in many ways since the agent would need
to understand the definition implicit in topologi-
cal containment of part and whole relationship.
Examples of interpretation of “in Ohio” may in-
clude or exclude highways that run along the state
border, that are wholly contained within the state,
or those crossing state boundaries. Thus, different
interpretations will result in different access and
processing requirements to different layers and
spatial joins. There is no standard or common
formal convention for the description and categori-
zation of semantic content in the query which can
be used for the global interoperability of spatial
agents. In addition, the lack of context-dependent
data views can lead to monolithic systems that
do not promote interoperability. For example, it
will be difficult for a local government or busi-
ness service provider to be able to update its land
records while serving them to other organizations
without transforming geometric and topological
semantics to make them compatible to the client’s
structure. A state highway company would have
to maintain and reflect the changes made by local
governments for its own data view to be able to
use that local government’s base map directly. At
present, there is no accepted convention, which
can be used to facilitate global interoperability
of spatial objects.
ONTOLOGY FOR SEMANTIC
INTEROPERABILITY
The objective of semantic interoperability is to
be able to attach meaning to information entities
or services and thereby draw inferences from the
semantic annotation. In spatial semantic interop-
erability, the integration goes beyond the process
of data translation or the conversion of geometric
primitives. The semantic interoperability eventu-
ally boils down to the problem of the identification
of semantically similar objects that belong to dif-
ferent databases and the resolution of their seman-
tic differences (Kashyap & Sheth, 1996). The use
of an ontology (Guarino & Giaretta, 1995) as a
framework for defining similarity among objects
has the benefit of a formal definition for concepts
in different metadata, a definition that could be
used to define axioms for semantic translation
between ontologies. The term “ontology” has its
root in the philosophical literature as the study of
being. In the domain of information systems and
AI, ontology has a somewhat different connotation
as an “explicit specification of a conceptualiza-
tion” (Gruber, 1993; Farquhar, Fikes, & Rice,
1996) and provides a more pragmatic definition:
Ontologies are explicit specifications of domain
conceptualization names and describe the entities
that may exist in that domain and relationships
among those entities. In other words, the tacit and
implicit knowledge hidden in a particular domain
is explicitly conceptualized in ontology (Guarino,
1997). Ontology is considered as a logical theory
accounting for the intended meaning of a formal
vocabulary while conceptualizations are the
formal structures of reality as perceived and or-
ganized by an agent. In spatial ontology, although
agents may have a shared vocabulary capable of
establishing relationships, or mapping between
corresponding instances, the conceptualization
of space as “object-based” and “field-based” may
be still implicit among agents.
Although spatial ontology is an established
concept, and is capable of providing a naturalistic
representation of spatial objects, in the sense of
“naive geography” (Egenhofer & Mark, 1995),
it is still a complex specification to be realized.
Spatial semantic interoperability goes beyond
simple data translation or the conversion of
geometric primitives based on a-priori ontology.
Classical ontology, offering limited expression of
spatial relations to simulate spatial processes, can
341

be used to express spatially complex phenomena,
is not well understood. Specifically, such expli-
cation requires an understanding of mereology
and mereotopology (Smith, 1996). The study of
parts and boundaries of the whole is an important
aspect of representing the multiple constituents of
a complex spatial object in relation to adjacency,
containment, selection, or the separateness of
constituent parts. Ontology in this sense needs
also to be viewed as a mediator for knowledge
exchange, to build a business application that
can enable data integration and avoid problems
such as inconsistency between ad-hoc ontologies
which might be built into the system (Fonseca &
Egenhofer, 1999). This approach to ontology in
geo-spatial service management would address
issues concerning knowledge sharing by creating
components from ontology in an object-oriented
fashion, using classical object-oriented concepts
such as multiple inheritances. Thus, spatial ontol-
ogy should allow overcoming the implicit seman-
tics of apparently disparate spatial data models by
providing a higher-level description to prepare a
cognitive correspondence among multiple com-
ponents. Such ontology is relevant not only for
spatial data models to unravel the representational
scheme of topological part-and-whole dichotomy
but also for the application models of GIS where
certain spatial operators (e.g., overlay, buffer, or
interpolation, etc.) must be executed in specific
sequences to reach a goal. Spatial application
developers should be able to combine classes
from diverse ontologies and create new classes
that represent the user needs. For example, a class
that represents a land parcel for a specific city
can be built from land parcel components speci-
fied in urban ontology, from polygons specified
in geometry ontology and from crisp boundary
specifications in boundary ontology (Fonseca &
Egenhofer, 1999).
342
Semantic Interoperability of Geospatial Services
CRAFTING ONTOLOGY-
SEMANTIC WEB INITIATIVES
While Web service description language (WSDL)
has become de facto standard for defining Web
services, its implication is yet to be understood for
geospatial services. In particular, XML encoding
of geospatial objects namely OGC’s (open GIS
consortium) geographic markup language (GML)
provides mechanisms for the encoding of spatial
feature data without regard to how the data is to
be presented. The spatial feature instance con-
tains feature properties, each as an XML element
whose name is the property name. Each of these
contains another element whose name is the type
of the property value or instance. This produces a
“layered” syntax in which properties and instances
can be interleaved. GML’s encoding convention
helps distinguish properties from instances. Thus,
a feature like settlement area might be composed
of households, a roadway, utility lines, and many
other features. Such structure-independent rep-
resentation of spatial features allows multiple
views of the same entity in a multimodal manner.
For instance, point features can assume different
symbol sets consistent with the predefined design
principle of the organizations. XML namespace
prefixes for the domain specific vocabularies,
for example, “urban hydrology” or “watershed
planning,” can be used to define feature types
from common geometric vocabulary. Since the
namespace defines a unique identifier, there is
little concern for naming conflicts. For instance,
the underground cable line and road network
databases are maintained by two different organi-
zations. In response to the user query to identify
the locations where the cable lines intersect with
highways, the two data sets need to be spatially
overlaid. However, the assigned meanings of
XML schema must still be inferred by a human.
A partial solution to these semantic issues can
be done using semantic markup of spatial Web ser-
vices and by using an expressive semantic markup
language that offers well-defined semantics of
Semantic Interoperability of Geospatial Services
spatial processing. A possible solution can be
found through the use of RDF (resource descrip-
tion framework) (Lassila & Swick, 2004). RDF
provides a metadata description to allow for the
creation of shareable and extensible vocabular-
ies. The so-called self-expressibility of RDF can
be achieved by using a triple (resource, property,
value). It provides a globally unique URI to de-
scribe a resource in a Web page in a serialized and
machine-readable format. Built on the top of RDF,
DAML (DARPA agent markup language) offers
semantics of higher order relationships by includ-
ing modeling constructs from descriptive logic.
DAML overcomes many of the shortcomings of
RDF semantics and is widely used for formulating
Web ontology language. Thus, users’ declarative
requests can be decomposed to initiate generic
procedures to compose complex spatial services
within the context of user-defined constraints. The
implication of these constructs in spatial databases
is likely to have a fundamental shift in interoper-
ability research, allowing distributed access to
multiple stakeholders’ data sets and integrating
them in a seamless Web interface. However, lack
of inference and logical layer of RDF makes it
difficult to express complex spatial semantics.
ONTOLOGY-BASED SPATIAL
RESOURCES INTEGRATION
SERVICES (OSIRIS)
So far, the feasibility of an ontology-based spatial
resource integration that combines the core spatial
reasoning with spatial data models and domain spe-
cific application models has been discussed. It has
been noted that the application of ontology-based
semantic mediation can lead to model integration.
In this section, the proposed OSIRIS architecture
that involves different levels of ontology to create
meta-models, especially for spatial services, has
been described. In addition, a specialized prototype
for decision making in spatial planning in a multi-
agent environment has been demonstrated. Later,
the OSIRIS framework will be used to illustrate
the collaborative aspect of multi-agent interaction
to solve complex spatial problems. The advantage
of using multi-agent collaboration in OSIRIS is
that it obviates the need for end-user analysts to
be able to decompose a problem to subproblems
or to map different models according to what they
actually mean. The proposed approach provides
a means for communicating different models in a
common semantic framework without the loss of
native autonomy. Having a comprehensive source
description of different modeling representation
has a unique advantage in understanding how the
models are organized in terms of interconnection
of schema and scope.
MULTI-AGENT AGENT
ARCHITECTURE OF OSIRIS
The OSIRIS architecture (Figure 1) is composed
of several integrated components. The model
manager is in charge of providing the overall
coordination of the agents by acting as a media-
tor for instantiating an appropriate model through
the ontology agent. It also maintains a catalog of
models and ontological semantics with the help
of ontology agents. The model manager exposes
the model schema to the ontology agent to be
used in response to specific user request. Using
the encapsulated description of the model from
the ontology agent, the model manager creates
appropriate instances. The integration of data and
corresponding models is coordinated by the model
manager which ensures a proper correspondence
through the ontology agent and resolves any dis-
agreements between data and models. Through the
interaction of ontology agent and user interface
agent, the model manager composes services or
custom models from discrete services to produce
a meaningful interconnection among them.
OSIRIS architecture incorporates spatial and
non-spatial databases as well as a collection of
application models (GIS resources). A wrapper
343

Figure 1. Multi-agent OSIRIS architecture
is used to encapsulate application models as well
as the database. The wrapper provides a compre-
hensive description of model parameters, scope,
input/output options, limitations, context, method
of use, and other necessary semantics. Encapsula-
tion agents encapsulate models and data to create
meta-models and created interfaces necessary for
the wrapper object. Meta-models may include
legacy models, domain models, generic proce-
dures, or high-level schema of interconnected
data. Encapsulation agents process queries origi-
nating from ontology agents and retrieve informa-
tion back to the system. In addition, these agents
translate query to the local query language, process
the query in a local system, and then forward
344
Semantic Interoperability of Geospatial Services
query results to the ontology agent. These descrip-
tions could be encoded in the extended RDF or
OWL-enabled ontology. Such descriptions pro-
mote consistency, exchangeability, and reuse of
modules through the shared repositories in a
distributed environment. Above the communica-
tion layer of the OSIRIS are two components
which ensure the consistent construction of the
agent platform. The internal platform message
transport (IPMT) provides a message routing
service for agents in this particular platform. The
IPMT can use either OMG’s IIOP (Internet inter-
ORB protocol) or HTTP protocol. The agent
platform security manager (APSM) is responsible
Semantic Interoperability of Geospatial Services
for managing the agent’s transport level security
by creating an audit log.
Models and Databases
The OSIRIS framework relies on a repository of
models and databases. The choice of models and
data set would depend on the type of system be-
ing considered in a particular case. For example,
a system for studying land slide hazard may
need composite of several application models,
algorithms including geoprocessing algorithms,
discrete event simulation models, and non-linear
dynamic models for estimating rainfall runoff.
In an e-government system, the set of models
and algorithms may include land use policies,
zoning regulations, and economic development
priorities. Encapsulation agent is responsible to
provide appropriate wrapper around such models
and make them available to an ontology agent for
semantic annotation and thereby help create a meta
model ontology. The creation of wrapper involves
interaction with the ontology repository through
the ontology agent to determine if an existing
ontological specification is adequate; otherwise
a new specification and taxonomy is created by
the encapsulation agent and the specification is
deposited in the ontology repository.
OSIRIS Model Manager
The model manager provides the overall coordi-
nation of the agents by acting as a mediator for
instantiating an appropriate model corresponding
to the meta-model ontology. It also maintains a
catalog of models. The export model schema to
be used in response to specific user request is ex-
posed to model manager by an ontology agent. The
integration of data and corresponding models is
coordinated by the model manager which ensures a
proper correspondence through the ontology agent
and resolves any disagreements between data and
models. The model manager composes services or
custom models from discrete services to produce
a meaningful interconnection among them.
A model manager can be divided into two-core
components: a) custom model manager and b)
generic model manager. A custom model manager
is designed to create and manage modules dynami-
cally from core ontology and extended domain
ontology while a generic model manager is used
to integrate legacy models through encapsulated
interfaces by maintaining reusable libraries of
models. Generic model managers offer services
to custom model managers as encapsulated meth-
ods and services of a particular class module and
embed them in the model hierarchy.
Ontology Agent
An ontology agent establishes correspondences
and interrelations among different domain knowl-
edge and data models and relations. It defines a
common semantic framework and formalism to
share data and models and identifies a set of model-
ing approaches and corresponding data sources to
resolve semantic ambiguities. An ontology agent
is responsible for ensuring the class consistency
and implicit relationships of application models
and data. It exposes higher level model schema of
legacy models and native data structure to model
manager and help generate meta-model catalog.
An ontology agent provides standardized
descriptions of model constructs in conformity
with ontological specifications. Its responsibility
also includes ontology deployment, validation of
specification leading to creation of new wrapper
objects in the database, and updating the ontology
repository. For example, consider an e-government
system that maintains databases regarding water/
sewer storm water utility. A network infrastructure
provider company may need access to the existing
water sewer database. This client’s request can be
realized by dynamically populating a template
schema of water utilities data model.
The ontology agent populates the template
schema from existing databases for which it
345

receives the necessary request from the model
manager or the user interface agent (UIA) of the
OSIRIS. The template schema can be created if
a user ontology matching the specific client’s
request exists. In case such a match is not avail-
able, the ontology agent can create a user ontology
through an interactive process between itself, the
client, and the UIA. The ontology agent invokes
the model manager to service the client request.
In the case of complex requests, the ontology
agent invokes the services of collaborative agent
to resolve ambiguities or semantic disagreement
between domain ontologies and user ontologies.
Embedded within the ontology agent is a reason-
ing support engine. At the implementation level, it
may include a reasoning engine like FaCT (Hor-
rocks, 1998) system that can translate ontology
into SHIQ (D) DL to indicate inconsistencies and
implicit subsumptions. In addition, axioms could
be added to make implicit subsumptions explicit.
Ontology Repository
The ontology repository consists of i) core ontol-
ogy, ii) domain ontology, iii) user ontology, and
iv) Web procedure/services ontology. In general,
these repositories contain a set of axioms (e.g.,
assertion of class subsumptions/equivalence),
and taxonomic class hierarchies. All of these on-
tologies are strongly related to one another. The
core ontology contains immutable and generic
concepts and top-level categorical abstractions of
physical relationships and corresponding semantic
descriptions and axioms. For example, in the case
of complex spatial data required for representing
water and sewer network, it will be necessary to
use core ontology that contain high-level defini-
tions of spatial entities and geometric primitives.
Domain ontologies are generally associated with
specific application domains such as geographic
information metadata-ISO 19115 (Islam, Bermu-
dez, Beran, Fellah, & Piasecki, 2006) and infra-
structure product ontology (Osman & El-Diraby,
2006). A domain model is incorporated into the
346
Semantic Interoperability of Geospatial Services
domain ontology following the semantic structure
of the core ontology.
In OSIRIS, core ontology is related to the
domain ontology to specify appropriate semantic
characterization of the domain model. An example
would be the relationship between the classes of
core ontology and the classes of domain ontology
that can be expressed using an user-defined seman-
tics such as: equivalentClassOf (DomainClass,
CoreClass) or aggregateClassOf (DomainClass,
CoreClass). In specific cases, core ontology may
include reusable UpperCyc (OpenCyc, 2006)
ontology which covers most of the top concepts
applicable to the domain of e-government. The
class relationships allow generating ontology in a
hierarchy with the child class having explicit at-
tributes to specialize the parent classes. Thus, the
orthogonal property of ontology can be maintained
by decomposing compound concepts into constitu-
ent parts. An ontology agent relies on Web services
and generic procedures to discover new services
and modules. Web procedure/services ontology
manages this function by defining the levels of
abstractions of external services and fitting them
with the repository structure. At an implementation
level, Web services can be expressed with a Web
service description language (WSDL).
Ontology Collaboration Agent (OCA)
OCA is designed to assert inter-ontology relation-
ships within ontology repository and compute
class hierarchy/consistency. It merges multiple
ontologies by mapping domain classes with core
ontology and provides semantic agreement of user
ontology with domain ontology. For example, a
road network classification by water/sewer net-
work authorities may not match with a specific
domain of a user’s ontologies. In such cases, the
collaboration agent first tries to match the dispa-
rate schema by mapping individual road network
classes with classes from core ontology. If the
mapping does not result in consistent well-formed
class hierarchy, the collaboration agent may pro-
Semantic Interoperability of Geospatial Services
ceed to modify user ontology through mediation
of user interface agent and ontology agent. It also
integrates specialized Web procedure services
ontology (WPSO) and provides Web resources for
custom model formulation. WPSO also includes
a set of triggers that gets activated in the event
of any change of Web-service composition or
changes in domain ontology. OCA also reacts to
these changes and carries out necessary updates
to user ontology.
User Interface Agent (UIA)
This component of OSIRIS is the primary inter-
mediary between the system user and OSIRIS.
It resolves ambiguities in the user expression in
formulating structured requests through an interac-
tive session using the user profile. The ontology
agent collaborates with the UIA to develop model-
ing profiles of the user following user needs and
strategies. Some of the tasks achieved through the
interaction between the ontology agent and the
UIA include the user modeling (Cali, Calvanese,
Colucci, Noia, & Donini, 2004; Rich, 1989), iden-
tification of trends in user requests, and extreme
or unusual requests. Embedded within the UIA
is a query formulation engine that has access to
native data semantics of user profiles via ontol-
ogy agent. Once the UIA completes the query
formulation, the resultant query is parsed by the
model manager for the identification of specific
model or service as requested by the UIA, and
finally validated by the user. The user ontology
in the ontology repository is updated through the
interaction among UIA, ontology agent, and the
collaborating agent. In the absence of the requested
service in the OSIRIS’meta-model catalog, custom
model is composed dynamically and validated
by an interactive session involving the UIA and
multiple agents in OSIRIS. The UIA also includes
a set of triggers to respond to events such as
changes of user ontology in ontology repository,
and meta-model catalog updates. As a practical
implementation issue, at the end of an interac-
tive session, the UIA generates and updates log
files of frequently requested services and creates
corresponding index structure to facilitate faster
retrieval of future request.
Agent Platform Security Manager
The agent platform security manager verifies
inputs from ontology agents and enforces compli-
ance with underlying system security protocol.
The nature of the compliance may differ based
on the application domain. For example, the se-
curity standards specified by FIPA (FIPA, 2006)
would be enforced by the security manager in an
application involving landslide hazards.
Internet Platform Message Transport
The Internet platform message transport is de-
signed to transfer a message from one system to
another. Such a transfer may involve technologies
such as Web services or CORBA. The intention of
the OSIRIS model is to be a high level protocol that
interacts with the underlying system’s network.
Multi-Agent Collaboration:
an Event Flow Scenario
In this section, two different scenarios for process-
ing user requests are illustrated: composition of
services using generic models and custom mod-
els. A generic model can be created by means of
searching the model catalog (see the description of
model manager). Model manager maintains reus-
able libraries of geo-processing models that can be
integrated with legacy models through the wrapper
objects in the database. For example, in order to
extend an existing water utility infrastructure, the
local government agency would need to identify
the optimum location and network of services
matching those of its customers’ locations. A user
request for identifying location for the proposed
water utility infrastructure would result in a service
request to OSIRIS. The resolution of this request
347

Figure 2. Message sequence of agent interactions for composing generic procedures
may consist of searching geocoded customer
location (geocoding combines map information
with street addresses to locate a point uniquely)
and combining the search results with the gravity
model (Haynes & Fotheringham, 1984) for loca-
tion allocation. The OSIRIS can accomplish this
task by initiating a series of message transports
through inter-agent communication. Figure 2
illustrates a scenario where a generic model is
formulated from the user’s declarative request.
A request for geocoding customer location is
initiated by a series of message transports and the
creation of the request object. While the model
manager (MM) provides access to meta-model
of geo-processing catalogue, the ontology agent
(OA) and the ontology collaboration agent (OCA)
provide semantic support to resolve semantic mis-
match between user request and domain ontology.
Additionally, the data encapsulation agent (DEA)
interprets the schema of existing spatial data
and corresponding meta-data and populates the
instances in the parameters of processing model.
348
Semantic Interoperability of Geospatial Services
Unlike the generic model formulation, custom
models are composed dynamically from core
ontology and extended domain ontology. The
composition process involves logical inferences
from core ontology to generate generic procedures
from domain ontology and WPSO (Figure 3). The
derivation of the logical inference is iterative and
requires user feedback for semantic refinement
of the model design. In the next section, an ex-
ample of custom model composition from axi-
omatic proposition in core ontology is described.
Composing Custom Models by
Collaborative Modeling: Example of
Logical Inference
The interaction of multi-agents in collaborative
environments can lead to the solution of complex
problems not otherwise amenable to analysis in
a stand-alone system [43]. The custom model
manager feature of OSIRIS exploits the interaction
of multi-agents in a collaborative environment in
modeling spatial applications. A simplified ap-
Semantic Interoperability of Geospatial Services
Figure 3. Composing custom procedure through logical inference in OSIRIS agent interactions
plication scenario of OSIRIS for custom model
management can be visualized in a spatial plan-
ning problem. Consider, for instance, an insurance
company is interested in finding out the regions
that are prone to landslides in the proposed ex-
tended water/sewer utility network. This problem
can be addressed in two different ways. First, the
domain expert model for the landslide hazard
zone modeling (a domain model developed by
geomorphologists) can be identified by OSIRIS
model manager as the candidate model to ad-
dress the issue of landslide hazard. The model
parameters, requirements, constraints, and schema
can be described using description logics. This
description would constitute a meta-model of the
actual landslide model that eventually becomes
the domain ontology for this particular problem.
The model may require different input classes,
for example, 3D digital elevation model (DEM),
soil map, moisture index map, and satellite im-
ages. Once the processing of the available input
is completed, output such as landslide risks and
zonation can be produced.
Suppose the model server does not have ac-
cess to the domain model for the landslide hazard
zonation model. This scenario would invoke the
custom modeling abilities of OSIRIS which then
proceeds to answer the landslide query indirectly
in some less rigorous way by using core ontology
and user templates. Let us assume that the user
input is only a DEM class. The system should be
able to reason that the DEM class can be used to
create a slope map class that would eventually
identify the locations of highly steep areas that
are likely to be landslide-prone. Creating a slope
map from the DEM is a standard and simple GIS
procedure and the system should have no problem
converting the map to a slope map using standard
functions from domain ontology. The core ontol-
ogy should be able to infer that higher the slope,
the greater is the probability of landslide and
specifies to the model manager the requirements
349

for generating the slope map. The common sense
understandings are encoded and inferred from the
knowledge of core ontology agent. This knowledge
includes generic axioms and contexts such as rocks
fall downward under gravity; rocks move faster
as the slope increases; and rocks do not move
in flat plains. The model can progressively lead
to more sophistication with enhanced inference
and reasoning. Since soil types are associated
with landslide phenomena, the system may infer
that it needs to create a soil class in addition to
DEM and therefore, would begin searching for
soil data before it actually processes the query.
While developing the “custom” model, the system
might “consult” the domain ontology and the user
ontology to gain further “insight” to solve similar
complex problem.
In the following example, consider the fol-
lowing 4 axioms: a plot is sameClassAs a parcel;
landslide is the result of (or subClassOf) steep
slope and high moisture content; and the slope
can be derived from DEM. Given a set of well-
formed axioms, the semantics can be defined via
a translation into an expressive description logic
(DL) which is equivalent to SHIQ (D) DL (Ian
Horrocks, Patel-Schneider, & Harmelen, 2003).
Plot Β Parcel
Landslide ⊆ steepSlope ∩ highMois-
ture
Slope ⊆ ∃hasSlopefunction.DEM
Using multiple equality/inclusion axioms, one
can derive following assertion.
steepPlot Β Plot ∩ hasHighval.Slope
moistPlot Β Plot ∩ hasHighval.Mois-
ture
Slope ⊆ ∃hasSlopefunction.DEM
Transitive role of classes or causality can be
captured as:
350
Semantic Interoperability of Geospatial Services
HeavyRain ⊆ ∃causes.Moisture
Moisture ⊆ ∃causes.Landslide
⇒ HeavyRain ⊆ ∃causes.Landslide
Inverse roles for example, causes/causedBy
relation provides similar to the power of adduc-
tive reasoning.
Moisture ∩ ∃causedBy.HeavyRain ⊆
Landslide
⇒ HeavyRain ⊆ ∃causes.Landslide
The inference also provides cardinality restric-
tions and adds consistency constraints.
RiskIndex ⊆ ∃hasValue.(High ∩ Low) ∧
≤1hasValue)
High ⊆ ¬ Low ⇒ HighLowRiskindex ⊆ ⊥
The inference allows one to express risk
score of landslide of a given area from the core
ontology. It should be noted that causal inference
can be derived from the core as well as domain
ontology of OSIRIS.
Semantic Matching of Spatial
Features in OSIRIS
The relationship between domain ontology and
core ontology in OSIRIS corresponds to the
relationship between the concepts of specializa-
tion and generalization. Just as the RDF schemas
allow the creation of subtypes using multiple
inheritance properties like subClassOf or sub-
PropertyOf, the domain classes can be derived
from core classes or ontology. In a spatial context,
by using GML’s core schema, one can develop
fundamental building blocks for features and fea-
ture geometry and define application or domain
schemas following RDF/DAML constraints. For
example, new feature types or collection types
can be defined from gml:AbstractFeatureType or
gml:AbstractFeatureCollection which could be
real world categories such as “roads” or “bridges”.
Semantic Interoperability of Geospatial Services
A user-defined complex geometric feature can
consist of many other geometric elements. Such
schema can be enriched by DAML+OIL con-
straints. The DAML+OIL enabled spatial agent
is capable of new classifications by inference and
reasoning. Using GML’s spatial primitives within
RDF, a domain specific application schema can
be expressed to describe the different associative,
causal, and functional relationships of the domain.
Suppose the water/sewer authority in an
e-government system defines a namespace
<myDef:River> using GML’s namespace and
type hierarchy.
< -- xmlns:myDef=”http://www.xyzcorp/
example.rdf -->
<myDef:River>
<gml:description> T1</
gml:description>
<myDef:classification>water body</
myDef:classification>
<gml:CenterLineOf>
<gml:LineString srsName=”GFD:987”/>
<gml:coordinate>0.00, 1000.0 1000.0,
0.0</gml:coordinate>
</gml:LineString>
</gml:CenterLineOf>
</myDef:River>
It is possible that there exists other definitions
for the river class, say <usgs:River>, defied by
some other agency such as USGS. The use of
a namespace prefix <myDef:River> allows the
user to specify the agent’s domain or to scale
specific vocabularies which essentially differenti-
ate other namespace of similar categories such as
<usgs:River>. DAML allows constructs such as
daml:equivalentTo or daml:samePropertyAs that
can be used to inherit specific properties from
existing type definitions and associated descrip-
tions. Thus, it is possible to have two distinct
spatial objects, having two different application
schemas, sharing the common geometric primitive
of GML and the semantics of the two data sets
from two different agencies. At the implementa-
tion level, this approach can be used to identify
spatial features and objects in different Web pages
and to initiate spatial reasoning among disparate
data sources. The serialized and machine-readable
semantic descriptions, as supported by DAML,
GML, and other, provide a means for ontology
agents to exchange spatial models and schema
on the Web.
Domain Ontology for Spatial
Application: An Example
Using features of the OSIRIS framework, do-
main ontology for urban environmental hydrol-
ogy tightly coupled with a Web-based “what-if”
scenario generator have bben developed. The
ontology expresses the widely used specifica-
tion of a “non-point pollution model” or what is
commonly known as “simple method” [41] for
estimating surface runoff or pollutant load. The
first step in markup of the domain components
is to describe each constituent component of
the pollution model. It consists of a declarative
description of the model’s properties. The objec-
tive is to model pollutant loads originating from
different changes of land use category. Depending
on the land use category, the planning authority
classifies a certain number of zones. Zones are
spatial as well as administrative features managed
by an environmental agency in accordance with
the environmental regulations. The definition of
the organization (in this case-environment and
conservation-and-wildlife-organizations) relevant
to zonal planning is derived from another top level
ontology namespace to ensure the semantic cor-
respondence. Land use classes have essentially
spatial features. A specific land use category
can result in several “pollutant” types which are
expressed as enumeration types—daml:one of
element as a closed lists of the daml:collection
parse type. Since the ontology is extensible, any
new ontology component can be added to describe
wider algorithm of the domain. An excerpt from
351
 Semantic Interoperability of Geospatial Services

an example of an urban hydrology developed <rdfs:subClassOf>

using DAML is shown. <daml:Restriction>
<daml:onProperty
<rdf:RDF rdf:resource=”#zoneID”/>
xmlns:rdf =”http://www. <daml:toClass
w3.org/1999/02/22-rdf-syntax-ns#” rdf:resource=”http://www.
xmlns:daml=”http://www.daml. w3.org/2000/10/XMLSchema#string”/>
org/2001/03/daml+oil#” <daml:maxcardinality>1</
xmlns:rdfs=”http://www.w3.org/ daml:maxcardinality>
TR/1999/PR-rdf-schema-19990303#”> </daml:Restriction>
xmlns=”http://cis.csuohio.edu/daml/ </rdfs:subClassOf>
ontologies/Non-point-pollution#” <rdfs:subclassOf>
> <daml:Restriction>
<daml:Ontology rdf:about=”http:// <daml:onProperty rdf:resource=”#e
opengis.net/gml/gml.rdfs#”> nvironmentalRegulation”/>
<daml:versionInfo>$1.0</ <daml:toCalss rdf:resource=”#Mana
daml:versionInfo> gementAgency”/>
<rdfs:comment>An ontology for </daml:Restriction>
Non-point Pollution modelling</ </rdfs:subClassOf>
rdfs:comment>
<daml:imports <daml:DataProperty rdf:ID=”environme
rdf:resource=”http://www.daml. ntalRegulation”>
org/2001/03/daml+oil#”/> <daml:DataProperty rdf:ID=”#zoneID”>
<daml:imports
rdf:resource=”http://opengis.net/gml/ <rdfs:Class rdf:ID=”LandUse”>
gml.rdfs#”/> <rdfs:subClassOf
rdf:resource=”Zone”>
</daml:Ontology> <rdfs:subClassOf>
<daml:Class <daml:Restriction>
rdf:ID=”ManagementAgency”> <rdfs:domain
<rdfs:subClassOf rdf:resource=”#LandUse”/>
rdf:resource=”http://www.daml. <daml:onProperty
org/cgi-bin/hyperdaml?http:// rdf:resource=”#name”>
opencyc.sourceforge.net/daml/ <daml:toClass rdf:resource=”http://
naics#Environment,-Conservation-and- www.w3.org/2000/10/
Wildlife-Organizations;813312”/> XMLSchema#string”/>
<rdfs:label>Urban Land Land manage- <daml:maxcardinality>1</
ment Agency</rdfs:label> daml:maxcardinality>
</daml:Class> </rdfs:subClassOf>
</daml:Restriction>
<rdfs:Class rdf:ID=”Zone”> <rdfs:subClassOf>
<rdfs:subClassOf <daml:Restriction>
rdf:resource=”http://opengis.net/gml/ <daml:onProperty
gml.rdfs#Polygon”/> rdf:resource=”#pollutant”/>

352
Semantic Interoperability of Geospatial Services

<daml:toClass to illustrate inheritance from a generic class. Us-

rdf:resource=”#Pollutant”/> ing the set of axioms provided by DAML+OIL,
<daml:Maxcardinality>1</ one can assert class subsumption, equivalence of
daml:Maxcardinality> class or property, and various constructors such
</daml:Restriction> as intersectionOf, and Maxcardinality to create
</rdfs:subClassOf> complex semantic constraints. An intelligent
agent with an inference engine can easily deduce
<daml:DatatypeProperty new knowledge about the environment from the
rdf:ID=”LandUseID”> ontology.
<daml:toClass rdf:resource=”http://
www.w3.org/2000/10/ Explorative Land Use Change
XMLSchema#string”/> Analysis: Simulations of
<rdfs:range rdf:resource=”http:// Multi-Agent Interaction
www.w3.org/2000/10/
XMLSchema#nonNegativeInteger”/> Given ontology, multiple agents can interact
<rdf:type rdf:resource=”http:// meaningfully. The following example illustrates
www.w3.org/2001/10/ an agent’s understanding of the domain model
daml+oil#UniqueProperty”/> that can be used to simulate different disicion
</daml:DatatypeProperty> scenarios visualized by map rendering. A “what-
if” type of pollution simulation model has been
<rdfs:Class rdf:ID=”Pollutant”> developed in response to the decision of changing
<daml:unionOf rdf:parseType=”daml:c land use from one category to another category
ollection”> (see Figure 4). This model can be used to simulate
<Pollutant rdf:ID=”BOD”/> the consequence of an agent’s decision, when for
<Pollutant example, an agent changes a certain category of
rdf:ID=”TotalNitrogen”/> land use from “vacant” to “landfill.” The result is
<Pollutant a corresponding increase or decrease in pollution
rdf:ID=”TotalPhosphorous”/> content. Every request to change in land use type
</daml:unionOf> results in recalculation of the mass export of pol-
</rdfs:Class> lutant and corresponding statistics. The resulting
pollution map can be visualized with multiple
<daml:DatatypeProperty theme overlay. The system logs individual user’s
rdf:ID=”pollutant”/> preferences to input into mediating algorithm to
resolve conflict among user preferences of land
<…..X..……> use choice. Built on top of the ESRI’s ArcIMS,
the system uses ArcXML (native XML encod-
The full XML serialization is not shown ing of spatial object) to communicate between
here due to space limitations. In the example, the custom middleware and the Web mapping
the daml:collection represents a DAML+OIL server. The services offered by the Web mapping
extension of RDF to provide a “shorthand” server are similar to the requirements of OGC’s
representation of structured list that defines implementation specification for a Web map
triples. It should be noted that the daml:Class service (WMS).
rdf:ID=”ManagementAgency” is derived from The map server contains registered model
an external URI of Opencyc (OpenCyc, 2006) components which are instantiated following a

353

Figure 4. Implementing map service for simulation of land use change
request from an agent. The server side application
processes the agent’s request and makes necessary
updates in the database to reflect the correspond-
ing changes of the pollutant coefficients. Every
request to change in land use type results in re-
calculation of the mass export of pollutants and
corresponding statistics. The processed result is
then sent back to the Web server and then to the
client agents. For a given planning zone, the en-
vironmental regulation of land use constraints is
stipulated by the management agency. The domain
ontology includes several such constraints:
Zone (low_density_residential) →Per
mittedLandUseCategory(multifamily_
dwelling)
∀x,y∃d HighDensityResZone(x)^Preserva
tionZone(y)→MinDistApart((x,y),d)
Similarly, other spatial contiguity constraints or
topologic constraints can be expressed in domain
ontology in relation to the neighboring zone or
land use. The system provides cartographic ren-
dering options for typical mapping manipulation
procedures such as selecting and identifying a
354
Semantic Interoperability of Geospatial Services
feature, visual queries, rendering legends cor-
responding to feature types (classification for
both continuous and unique data type), multiple
themes, and overlays. The solution space of the
explorative scenario generated by the simulation
process can be further resolved using different a
mediation algorithm in an evolutionary process.
For instance, genetic algorithms can be used to
construct links between an agent’s template and a
possible solution space and let the system evolve
until a mutually acceptable solution converges.
CONCLUSION
In this article, how the use of semantic reason-
ing can be used to aggregate and reason over
geographic features taken from multiple sources
has been demonstrated. It has been noted that a
semantic layer is essential to fulfill the objec-
tives of e-government’s geospatial portal. To the
best of knowledge, there is no current standard
specification for re-use of spatial models which
could enable high level effective communication
among different agents. There is still a need for a
Semantic Interoperability of Geospatial Services
generic formalism to link geo-spatial models to
the domain specific application classes. Using
a multi-agent paradigm, a promising direction
to solve complex spatial problems that would
be otherwise difficult to solve has been demon-
strated. The OSIRIS framework holds that the
description of the infrastructure of spatial data is
essential for ontology-assisted interoperability of
heterogeneous sources. It has been demonstrated
that ontology can be used to provide a common
context for the semantic grounding for spatial
application models which agents can relate to
their native terminologies and thereby enable ac-
cess to multiple services. Spatial services can be
composed or manipulated along with other type
of services, including Web services. The imple-
mentation of the sample prototype of OSIRIS
framework indicates that a scaleable Web based
spatial ontology can be developed using ontology
modeling language to enable real world spatial
decision-making. However, further research is
needed to understand the requirements of ontol-
ogy language to represent the semantic structure
for complex spatial and temporal objects. This is
especially important for ontology matching and
merging of complex spatial systems. In particu-
lar, the implication of imprecision related to the
finite resolution observation of an agent is not
well addressed in current ontology research. A
robust specification for expressing topology and
mereological aspect of spatial reasoning in a se-
mantic language is still needed. Further research
is necessary to identify the mechanism by which
spatial representation and different level of explic-
itness at multiple scales affects an agent’s logical
reasoning and comprehension of spatial processes.
REFERENCES
Cali, A., Calvanese, D., Colucci, S., Noia, T. D.,
& Donini, F. M. (2004). A description logic based
approach for matching user profiles. Unpublished
manuscript.
CEC. (2003). Linking up Europe: the importance of
interoperability for e-government services- com-
mission of the European communities (Commis-
sion Staff Working Paper, No. 801). Commission
of the European Communities & Interchange of
Data between Administrations (IDA).
CEC. (2006). European interoperability frame-
work for pan-European e-government services-
Communication from the commission to the
council and the European parliament. Brussels:
Commission of the European Communities.
Craglla, M., & Signoretta, P. (2000). From global
to local: the development of local geographic
information strategies in the United Kingdom.
Environment and Planning, B(27), 777–788.
Cummens, P. (2003). Geospatial one-stop portal is
key to President’s e-government strategy. ArcNews
Online, Summer.
Egenhofer, M., & Mark, D. (1995). Naive geog-
raphy. Paper presented at the International Con-
ference COSIT ‘95. Lecture Notes in Computer
Science, Berlin.
Farquhar, A., Fikes, R., & Rice, J. (1996). The on-
tolingua server: a tool for collaborative ontology
construction. Stanford, CA: Knowledge Systems
Laboratory-Stanford University.
FGDC. (2006). Geospatial metadata standards.
Retrieved September 2006, from http://www.
fgdc.gov/metadata/geospatial-metadata-standards
FIPA. (2006). The foundation for intelligent
physical agents. Retrieved May, 2006, from http://
www.fipa.org/
Fonseca, F., & Egenhofer, M. (1999). Ontology-
driven geographic information systems. In C. B.
Medeiros (Ed.), 7th ACM Symposium on Advances
in Geographic Information Systems (pp. 14-19).
Kansas City: ACM Press.
355

GAO. (2003). Geographic information systems:
Challenges to effective data sharing (Testimony
before the subcommittee on technology, informa-
tion policy, intergovernmental relations and the
census, committee on government reform. House
of Representatives No. GAO-03-874T). United
States General Accounting Office.
Goh, C. H. (1997). Representing and reason-
ing about semantic conflicts in heterogeneous
information sources. Boston: Sloan School of
Management, MIT.
Gruber, T. R. (1993). A translation approach to
portable ontology specifications. Knowledge
Acquisition, 5(2). doi:10.1006/knac.1993.1008
Guarino, N. (1997). Semantic matching: Formal
ontological distinctions for information organiza-
tion, extraction, and integration. In M. Pazienza
(Ed.), Information extraction: A multidisciplinary
approach to an emerging information technology
(pp. 139-170). Frascati, Italy: International Sum-
mer School.
Guarino, N., & Giaretta, P. (1995). Ontologies
and knowledge bases: Towards a terminologi-
cal clarification. In N. Mars (Ed.), Towards very
large knowledge bases: Knowledge building and
knowledge sharing (pp. 25-32).
Haynes, K. A., & Fotheringham, A. S. (1984).
Gravity and spatial interaction models. Beverly
Hills, California: Sage Publications.
He, M., Jennings, N. R., & Leung, H. F. (2003).
On agent-mediated electronic commerce.
IEEE Transactions on Knowledge and Data
Engineering, 15(4), 985–1003. doi:10.1109/
TKDE.2003.1209014
Horrocks, I. (1998, May 5-- 8). The FaCT sys-
tem. Paper presented at the TABLEAUX ‘98, In
Automated Reasoning with Analytic Tableaux
and Related Method, International Conference
Proceedings, Oisterwijk, The Netherlands.
356
Semantic Interoperability of Geospatial Services
Horrocks, I., Patel-Schneider, P. F., & Harmelen,
F. v. (2003). From SHIQ and RDF to OWL: The
making of a Web ontology language. Journal of
Web Semantics, 1(1), 7–26. doi:10.1016/j.web-
sem.2003.07.001
Islam, A. S., Bermudez, L., Beran, B., Fellah, S.,
& Piasecki, M. (2006). Ontology for geographic
information—metadata (ISO 19115:2003). Re-
trieved May, 2006, from http://loki.cae.drexel.
edu/~wbs/ontology/iso-19115.htm
Kashyap, V., & Sheth, A. (1996). Semantic het-
erogeneity in global information system: The
role of metadata, context and ontologies. In M.
Papazoglou & G. Schlageter (Eds.), Cooperative
information systems: Current trends and directions
(pp. 139-178). London: Academic Press.
Lassila, O., & Swick, R. (2004). Resource de-
scription framework (RDF) model and syntax
specification. Retrieved from http://www.w3.org/
TR/REC-rdf-syntax/
Malucelli, A., Palzer, D., & Oliveira, E. (2006).
Ontology-based services to help solving the het-
erogeneity problem in e-commerce negotiations.
Electronic Commerce Research and Applications,
5, 29–43. doi:10.1016/j.elerap.2005.08.002
Nedovic-Budic, Z., & Pinto, J. K. (1999). Interor-
ganizational GIS: Issues and prospects. The Annals
of Regional Science, 33, 183–195. doi:10.1007/
s001680050100
Nyerges, T. (1989). Information integration for
multipurpose land information systems. URISA,
1, 27–38.
OpenCyc. (2006). OpenCyc 1.0. Retrieved May,
2006, from http://www.cyc.com/cyc/opencyc/
Osman, H., & El-Diraby, T. E. (2006, June 14-16).
Interoperable decision support model for routing
buried urban infrastructure. Paper presented at
the Joint International Conference on Comput-
ing i& Decision Making in Civil and Building
Engineering, Montreal.
Semantic Interoperability of Geospatial Services
Padmanabhuni, S. (2004). Semantic interoper-
ability for service oriented architecture (White
Paper): Infosys Technologies Limited.
Peng, Z. R., & Tsou, M. H. (2003). Internet GIS:
Distributed geographic information services for
the Internet and wireless networks. John Wiley.
Radwan, M., Bishir, Y., Emara, B., Saleh, A., &
Sabrah, R. (2005, April 16-21). Online cadastre
portal services in the framework of e-government
to support real state industry in Egypt. Paper
presented at the 8th International Conference
of Global Spatial Data Infrastructure (GSDI-8),
Cairo, Egypt.
Rich, E. (1989). Stereotypes and user modeling.
In A. Kobsa & W. Wahlster (Eds.), User models
in dialog systems. Springer.
Sikder, I., & Gangapadhayay, A. (2002). Design
and implementation of a Web-based collaborative
spatial decision support system: Organizational
and managerial implications. Information Re-
sources Management Journal, 15(4), 34–49.
This work was previously published in International Journal of Intelligent Information Technologies, Vol. 4, Issue 1, edited by
V. Sugumaran, pp. 31-51, copyright 2008 by IGI Publishing (an imprint of IGI Global).
Smith, B. (1996). Mereotopology: A theory
of parts and boundaries. Data & Knowledge
Engineering, 20, 287–303. doi:10.1016/S0169-
023X(96)00015-8
Vetere, G., & Lenzerini, M. (2005). Models for
semantic interoperability in service oriented ar-
chitectures. IBM Systems Journal, 44.
Warnecke, L., Beattie, J., Cheryl, K., & Lyday,
W. (1998). Geographic information technology
in cities and counties: A nationwide assessment.
Washington, DC: American Forests.
Wayne, L. (2005). Metadata in action: expand-
ing the utility of geospatial metadata. Federal
Geographic Data Committee.
Yeap, W. K., & Handley, C. C. (1990). Four im-
portant issues in cognitive mapping. Otago.AI Lab.
ENDNOTES
1
E-government Act of 2002, P.L. 107-347
(December 17, 2002)
357
358

Chapter 22
New Discovery
Methodologies in GIS:
Improving the Information
Retrieval Process

Nieves R. Brisaboa
University of A Coruña, Spain

Miguel R. Luaces
University of A Coruña, Spain

Diego Seco
University of A Coruña, Spain

ABSTRACT
In the last decade, the availability of on-line resources, and also the number of users accessing those
resources, has grown exponentially. The information retrieval process, which aims at the improvement
of the access to such resources, has been the focus of interest of many researchers. The presence of geo-
graphic data in these repositories of information is surprisingly high (for example, note that most of the
web pages about business contain information about the locations of their offices). In order to properly
manage this geographic data, the information retrieval process has been extended using architectures,
data structures, and other techniques developed by the GIS community. This has meant the beginning
of a new research field called Geographic Information Retrieval. In this chapter, the authors present a
study of the state-of-the-art of this new field, and they also highlight the main open problems that will
concentrate efforts during the next years.

DOI: 10.4018/978-1-4666-2038-4.ch022

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
New Discovery Methodologies in GIS
INTRODUCTION
The need to manage information has been one of
the key factors behind the consolidation of infor-
mation technology as an essential driving force for
the development of our society. Over the years,
many system architectures, index structures, and
other components have been proposed with the
fundamental goal of allowing efficient access to
information stored in huge document databases.
The research field that focuses on this goal is
called Information Retrieval (IR) (Baeza-Yates &
Ribeiro-Neto, 1999) and it started with the work
of Salton (1963). This research field has recently
undergone a spectacular development motivated
by the growth of the Internet and the need to search
the Web. A very important feature of IR is that it
deals with the problem of retrieving information
by its content rather than its metadata. Thus, there
are a number of techniques for retrieving docu-
ments of various types: texts, images, sound and
video files, etc.
Textual information often includes geographic
references in the text (for example, press releases
usually mention the place where the events hap-
pened). Taking these geographic references into
account provides added value to classic informa-
tion retrieval systems. The research on Geographic
Information Systems (GIS) (Worboys, 2004) has
dedicated much effort to study the special fea-
tures of geographic information and to develop
systems able to use and take advantage of them.
This field has received much attention in recent
years due to recent improvements in hardware
that have made possible the development of such
systems by many organizations. In addition, two
international organizations ISO (ISO/IEC, 2002)
and the Open Geospatial Consortium (OGC,
2003) are undertaking a major collaborative
effort to define standards and specifications to
develop interoperable systems. At the European
level, the INSPIRE (Infrastructure for Spatial
Information in Europe) directive (European Com-
mission, 2011) has enabled a breakthrough in the
field of corporate GIS and it remarks the future
importance of geographic information. Thanks
to these initiatives, many public organizations
are working in the development of spatial data
infrastructures (GSDI, 2011) that enable them to
share their spatial information.
These two research areas have progressed
independently over the years. On the one hand,
the index structures and techniques from the IR
field do not take into account the spatial nature
of geographic references that appear in text docu-
ments. On the other hand, spatial index structures
are not directly applicable in information retrieval
systems. However, users increasingly demand
services that allow them to locate the informa-
tion in its spatial context and even to access this
information using queries that take into account the
spatial information. These demands have caused
that researchers in each area have began to pay at-
tention to the other one resulting in a new research
field called Geographic Information Retrieval
(GIR). The aim of this field is to propose new
system architectures, index structures, and other
components in order to develop systems to retrieve
documents both thematically and geographically
relevant in response to queries of the form <sub-
ject, place>. An example of the type of queries
studied in this new field is the following: “Ph.D.
dissertations regarding geographic information
systems published in Spain.” The reader familiar
with classic information retrieval systems knows
that the relevance of the documents in a textual
search engine is based on the frequency of the
words that appear in the text of the documents.
Therefore, if the word Spain does not appear ex-
plicitly in a document its relevance will be low
with respect to this query. This happens even if the
word Madrid appears in the document (or any other
autonomous region, province or city of Spain)
because traditional IR systems are not prepared
to take into account the special characteristics
of the geographic information space (e.g., the
contained by spatial relationship between Madrid
and Spain). Query expansion techniques in clas-
359

sical IR systems reformulate queries by adding
new terms to the original query in an attempt to
provide a better context (Baeza-Yates & Ribeiro-
Neto, 1999) (for instance, in the previous example
we could expand the query using the term thesis
that is related with the term dissertation). Some
examples of these query expansion techniques are:
term reweighting, local clustering, local context
analysis, or those based on thesaurus. However,
all the particularities of the geographic references
are not properly represented by any of them.
Among the topics of interest in the area of
geographic information retrieval are the definition
of system architectures, index structures, and other
components that allow to model, capture, store,
manipulate, access, retrieve, analyze, and display
information efficiently. In addition, these tasks
involve additional difficulties over the same tasks
in the area of IR because of the special features
and requirements of geographic information.
Despite the common geographical nature of
the information, there are two fundamental differ-
ences on the requirements between GIS and GIR
systems that must be taken into account. First, the
spatial component of the queries in GIR system
is much simpler than the queries that are usually
posed to GIS because the latter involve complex
spatial relationships. For example, a typical query
in GIS can be monuments located in municipali-
ties adjacent to the location of a particular hotel.
In this case, the municipality where the hotel is
located must be found, and then the monuments
in adjacent municipalities must be retrieved. In
contrast, a typical query to a GIR system can be
monuments in London where the only spatial re-
lationship to check is whether the spatial scope of
the document lies within the geometry associated
with London.
The second difference is related to the type
of results expected for queries to both systems.
In GIS queries, the expected results are similar
to those of traditional databases, that is, objects
in the database either belong to the result or not.
However, in GIR systems, the expected results
360
New Discovery Methodologies in GIS
are similar to those of IR systems, that is, objects
belong to the result with a certain probability. Fol-
lowing the example queries, in the GIS query only
monuments in the municipalities adjacent are part
of the results. On the other hand, in a GIR system
query, monuments located in a city very close to
London may be relevant to the user.
In this chapter, we review the most important
contributions made in recent years to the field of
geographic information retrieval as part of the
description of our system architecture and index
structure (Brisaboa, et al., 2010).
SYSTEM ARCHITECTURE
FOR GEOGRAPHIC
INFORMATION RETRIEVAL
Figure 1 shows our proposal for the system ar-
chitecture of a geographic information retrieval
system. The architecture can be divided into
three independent layers: the index construction
workflow, the processing services, and the user
interfaces. The influence of GIS architectures and
spatial data infrastructures can be clearly identi-
fied. This influence is also reflected in the use of
the standards WMS (OGC, 2002) (map generation)
and SFS (OGC, 2006) (geographic data storage).
The bottom part of the figure shows the index
construction workflow, which in turn consists of
three modules: the document abstraction module,
the index structure, and the index construction
module.
The processing services are shown in the
middle of the figure. The Geographic Space Ontol-
ogy Service used in the spatial index construction
is shown on the left side. This service is used
extensively in the index construction module. On
the right side, one can see the two services that
are used to solve queries. The rightmost one is the
query evaluation service, which receives queries
and uses the index structure to solve them. The
other service is a Web Map Service following the
OGC specification (OGC, 2002) that is used to
New Discovery Methodologies in GIS
Figure 1. System architecture
create cartographic representations of the query
results. On top of these services a Geographic
Information Retrieval Module is in charge of
coordinating the task performed by each service
to respond to the user requests.
The topmost layer of the architecture shows the
two user interfaces that exist in the architecture:
the Administration User Interface and the Query
User Interface.
INDEX CONSTRUCTION
WORKFLOW
Document Abstraction
Given that the system must be generic, it must sup-
port indexing several kinds of documents. These
documents will be different not only because they
may be stored using different file formats (plain
text, XML, etc.), but also because their content
schema may be different. The set of attributes
that have to be stored in the index may be differ-
ent in each document collection. For instance, a
document collection may have a set of attributes
(such as document ID, author, and document text),
whereas other document collection may have a
different set (such as document ID, summary, text,
author, and source).
To solve this problem, we have defined an
abstraction for documents similar to the one used
in the Lucene text search engine (Apache, 2011).
We have extended this idea adding the spatial in-
dexing possibility. In our abstraction, a document
consists of a set of fields, each one with a value
that is extracted from the document text. Each
field can either be stored, indexed, or both. If a
field is stored, its contents are stored in the index
361

structure and they can be retrieved by a query. If
a field is indexed, then this field is used to build
the index structure. Furthermore, a field can be
indexed textually, spatially, or in both indexes.
Indexing
Some work has been done to combine textual
indexes and spatial indexes in structures able to
solve the queries of interest in GIR systems. These
structures can be broadly classified into hybrid
structures (i.e., textual and spatial indexes are kept
separate) and double-index structures (i.e., both
indexes are merged in one single structure). The
index structure proposed in the SPIRIT project
(Vaid, et al., 2005) is based on the combination
of a grid (Nievergelt, et al., 1981) and an inverted
index. In Vaid et al. (2005), the authors conclude
that keeping separate text and spatial indexes,
instead of merging both in one, results in less
storage costs but it could lead to higher response
times. Keeping both indexes separated has many
advantages (Martins, et al., 2005). First of all, all
textual queries can be efficiently processed by
the textual index and all spatial queries can be
efficiently processed by the spatial one. Moreover,
queries combining textual and spatial aspects are
supported. Updates in each index are handled in-
dependently, which makes easier the addition and
removal of data. Finally, specific optimizations can
be applied to each individual indexing structure.
Figure 2. GIR index structures
362
New Discovery Methodologies in GIS
In more recent works (Martins, et al., 2005;
Chen, et al., 2006), the authors survey the work
in the SPIRIT project and propose improvements
to the system and the algorithms defined. In their
work, two naive algorithms are proposed: Text-
First and Geo-First. Both algorithms use the
same strategy: one index is first used to filter the
documents (textual index in Text-First and spa-
tial index in Geo-First), the resulting documents
are sorted by their identifiers and then filtered
using the other index (spatial index in Text-First
and textual index in Geo-First). These two naive
algorithms provide a broad classification of GIR
index structures. Note that it can also be used to
classify hybrid structures. Figure 2 shows the three
basic structures according to this classification.
The left most structure belongs to the double-index
class and it can be both Text-First and Geo-First
depending on the algorithm used to solve the
queries. The other two are hybrid structures. The
first one belongs to the Geo-First class (the spatial
index is always accessed first) and the second one
belongs to the Text-First class (the textual index
is always accessed first).
In Zhou et al. (2005), the use of an inverted
index and an R-tree is proposed. Authors combine
both structures in the three ways described above
and they conclude that keeping both indexes
separated is less efficient than combining them
(a similar conclusion had been presented in Vaid
et al. [2005]) and that the use of the R-tree out-
performs the efficiency of the grid based structures.
New Discovery Methodologies in GIS
In Chen et al. (2006), authors propose the use
of space-filling curves (Morton, 1966; Böhm,
et al., 1999) and compare the resulting structure
both with the grid based and the R-tree based
structures. Space-filling curves are based on the
storage of the spatial objects according to the
order determined by a filling curve. According
to their experiments, the use of the space-filling
curves outperforms both the grid and the R-tree
based approaches.
Finally, in the STEWARD project (Lieberman,
et al., 2007), a double-index structure based on an
inverted index and a Quad-tree (Nelson & Samet,
1986) is presented. In addition, the authors propose
the use of a query scheduler in charge of choosing
a Text-First or a Geo-First algorithm according to
which index may return fewer results (this decision
is based on statistics collected by the system).
Nevertheless, none of these approaches
take into account the relationships between the
geographic objects that they are indexing. A
structure that can properly describe the specific
characteristics of geographic space is an ontol-
ogy, which is a formal explicit specification of
a shared conceptualization (Gruber, 1993). An
ontology provides a vocabulary of classes and
relations to describe a given scope. In Brisaboa et
al. (2010), we present an index structure based on
an ontology of the geographic space that describes
the concepts in our domain and the relationships
that hold between them. Our spatial ontology is
described in OWL-DL (W3C, 2011) and it can be
downloaded from the following URL: http://lbd.
udc.es/ontologies/spatialrelations. OWL classes
can be interpreted as sets that contain individu-
als (also known as instances). Individuals can be
considered instances of classes. Our ontology
describes eight classes of interest: SpatialTh-
ing, GeographicalThing, GeographicalRegion,
GeopoliticalEntity, PopulatedPlace, Region,
Country, and Continent. In our ontology there
are hierarchical relations among SpatialThing,
GeographicalThing, GeographicalRegion, Geo-
politicalEntity because:
• GeopoliticalEntity is subclass of
GeographicalRegion
• GeographicalRegion is subclass of
GeographicalThing and
• GeographicalThing is subclass of
SpatialThing.
That is, these four classes are organized into a
superclass-subclass hierarchy, which is also known
as taxonomy. Subclasses specialize (are subsumed
by) their superclasses. GeopoliticalEntity has four
subclasses: PopulatedPlace, Country, Continent,
and Region. All the individuals are members of
these subclasses. These four subclasses have an
additional necessarily asserted condition regard-
ing their relations with each other. They are con-
nected by the property spatiallyContainedBy that
describes the existence of a spatial relationship
among them. For instance, all the individuals of
class PopulatedPlace are spatiallyContainedBy
individuals of class Region (described in OWL
as PopulatedPlace spatiallyContainedBy only
[AllValuesFrom] Region). Figure 3 shows an ex-
ample of these relationships. Ontology classes are
represented as circles, individuals as rectangles,
and the relationships as labeled lines.
After having defined this ontology, we can
define an spatial index structure based on it. This
structure is a tree with four levels, one for each of
the subclasses of GeopoliticalEntity. The top-most
level contains a node for each of the instances of
the class Continent. Each node in this level refer-
ences the instances of the class Country that are
connected by the spatiallyContainedBy relation-
ship. The levels of Region and PopulatedPlace
are built using the same strategy. That is, the
structure of the tree follows the taxonomy of the
ontology. Figure 4 shows the spatial index structure
built from the instances shown in Figure 3.
The main advantage of this spatial index
structure over other alternatives is that intermedi-
ate nodes in the structure have a meaning in the
geographic space and they can have additional
information associated. For instance, we can as-
363

Figure 3. Ontology instances
sociate a list of documents that reference a given
Country and use this list of documents to solve
combined textual and spatial queries. Moreover,
given that there is a superclass-subclass relation-
ship between the levels, the bottom levels can
inherit the properties of the top levels. Particu-
larly, the documents associated to a node in the
Figure 4. Example of the index structure
364
New Discovery Methodologies in GIS
structure also refer to all nodes in its subtree.
Furthermore, the index structure is general in the
sense that the ontology of geographic space can
be adapted to each particular application. For
example, if a particular application uses a re-
stricted area of the geographic space where the
classes Continent and Country are not necessary
New Discovery Methodologies in GIS
and, on the other hand, the classes Province,
Municipality, City, and Suburb are needed, we
could define a different ontology of space and
base the index structure on it as long as the rela-
tionship spatiallyContainedBy still holds between
the classes. Finally, we could define additional
spatial relationships in the ontology such as spa-
tiallyAdjacent and maintain these relationships
in the index structure to improve the query capa-
bilities of the system.
This structure automatically performs query
expansion over the geographic component of a
query. In classical information retrieval systems,
the use of ontologies to perform query expansion
is well known as an ontology can be regarded as
a generalization of a thesaurus. However, our
structure does not use an additional geographic
ontology but it is defined using one. Thus, the query
expansion process is intrinsic to the nature of the
index and not an additional process. Furthermore,
our whole structure combines this geographic
ontology-based index with a textual index (an
inverted index), and classical query expansion
techniques can be used over the textual compo-
nent achieving full query expansion (in the sense
that both the textual and geographic components
of a query are expanded). Considering again the
example Ph.D. dissertations in Spain, using this
full query expansion technique we could retrieve
documents containing the terms thesis and Madrid.
Geo-Referencing of Documents
This is probably the most complex stage of the
workflow. In this stage, a geographic footprint is
assigned to each document allowing its spatial
indexing. A geographic footprint shows the geo-
graphic scope of the document and it can be set
as a list of geographic coordinates, a bounding
box grouping that coordinates, etc. For example,
if the cities of London and Liverpool are cited in
a document, the geographic coordinates of these
cities or the minimum bounding box containing
them can be used as geographic footprint of the
document. This stage comprises two steps. First,
the system analyses the document fields that are
spatially indexable and extracts candidate location
names from the text (i.e., discovery of location
names). In a second step, these candidate locations
are processed in order to determine whether the
candidates are real location names, and, in this
case, to compute their geographic locations (i.e.,
translation of location names to a geographic
model).
The main problem that can happen at this point
is the ambiguity of the geographic references. A
recent research by Garbin and Mani (2005) claims
that more than the 67% of the geographic refer-
ences cited in texts are ambiguous. Furthermore,
in Tjong et al. (2003) two kinds of ambiguity are
presented. First, a location name can be ambiguous
(polysemy). For instance, “London” is the capital
of the United Kingdom and it is a city in Ontario,
Canada too. Second, there can be multiple names
for the same geographic location, such as “Los
Angeles” and “LA.” A third class of ambiguity
could be considered. This is due to the use of
the same word to refer both a place name and a
organization, company, or person (e.g., Santiago).
Discovery of Location Names
Unlike geographic information systems, infor-
mation in GIR systems is not structured. It is
not possible to know a priori where geographic
references are stored, nor their categories (e.g.,
city, state, country, etc.). In this kind of systems,
geographic references are contained in the text
of the indexed documents. Therefore, these texts
have to be analyzed in order to discover the geo-
graphic references.
In this analysis, all the spatially indexable
fields are processed in order to discover the place
names contained within. There are two Linguistic
Analysis techniques that are widely used for this:
Part-Of-Speech tagging (Brill, 1992) and Named-
Entity Recognition (Chinchor & Robinson, 1997;
Pustejovsky, et al., 2005). On the one hand, Part-
365

Of-Speech tagging is a process whereby tokens are
sequentially tagged with syntactic labels, such as
verb or gerund. On the other hand, Named-Entity
Recognition is the process of finding mentions of
predefined categories such as the names of per-
sons, organizations, locations, etc. Combine both
techniques is a good solution to discover possible
place names contained in the text of documents. In
our prototype, we use the Natural Language Tool
LingPipe (Alias-i, 2011) to find locations. It is a
suite of Java libraries for the linguistic analysis
of human language free for research purposes
that provides both Part-Of-Speech tagging and
Named-Entity Recognition. LingPipe involves the
supervised training of a statistical model to rec-
ognize entities. The training data must be labeled
with all of the entities of interest and their types.
In spite of the good performance of these lin-
guistic analysis techniques, when the discovered
location names are translated to a geographic
model many problems related with the ambiguity
of the location names arise. Although these prob-
lems mainly affect the next step (i.e., translation
of location names), some issues are related with
this step. First, the system must determine if the
discovered location names are true place names.
Gazetteers have been widely used for this purpose.
A Gazetteer is a geographical dictionary that
contains, in addition to location names, alterna-
tive names, populations, location of places, and
other information related to the location. Although
some years ago the availability of these gazetteers
was very scarce (Petasis, et al., 2000), nowadays
there are many resources that provide this kind
of information. A more complex problem is the
disambiguation of a location name once it has
produced a true positive in the gazetteer. Many
clues within the whole text of the documents
are used by human beings to disambiguate each
location name cited in the text. For example, if
the location name Santiago is cited in a document
near other location names, such as A Coruña, we
will assume that the cited place is Santiago de
Compostela. However, if it is cited near Atacama,
366
New Discovery Methodologies in GIS
we will assume that the cited place is Santiago de
Chile. Perform this process in an automatic, and
even semi-automatic, way becomes a challenge
for the GIR community. Bruno Martins (2008)
describes in his Ph.D. some basic principles to
guide this automatic disambiguation process:
• One referent per discourse: A location
name cited several times in the same text
is likely to mean the same place. For ex-
ample, if Santiago is cited several times in
a document it should refer always either
Santiago de Compostela or Santiago de
Chile.
• Related referents per discourse:
Geographic references appearing in the
same document tend to refer to related lo-
cations. In our previous example, we use
the place name A Coruña or Atacama to
disambiguate the place name Santiago.
• Default senses: Important places are
more likely to be referenced. Therefore,
if no clues are available to disambiguate
a location name, the most important place
should be assigned. For example, countries
are most important than cities, a capital is
more important than each other city, a city
is more important than a street, etc.
In the SPIRIT project (Jones, et al., 2001; Jones,
et al., 2003; Jones, et al., 2004; Fu, et al., 2005),
a spatial ontology is used instead of a gazetteer.
Therefore, searches in the ontology both check
if the location name is a true place and provide a
disambiguation based on the height of the node
in the ontology. This follows the default senses
principle (the height of the regions is less than the
height of the cities). Besides this project, the most
important works in the area are: Web-a-where
(Amitay, et al., 2004), which uses spatial contain-
ers in order to identify locations in documents;
MetaCarta (the commercial system described by
Rauch et al. [2003]), which uses a natural language
New Discovery Methodologies in GIS
processing method; and STEWARD (Lieberman,
et al., 2007), which uses an hybrid approach.
Translation of Location Names
Once the location names have been located and
disambiguated, they have to be translated to a
geographic model, i.e., a geographic footprint
has to be assigned to each document in order to
make it indexable by a spatial index. There are
some differences in the methods proposed in the
bibliography. The first one is the type of geographic
object used to represent those footprints. Some of
the common options chosen for this purpose are:
geographic points (the geographic coordinates
of all the places cited in document), minimum
bounding boxes (the boxes of minimum extent that
cover the geographic positions of all the places
cited in the document), and the centroid of such
bounding boxes. A second feature that makes
proposed methods different is the uniqueness
of the footprint. Although documents are better
described when several footprints are allowed,
because many distant places can be cited in the
same document (consider for example a document
about the evolution of the world economy), many
of the approaches in the bibliography suggest the
use of a single footprint. The usage of one footprint
simplifies the process and improves the perfor-
mance of both the indexing and querying process.
One of the pioneering projects, previous to the
SPIRIT project, aiming at the geo-referencing of
contents in digital libraries is GIPSY (Georefer-
enced Information Processing SYstem) (Larson,
1995). In this project, each location name is trans-
lated to a geographic representation (for example,
a polygon) and a weighting value is assigned to it.
These values depend on features intrinsic to the
content of the documents (e.g., frequency of the
location name in the text of the document). Then,
all these geographic representations are combined
in three-dimensional topographic representations
that consider the weighting values. Finally, a
threshold determines the minimum elevation of
the topographic representation that makes the
area relevant.
In the SPIRIT project, each location name
cited in the text of a document is translated to a
bounding box and the footprint of a document
consists of several bounding boxes (one for each
location name cited in its text). This schema was
also used by Zhou et al. (2005). In Smith and Crane
(2001), authors propose the use of a set o points
as the footprint of a document. Thus, the footprint
of a document consists of the geographic coordi-
nates of all the locations cited in its text. These
coordinates are weighted by the frequency of the
location name in the document. Then, the centroid
of this set of points and its standard deviation are
computed according to the weights of the points,
and all the points that are more than twice from
the centroid are prune (remaining points make up
the footprint of the document).
Finally, as we mentioned in the previous sec-
tion, in the Web-a-where (Amitay, et al., 2004)
project the disambiguation process is based on
spatial containers. These containers are defined
according to the topological relationships that
exist in a gazetteer. Most of these relationships
belong to the class part-of (for example, Galicia
is part-of Spain). Once place names have been
disambiguated, all the related places are merged
in a taxonomy. The levels in this taxonomy are
ordered according to their relevance and those
levels which height is less than a threshold make
up the footprint of the document.
In our prototype, we have developed a service
based on an ontology of the geographic space
that is built using a Gazetteer (Geonames, 2011)
and a Geometry Supplier (NIMA, 2011). This
service uses information available in the gazetteer
(such as, place type, population, capital, etc.). All
these data are combined to compute the intrinsic
importance of each place.
367

QUERY EVALUATION SERVICE
The Query Evaluation Service is the component
in charge of using the index structure to answer
the queries posed by the users. Moreover, in
order to return a useful result, this service must
also provide a relevance ranking of the results.
In this section, we describe the types of queries
in GIR systems, the algorithms to solve them
and the equations used to compute the relevance
of the result for each of such types: pure textual
queries, pure spatial queries, and hybrid queries.
Note that the concept of relevance of a docu-
ment, although well known in the field of IR,
had not been introduced in the GIS before of the
arisen of the GIR systems. In Godoy and Rodrí-
guez (2004) some qualitative measures for the
spatial relevance of a document are introduced
based on concepts such as bounding boxes, dis-
tances, overlapping, and relative sizes. In Jones
et al. (2001), an hybrid approach combining the
distance in an ontology of the geographic space
and the Euclidean distance in the geographic space
is presented. Furthermore, in the context of the
Tumba project (Martins, et al., 2005; Andrade
& Silva, 2006), authors use some well-known
semantic concepts (e.g., adjacency, connectivity,
etc.) to calculate this relevance when an ontology
is available in the system. A different approach,
which does not assume the use of an ontology, but
based on similar concepts is presented in Zhou
et al. (2005).
Pure Textual Queries
These are queries such as “retrieve all documents
where the words hotel and sea appear.” The textual
index that is part of the index structure is used to
solve them. In our prototype we use Lucene to
implement this textual index, and thus, the rel-
evance ranking depends on its scoring. Lucene
scoring uses a combination of the vector space
model and the boolean model of information
retrieval (Baeza-Yates & Ribeiro-Neto, 1999).
368
New Discovery Methodologies in GIS
All scores are guaranteed to be 1.0 or less. More
information about the Lucene scoring can be found
in Gospodnetic and Hatcher (2005) and Apache
Lucene (2011).
Pure Spatial Queries
An example of this type of queries is “retrieve all
documents that refer to the following geographic
area.” The geographic area in the query can be a
point, a query window, or even a complex object
such as a polygon. The spatial index that is part
of the index structure is used to solve them. Given
that a document in the result set of a query can
include geographic references to one or more lo-
cation names relevant to the query, the relevance
of the document d with respect to the query q due
to each location name l has to be computed. We
denote this relevance as toponymRelevanceq,d,l. We
guarantee that this value is 1.0 or less in order to
make the integration of both spatial and textual
relevancies easier. In van Kreveld et al. (2005) both
spatial and textual relevancies are also normalized
to values between 0 and 1. Finally, we compute
the relevance of the document d with respect to
the query q as the maximum relevance due to any
location name (Equation 1).
spatialRelevanceq ,d
= max{toponymRelevanceq ,d ,l }
(1)
The computation of toponymRelevanceq,d,l for
queries specified selecting a node in the spatial in-
dex is a simplification of the previous one because
in this case we have the certainty that the query
refers to a specific node in the tree. Therefore, the
documents associated to this node have relevance
1.0. The relevance of a document associated to
the nodes in the subtree is computed using the
previous equation. This is reflected in Equation 2.
New Discovery Methodologies in GIS
toponymRelevanceq,d,l =
1 if l is specifiied in the query

 0.5 distance
 otherwise
importance
(2)
The sketch of the index structure shown in
Figure 5 is useful to understand the difference
between both types of queries. Each node in the
figure is annotated with its importance between
parentheses. On the one hand, when the user
specifies a query using the location name England,
the relevance of a document due to England (an
important city of Arkansas) will be higher than
the relevance due to England (a small city of
Oppland Fylke), and lower than the relevance
due to England (a part of the United Kingdom).
Concrete values of relevance are 0.5 for England
in Arkansas, 0.33 for England in Oppland, and 1.0
for England in the United Kingdom. Moreover,
the relevance of the document due to important
cities of England (UK) like London or Liverpool
is 0.5. This value is high enough to be taken into
consideration. On the other hand, when the query
is specified selecting the node for England in
Arkansas the relevance of a document due to this
node is 1.0 because the user explicitly indicates
the interest about documents with geographic
references to that location.
Finally, in the case of queries specified using
a query window the nodes are selected using the
Figure 5. Queries specified using a location name vs. queries specified selecting a node
classical algorithm of spatial indexes. Therefore,
the computation of toponymRelevanceq,d,l must be
performed using the distance (dcq,l) and the over-
lap area (oaq,l) between the query window and the
location name. Equation 3 defines this computa-
tion. We use parameters wdc and woa to weight the
relevance of each factor and we use the importance
of the location name to assign more relevance to
the most important nodes that reference the loca-
tion name.
wdc × dcq ,l + woa × oaq ,l
toponymRelevanceq ,d ,l =
importance
(3)
Equation 4 defines how to calculate the rel-
evance due to the distance to the query window.
centerDistanceq,l represents the Euclidean distance
between the location name l and the query window
q. Similarly, cornerDistanceq is a weight factor
that represents the maximum distance to the center
of the window.
centerDistanceq ,l
dcq ,l = 1 − (4)
cornerDistanceq
The relevance due to the overlap area with the
query window is calculated according to Equation
5. When the geometry stored in the node is a point
(leaf node), the overlap area is not significant.
Thus, we use 1/[area(q)+1]0.15. This value depends
369
 New Discovery Methodologies in GIS

only on the query window and is inversely propor- Moreover, three distances used to compute the
tional to its area. The concrete equation has been relevance due the distance to the query window
constructed based on the average area of the nodes (Equation 4) are shown. The weight factor cor-
in each level of the ontology of geographic space. ner distance is depicted as a solid line, and the
distances from Rome and Abruzzi to the center
oaq,l = of the query window are depicted as dotted lines.

 1

 if l is a po
oint

  
0.15
Textual Queries over a Geographic
 area(q) + 1
   Area

  area(l ∩ q) area(l ⊗ q) 

max  0, −  otherwise

  

 
 area(q) area(l) In this case, a geographic area of interest is given
(5) in addition to the set of words. An example is
“retrieve all documents with the word hotel that
Figure 6 uses an example query window in cen- refer to the following geographic area.” As in the
tral Italy to clarify the aforementioned equations. previous case, the geographic area in the query
The bounding boxes of two regions, Umbria and can be a point, a query window, or a complex
Abruzzi, and a populated place, Rome, are shown object. Both the textual and spatial indexes are
in this figure. These bounding boxes as well as used to solve them. Hence, we use the previous
the query window q are used to compute the area equations to compute the spatial and textual
of their respective entities (i.e., area(Umbria), relevance. Equation 6 defines how we combine
area(Abruzzi), and area(q)). The region of both relevance rankings. The weighted sum of the
Umbria is used to illustrate the relevance due to spatial and textual ranking values is one of the
the overlap area (Equation 5). This relevance is simplest methods and is commonly used (Mar-
computed using the area of the intersection of the tins, et al., 2005; Zhou, et al., 2005; Andrade &
region with the query and the area in the part of Silva, 2006). Furthermore, it is the base of more
the region that does not intersect with the query. complex ranking methods (Yu & Cai, 2007). We

Figure 6. Queries specified using a query window

370
New Discovery Methodologies in GIS

assume wt = 1-ws and calculate wt to normalize the USER INTERFACES

differences between textual and spatial rankings.
The system has two different user interfaces: an
relevanceq ,d = wt × textualRelevanceq ,d administration user interface and a query user
interface. The administration user interface can
+ ws × spatialRelevanceq ,d
be used to manage the document collection. The
(6)
main functionalities are: creation of indexes,
addition of documents to indexes, loading and
Textual Queries with Place Names
storing indexes, etc. Figure 7 shows a screen-shot
of the query user interface. This interface was
In this type of queries, some of the words are
developed as a web application using the Open
place names. For instance, “retrieve all documents
Layers API (OSGeo, 2011). This API provides
with the word hotel that refer to Spain.” Both the
a number of utilities for manipulating maps and
spatial and textual indexes are used to solve them
adding content to the map.
and, as in the previous case, the weighted sum
In the previous section, we have presented the
can be used to combine the relevancies obtained
types of queries that can be solved with this sys-
in each index.
tem. These queries have two different aspects: a
Our index structure presents an improvement
textual aspect and a spatial aspect. In our prototype,
over the rest of proposals: it can easily perform
the query user interface allows the user to indicate
query expansion on geographic references because
both aspects. The spatial context can be introduced
the index structure is built from an ontology of the
in three ways that are mutually exclusive:
geographic space. Consider the following query
“retrieve all documents that refer to Spain.” The
• Typing the location name: In this case,
query evaluation service will discover that Spain
the user types the location name in a text
is a geographic reference and then the internal
box. This is the most inefficient way be-
node that represents the geographic object Spain
cause the system has to obtain all the geo-
will be quickly located. Then, all the documents
graphic references associated with the
associated to this node are part of the query result.
place name typed by the user, which is a
Moreover, all the children of this node are geo-
time-expensive process.
graphic objects that are contained within Spain
• Selecting the location name in a tree: In
(for instance, the city of Madrid). Therefore, all
this case, the user sequentially selects a
the documents referenced by the subtree are also
continent, a country within this continent, a
part of the result of the query. The consequence is
region within the country, and a populated
that the index structure has been used to expand
place within the region. If the user wants
the query because the result contains not only
to specify a location name of a higher level
those documents that include the term Spain, but
than a populated place, it is not necessary
also all the documents that contain the name of
to fill in all the levels. The operation is very
a geographic object included in Spain (e.g., all
easy and intuitive because the interface is
the cities and regions of Spain). This geographic
implemented with a custom-developed
query expansion is complementary to other clas-
component using the AJAX technology
sical techniques expanding the textual component
that retrieves in the background the loca-
of the query.
tion names for the next level. When the
user selects a place in the component, the

371

Figure 7. Query User Interface
map on the right zooms in automatically to
the selected place.
• Selecting the spatial context of interest
in the map: The user can navigate using
the map on the right to visualize the spatial
context of interest. After that, a rectangle
can be drawn over it. The system will use
this rectangle as the query window if the
user did not type a place name or did not
select a location name.
CONCLUSION
Nowadays, Geographic Information Systems
constitute a consolidated area in computer sci-
ence. Many impressive research results have been
presented and, more importantly, an effective
technology transfer has improved the manage-
ment of geographic information in traditional
information systems. For example Spatial Data
Infrastructures are prominent examples of this
technology transfer.
372
New Discovery Methodologies in GIS
The work presented in this paper can also be
easily integrated in a spatial data infrastructure.
First, the query processing functionality can be
implemented as a Web Processing Service (OGC,
2007), which aims at the standardization of the
way that GIS calculations are made available
on the Internet. This service can then be used to
index document collections such as administra-
tive archives.
Furthermore, some of the internal components
of the architecture could also be implemented us-
ing Web Processing Services. As an example, in
Ladra et al. (2008) we show how the Geographic
Space Ontology Service, which is enclosed in
the intermediate layer of the architecture, can be
integrated in a Spatial Data Infrastructure using
the OGC Web Processing Service. Moreover, even
though our prototype uses a database as Gazetteer
Service, other implementations could easily use a
Gazetteer from a spatial data infrastructure.
Finally, the GIR architecture proposed in
this chapter is a perfect complement to OGC
catalogues. A GIR system like the one described
New Discovery Methodologies in GIS
in this chapter can be built over a collection of
OGC catalogues. In this case, instead of index-
ing digital documents, the system would index
metadata records which are composed of textual
descriptions and geographic references of the da-
tasets and services. The user interface of the GIR
system would allow a user to query the catalogues
with keywords and a geographic reference and
it would return a list of metadata records ranked
by their relevance. An additional advantage of a
GIR system is that an OGC catalogue is built by
a human being that decides what is relevant and
that categorizes the documents using thesaurus.
Furthermore, an OGC catalogue is oriented to
structured searches applying filters to metadata
fields. On the other hand, a GIR system catego-
rizes the documents automatically deciding what
is relevant using the contents and allowing non-
structured searches.
The application of this research to improve
the task of information retrieval turns out to be a
rather challenging problem. Due the importance
of this task (millions of users perform queries
on-line each day), many research efforts have
been devoted to this new research topic and a
new research area, named Geographic Informa-
tion Retrieval, has emerged covering the topics
in the intersection between GIS and IR. In this
chapter, we have presented the state-of-the-art in
this new field. Our own architecture (Brisaboa, et
al., 2010) was used as the framework that embraces
the majority of the topics in the area.
Many new research topics have emerged in
this young area. First, Geo-referencing tech-
niques must be improved to solve the ambiguity
problems. This is a crucial task as its influence
in the precision and recall of the system is ex-
tremely high. Second, the development of new
index structures is also very important as they
are a key factor in the performance of the GIR
systems. The use of ontologies in these structures,
enhancing their semantic, represents an exciting
problem that was just sketched with some initial
proposals. As we shown in Brisaboa et al. (2010),
structures considering the semantic of the space
present valuable improvements for GIR systems
over classical spatial index structures (e.g., query
expansion, relevance ranking, etc.). Efficient
implementations of these structures will also be
crucial in their applicability. Finally, some efforts
must be devoted to improve the system usability.
The spatial component of GIR systems entails the
need of user interfaces that allow users to properly
express the spatial scope and represent the results
in a user-friendly way.
REFERENCES
W3C. (2011). World wide consortium - OWL
2 web ontology language document overview.
Retrieved June 9, 2011 from http://www.w3.org
/TR/owl2-overview.
Alias-i. (2011). LingPipe, natural language tool.
Retrieved June 9, 2011 from http://www.alias-i.
com /lingpipe.
Amitay, E. Har’El, N., Sivan, R., & Soffer, A.
(2004). Web-a-where: Geotagging web content.
In Proceedings of the ACM Special Interest Group
on Information Retrieval Conference, SIGIR 2004,
(pp. 273-280). ACM Press.
Andrade, L., & Silva, M. J. (2006). Relevance
ranking for geographic IR. In Proceedings of the
Workshop on Geographic Information Retrieval,
GIR 2006. ACM Press.
Apache. (2011). The Apache Lucene project. Re-
trieved June 9, 2011 from http://lucene.apache.org.
Apache Lucene. (2011). Apache Lucene - Scoring.
Retrieved June 9, 2011 from http://lucene.apache.
org /java/2_2_0/scoring.html.
Baeza-Yates, R., & Ribeiro-Neto, B. (1999).
Modern information retrieval. Boston, MA: Ad-
dison Wesley.
373

Böhm, C., Klump, G., & Kriegel, H.-P. (1999).
XZ-ordering: A space-filling curve for objects with
spatial extensión. In Proceedings of the Advances
in Spatial Databases Conference, SSD 1999, (pp.
75-90). Springer.
Brill, E. (1992). A simple rule-based part of speech
tagger. In Proceedings of the Conference on Ap-
plied Natural Language Processing, ANLP 1992,
(pp. 152-155). Association for Computational
Linguistics.
Brisaboa, N. R., Luaces, M. R., Places, A. S., &
Seco, D. (2010). Exploiting geographic refer-
ences of documents in a geographical information
retrieval system using an ontology-based index.
GeoInformatica, 14(3), 307–331. doi:10.1007/
s10707-010-0106-3
Chen, Y.-Y., Suel, T., & Markowetz, A. (2006).
Efficient query processing in geographic web
search engines. In Proceedings of the ACM Special
Interest Group on Management of Data, SIGMOD
2006, (pp. 277-288). ACM Press.
Chinchor, N., & Robinson, P. (1997). MUC-7
named entity task definition. In Proceedings of
the Message Understanding Conference. ACM.
European Commission. (2011). INSPIRE direc-
tive. Retrieved June 9, 2011 from http://www.
ec-gis.org /inspire.
Fu, G., Jones, C. B., & Abdelmoty, A. I. (2005). On-
tology-based spatial query expansion in informa-
tion retrieval. In [Springer.]. Proceedings of On the
Move to Meaningful Internet Systems, ODBASE,
2005, 1466–1482. doi:10.1007/11575801_33
Garbin, E., & Mani, I. (2005). Disambiguating top-
onyms in news. In Proceedings of the Conference
on Human Language Technology and Empirical
Methods in Natural Language Processing, HLT-
EMNLP 2005, (pp. 363-370). The Association for
Computational Linguistics.
374
New Discovery Methodologies in GIS
Geonames. (2011). GeoNames gazetteer. Re-
trieved June 9, 2011 from http://www.geonames.
org.
Godoy, F., & Rodríguez, A. (2004). Defining
and comparing content measures of topologi-
cal relations. GeoInformatica, 8(4), 347–371.
doi:10.1023/B:GEIN.0000040831.81391.1d
Gospodnetic, O., & Hatcher, E. (2005). Lucene
in action. New York, NY: Manning.
Gruber, T. R. (1993). A translation approach to por-
table ontology specifications. Knowledge Acquisi-
tion, 5(2), 199–220. doi:10.1006/knac.1993.1008
GSDI. (2011). Global spatial data infrastructure
association, online documentation. Retrieved June
9, 2011 from http://www.gsdi.org.
ISO/IEC. (2002). Geographic information - Refer-
ence model. ISO 19101. Geneva, Switzerland: ISO.
Jones, C. B., Abdelmoty, A. I., & Fu, G. (2003).
Maintaining ontologies for geographical infor-
mation retrieval on the web. In On The Move to
Meaningful Internet Systems, ODBASE 2003, (pp.
934-951). Springer.
Jones, C. B., Abdelmoty, A. I., Fu, G., & Vaid,
S. (2004). The SPIRIT spatial search engine:
Architecture, ontologies and spatial indexing. In
Proceedings of the International Conference on
Geographic Information Science, (pp. 125-139).
Springer.
Jones, C. B., Alani, H., & Tudhope, D. (2001).
Geographical information retrieval with ontolo-
gies of place. In Proceedings of the International
Conference on Spatial Information Theory, COSIT
2001, (pp. 322-335). Springer.
New Discovery Methodologies in GIS
Jones, C. B., Purves, R., Ruas, A., Sanderson, M.,
Sester, M., van Kreveld, M. J., & Weibel, R. (2002).
Spatial information retrieval and geographical
ontologies an overview of the SPIRIT project. In
Proceedings of the ACM Special Interest Group
on Information Retrieval Conference, SIGIR 2002,
(pp. 387-388). ACM Press.
Ladra, S., Luaces, M. R., Pedreira, O., & Seco, D.
(2008). A toponym resolution service following
the OGC WPS standard. In Proceedings of the
International Sympoium on Web and Wireless
Geographic Information System, W2GIS 2008,
(pp. 75-85). Springer.
Larson, R. R. (1995). Geographic information
retrieval and spatial browsing. In Geographic
Information Systems and Libraries: Patrons,
Maps, and Spatial Information (pp. 81–123).
Champaign-Urbana, IL: University of Illinois.
Lieberman, M. D., Samet, H., Sankaranarayanan,
J., & Sperling, J. (2007). STEWARD: Architecture
of a spatio-textual search engine. In Proceedings
of the ACM International Symposium on Advances
in GIS, ACMGIS 2007, (pp. 186-193). ACM Press.
Martins, B. (2008). Geographically aware web
text mining. Unpublished Doctoral Dissertation.
Lisbon, Portugal: Universidade de Lisboa.
Martins, B., Silva, M. J., & Andrade, L. (2005).
Indexing and ranking in geo-IR systems. In
Proceedings of the Workshop on Geographic
Information Retrieval, GIR 2005, (pp. 31-34).
ACM Press.
Morton, G. M. (1966). A computer oriented
geodetic data base and a new technique in file se-
quencing. Technical Report. Palo. Alto, CA: IBM.
Nelson, R. C., & Samet, H. (1986). A consistent
hierarchical representation for vector data. In Pro-
ceedings of the Annual Conference on Computer
Graphics and Interactive Techniques, SIGGRAPH
1986, (pp. 197-206). ACM Press.
Nievergelt, J., Hinterberger, H., & Sevcik, K. C.
(1981). The grid file: An adaptable, symmetric
multi-key file structure. In [Springer.]. Proceed-
ings of the European Cooperation in Informatics,
ECI, 1981, 236–251.
NIMA. (2011). National imagery and mapping
agency, vector map level 0. Retrieved June 9, 2011
from http://www.mapability.com.
OGC. (2002). OpenGIS web map service
implementation specification. OpenGIS Project
Document 01-068r3. New York, NY: Open GIS
Consortium, Inc.
OGC. (2003). OpenGIS reference model. OpenGIS
Project Document 03-040. New York, NY: Open
GIS Consortium, Inc.
OGC. (2006). OpenGIS implementation specifica-
tion for geographic information - Simple feature
access - Part 2: SQL option. OpenGIS Project
Document 06-104r3. New York, NY: Open GIS
Consortium, Inc.
OGC. (2007). OpenGIS web processing service
implementation specification. OpenGIS Project
Document 05-007r7. New York, NY: Open GIS
Consortium, Inc.
OSGeo. (2011). Open layers API. Retrieved June
9, 2011 from http://openlayers.org.
Petasis, G., Cucchiarelli, A., Velardi, P., Paliouras,
G., Karkaletsis, V., & Spyropoulos, C. D. (2000).
Automatic adaptation of proper noun dictionar-
ies through cooperation of machine learning and
probabilistic methods. In Proceedings of the ACM
Special Interest Group on Information Retrieval
Conference, SIGIR 2000, (pp. 128-135). ACM
Press.
Pustejovsky, J., Knippen, R., Littman, J., & Saurí,
R. (2005). Temporal and event information in
natural language text. Computers and the Hu-
manities, 39, 123–164.
375

Rauch, E., Bukatin, M., & Baker, K. (2003). A
confidence-based framework for disambiguating
geographic terms. In Proceedings of the Human
Language Technology Conference, HLT-NAACL
2003, (pp. 50-54). Association for Computational
Linguistics.
Salton, G. (1963). Associative document re-
trieval techniques using bibliographic informa-
tion. Journal of the ACM, 10(4), 440–457.
doi:10.1145/321186.321188
Smith, D. A., & Crane, G. (2001). Disambiguating
geographic names in a historical digital library.
In Proceedings of the European Conference on
Research and Advanced Technology for Digital
Libraries, ECDL 2001, (pp. 127-136). Springer.
Tjong, E. F., Sang, K., & Meulder, F. D. (2003).
Introduction to the CoNLL-03 shared task:
Language-independent named entity recognition.
In Proceedings of the Conference on Natural
Language Learning, CoNLL 2003, (pp. 142-147).
Springer.
Vaid, S., Jones, C. B., Joho, H., & Sanderson, M.
(2005). Spatio-textual indexing for geographical
search on the web. In Proceedings of the Sympo-
sium on Spatial and Temporal Databases, SSTD
2005, (pp. 218-235). Springer.
van Kreveld, M., Reinbacher, I., Arampatzis,
A., & van Zwol, R. (2005). Multi-dimensional
scattered ranking methods for geographic infor-
mation retrieval. GeoInformatica, 9(1), 61–84.
doi:10.1007/s10707-004-5622-6
Worboys, M. F. (2004). GIS: A computing per-
spectiva. Boca Raton, FL: CRC.
This work was previously published in Discovery of Geospatial Resources: Methodologies, Technologies, and Emergent
Applications, edited by Laura Díaz, Carlos Granell and Joaquín Huerta, pp. 37-55, copyright 2012 by Information Science
Reference (an imprint of IGI Global).
376
New Discovery Methodologies in GIS
Yu, B., & Cai, G. (2007). A query-aware docu-
ment ranking method for geographic information
retrieval. In Proceedings of the Workshop on
Geographic Information Retrieval, GIR 2007.
ACM Press.
Zhou, Y., Xie, X., Wang, C., Gong, Y., & Ma, W.
Y. (2005). Hybrid index structures for location-
based web search. In Proceedings of the ACM
International Conference on Information and
Knowledge Management, CIKM 2005, (pp. 155-
162). ACM Press.
KEY TERMS AND DEFINITIONS
Architecture: Formal description of the set
of services and structures that compose a system.
Geographic Information Retrieval (GIR):
Research area related with the properly manage-
ment of the geographic information available in
IR repositories.
Geo-Reference: Geographic references con-
tained in repositories of information in textual
form (e.g., place names, postal codes, etc.)
Geo-Referencing Process: Common task in
GIR involving the location of geo-references and
its translation to a formal model of the geographic
space.
Information Retrieval (IR): Research area
related with the access to non-structured reposi-
tories of information.
Relevance: Measure of the importance of a
resource regarding to a specific query.
Workflow: Sequence of tasks that have to be
performed to achieve a goal.
 377

Chapter 23
A Multidimensional Model
for Correct Aggregation of
Geographic Measures
Sandro Bimonte
Cemagref, UR TSCF, France

Marlène Villanova-Oliver
Laboratoire d’Informatique de Grenoble, France

Jerome Gensel
Laboratoire d’Informatique de Grenoble, France

ABSTRACT
Spatial OLAP refers to the integration of spatial data in multidimensional applications at physical, logical
and conceptual levels. The multidimensional aggregation of geographic objects (geographic measures)
exhibits theoretical and implementation problems. In this chapter, the authors present a panorama of
aggregation issues in multidimensional, geostatistic, GIS and Spatial OLAP models. Then, they illustrate
how overlapping geometries and dependency of spatial and alphanumeric aggregation are necessary
for correctly aggregating geographic measures. Consequently, they present an extension of the logical
multidimensional model GeoCube (Bimonte et al., 2006) to deal with these issues.

INTRODUCTION of facts and dimensions. Facts are described by

A Data Warehouse (DW) is a centralized reposi-
tory of data acquired from external data sources
and organized following the multidimensional
model (Kimball, 1996) in order to be analyzed by
On-Line Analytical Processing (OLAP) systems.
Multidimensional models rely on the concepts
DOI: 10.4018/978-1-4666-2038-4.ch023
values called measures. Dimensions, structured
in hierarchies, permit to analyze facts accord-
ing to different analysis axes and at different
levels of detail. An instance of a dimension is a
set of members organized according to the hier-
archies. An instance of the conceptual model is
represented by a hypercube whose axes are the
dimension members at the finest levels. Each cell
of a hypercube contains the value of the detailed

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 A Multidimensional Model for Correct Aggregation of Geographic Measures
measure. This basic cube (also called facts table)
is then enhanced with cells that contain aggregated
values of the measures for each combination of
higher level’s members. Aggregation operators
applied on the measures must be specified in the
conceptual model and depend on the semantics
of the application. The classical functions used
to aggregate numeric measures are the standard
SQL operations “COUNT”, “SUM”, “MIN”,
“MAX” and “AVG”. The multidimensional
model allows pre-computation and fast access to
summarized data in support of multidimensional
analysis through OLAP operators which permit to
explore the hypercube. Drill operators (Roll-Up
and Drill-Down) permit to navigate in the dimen-
sions hierarchies aggregating measures. Cutting
operators (Slice and Dice) select and project a part
of the hypercube. The multidimensional model
and OLAP operators have been formalized in
some logical models (Abello et al., 2006) as a
support to correct aggregation of measures which
plays a central role in multidimensional analysis
(Pedersen et al., 2001). They define constraints
on the aggregation functions in compliance with
the semantics of the measure and explicit the
dimensions that can be used in the multidimen-
sional queries.
Most of 80% of transactional data contain
spatial information, which represents the form
and the location on the earth surface of real world
objects (Franklin, 1992). The heterogeneity of
physical spaces and the strong spatial correlation
of thematic data (Anselin, 1989) are not taken into
account into multidimensional models. Then, a
new kind of systems have been developed, which
intended to integrate the spatial component of the
geographic information into multidimensional
analysis: Spatial OLAP (SOLAP) (Bédard et al.,
2001). Spatial OLAP allows decision-makers to
explore, analyze and understand huge volume of
geo-spatial datasets, in order to discover unknown
and hidden knowledge, patterns and relations.
This useful information can help spatial analysts
and decision-makers to validate and reformulate
378
decisional hypothesis, and to guide their spatial
decision making processes. SOLAP technologies
have been usefully applied in several domains:
geo-marketing, urban, health, environment, crisis
management, etc. (Bédard et al., 2001) as they
allow non-computer science users to exploit data-
bases, statistical analysis and spatial analysis tools
without mastering complex query languages and
Geographic Information Systems functionalities,
and understanding underlying complex spatial
datasets. SOLAP redefines main multidimensional
concepts: spatial dimensions, spatial measures and
spatial aggregation functions. In this approach,
spatial measures are not numerical values, but
spatial objects (geometries) which are aggregated
using spatial aggregation functions (union, inter-
section, etc.) (Shekar et al., 2001). As shown in
this work, SOLAP models only partially support
dependency of spatial and numerical values,
which can lead to wrong aggregation of spatial
and numerical measures (geographic measures).
In this paper, we identify a three-step ag-
gregation process for the correct aggregation
of geographic measures, and we formalize it by
providing an extension of the logical multidimen-
sional model, GeoCube (Bimonte et al., 2006). The
model provides a set of rules to ensure the valid
aggregation of geographic measures.
This paper is organized as follows. In the Sec-
tion ”Background”, we introduce main concepts
of geographic data and Spatial OLAP. Section
“Related Work” discusses aggregation issues in
multidimensional databases, GIS, geostatistic
and Spatial OLAP domains. We investigate the
problem of the correct aggregation of geographic
measures in the Section “Research Motivations”.
The extension of the multidimensional model Geo-
Cube is presented in Section “Correct Geographic
Multidimensional Aggregation”. Conclusions and
discussions about implementation issues are given
in the Section “Conclusion and Discussion”.
A Multidimensional Model for Correct Aggregation of Geographic Measures
BACKGROUND
Geographic Data
Geographic data are objects or phenomena of the
real world described by both a spatial component
and a semantic component (Degrene & Salgé,
1997). The spatial component gives the object’s
geometry and its position on the earth’s surface
(e.g. shape of a department). The semantic com-
ponent is a set of descriptive properties (e.g. name
and population of a department) and of spatial,
thematic and map generalization relationships
that the object shares with other (geographic)
objects (e.g. a department belongs to a region).
Map generalization is the process of obtaining
less detailed maps at different scales (Weibel &
Dutton, 2001).
Geographic Information Systems (GIS) are
main tools to handle geographic information
(Longley et al., 2001). GIS allow storing, visual-
izing and analyzing geographic data. GIS store
geographic data using two different models: vector
and raster models (Rigaux et al., 2002). The raster
model represents the space as a single continu-
ous space by means of a grid. Thematic data are
associated with each cell of the grid. The vector
model represents the space through geographic
objects. A geographic object is a set of spatial
objects composed by points. Alphanumeric at-
tributes are associated with each spatial object.
Querying and analyzing methods for geographic
data extends classical ones, explicitly exploiting
its spatial component (Longley et al., 2001). Some
models, representing the geographic informa-
tion as a set of uniform geographic objects have
been proposed in literature (Erwig & Schneider,
2000; Guting, 1998; Guting & Schneider, 1995,
Voisard & David, 2002). They provide classical
GIS operators such as spatial selection, overlay,
merge, fusion, etc (Longley et al., 2001). These
operators allow selecting geographic objects us-
ing thematic and spatial predicates (i.e. “Which
cities are located at 50 Km far from Paris?”) and
creating new geographic objects by splitting and/
or aggregating geographic objects. Aggregation of
geographic objects is performed using spatial ag-
gregation functions (i.e. union, centroid, etc.) (Tao
& Papadias, 2005) to aggregate the spatial com-
ponent, and alphanumeric aggregation functions
to aggregate alphanumeric attributes (SUM, AVG,
etc.) (Rigaux et al., 2002). Spatial disaggregation
functions are used to split geographic objects. A
spatial disaggregation is a spatial function that
does not preserve all the original geometries into
its final result (i.e. intersection, centroid, etc.).
Spatial OLAP
Integration of spatial data in OLAP leads to the
definition of Spatial OLAP (SOLAP). (Bédard
et al., 2001) define SOLAP as “a visual platform
built especially to support rapid and easy spatio-
temporal analysis and exploration of data fol-
lowing a multidimensional approach comprised
of aggregation levels available in cartographic
displays as well as in tabular and diagram dis-
plays”. SOLAP systems integrates OLAP and GIS
functionalities into a unique framework allowing
visualizing measures on maps, and triggering
OLAP and spatial analysis operators by simply
interacting with the cartographic interactive com-
ponent of the SOLAP client (Rivest et al., 2005;
Escribano et al., 2007).
In order to exploit the spatial component into
OLAP systems, SOLAP re-think main OLAP con-
cepts: dimensions and measures. SOLAP models,
using the vector model, introduce geographic
information as analysis axes thanks to spatial
dimensions (Bédard et al., 2001; Fidalgo et al.,
2004; Malinowski & Zimányi, 2005). A spatial
dimension is described by spatial hierarchies that
are characterized by the presence of the geometric
attribute at the different levels (spatial levels). Spa-
tial hierarchies, usually very complex, can present
spatial and classical levels. Members of different
spatial levels are related by topologic inclusion or
intersection relationships (Malinowski & Zimányi,
379
 A Multidimensional Model for Correct Aggregation of Geographic Measures
2005). An example of spatial hierarchy is French
administrative organization, which groups cities
into departments and departments into regions. A
topological inclusion relationship exists between
these geographic objects.
Spatial data can be also used as analysis subject
using the concept of spatial measure. Different
definitions of spatial measure can be found in
literature: a collection of pointers to spatial objects
or the result of topological or metric operators
(Bédard et al., 2001; Malinowski & Zimányi, 2004;
Stefanovic et al., 2000). In (Bimonte et al., 2006),
the concept of geographic measure is introduced
as a geographic object representing a real world
entity and described by some alphanumerical at-
tributes and geometry. Aggregation of geographic
measures corresponds to the aggregation of their
alphanumeric and spatial attributes by means of
classical, spatial and user-defined aggregation
functions.
An example of SOLAP application concerns
the monitoring of natural phenomena in a park
(Figure 1). The dimensions of that spatio-multi-
dimensional application are time and the type of
phenomenon (e.g. fire, landslide, etc.). “Time”
dimension is described by a calendar hierarchy
(“Day” < “Month” < “Year”). “Natural Phenom-
Figure 1. Multidimensional application with a geographic measure
380
ena” dimension groups phenomena (i.e. fire,
avalanche, etc.) into categories of phenomena
(i.e. dangerous, etc.).
The measure is the damaged zone, which is
described by a geometry, the number of damaged
trees and the area. The area is a derived measure
as it is calculated using other measures (Abello
et al., 2006). Then, to calculate the area for the
aggregated geographic object, no aggregation
function is applied. It is calculated using the
geometry resulting from the spatial aggregation.
Figure 2 shows an example of the facts table
and the cartographic representation of two geo-
graphic measures.
The model allows users to answer queries like:
“What regions have been damaged by fires during
1978 ?”. This query implies the aggregation of
the geometry and the number of damaged trees.
Some works study the aggregation of spatial
and numerical measures in spatial data warehous-
es. Performance is fundamental for SOLAP analy-
sis. Then, several authors, focusing on numerical
measures and spatial dimensions, integrate spatio-
temporal indexes with pre-aggregation techniques
(i.e. Tao & Papadias, 2005; Rao et al., 2003 Zhang
& Tsotras, 2005). These methods speed-up ag-
gregation computation on evolving user-defined
A Multidimensional Model for Correct Aggregation of Geographic Measures
Figure 2. Facts table and cartographic representation of geographic measures
spatial hierarchies. While numerical aggregation
functions have been well defined (i.e. SUM, MIN,
MAX, etc.), a standard set of spatial aggregation
functions for spatial measures (geometries) has
not been defined yet. (Shekar et al., 2001) classify
spatial aggregations as distributive, algebraic, and
holistic in order to grant summarazability (c.f.
Sec “Multidimensional Models”) in spatial data
warehouse. Lopez & Snodgrass (2005), and Silva
et al., (2008) formally define a set of spatial and
numerical aggregation functions (i.e. AvgArea,
Count_AtNorth, etc.). In addition to the conceptual
aspects of spatial aggregation, some efforts have
been done for improving queries computation
in large spatial data warehouses. They provide
indexes materialization, selection of aggregated
measures and computational geometry algorithms
(i.e. Stefanovic et al., 2000; Han et al., 1998).
Finally, some models have been proposed for
spatial data warehouses, which formally define
SOLAP main concepts: spatial dimensions, spa-
tial measures, multidimensional operators and
aggregation functions. In particular, the correct
aggregation of geographic information is a quite
complex task, which raises some unresolved
problems as described in the next sections.
RELATED WORK
In this Section, we investigate problems and
solutions related to aggregation of geographic
data according to different models: logical mul-
tidimensional models, GIS models, geostatistic
models and logical SOLAP models.
Multidimensional Models
The correct aggregation of measures is of crucial
importance in the multidimensional analysis
process. The correctness of aggregation depends
on both the semantics of the measure and the
multidimensional structure of the data warehouse
(Horner et al., 2004; Pedersen et al., 2001). For
example, the sum of pollution values has no
sense, while the max or the min operators can
safely be applied. Also, it is possible to sum
the populations of cities of a region, but to sum
these values on the time dimension is not correct
(the same inhabitants would be counted several
times). This problem is known in OLAP literature
as Additivity (Kimball, 1996). A measure is: (1)
Additive if the sum operator can be applied on all
dimensions (2) Semi-additive if the sum can be
applied on a subset of dimensions (3) Non additive
if the sum makes no sense whatever the dimen-
sion. The correct aggregation means providing
aggregation constraints (i.e. a control on the type
of the aggregation, considering the semantics of
the measure (i.e. the nature of the measure) and
used dimensions). Some logical multidimensional
models define explicitly the type of aggregation
functions that can be applied to measures (Abello
et al., 2006; Lenher, 1998; Pedersen et al., 2001;
Trujillo et al., 2000). These models, based on
the definitions introduced by Rafanelli & Ricci
381
 A Multidimensional Model for Correct Aggregation of Geographic Measures
(1983), classify measures according to three dif-
ferent types of aggregation functions that can be
applied to them: ∑ (data can be added together,
e.g. population), ω (data can be used for average
calculations, e.g. temperature) or c (constant data
implying no application of aggregation operators,
e.g. name). Considering only the SQL standard
aggregation functions (AF) applying to each type
of data (c, ω and ∑), an inclusion relationship
exists so that AFc = {COUNT} ⊂ AFω = {AVG,
MIN, MAX, COUNT} ⊂ AF∑ = {SUM, AVG,
MIN, MAX, COUNT}. This inclusion relation-
ship allows deducing that “∑ data” which can be
summed can also be averaged and identified as the
minimum or the maximum of a set (as “ω data”)
and that “ω data” can also be counted as “c data”.
By transitivity, “∑ data” can also be counted.
Geostatistic Models
The aggregation of geographic information is cru-
cial in spatial analysis especially in the geostatistic
domain. Spatial aggregation operators (i.e. union,
convex hull, etc.) and the management of aggre-
gations applied to alphanumeric attributes have
been widely discussed. Different frameworks,
dealing with the type of alphanumeric attributes
and with their aggregation functions have been
proposed to address the issue of disjoint spatial
units aggregation into bigger spatial units (Charre
et al., 1997; Chrisman, 1974; Chrisman, 1998).
Alphanumeric attributes are grouped into classes
and they are associated with particular aggregation
rules. For example, a rule defines that attributes
representing relative values such as the tempera-
ture or the population density, can be aggregated
using a weighted average. Another rule expresses
that attributes used to represent raw quantitative
values, such as the number of inhabitants, can be
aggregated using the SUM operator. For the disag-
gregation process (moving to smaller units), the
sum is forbidden as it makes non-sense (Charre
et al., 1997; Chrisman, 1974; Egenhofer & Frank,
1986). Finally, we underline that the concept of
382
analysis dimension is not present. Indeed, all the
aggregation rules are defined exclusively on the
spatial dimension, while we claim that performing
analysis on other dimensions (e.g. time dimension)
can reveal itself interesting.
GIS Models
Spatial aggregation operators have been formal-
ized and implemented in Geographic Information
Systems (GIS) (Rigaux & Scholl, 1995; Voisard
& David, 2002; Longley et al., 2001). In all these
approaches, unlike the geostatistic solutions, the
semantics of alphanumeric attributes are not taken
into account. This lack is also evident from the
implementation point of view. Nowadays, com-
mercial GIS systems (i.e. ArcGIS, MapInfo, etc.)
implement only union and splitting operators. They
provide a simple control on the aggregations ap-
plied to alphanumeric attributes using only their
type (i.e. numeric, alphanumeric, etc.) but not their
semantics. These tools propose some numeric ag-
gregation functions (i.e. sum and average) for the
numeric attributes and some particular methods, as
for example the usage of a default value or of the
count operator, for textual attributes. They do not
consider the semantics of the attributes. Therefore,
it is possible to sum temperatures or population
densities, which makes no sense. Moreover, all
other spatial aggregation operators (i.e. the convex
hull, etc.) only create new geometries without
aggregating alphanumeric attributes.
Spatial OLAP Models
Some SOLAP logical models have been proposed
in literature (Ahmed & Miquel, 2005; Damiani &
Spaccapietra, 2006; Jensen et al., 2004; Pourab-
bas, 2001; Sampaio et al, 2006, Silva et al., 2008).
Ahmed & Miquel (2005) define a spatial
dimension according to the raster model as an
infinite set of members. They simulate this “infi-
nite” set using interpolation functions to calculate
measures in every point of the analysis region
A Multidimensional Model for Correct Aggregation of Geographic Measures
represented by spatial dimensions. Pourrubas
(2003) provides a formal framework to integrate
multidimensional and geographic databases. The
model handles spatial dimensions in which mem-
bers of different levels are related by topological
inclusion relationship. These models only handle
numerical SQL aggregation functions. Sampaio et
al. (2006) define a logical model for spatial data
warehouse using the object-oriented approach.
Spatial measures and numerical measures can be
aggregated using SQL spatial and numerical ag-
gregation functions. The model does not provide
any support for correct aggregation. Damiani &
Spaccapietra (2006) define a model in which
all levels of a spatial dimension can be used as
measures, allowing multi-granular analysis. The
model aims at supporting measures as geographic
objects, but it does not explicitly represent spatial
and alphanumeric attributes. Then, no aggregation
constraint is defined on measures (attributes of
geographic objects).
Silva et al., (2008) define formally the ele-
ments of a spatial data warehouse with particular
attention to aggregation functions for spatial data.
They introduce a set of aggregation functions that
combine numerical and spatial functions. These
aggregation functions are classified according to
whether numerical aggregation is scalar, distribu-
tive or holistic, and whether spatial aggregation
is unary or n-ary function. However, these ag-
gregation functions are not associated with any
aggregation constraints and numerical functions
are applied exclusively to metric values of spatial
data (i.e. perimeter, etc.).
Only, Jensen et al., (2004) and Pedersen et al.,
(2001) define logical multidimensional models
for spatial data warehouses taking into account
aggregation constraints. In particular, Jensen et
al., (2004) propose a model for location-based
services. It supports partial inclusion of spatial
dimension members granting imprecise measures.
Measures are numerical and alphanumeric values.
The model extends (Pedersen et al., 2001) and it
grants correct aggregation of non-spatial measures
by classifying measures according valid aggrega-
tion functions (c.f. the Section “Multidimensional
Models”). Nevertheless, the model does not ex-
plicitly represent spatial measures nor associated
aggregation constraints.
Pedersen & Tryfona (2001) investigate pre-
aggregation in multidimensional applications
with spatial measures. In particular, they study
the pre-aggregation of alphanumeric attributes
associated with spatial measures. In other words,
they investigate the correct aggregation of alpha-
numeric attributes of geographic objects according
to their spatial components. The model represents
measures as bottom levels of dimensions. Ag-
gregated measures are less detailed dimensions
levels. Since a topological inclusion relationship
always exists between spatial objects at different
levels, the model does not allow applying spatial
disaggregation functions. Then, only union can
be used. Spatial disaggregation functions, such
as intersection or centroid, can not be applied.
To conclude, few SOLAP models consider
aggregation constraints on spatial and numerical
measures without correctly support aggregation
of geographic measures. Indeed, as shown in the
next section, correct aggregation of geographic
measures should take into account: semantics
of measures, used dimensions, overlapping ge-
ometries and dependency between spatial and
alphanumeric functions.
RESEARCH MOTIVATIONS
The aggregation of geographic measures raises
several problems from both theoretical and
implementation points of view. In particular,
alphanumeric and spatial functions used for the
aggregation are dependent: used spatial func-
tion dictates alphanumeric functions that are
allowed for non-spatial attributes. This implies
a redefinition of the OLAP additivity concept
for spatio-multidimensional databases. Existing
spatio-multidimensional models do not complete-
383
 A Multidimensional Model for Correct Aggregation of Geographic Measures
ly investigate this issue, which covers different
research domains: multidimensional databases,
geostatistic models and Geographic Information
Systems (GIS) as discussed in the previous sec-
tion. Let us consider the application of Figure 1,
and let us suppose we need to get information at
the year level. A Roll-Up operation on the year
level aggregates the geographic measures Zone A
and Zone B. In such a situation, how to aggregate
these geographic objects?
Using union to aggregate the geometries, que-
ries like: “Where and how many trees have been
damaged by some fires during 1978?” (Query 1)
can be answered.
To calculate the number of damaged trees
aggregating on the time dimension, we propose
to apply the average operator, as the number of
trees is non-additive, and then to sum all these
averages in order to have the total of trees in the
merged zone. More in details, Zone A (resp. Zone
B) is first splitted into two zones (Zone 1 of A and
Zone 2 of A (resp. Zone 1 of B, and Zone 2 of
B)), and then its alphanumeric attributes values
are recalculated (trees and area) (Figure 3). Please
note that a user-defined function is used to calculate
alphanumeric non-derived attributes of geographic
Figure 3. Decomposition of geographic measure (Step 1)
384
measure. The number of trees is calculated as a
weighted average on the surface. Areas of Zone
1 of A and of Zone 2 of A are the same (55), and
a weighted average on the surface is used for the
number of trees. Then, this latter is the same (10)
for the two new zones. By the same way, spatial
and alphanumeric attributes values of Zone 1 of
B and Zone 2 of B are calculated.
Then, using these geographic objects, we
calculate another set of geographic objects apply-
ing the average to the number of trees for the
geographic objects with the same geometry and
coordinates (Figure 4). We use average because
the aggregation is made on the time dimension,
which requires not counting several times the
same tree. For instance in Figure 3, Zone 2 of A
and Zone 1 of B have the same geometry and
coordinates, and some trees (note that the number
of trees varies in time) (Figure 3). Then, from
these two geographic objects, we create a new
geographic object Zone V2 whose geometry is
the same as the ones of Zone 2 of A and Zone 1
of B and whose number of trees is the average
(8=(10+6)/2) (Figure 4).
Finally, we apply union to these zones and sum
the numbers of trees (Figure 5). In particular, the
A Multidimensional Model for Correct Aggregation of Geographic Measures
Figure 4. Vertical aggregation of trees using average
geometry of Zone E is the union of the geometries
of Zone V1, Zone V2 and Zone V3. The number
of trees is the sum of the number of trees of Zone
V1, Zone V2 and Zone V3 (24= 10+8+6). Without
this approach, damaged trees for the Zone E will
be erroneously 32 as it is the union of Zone A and
Zone B and as the associated sum operation does
not consider that the damaged trees of the Zone 2
of A, Zone 1 of B are counted twice when aggre-
gating on the time dimension. Finally, the result
of the Roll-Up operator is shown in Table 1.
Let us use another spatial function: intersec-
tion. In the OLAP context, intersection (Shekar
et al., 2001) is a spatial aggregation function, but
it should be better to considered it as a spatial
disaggregation function following the geostatistic
approach (c.f. Sec. “Geographic Data” and “Geo-
statistic Models”).
Table 1. Roll-up on the time dimension
Year Phenomenon Aggregated Zone
1978 Fire Zone E
… … …
Figure 5. Aggregation of park’s zones on the time
dimension
In this case, the multidimensional query is:
“Where and how many trees have been damaged
by all fires during 1978?” (Query 2). The number
of trees cannot be calculated using the sum of the
averages because it is not representative of the
number of trees in the intersected zone. Therefore,
instead of applying sum, we apply a weighted
average on the area (Figure 6). Consequently,
the number of trees of Zone E is the same than
the one of Zone V2 (8). Damages trees are not
counted twice.
These two examples show that:
1. aggregation process has to take into account
overlapping geometries and
2. alphanumeric aggregation functions applied
to the descriptive attributes of geographic
measures depend on their semantics (∑:
measures that can be summed, ω: measures
that can be averaged, c: measures that can be
only counted), on used dimensions (as for
classical OLAP measures), and on spatial
function (as in geostatistic models). In other
Figure 6. Aggregation of park’s zones on the time
dimension
385
 A Multidimensional Model for Correct Aggregation of Geographic Measures
words, additive measures cannot be added
if a spatial disaggregation function is used.
Finally, the aggregation process of geographic
measures can be seen as a three-step process using
two alphanumeric aggregation functions (noted
φ and κ) for each alphanumeric attribute and a
spatial function (spatial aggregation or spatial
disaggregation) for the spatial attribute.
The three steps are:
• Step 1-Calculate disjoint geographic ob-
jects: Calculate the set of geographic ob-
jects whose geometries are obtained using
the intersection operator on the geographic
measures (i.e. Figure 3). Alphanumeric at-
tributes are calculated using user-defined
functions. Geographic objects with the
same geometry and coordinates could be
created.
• Step 2-Vertical Aggregation: Geographic
objects created at step 1 with the same ge-
ometry and coordinates are unified to create
one geographic object, by applying an ag-
gregation function (κ) to the alphanumeric
attributes (i.e. Figure 4). These aggrega-
tions have to be coherent with the seman-
tics of the measures and with the dimen-
sions used in the multidimensional query
(Semantic of the measures “∑, ω, c data”
cf. Sec. “Multidimensional Models”).
• Step 3-Horizontal Aggregation:
Aggregate the geometries of original geo-
graphic measures. Aggregate the alphanu-
meric attributes of the geographic objects
created at step 2 (i.e. Figures 5 and 6). The
alphanumeric aggregations (φ) have to be
coherent with the semantics of the measure
(∑, ω, c) and with the used spatial function.
It is important to underline that this process
can be applied also to lines, points, etc. To best of
our knowledge, no work deals with this problem.
386
As discussed in Section “Related Work”, exist-
ing spatio-multidimensional models (Abello et al.,
2006; Damiani & Spaccapietra, 2006; Pourabbas,
2001, Sampaio et al., 2006, Silva et al., 2008)
do not introduce any constraint on spatial and
alphanumeric functions.
Jensen et al. (2004) do not extend aggregation
constraints for spatial measures. Moreover, since
the model provides a symmetrical representation
of measures and dimensions and since it does
not support dimensions attribute, each attribute
of a geographic object has to be represented as a
dimension which implies aggregating attributes
separately. This means that the model cannot sup-
port the dependency of spatial and alphanumeric
functions in the aggregation process.
Only Pedersen & Tryfona, (2001) model the
dependency of spatial and alphanumeric functions,
but as described above, they restrict spatial func-
tions to union, discarding spatial disaggregation
functions. This limits the analysis capabilities of
the model because, as shown by our case study,
spatial disaggregation functions could be neces-
sary for spatial analysis.
CORRECT GEOGRAPHIC
MULTIDIMENSIONAL AGGREGATION
In this section we introduce the base concepts
of the logical multidimensional model GeoCube
(Bimonte et al., 2006) and its extension to cor-
rectly aggregate geographic measures.
An Overview of the GeoCube Model
Bimonte et al., (2006) present a logical multidi-
mensional model called GeoCube, which supports
measures and dimension members as geographic
objects. GeoCube allows the usage of a set of
spatial and/or alphanumeric attributes as one
single complex measure (geographic measures).
It also proposes user-defined aggregation func-
tions for each attribute of the geographic object
A Multidimensional Model for Correct Aggregation of Geographic Measures
representing the aggregated measure. GeoCube
also provides an algebra that redefines common
OLAP operators.
Data Model
The main concepts of multidimensional data model
are: Entity, Hierarchy and Base Cube. The con-
cepts of Entity Schema and Entity Instance permit
to represent indifferently the data of the analysis
universe: dimension members and measures. An
Entity is a set of attributes and functions used to
represent derived attributes (data calculated using
other data). Derived attributes are necessary to
model metric attributes (i.e. area, perimeter, etc.)
of geographic objects.
Entity Schemas and their instances are orga-
nized into hierarchies (Hierarchy Schema and
Hierarchy Instance). The Base Cube represents
the facts table.
In what follows, we only provide the defini-
tions that are necessary to describe the framework
we propose.
Definitions-Entity Schema and Entity Instance:
An Entity Schema Se is a tuple 〈a1, …an,
[F]〉 where:
◦◦ ai is an attribute defined on a domain
dom(ai)
◦◦ F, if it exists, is a tuple 〈f1,…fm〉 where
fi is a function defined on a sub-set of
attributes ar, …ak.
An Instance of an Entity Schema Se is a tuple
ti such as
• If F exists then ti = 〈val(a1), val(an),
val(b1),…val(bm)〉 where val(ai) ∈ dom(ai)
and val(bj) = fj(val(ar),…val(ak))
• If F does not exist, ti = 〈val(a1),…val(an)〉
where val(ai) ∈ dom(ai)
F permits to model derived measures and/or
dimension attributes.
Definition-Geographic Entity Schema: An
Entity Schema Se is a Geographic Entity
Schema if the domain of one attribute is a
set of spatial objects.
Example 1: In the case study presented previously,
the geographic measure representing the
zones is Szone = 〈geometry, nbDamagedTrees,
farea〉 where farea: dom(geometry) → N is a
function to calculate the area of a zone. An
instance of Szone is 〈pt01, 20, 110〉 (Zone A)
(Figure 2).
Entities are organized in hierarchies thanks
to the concepts of Hierarchy Schema and Hier-
archy Instance. A level of a hierarchy is an Entity
Schema and a member is an Entity Instance. The
Hierarchy Schema organizes levels into a lattice.
Thanks to the Hierarchy Instance, the members’
levels are organized in a tree structure. The root
is an instance of the Entity Schema which repre-
sents the top level of the lattice represented by
the Hierarchy Schema. Leafs are the instances of
the Entity Schema which represents the bottom
level of the lattice.
This definition allows modeling non-balanced
and non-strict hierarchies, which are necessary
for spatio-multidimensional applications (Ma-
linowski & Zimányi, 2005).
Definition-Hierarchy: A Hierarchy Schema is a
tuple Hh = 〈L h, ⌊h, ⌈h, ⇞h〉 where:
◦◦ Lh is a set of Entity Schemas,
◦◦ ⌊h and ⌈h are two Entity Schema and
⌈h contains one instance (‘all’),
◦◦ /h is a partial order defined on the
levels of the hierarchy (L h ∪ ⌊h ∪ ⌈h)
and /h is a lattice where ⌊h and ⌈h are
respectively the bottom and the top
levels of the order.
We call Entity Schema levels (L(Hh)) the entities
belonging to the set {L h ∪ ⌊h ∪ ⌈h}. An Instance
of a Hierarchy Hh is a partial order ↑h defined on
L(Hh), such as:
387
 A Multidimensional Model for Correct Aggregation of Geographic Measures
• if ti ↑h tj then Si /h Sj, where ti ∈ I(Si) and
tj ∈ I(Sj)
• ∀ ti not belonging to the top level, then ∃ tj
such as ti ↑h tj
• ∀ ti not belonging to the bottom level, ∃ tj
such as tj ↑h ti
Schema and data of the spatio-multidimension-
al application are represented by Base Cube. Base
Cube Schema defines dimensions (Hierarchies)
and measures (e.g. the spatio-multidimensional
application schema of Figure 1). The instance of
the Base Cube represents facts table data (e.g. data
of Figure 2). In our approach, dimensions and mea-
sures are (geographic) objects, for example phe-
nomena and zones. Then, our model must define
measures in the same way as dimensions levels.
Following the approach of Vassiliadis (1998), we
define a Base Cube as a tuple of Hierarchies and
a boolean function which represents the tuples
of the facts table. Here, the bottom levels of the
hierarchies are (geographic) objects. They can be
used as dimensions and measures. This definition
allows defining measures as (geographic) objects,
and not as numerical value.
Example 2: The multidimensional model of Fig-
ure 1 is represented by the Base Cube Schema
BCnaturalrisks = 〈Hnatural_phenomena, Htime, Hzone, δ〉
where Hnatural_phenomena, Htime and Hzone are the
hierarchies representing the phenomena,
time and zones. δ: I(Sphenomen)×I(Sday)×I(Szone)
is a boolean function defined on the bottom
levels of the hierarchies. The instance of BC-
naturalrisks
represents the facts table (Figure 2).
In this application, each bottom level of a
hierarchy (object and/or geographic object) can
be used as measure.
Multidimensional Query Model
Base Cube represents the basic cuboid, which is
represented by measures values associated with
388
the most detailed levels of all hierarchies. These
measures are not aggregated. Then, in order to
represent a multidimensional query (i.e. cuboid),
we introduce the concepts of Aggregation Mode
and View. A multidimensional query defines the
dimension levels used (e.g. “Type” and “Year”),
the (geographic) object used as measure (“Zone”)
and a set of aggregation functions (Aggregation
Mode) to aggregate its attributes (e.g. union for
geometry and sum for number of trees) (Figure
7). For simplicity, we provide only the definition
of the Aggregation Mode and we give a example
of View.
Aggregation Mode defines a function for each
non-derived attribute of the (Geographic) Entity
representing the measure. The result of the ag-
gregation is another (Geographic) Entity whose
attributes values are calculated using these func-
tions.
Definition-Aggregation Mode: An Aggregation
Mode Θk is a tuple 〈Sa, Sb, Φ〉 where:
◦◦ Sa is an Entity Schema 〈a1,…am,
[Fa]〉 (the detailed measure)
◦◦ Sb is an Entity Schema 〈b1,…bp,
[Fb]〉 (the aggregated measure)
◦◦ Φ a set of p ad-hoc aggregation func-
tions ϕi
An Aggregation Mode defined on geographic
objects (geographic measures) is called Geo-
graphic Aggregation Mode.
Definition-Geographic Aggregation Mode:
An Aggregation Mode Θk =〈Sa, Sb, Φ〉 is a
Geographic Aggregation Mode if Sa and Sb
are Geographic Entity Schemas.
Example 3: Let the aggregation of the geo-
graphic Measures Szone be represented by
the Entity Schema Szone_agg = 〈geometry,
nbDamagedTrees, F〉.
The area is a metric attribute of the damaged
zone and so it can be directly and univocally
A Multidimensional Model for Correct Aggregation of Geographic Measures
Figure 7. Geographic aggregation mode: Zones of a park
calculated using the geometry. For this reason, it
has to be considered as a derived measure. Then,
an example of Geographic Aggregation Mode for
the zones is Θzone= 〈Szone, Szone_agg, Φunion〉 where
Φunion is the set of functions:
1. ϕ g e o m e t r y : d o m ( S z o n e . g e o m e t r y ) →
n
dom(Szone_agg.geometry), i.e. the Geometric
Union
2. ϕnbtrees:dom(Szone.geometry, Szone.nbDam-
agedTrees) n → dom(S zone_agg .nbDam-
agedTrees), i.e. the sum
An example of aggregation of two geographic
measures (Zone 3 and Zone 4) using Θzone is shown
in Figure 7. Szone is a schema entity with attributes
a1 (the number of trees damaged by fires), a2 (the
geometry of the region) and a a3 area attribute a3
(a3 is derived from the geometry). Zone 3 and Zone
4 in the figure are two Szone entities. Szone-aggreg is
another schema entity, which represents the result
of aggregating Zone 3 and Zone 4 (as shown in
the dark grey area in the right panel of the figure).
The attribute b1 in Szone-aggreg is the result of apply-
ing the operator ϕ1 (sum) to attribute a1, and the
attribute b2 is the result of applying the operator
ϕ2 (union) to attribute a2.
GeoCube introduces the concept of View to
represent the result of a multidimensional query.
A View is a tuple of a Base Cube (the facts table),
the set of Entity Schemas used for the multidimen-
sional query, and an Aggregation Mode which
defines the Entity used as measure and aggregation
functions to aggregate it.
Example 4: The multidimensional query Where
and how many trees (which zones) have
been damaged by some fires during 1978?”
(Query 1, cf. Sec. “Research Motivations”)
is represented by the View Vzones-year = 〈BC-
naturalrisks
, 〈Sphenomenon, Syear〉, Θzone, γ〉. This query
aggregates geographic measures along the
time dimenson. We note that the View is
composed of the Base Cube, the levels
used in the query Sphenomenon (phenomena)
and Sye(years) and the Aggregation Mode
Θzone which defines Szone as measure and
establishes how to aggregate it. The instance
of Vzones-year is shown in Table 1.
Correct Geographic Aggregation
In order to formalize the constraints that ensure
the correct aggregation of geographic measures
according to the algorithm described in the Sec-
tion “Research Motivations”, we formalize the
semantics of alphanumeric attributes of geo-
graphic measures and the dependency of spatial
and alphanumeric functions, and then we introduce
the constraints.
389
 A Multidimensional Model for Correct Aggregation of Geographic Measures
Semantics of Alphanumeric Attributes
of Geographic Measures
Following the approach to ensure correct aggrega-
tion, described in the Section “Multidimensional
Models”, we introduce the functions Dimen-
sionsMeasure and SemanticsMeasure.
They classify aggregation functions of at-
tributes of geographic objects in three different
groups, called aggregation types: AF∑, AFω and
AFc, where AF∑ > AFω > AFc, “>” being a total
order. Functions classified as AF∑ can be applied
to measures that can be summed, AFω are func-
tions that can be applied to measures that can be
averaged, and AFc are functions that can be applied
to measures that can be only counted.
SemanticsMeasure takes into account the
semantics of attributes. It takes as input an al-
phanumeric attribute and returns an aggregation
type (AF∑, AFω, AFc). For example, functions that
can be applied to the number of trees are AF∑, as
number of trees can be summed.
Definition-SemanticsMeasure: For each Entity
Schema Se, we assume a function Seman-
ticsType: A(Se) → {AF∑, AFω, AFc} which
returns the aggregation type.
DimensionsMeasure takes into account the se-
mantics of attributes, and used dimensions levels.
It takes as input an alphanumeric attribute and the
dimensions levels of the View, and it returns an
aggregation type (AF∑, AFω or AFc). For example,
functions that can be applied to the number of trees
aggregating on the time dimension are AFω, as the
number of trees cannot be summed on the time
dimension (some trees must not be counted twice.
Definition-DimensionsMeasure: For each Entity
Schema Se and set of levels of a View Ve,
we assume a function DimensionsMeasure:
A(Se)× L(Ve) → {AF∑, AFω, AFc} which
returns the aggregation type.
390
Example 5: In our example, SemanticsMeasure(
nbDamagedTrees) = AF∑, because the sum
of number of trees makes sense, and Dim
ensionsMeasure(nbDamagedTrees, Syear,
Sphenomenon)= AFω, because number of trees
is not additive on the time dimension (the
level Syear, which is not the most detailed
level of the “Time” dimension, is used in
the multidimensional query).
Dependency of Spatial and
Alphanumeric Functions
In order to model dependency of spatial and
alphanumeric functions, we introduce a function
SpatialType, which takes as inputs the type of
spatial function (AFU, AFΩ) and the semantics of
the alphanumeric attribute (SemanticsMeasure),
and returns the aggregation type (AF∑, AFω, or
AFc) that can safely be applied to the alphanumeric
attribute. Let us note spatial aggregation functions
as AFU, and spatial disaggregation functions as
AFΩ. Then, spatial functions classified as AFU al-
lows summing additive data (e.g. number of trees)
because they preserve the set of input geometries
in the result (i.e. union, convex hull, etc.). AFΩ do
not allow summing data as they leave out some
spatial objects (i.e. centroid, intersection, etc.).
Definition-SpatialType function: We define a
function SpatialType: {AFU, AFΩ} × { AF∑,
AFω, AFc} → { AF∑, AFω, AFc } as defined
in Table 2.
The function SpatialType models the depen-
dency of spatial and alphanumeric functions as
defined by the Step 3 of our algorithm. In par-
Table 2. SpatialType function
SpatialType AF∑ AFω AFc
AFU AF∑ AFω AFc
AFΩ AFω AFω AFc
A Multidimensional Model for Correct Aggregation of Geographic Measures
ticular, when using spatial aggregations (AFU), it
is possible to sum additive alphanumeric attributes
(AF ∑ ) values of geographic measure
(SpatialType(AFU, AF∑) = AF∑). For example,
when using the union, it is possible to sum num-
ber of trees as it is additive (AF ∑ ):
SpatialType(UNION, AF∑) = AF∑ (see Figure 5)..
When using spatial disaggregations (AFΩ), it is
not possible to sum alphanumeric attributes values
of geographic measures, but it should be possible
to use average (AF ω ) or count (AF c )
(SpatialType(AFΩ, AF∑)= AFω). For example,
when using intersection, it is not possible to use
sum, even if number of trees is additive (AF∑):
SpatialType(INTERSECTION, AF∑) = AFω (see
Figure 6)..
Finally, if the alphanumeric attribute is not
additive (AFω or AFc) and spatial aggregation is
used, then it is not possible to add alphanumeric
attribute values (SpatialType(AFU, AFω) = AFω
and SpatialType(AFU, AFc)= AFc).
Correct Geographic Aggregation Mode
Before defining the Correct Geographic Aggrega-
tion Mode, we need to introduce the concept of
overlay. The GIS operator overlay takes as inputs
two maps and overlays them one on the top of the
other to form a new map. The Union Overlay is an
operator that takes as inputs 2 maps and returns a
map whose geometries are the set of all disjoint
geometries obtained by applying the topological
intersection among all features. Figure 8 shows an
example of overlay of two maps whose geometries
are polygons.
Figure 8. Union overlay
Using the definitions previously given, we
formalize the three steps of our algorithm by
defining two functions (DisjGeoObjects and
OverlayGeoObjects) and two constraints (Vertical
Aggregation Constraint and Horizontal Aggrega-
tion Constraint) for the Geographic Aggregation
Mode.
These two constraints allow the Geographic
Aggregation Mode ensuring a correct geographic
aggregation of geographic measures.
In particular, each alphanumeric aggregation
function ϕ of the Geographic Aggregation Mode
is decomposed in two functions: κ and φ used
to calculate attributes values of the geographic
objects created at Step 2 and 3 respectively.
Then, the two constraints are defined on κ and φ
according to the requirements of the Vertical and
the Horizontal steps respectively.
The Vertical Aggregation Constraint defines
correct functions for κ considering disjoint
geographic measures using the function Dis-
jGeoObjects (see Figure 3), and taking into
account semantics and used dimensions (i.e. Di-
mensionsMeasure) of the alphanumeric attributes
(e.g. number of trees cannot be summed on the
time dimension).
For example, in order to calculate geographic
objects of Figure 3 resulting from an aggregation
on the time dimension, the Vertical Aggrega-
tion Constraint does not allow using the sum
(κnbDamagedTrees = AVG) because number of trees is
not additive on the time dimension (κnbDamagedTrees∈
AFω = Min(DimensionType(nbDamagedTrees,
Syear,Stype), where DimensionType(nbDamagedT
rees,Syear,Stype)= AFω).
391
 A Multidimensional Model for Correct Aggregation of Geographic Measures
The Horizontal Aggregation Constraint de-
fines correct functions for φ taking into account the
semantics of the alphanumeric attributes thanks to
the function SemanticsType (e.g.. number of trees
can be summed), and the dependency between
spatial and alphanumeric functions thanks to the
SpatialType function (when using intersection, it
is not possible to sum the number of trees).
For example, in order to aggregate geographic
objects resulting from Vertical Aggregation Step of
Figure 3, the Horizontal Aggregation Constraint
does not allow using the sum for number of trees.
Therefore, φnbDamagedTrees = Weighted Average on
surface because a spatial disaggregation (inter-
section) is used for geometry (Figure 6).Indeed,
sum is not representative of number of trees in
the intersection region (φnbDamagedTrees ∈ AFω =
SpatialType(AFΩ, Min(SemanticType(nbDama
gedTrees))).
By this way, user chooses the functions φ and
κ, and the model avoids building any non-sense
aggregated geographic measures.
In the following, we formally introduce the
Correct Geographic Aggregation Mode and
then the functions DisjGeoObjects and Overlay-
GeoObject.
Definition-Correct Geographic Aggregation
Mode: Let:
◦◦ Ve be a View (See example 4:
Vzones-year = 〈BCnaturalrisks,
〈Sphenomenon, Syear〉, Θzone, γ〉)
◦◦ t1 …,tn be instances of the Geographic
Entity which must be aggregated: in-
put instances (See Zone A and Zone
B, Figure 2)
◦◦ tiD1, … tiDr be the instances calcu-
lated using DisjGeoObjects on input
instances: vertical instances (See
Zone 1 of A, Zone 2 of A, Zone 1 of B
and Zone 2 of B, Figure 3)
◦◦ td1 …,tdm be the instances calcu-
lated using OverlayGeoObjects on
input instances and tdi have the same
392
geometry and coordinates than tiD1,
… tiDr of the vertical instances (See
Zone V1, Zone V2 and Zone V3,
Figure 4)
◦◦ SpatialAggregationType be the spa-
tial function type {AFU, AFΩ}
◦◦ geomAggregate be the result of the
spatial aggregation on the geom-
etries of t1 …,tn
◦◦ The alphanumeric aggregation func-
tion ϕi of Φe be defined by means of:
◦◦ φi (tdi.am, … tdi.al, tdi.geom)
(Horizontal Aggregation)
◦◦ κi (tiDi.am,…,tiDi.al) (Vertical
Aggregation) such as:
1. ϕi = φ1
2. tdi.ai=κ1r(tiDi.am,…,tiDi.al)
then the Geographic Aggregation Mode Φe is
correct if are respected the following constraints:
1. Vertical Aggregation Constraint: κ i ∈
Min(DimensionsMeasure(am, Sl1,…Slm),…
,DimensionsMeasure(al, Sl1, …, Slm))
2. Horizontal Aggregation Constraint:
φ i ∈ SpatialType(SpatialAggregation
Type, Min(SemanticsMeasure(a m ),…,
SemanticsMeasure(al))
Definition-DisjGeoObjects: DisjGeoObjects is a
function which takes as inputs n Geographic
Entity Instances and returns l ≥ n Geographic
Entity Instances whose geometries are
obtained using the geometric intersection
operator.
Figure 3 shows the results of the DisjGeoOb-
jects function on geographic objects of Figure 2.
Definition-OverlayGeoObjects: Overlay-
GeoObjects is a function which takes as
inputs n Geographic Entity Instances repre-
senting n maps and returns m≥ n Geographic
Entity Instances whose geometries are
obtained using the Union Overlay.
A Multidimensional Model for Correct Aggregation of Geographic Measures
Figure 4 shows the results of the Overlay-
GeoObjects function on geographic objects of
Figure 3.
We present now the two Geographic Aggre-
gation Modes for our case study: the first uses
union and the second intersection. We show how
the Geographic Aggregation Mode constraints
oblige the user to use sum only in the first case.
Moreover, these examples show how the user is
forced to aggregate the number of trees consider-
ing the semantics of the attributes and the used
dimensions.
Example 6: Let us consider the query “Where
and how many trees have been damaged by
some fires during 1978?” (Query 1, cf. Sec
“Research Motivations”). We define a View
with a Correct Geographic Aggregation
Mode, which uses average for the Vertical
Aggregation and sum and union for the
Horizontal Aggregation (Figure 9).
Formally, let Vzones-year = 〈BCnaturalrisks, 〈Sphenomenon,
Syear〉, Θzone, γ〉 represent the Query 1 where:
• DimensionsMeasure(nbDamagedTrees,
Syear, Stype) = AFω
• SemanticsMeasure(nbDamagedTrees) =
AF∑
Figure 9. Correct Geographic Aggregation Mode on Time dimension and using spatial union
• Geographic Aggregation Mode Θzone =
〈Szone, Szone_agg, Φunion〉 where Φunion is (Figure
5)
1. ϕ geometry: dom(S zone.geometry) n →
dom(Szone_agg.geometry) is Union
• ϕgeometry is AFU (spatial aggregation)
2. ϕnbDamagedtrees: dom(Szone.geometry, Szone.
nbDamagedTrees)n → dom(Szone_agg.nb-
DamagedTrees)= φnbDamagedTrees where:
• φnbDamagedTrees is SUM
• κnbDamagedTrees = Average
The multidimensional query Vzones-year defines
a Correct Geographic Aggregation Mode because
the Geographic Aggregation Mode Constraints
are satisfied:
1. κnbDamagedTrees∈ AFω =Min(DimensionType(n
bDamagedTrees,Syear,Stype)
(It is not possible to apply the sum operator
to numbers of trees because this mea-
sure is not additive on time dimension)
2. φnbDamagedTrees∈ AF∑ =SpatialType(AFU,Min(
SemanticsType(nbDamagedTrees)))
(It is possible to apply the sum to the num-
bers of trees for the geographic objects
resulting from VerticalAggregation
because the spatial function is union
and it is a spatial aggregation)
393
 A Multidimensional Model for Correct Aggregation of Geographic Measures
The Geographic Aggregation Mode Θzone uses
spatial Union (ϕgeometry) to aggregate geometry.
The aggregation of the number of damaged trees
is performed using the AVG for the vertical ag-
gregation (κnbDamagedTrees), as the number of damaged
trees is not additive on time dimension (Dimensio
nsMeasure(nbDamagedTrees, Syear, Stype) = AFω),
and the SUM (φnbDamagedTrees) for the horizontal ag-
gregation, as ϕgeometry is a spatial aggregation (AFU).
Example 7: Let us suppose now we want to
answer to the query “Where and how many
trees have been damaged by all fires during
1978?” (Query 2, cf. Sec “Research Moti-
vations”). We define a View with a Correct
Geographic Aggregation Mode, which uses
an interpolation function and intersection for
the Horizontal Aggregation, and average for
the Vertical Aggregation (Figure 10).
Let Θzone= 〈Szone, Szone_agg, Φintersect〉 (Query 2)
where Φintersect is:
1. ϕ g e o m e t r y : d o m ( S z o n e . g e o m e t r y ) n →
dom(Szone_agg.geometry) is Intersection
◦◦ ϕgeometry is AFΩ (spatial disaggregation)
2. ϕ nbDamagedtrees: dom(S zone.geometry, S zone.
nbDamagedTrees)n → dom(Szone_agg.nbDam-
agedTrees) = φnbTrees where:
Figure 10. Correct geographic aggregation mode on time dimension and using spatial intersection
394
◦◦ φnbDamagedTrees is Weighted Average on
surface
◦◦ κnbDamagedTrees = Average
The multidimensional query Vzones-year uses a
Correct Geographic Aggregation Mode because
the Geographic Aggregation Mode constraints
are satisfied:
1. κnbDamagedTrees ∈ AFω = Min(DimensionType(
nbDamagedTrees, Syear, Stype))
(The sum cannot be applied to numbers of
trees because this measure is not additive
on time dimension)
2. φnbDamagedTrees ∈ AFω = SpatialType(AFΩ, Mi
n(SemanticType(nbDamagedTrees)))
(The sum cannot be applied to numbers of
trees of the geographic objects resulting
from VerticalAggregation because the spatial
function is intersection and it is a spatial
disaggregation).
The Geographic Aggregation Mode Θzone uses
spatial Intersection (ϕgeometry) to aggregate geom-
etry. The aggregation of the number of damaged
trees is performed using the AVG for the vertical
aggregation (κnbDamagedTrees), as the number of dam-
aged trees are not additive on time dimension (D
imensionsMeasure(nbDamagedTrees, Syear, Stype)
= AFω), and the Weighted Average on surface
A Multidimensional Model for Correct Aggregation of Geographic Measures
(φnbDamagedTrees) for the horizontal aggregation, as
ϕgeometry is a spatial disaggregation (AFΩ).
We have presented and illustrated the approach
we propose for Correct Geographic Aggregation
Mode. To complete our case study, we have to
recall that no aggregation function is provided to
get the value of the attribute representing the area
of the damaged zone. Indeed, the area is a derived
measure calculated using the geometry resulting
from the spatial aggregation of the component
geographic objects.
CONCLUSION AND DISCUSSION
Spatial OLAP refers to the integration of spatial
data in OLAP. Correct aggregation is crucial
in multidimensional analysis. In this paper, we
provide an overview of solutions for aggrega-
tion of geographic objects in multidimensional,
geostatistic, GIS and SOLAP models. By intro-
ducing a case study concerning the monitoring of
natural phenomena, we define the aggregation of
geographic measures as a –three-step process in
which two constraints on aggregation functions are
defined. They extend classical OLAP aggregation
constraints with dependence between spatial and
alphanumeric aggregation functions applied to the
attributes of geographic measures. We present an
extension of the logical multidimensional model
GeoCube (Bimonte et al., 2006), which formal-
izes our approach. Alphanumeric aggregation
functions are decomposed into two aggregation
functions and two constraints are applied to them to
take into account the additivity of the measures and
the dependency of spatial and alphanumeric ag-
gregation functions. By this way, GeoCube ensures
the correct aggregation of geographic measures.
This requirement is not supported by existing
spatio-multidimensional model. Indeed, most of
SOLAP models do not provide any aggregation
constraint for aggregation functions. Jensen et al.
(2004) define aggregation constraints, but they
reduce spatial measures to classical data without
taking into account the spatial component of the
geographic information. Finally, only Pedersen
& Tryfona (2001) provide a model that explicitly
supports aggregation constraints for spatial and
alphanumeric data. The model supports overlap-
ping geometries and aggregation constraints as
well as our 3-step framework. However, it sup-
ports only partially the dependency of spatial
and alphanumeric functions as it does not allow
disaggregation functions. We have addressed this
issue in this paper
Currently, we are working on the implemen-
tation of our framework in the SOLAP system
GeWOlap (Bimonte et al., 2006b). We are imple-
menting the three steps of our framework using
PL/SQL functions and user-defined aggregation
functions provided by Oracle. Semantics of mea-
sures (valid aggregation functions) are defined
in XML files representing the multidimensional
application in the ROLAP Server Mondrian.
Semantics of measures have to be parsed before
any OLAP query computation in order to verify
dependencies between spatial and alphanumeric
functions. The main issues of this approach deal
with performances in large spatial data ware-
houses. Developing ad-hoc pre-aggregation and
indexing techniques for geographic measures are
our future work.
ACKNOWLEDGMENT
Authors wish to thank Pr. Pierre Dumolard for
precious discussions and material on GIS and
geostatistics.
REFERENCES
Abelló, A., Samos, J., & Saltor, F. (2006). YAM2:
a multidimensional conceptual model extending
UML. Information Systems, 31(6), 541–567.
doi:10.1016/j.is.2004.12.002
395
 A Multidimensional Model for Correct Aggregation of Geographic Measures

Ahmed, T., & Miquel, M. (2005). Multidimen- Degrene, J., & Salge, F. (1997). Les systèmes
sional Structures Dedicated to Continuous Spatio- d’information géographique, (2nd Ed.). Paris:
temporal Phenomena. In Proceedings of the 22th Presses Universitaires de France.
BNCOD (pp. 29-40). Berlin: Springer.
Egenhofer, M., & Frank, A. U. (1986). Connec-
Anselin, L. (1989). What Is Special about Spatial tion between Local and Regional: Additional
Data? Alternative Perspectives on Spatial Data Intelligence Needed. In Proceedings of the 18th
Analysis, (Technical Report). Santa Barbara, CA: International Congress of figure
National Center for Geographic Information and
Egenhofer, M., & Frank, A. U. (1992). Object-
Analysis.
Oriented Modeling for GIS. URISA Journal,
Bédard, Y., Merrett, T., & Han, J. (2001). Funda- 4(2), 3–19.
ments of Spatial Data Warehousing for Geographic
Erwig, M., & Schneider, M. (2000). Formaliza-
Knowledge Discovery. In Geographic Data
tion of advanced map operations. In Proceedings
Mining and Knowledge Discovery (pp. 53-73).
of the 9th Int. Symp. on Spatial Data Handling
London: Taylor & Francis.
(pp. 3-17).
Bimonte, S., Tchounikine, A., & Miquel, M.
Escribano, A., Gomez, L., Kuijpers, B., & Vais-
(2006). GeoCube, a Multidimensional Model
man, A. (2007). Piet: a GIS-OLAP implementa-
and Navigation Operators Handling Complex
tion. In Proceedings of the ACM 10th International
Measures: Application in Spatial OLAP. In Pro-
Workshop on Data Warehousing and OLAP (pp.
ceedings of the 4th ADVIS (pp. 100-109). Berlin:
73-80). New York: ACM Press.
Springer-Verlag.
Fidalgo, R., Times, V., Silva, J., & Souza, F. (2004).
Bimonte, S., Wehrle, P., Tchounikine, A., &
GeoDWFrame: A Framework for Guiding the
Miquel, M. (2006). GeWOlap: A Web Based Spa-
Design of Geographical Dimensional Schemas.
tial OLAP Proposal. In Meersman, R., Tari, Z., &
In Proceedings of 6th International Conference
Herrero, P. (Eds.), In Proceedings of the Workshop
on Data Warehousing and Knowledge Discovery
on Semantic-Based Geographical Information
(pp. 26-37). Berlin: Springer.
Systems (pp. 1596-1605). Berlin: Springer.
Franklin, C. (1992). An Introduction to Geographic
Charre, J., Dumolard, P., & Le Berre, I. (1997).
Information Systems: Linking Maps to databases.
Initiation aux pratiques statistiques en geographie.
Database, 15(2), 13–21.
Paris: Masson.
Guting, R. (1998). Geo-Relational Algebra: A
Chrisman, N. (1974). Attributes of geographic
model and query language for geometric data-
entities. Technical Report, Harvard Laboratory
base systems. In Proceedings of the EDBT (pp.
for Computer Graphics, Cambridge, MA.
506-527).
Chrisman,N.(1998).RethinkingLevelsofMeasure-
Güting, R., & Schneider, M. (1995). Realm-based
ment in Cartography. Cartography and GIS, 25(4),
spatial data types: the ROSE algebra. The VLDB
231–242. doi:10.1559/152304098782383043
Journal, 4(2), 243–286. doi:10.1007/BF01237921
Damiani, M., & Spaccapietra, S. (2006). Spatial
Data Warehouse Modeling. In Processing and
Managing Complex Data for Decision Support
(pp. 1-27). Hershey, PA: IGP.

396
A Multidimensional Model for Correct Aggregation of Geographic Measures
Han, J., Stefanovic, N., & Koperski, K. (1998).
Selective Materialization: An Efficient Method for
Spatial Data Cube Construction. In Proceedings
of 6th PAKDD Conference (pp. 144-158). Berlin:
Springer.
Horner, J., Song, I., & Chen, P. (2004). An Analysis
of Additivity in OLAP Systems. In Proceedings
of the 7th DOLAP (pp. 83-91). New York: ACM
Press.
Jensen, C., Kligys, A., Pedersen, T., & Timko,
I. (2004). Multidimensional data modeling for
location-based services. The VLDB Journal, 13(1),
1–21. doi:10.1007/s00778-003-0091-3
Kimball, R. (1996). The Data Warehouse Toolkit.
New York: John Wiley.
Lehner, W. (1998). Modeling large scale OLAP
scenarios. In Proceedings of the 6th EDBT (pp.
153–167).
Longley, P., Goodchild, M., Maguire, D., & Rhind,
D. (2001). Geographic Information Systems and
Science. New York: John Wiley & Sons.
Lopez, I., & Snodgrass, R. (2005). Spatiotemporal
aggregate computation: A survey. IEEE Transac-
tions on Knowledge and Data Engineering, 17(2),
271–286. doi:10.1109/TKDE.2005.34
Malinowski, E., & Zimányi, E. (2004). Represent-
ing spatiality in a conceptual multidimensional
model. In Proceedings of the 12th ACM GIS (pp.
12-22). New York: ACM Press.
Malinowski, E., & Zimányi, E. (2005). Spatial
Hierarchies and Topological Relationships in
SpatialMultiDimER model. In Proceedings of the
22th British National Conference on Databases
(pp. 17-22). Berlin: Springer.
Pedersen, T. B., Jensen, C., & Dyreson, C. (2001).
A foundation for capturing and querying complex
multidimensional data. Journal of Information
Systems, 26(5), 383–423. doi:10.1016/S0306-
4379(01)00023-0
Pedersen, T. B., & Tryfona, N. (2001). Pre-aggre-
gation in Spatial DataWarehouses. In Proceedings
of the 7th International Symposium on Spatial and
Temporal Databases (pp. 460-478).
Pourabbas, E. (2003). Cooperation with Geo-
graphic Databases. In Multidimensional data-
bases: problems and solutions (pp. 393-432).
Hershey, PA: IGP
Rafanelli, M., & Ricci, F. (1983). Proposal of a
logical model for statistical databases. In Pro-
ceedings of the 2nd International Workshop on
Statistical and Scientific Database Management,
(pp. 264-272).
Rao, F., Zhang, L., Yu, X., & Li, Y. & Chen, Y.
(2003). Spatial hierarchy and OLAP-favored
search in spatial data warehouse. In Proceedings
of the 6th ACM International Workshop on Data
Warehousing and OLAP (pp. 48-55). New York:
ACM Press.
Rigaux, P., & Scholl, M. (1995). Multi-Scale
Partitions: Application to Spatial and Statistical
Databases. In Proceedings of the 4th International
Symposium on Advances in Spatial Databases,
(pp. 170-183). Berlin: Springer-Verlag
Rigaux, P., Scholl, M., & Voisard, A. (2002).
Spatial databases with applications to Gis. New
York: Academic Press.
Rivest, S., Bédard, Y., Proulx, M., Nadeaum, M.,
Hubert, F., & Pastor, J. (2005). SOLAP: Merging
Business Intelligence with Geospatial Technology
for Interactive Spatio-Temporal Exploration and
Analysis of Data. Journal of International Society
for Photogrammetry and Remote Sensing, 60(1),
17–33. doi:10.1016/j.isprsjprs.2005.10.002
Sampaio, M. C., Sousa, A. G., & Baptista, C.
(2006) Towards a logical multidimensional model
for spatial data warehousing and olap. In Proceed-
ings of the 9th ACM international workshop on
Data warehousing and OLAP (pp. 83-90). New
York: ACM Press.
397
 A Multidimensional Model for Correct Aggregation of Geographic Measures
Shekar, S., Lu, C., Tan, X., Chang, S., & Vatsrai,
R. (2001). Map Cube: A Visualization Tool for
Spatial Data Warehouses. In Geographic Data
Mining and Knowledge Discovery (pp. 74-90).
London: Taylor & Francis
Silva, J., Times, V., Salgado, A., Souza, C., Fidalgo,
R., & Oliveira, A. (2008). A set of aggregation
functions for spatial measures. In Proceedings of
the ACM 14th International Workshop on Data
Warehousing and OLAP (pp. 25-32). New York:
ACM Press.
Stefanovic, N., Han, J., & Koperski, K. (2000).
Object-Based Selective Materialization for Ef-
ficient Implementation of Spatial Data Cubes.
IEEE TKDE, 12(6), 938–958.
Tao, Y., & Papadias, D. (2005). Historical
spatio-temporal aggregation. ACM Transac-
tions on Information Systems, 23(1), 61–102.
doi:10.1145/1055709.1055713
Trujillo, J., Palomar, M., & Gómez, J. (2000). An
Object Oriented Approach to Multidimensional
Databases & OLAP Operations. International
Journal of Computer and Information Science,
2(3), 75–85.
Vassiliadis (1998). Modeling Multidimensional
Databases, Cubes and Cube Operations. In Pro-
ceedings of 10th International Conference on Sci-
entific and Statistical Database Management (pp.
53-62). Alamitos, CA: IEEE Computer Society.
Voisard, A., & David, B. (2002). A Database
Perspective on Geospatial Data Modeling. IEEE
TKDE, 14(2), 226–243.
This work was previously published in Evolving Application Domains of Data Warehousing and Mining: Trends and Solu-
tions, edited by Pedro Nuno San-Banto Furtado, pp. 162-183, copyright 2010 by Information Science Reference (an imprint
of IGI Global).
398
Weibel, R., & Dutton, G. (2001). Generalizing
Spatial Data and Dealing with Multiple Rep-
resentations. Geographic Information Systems
and Science. New York: John Wiley & Sons (pp.
125-155)
Zhang, D., & Tsotras, V. (2005). Optimizing spa-
tial Min/Max aggregations. The VLDB Journal,
14(3), 170–181. doi:10.1007/s00778-004-0142-4
KEY TERMS AND DEFINITIONS
Geographic Information System: Informa-
tion system for capturing, storing, analyzing,
managing and presenting data that are spatially
referenced
Multidimensional Model: Conceptual model
for the multidimensional analysis of huge datasets,
based on the concepts of dimensions, facts and
measures.
Spatial Data Warehouse: Subject-oriented,
non volatile, time variant and integrated reposi-
tory of spatial data that is designed to facilitate
reporting and spatial analysis.
Spatial Decision Support System: Interac-
tive, computer-based system designed to support
a user or group of users in achieving a higher
effectiveness of decision making while solving a
semi-structured spatial problem
Spatial OLAP: Visual platform built espe-
cially to support rapid and easy spatio-temporal
analysis and exploration of data following a mul-
tidimensional approach comprised of aggregation
levels available in cartographic displays as well
as in tabular and diagram displays.
 399

Chapter 24
Location Privacy in
Automotive Telematics
Muhammad Usman Iqbal
University of New South Wales, Australia

Samsung Lim
University of New South Wales, Australia

ABSTRACT
Over the past few decades, the technologies of mobile communication, positioning, and computing have
gradually converged. The automobile has been a natural platform for this convergence where satellite-
based positioning, wireless communication and on-board computing work in tandem offering various
services to motorists. While there are many opportunities with these novel services, significant risks to
the location privacy of motorists also exist as a result of the fast-paced technological evolution. These
risks must be confronted if trust and confidence are to prevail between motorists and service providers.
This chapter provides an overview of the current situation of location privacy in automotive telematics
by exploring possible abuses and existing approaches to curb these abuses followed by a discussion of
possible privacy-strengthening measures.

INTRODUCTION filter. These services can be triggered by an event,

The proliferation of location-aware computing
devices promises an array of “quality-of-life
enhancing” applications. These services include
in-car navigation, roadside assistance, infotain-
ment, emergency response services, vehicle
diagnostics and prognostics. The key idea is to
provide services using “location” as a geographic
DOI: 10.4018/978-1-4666-2038-4.ch024
for example, the location of the vehicle can be
transmitted to an emergency response center on
deployment of air bags. Some services can be
explicitly requested by the driver, for example,
in-car navigation or road side assistance. While
other applications can be quietly running at all
times, passing on real-time information of the
vehicle’s movements such as Global Positioning
System (GPS) enabled Pay-As-You-Drive (PAYD)
insurance (Grush, 2005).

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

Although location data is critical to the opera-
tion of such applications, there is a precarious
balance between the necessary dissemination of
location information and the potential for abuse
of this private information. Spatio-temporal (loca-
tion in time) information continuously monitored
(and logged) about the places a person visits can
reveal a lot about one’s persona. Given the current
capabilities of inference by combining disparate
sources of information, a lot can be inferred about
an individual. These derived profiles can then
be used to make judgments about a person or
used for unsolicited marketing by location-based
marketers. Orwell (1949), in his criticism against
totalitarianism, would have most likely referred
to these “Small Brothers” (location-based retail
marketers) had he known about these inference
attacks.
In the next few sections a background on loca-
tion privacy is presented, some possible privacy
abuses of telematics services are discussed, and
existing approaches to curb these abuses are
investigated. The chapter then suggests possible
measures to strengthen location privacy.
BACKGROUND
Before delving into the core issue of location
privacy, it is important to agree on a definition of
privacy itself. Much of the literature pertaining
to privacy refers to Westin’s precise definition.
In the context of telematics, location privacy is a
special case of privacy, relating to the privacy of
location information of the vehicle, and ultimately
the user of the vehicle.
Privacy is the claim of individuals, groups and
institutions to determine for themselves, when,
how and to what extent information about them
is communicated to others. (Westin, 1967)
400
Location Privacy in Automotive Telematics
How Positioning Systems
can be Privacy Invasive?
Positioning systems can be categorized into either
being ‘Self-positioning’ or ‘Remote-positioning’.
In Self-positioning systems, the vehicle is either
fitted with a GPS receiver or Dead-Reckoning
system (based on one or more gyroscopes, a
compass and odometer) to locate where it is on
the road. Remote-positioning systems require a
central site to determine the location of the vehicle
(Drane and Rizos, 1997). The result is a set of
coordinates (or position) of the vehicle expressed
in relation to a reference frame or datum. Self-
positioning systems inherently protect location
privacy because they do not report the location
of the vehicle to any other system. On the other
hand, remote-positioning systems track, compute
and retain the location information at the central
monitoring site and creates a risk to the individual’s
privacy. Self-positioning systems also pose a pri-
vacy risk if they report the vehicle’s GPS-derived
location to a server through the communications
infrastructure.
PRIVACY ATTACKS
ACME Rent a Car Company
Most readers would be familiar with the highly
publicized abuse of GPS technology where
ACME charged its customers $150 for speeding
occurrences of more than 80mph. A customer
took ACME to court and won on grounds that
the company failed to clearly explain how the
location tracking system would be used (Ayres
and Nalebuff, 2001). This is an obvious scenario
of how personal information can be exploited. It
is not unreasonable to imagine that an ordinary
car trip can become an Orwellian ordeal when
one’s location information can be used in ways
not imagined.
Location Privacy in Automotive Telematics
Location-Based Spam
Figure 1 illustrates a possible threat scenario
where a vehicle is equipped with an on-board GPS
receiver and the vehicle periodically transmits its
location data to a tracking server. The tracking
server is connected to various service providers
which have been authorized by the driver to ac-
cess location data in order to provide telematics
services. The service providers are not necessarily
trusted and it is not unreasonable to expect loca-
tion information of individuals being sold on the
market (much like email address lists).
Profiling Driving Behavior
Greaves and De Gruyter (2002) discuss how a
driving profile of a person can be derived from
GPS track data. They sought an understanding
of driving behaviors in real-world scenarios by
fitting low-cost GPS receivers to vehicles, and
logging the vehicle movements. Consequently,
they were able to identify driving styles from
this data. Imagine a PAYD insurance provider
accessing this information, in order to identify
an individual with an ‘aggressive’ driving style.
Figure 1. A typical privacy threat scenario
Electronic Toll Collection
Electronic toll collection seeks to alleviate traffic
congestion at toll gates, and provides a convenient
method for drivers to pay tolls. Such schemes
typically require the car to have an electronic tag
attached to the front windscreen. Tolls are deducted
from the vehicle’s account when the scanner
senses the toll tag. Electronic toll can become
privacy invasive, for example, if the toll system
passes the entry and exit times of the vehicle to
law enforcement agencies giving them the ability
to issue speeding tickets if the distance is traveled
in too short a time (Langheinrich, 2005).
PRIVACY DEFENSES
In the previous section location privacy threats
provoking some serious ambivalence about the
social and ethical telematics issues were discussed.
There are some countermeasures that can be taken.
The first and most simple one would be an opt-out
approach. This would result in a denial of service
for the vehicle driver. The more challenging issue
is how to preserve location privacy while at the
same time maximizing the benefits of telematics
services.
Legislation and Regulation
Location privacy can be considered to be a special
case of information privacy. However, because
this area of the law is in its embryonic stages,
one can consider ‘location’ and ‘information’ as
being synonymous.
In the United States, legislation to protect
location information arises primarily from the
Telecommunications Act of 1996 and the 1998
E911 amendments. As a result, there is ambiguity
about the so-called “opt-in” or “opt-out” approach
for customer consent. However, a bill specifi-
cally addressing location privacy, the Wireless
Location Privacy Protection Act of 2005, which
401

required location-based services (LBSs)to give
their informed consent for disclosure of loca-
tion information, was referred to the U.S. Senate
(Ackerman et al, 2003).
In Australia, the Privacy Act of 1988 (“Privacy
Act 1988 (Commonwealth)”, 2005) deals with
consumers’ privacy. Besides legislation, Standards
Australia has published a guideline suggesting
procedures for toll operators and electronic parking
operators to protect the personal privacy of their
customers (Standards-Australia, 2000).
Japan and the European Union have well
established laws for protecting consumers’ loca-
tion privacy (Ackerman et al, 2003).One issue
that should be emphasized is that legislation is
not the only defense against (location) privacy
attacks. The corporate world is very good at
obscuring questionable practices with fine print
in a service agreement or contract (Schilit et al,
2003). Therefore there has to be enforcement of
laws as well as open audit of privacy practices.
Policy-Based and Rule-
Based Protection
Privacy protection regulation concludes that “user
consent” is an essential requirement. If the growth
in telematics services proceeds as predicted, then
it would be difficult for a member of the public to
keep track of all details. Secondly, constant explicit
consent requirements can become a source of
driver distraction. Hence an analogy can be drawn
from the internet, where the Platform for Privacy
Preferences (P3P) is used to manage web server
privacy policies in Extensible Markup Language
(XML) machine readable format. Typically these
operate by comparing user profile rules of a web
client with the rules on a particular web server.
Such is the importance of location privacy that
there are already efforts to extend the P3P for
location rules (Morris, 2002). This means that
rules like “Allow Alice to access my location on
a weekday” can be created. Duri et al (2002) pro-
posed an end-to-end framework based on a similar
402
Location Privacy in Automotive Telematics
principle that provides both security and privacy
within the telematics environment. There is one
problem with these implementations, since the
policies serve as a mutual contract; the driver has
to trust the organization to abide by the policies.
The Internet Engineering Task Force (IETF),
the standards body responsible for developing
internet standards, has also realized the importance
of location privacy. It has proposed the Geographi-
cal Location/Privacy Charter, referred to simply as
geopriv. This standard seeks to preserve location
privacy of mobile internet hosts (IETF, 2006).
Synnes et al (2003) have implemented secure
location privacy, using a similar approach of using
rules to implement policies. In the near future, it is
not hard to imagine automobiles having Internet
Protocol (IP) addresses and ultimately using the
geopriv solution to implement privacy policies.
Identity and Location Decoupling
One conclusion that can be drawn is that the
vehicle can be uniquely identified when it com-
municates with a particular Telematics Service
Provider (TSP). Therefore, decoupling of identity
and vehicle location is essential at retention of
data. This can be regulated through policy, and
laws such as discussed above. Herborn et al (2005)
have studied this concept in pervasive comput-
ing networks. They argue that decoupling these
data from each other would have more benefits.
Name ‘hijacking’ would simply not be possible.
The issue here is that for decoupling the identity
and other data to work, a robust scheme to resolve
naming would be required. This, however, is still
an open research issue.
Anonymous Access
Researchers in the field of LBSs have looked at
anonymous solutions to location privacy. The basic
idea here is to access the services anonymously.
Unfortunately, this cannot be regarded as a com-
plete solution given the inference capabilities of
Location Privacy in Automotive Telematics
Geographical Information Systems (GIS) and
advanced surveillance techniques, as discussed
already (Gruteser and Hoh, 2005). An adversary
can apply data-matching techniques to indepen-
dent samples of anonymous data collected, and
map them on predictable paths such as roads,
and infer the identity of an individual based on
where one is.
Drawing from techniques used by census
boards and electoral commissions to obscure
data so that individuals are not identified, another
methodology similar to anonymous access has
been proposed. It is called “k-anonymous access”.
This means that when the location information of
a subject is requested, it will only be responded
to if there are k other nodes present in the vicinity
(Gruteser and Grunwald, 2003). This approach can
give good protection against privacy attacks if the
value of k is set to a high number, however this
would affect the quality of LBSs. In this approach,
k is a variable that could only be altered globally.
A second approach deals with k on a per node
basis. This means that each user can specify his
or her privacy variable (Gedik & Liu, 2005). This
approach appears to more realistically simulate
user privacy preferences in the real world.
Apart from being identified through map-
matching techniques, there is one additional
problem that can affect the correct operation of
telematics services using anonymous techniques.
Existing approaches discussed here are aimed
at solving the location privacy problem in the
context of LBSs. Telematics can be considered to
be a special case of LBSs, the authors, however,
argue that it needs a totally different mindset for
addressing privacy problems of the mobile public
mainly because of differences such as higher aver-
age speeds, predictable paths, and the magnanimity
of the number of users.
Obfuscation
The term “obfuscation” means the process of
confusing or obscuring. It has been identified as
one possible approach to protect location privacy
in location-aware computing (Duckham and
Kulik, 2005). This deliberate degradation of loca-
tion information is performed by the individual,
through deciding which service would require
what ‘granularity’ of information, often referred
to as the “need to know principle”. Snekkenes
(2001) constructed rules for implementing privacy
policies using this principle. He emphasized that
different services require different resolutions, or
accuracy, of location information. The advantage
of obfuscation over anonymity is that it allows
authentication and customization of the services.
However, it still is not the ideal remedy when high
accuracy of reported location, instead of deliberate
degradation, is required.
Privacy Aware Designs
While defenses discussed above propose measures
for limiting disclosure of location information,
others have sought to understand privacy aware
designs. The success of future LBSs depends on
designing systems with privacy in mind, not just
it being an “afterthought”. Langheinrich (2005)
discuss the need for anonymous location infra-
structures and transparency protocols allowing
customers to understand and track how their data
is collected and used. Kobsa and Telztrow (2006)
argue that clearly explaining privacy policies at
subscription would encourage users to disclose
information and create a sense of trust. They
conducted experiments to prove this comparing
their privacy friendly systems to the traditional
data collection systems.
Other examples of privacy aware designs in-
clude work by Coroama and Langheinrich (2006)
where they implemented a GPS based PAYD insur-
ance system depicting real-time risk assessment
of actual road conditions. Their system calculates
premiums on board the vehicle guaranteeing pri-
vacy of owners. There is periodic transmission of
aggregated information to the insurance provider
for bill generation. Iqbal and Lim (2006) extended
403

this idea further and proposed a GPS-based insur-
ance product that preserves location privacy by
computing distances traveled on the onboard unit.
They additionally safeguarded “spend privacy”
by proposing smart card based anonymous pay-
ment systems. Their approach was to redesign a
closed system curtailing redundant exchange of
location data.
CONCLUSION
Location privacy protection in telematics is indeed
a social issue. The authors have reviewed in this
short article location privacy threats and possible
countermeasures. Each countermeasure to protect
privacy has its own implications, and it is clear
that no general panacea exists. This suggests that
a combination of several different approaches may
be the best solution.
The reader might feel that the authors have
taken a pessimistic view of privacy issues. It is
acknowledged that location disclosure would be
necessary in life threatening scenarios, or where
law enforcement officials need access to this
information. This critical information, however,
like other worthwhile liberties needs to be pro-
tected by law. Under normal circumstances, only
the location information subject has the right to
exercise control of one’s personal information.
Development in telematics is through a coop-
eration of companies that are involved in transport
management, vehicle manufacture or information
technology services. The current approach recog-
nizes privacy to be a “non-technical barrier” to
the implementation of ITS (US-Department-of-
Transportation, 1997). Since research in transport
telematics is in its nascent stages, it is important
to understand that these issues are not merely
social hindrances. Once such scenarios become
commonplace, the general user may be reluctant
to use these telematics services at all. Therefore,
it is important to dispel these privacy concerns
right from the beginning, and focus on “building
404
Location Privacy in Automotive Telematics
in” privacy protection within such systems so that
as new applications become available, appropriate
privacy measures are integral to them.
REFERENCES
Ackerman, L., Kempf, J., & Miki, T. (2003).
Wireless location privacy: A report on law and
policy in the United States, the European Union,
and Japan: DoCoMo USA Labs Technical Report
DCL-TR2003-001.
Ayres, I., & Nalebuff, B. (2001). Connecticut’s
speeder-friendly crackdown. The New York Times.
Coroama, V., & Langheinrich, M. (2006). Per-
sonalized Vehicle Insurance Rates – A Case
for Client-Side Personalization in Ubiquitous
Computing. Paper presented at the Workshop on
Privacy-Enhanced Personalization at CHI 2006,
Montréal, Canada, 22 April, 2006.
Drane, C., & Rizos, C. (1997). Role of Position-
ing Systems in ITS. In Positioning Systems in
Intelligent Transportation Systems (pp. 298-299).
Boston: Artech House, Inc.
Duckham, M., & Kulik, L. (2005). A formal
model of obfuscation and negotiation for loca-
tion privacy. Lecture Notes in Computer Science,
3468, 152–170.
Duri, S., Gruteser, M., Liu, X., Moskowitz, P.,
Perez, R., Singh, M., et al. (2002). Framework
for security and privacy in automotive telemat-
ics. Paper presented at the Proceedings of the 2nd
international workshop on Mobile commerce,
Atlanta, Georgia, USA.
Gedik, B., & Liu, L. (2005). Location Privacy in
Mobile Systems: A Personalized Anonymization
Model. Paper presented at the 25th Institute of
Electrical and Electronic Engineers, Inc. (IEEE)
International Conference on Distributed Comput-
ing Systems (ICDCS’05).
Location Privacy in Automotive Telematics

Greaves, S. P., & De Gruyter, C. (2002). Profiling Morris, J. (2002). P3P and the Privacy of Loca-
driving behaviour using passive Global Position- tion Information. Paper presented at the World
ing System (GPS) technology. Paper presented at Wide Web Consortium (W3C) Workshop on the
the Institute of Transportation Engineers Interna- Future of Platform for Privacy Preferences (P3P).
tional Conference, Melbourne, Australia.
Orwell, G. (1949). Nineteen Eighty-four. New
Grush, B. (2005). Optimizing GNSS-Based Mo- York: Harcourt, Brace.
bility Pricing for Road-Use, Parking, and PAYD
Privacy Act 1988 (Commonwealth). (2005).
Insurance. Paper presented at the 4th European
(Vol. Act No. 119 of 1988 as amended): Office
Traffic Congress, Salzburg, Austria.
of Legislative Drafting and Publishing (OLDP),
Gruteser, M., & Grunwald, D. (2003). Anonymous Australia.
Usage of Location-Based Services through Spatial
Schilit, B., Hong, J., & Gruteser, M. (2003). Wire-
and Temporal Cloaking. Paper presented at the
less location privacy protection. IEEE Computer
First International Conference on Mobile Systems,
Magazine, 36(12), 135–137.
Applications, and Services.
Snekkenes, E. (2001). Concepts for personal
Gruteser, M., & Hoh, B. (2005). On the Anonymity
location privacy policies. Paper presented at the
of Periodic Location Samples. Paper presented at
Proceedings of the 3rd ACM conference on Elec-
the 2nd International Conference on Security in
tronic Commerce, Tampa, Florida, USA.
Pervasive Computing, Boppard, Germany.
Standards-Australia. (2000). AS 4721-2000: Per-
Herborn, S., Boreli, R., & Seneviratne, A. (2005).
sonal privacy practices for the electronic tolling
Identity Location Decoupling in Pervasive Com-
industry: Standards Australia.
puting Networks. Paper presented at the 19th
International Conference on Advanced Informa- Synnes, K., Nord, J., & Parnes, P. (2003). Location
tion Networking and Applications (AINA ‘05). Privacy in the Alipes Platform. Paper presented at
the 36th Annual Hawaii International Conference
IETF. (2006). Internet Engineering Task Force
on System Sciences (HICSS’03).
(IETF) Geographic Location/ Privacy (geopriv).
Retrieved 27th March, 2006, from http://www. US-Department-of-Transportation. (1997). A
ietf.org/html.charters/geopriv-charter.html Report to Congress: Nontechnical Constraints
and Barriers to the Implementation of Intelligent
Iqbal, M. U., & Lim, S. (2006). A privacy preserving
Transportation Systems. Washington, D.C: US
GPS-based Pay-as-You-Drive insurance scheme.
Department of Transportation.
Symposium on GPS/GNSS (IGNSS2006), Surfers
Paradise, Australia, 17-21 July, 2006. Vidales, P., & Stajano, F. (2002). The Sentient
Car: Context-Aware Automotive Telematics. Paper
Kobsa, A., & Teltzrow, M. (2006). Convincing
presented at the LBS-2002.
Users to Disclose Personal Data. In A. Kobsa, R.
Chellappa, & S. Spiekermann, (Eds.), Proceed- Westin, A. F. (1967). In PRIVACY AND FREE-
ings of PEP06, CHI 2006 Workshop on Privacy- DOM (pp. 7). New York: Atheneum.
Enhanced Personalization, Montréal, Canada.
Whelan, R. (1995). Smart Highways, Smart Cars.
Langheinrich, M. (2005). Personal Privacy in Boston: Artech House, Inc.
Ubiquitous Computing – Tools and System Sup-
port. PhD thesis No. 16100, ETH Zurich, Zurich,
Switzerland.

405

KEY TERMS AND DEFINITIONS
Context Aware Computing: The process of
customization of software and services to user
preferences. The computing mechanism changes
based on the context, in telematics perspective,
location is a context for customization.
Electronic Tolls: Electronic payment systems
designed to identify an electronic tag mounted on
a vehicle to deduct the toll charges electronically
from the vehicle owner’s account.
In-Car Navigation: Usually a voice-activated
system with a liquid crystal display (LCD) screen
displaying maps and a combination of on-board
GPS receivers, accelerometers, compass and gy-
roscopes for positioning the vehicle on the map.
Intelligent Transportation Systems: Tools,
software, hardware and services designed for the
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 293-301,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
406
Location Privacy in Automotive Telematics
efficient movement of road transportation and
provision of travel information to the vehicles.
Location Privacy: Location privacy is the
ability of an individual to control access to his/
her current and past location information.
Obfuscation: Obfuscation is the deliberate
degradation of location information by respond-
ing in a less granular fashion about requested
location data.
Telematics Service Provider: Telematics Ser-
vice providers offer services to vehicle drivers for
either a subscription fee or any other arrangement.
These can be emergency services or informational
services to improve the driving experience.
Vehicle Prognostics: Factory installed sys-
tems monitoring and reporting health of vehicle
equipment to owner and manufacturer periodi-
cally.
 407

Chapter 25
Diagnostic Modeling of
Digital Systems with Multi-
Level Decision Diagrams
Raimund Ubar
Tallinn University of Technology, Estonia

Jaan Raik
Tallinn University of Technology, Estonia

Artur Jutman
Tallinn University of Technology, Estonia

Maksim Jenihhin
Tallinn University of Technology, Estonia

ABSTRACT
In order to cope with the complexity of today’s digital systems in diagnostic modeling, hierarchical
multi-level approaches should be used. In this chapter, the possibilities of using Decision Diagrams
(DD) for uniform diagnostic modeling of digital systems at different levels of abstraction are discussed.
DDs can be used for modeling the functions and faults of systems at logic, register transfer and behavior
like instruction set architecture levels. The authors differentiate two general types of DDs – logic level
binary DDs (BDD) and high level DDs (HLDD). Special classes of BDDs are described: structurally
synthesized BDDs (SSBDD) and structurally synthesized BDDs with multiple inputs (SSMIBDD). A
method of iterative synthesis of SSBDDs and SSMIBDDs is discussed. Three methods for synthesis of
HLDDs for representing digital systems at higher levels are described: iterative superposition of HLDDs
for high-level structural representations of systems, symbolic execution of procedural descriptions for
functional representations of systems, and creation of vector HLDDs (VHLDD) on the basis of using
shared HLDDs for compact representing of a given set of high level functions. The nodes in DDs can be
modeled as generic locations of faults. For more precise general specification of faults different logic
constraints are used. A functional fault model to map the low level faults to higher levels, particularly,
to map physical defects from transistor level to logic level is discussed.

DOI: 10.4018/978-1-4666-2038-4.ch025

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
INTRODUCTION
The most important question in testing today’s
digital systems is: how to improve the test
generation and fault simulation efficiency and
quality at continuously increasing complexities
of systems? Two main trends can be observed:
defect-orientation and high-level modeling. To
follow the both trends simultaneously, a hierar-
chical approach seems to be the only possibility.
One of the attractive ways to manage hierarchy
in diagnostic modeling (test generation, fault
simulation, fault location) in a uniform way on
different levels of abstraction is to use decision
diagrams (DD).
Traditional flat low-level test methods and
tools for complex digital systems have lost their
importance, other approaches based mainly on
higher level functional and behavioral methods
are gaining more popularity (Lee, Patel, 1997;
Makris, Collins, Orailoglu & Vishakantaiah, 2000;
Fummi, & Sciuto, 2000; Vedula & Abraham,
2002; Mirkhani, Lavasani & Navabi, 2002; Al-
Yamani, & McCluskey, 2004; Kundu, 2004; Yi,
& Hayes, 2006; Misera, Vierhaus, Breitenfeld &
Sieber, 2006; Alizadeh, & Fujita, 2010; Misera
& Urban, 2010). Hierarchical mixed- or multi-
level approaches have also been used both, for
test generation (Lee, Patel, 1997; Makris, Collins,
Orailoglu & Vishakantaiah, 2000; Vedula & Abra-
ham, 2002; Ravi, Jha, 2001; Ichihara, Okamoto,
Inoue, Hosokawa & Fujiwara, 2005), and fault
simulation (Mirkhani, Lavasani & Navabi, 2002;
Kundu, 2004; Misera, Vierhaus, Breitenfeld &
Sieber, 2006; Misera & Urban, 2010). A general
idea in these methodologies is detailed low-level
fault simulation in one module of a system, while
propagating the effects through other modules
modeled at a higher level abstraction.
The trend towards higher level modeling moves
us away from the real life of defects and, hence,
from accuracy of testing. To handle adequately
defects in deep-submicron technologies, new
fault models and defect-oriented test methods
408
have been introduced for test generation (Blanton
& Hayes, 2003) and fault diagnosis (Mahlstedt,
Alt & Hollenbeck, 1995; Holst & Wunderlich,
2008). On the other hand, the defect-orientation
is increasing the complexity of the diagnostic
modeling task even more. To get out from the
deadlock, these two opposite trends – high-level
modeling and defect-orientation – should be com-
bined into a hierarchical approach. The advantage
of hierarchical diagnostic modeling compared to
the high-level functional modeling lies in the pos-
sibility of constructing and analyzing test plans
on higher levels, and modeling faults on more
detailed lower levels.
BACKGROUND
The difficulties in developing of analytical multi-
level and hierarchical approaches to digital test
generation and fault simulation lay in using dif-
ferent languages and models for different levels
of abstractions. Most frequent examples are logic
expressions for combinational circuits, state transi-
tion diagrams for finite state machines (FSM), ab-
stract execution graphs, system graphs, instruction
set architecture (ISA) descriptions, flow-charts,
hardware description languages (HDL, VHDL,
Verilog, System C etc.), Petri nets for system level
description etc. All these models need dedicated
for the given language manipulation algorithms
and fault models which are difficult to merge into
hierarchical test methods. HDL based modeling
methods which are efficient for fault simulation
lack the capability of analytical reasoning and
analysis that is needed in test generation and
fault diagnosis.
Excellent opportunities for multi-level and
hierarchical diagnostic modeling of digital systems
provide decision diagrams (DD) because of their
uniform cover of different levels of abstraction, and
because of their capability for uniform graph-based
fault analysis and diagnostic reasoning (Lee, 1959;
Ubar, 1976; Akers, 1978; Plakk & Ubar, 1980;
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Bryant, 1986; Minato, 1996; Sasao, & Fujita,
1996; Ubar, 1996; Drechsler & Becker, 1998).
Within the last two decades BDDs have be-
come the state-of-the-art data structure in VLSI
CAD for representation and manipulation of
Boolean functions. They were first introduced
for logic simulation in (Lee, 1959), and for test
generation in (Ubar, 1976, Akers, 1978). In 1986,
Bryant proposed a new data structure called
reduced ordered BDDs (ROBDDs). He showed
simplicity of the graph manipulation and proved
the model canonicity that made BDDs one of the
most popular representations of Boolean functions
(Minato, 1996; Sasao, & Fujita, 1996; Drechsler &
Becker, 1998). Different types of BDDs have been
proposed and investigated during decades such as
shared or multi-rooted BDDs (Minato, Ishiura, &
Yajima, 1990), ternary decision diagrams (TDD),
or in more general, multi-valued decision diagrams
(MDD) (Srinivasan, Kam, Malik, & Bryant, 1990),
edge-valued binary decision diagrams (EVBDD)
(Minato, Ishiura, & Yajima, 1990), functional
decision diagrams (FDD) (Kebschull, Schubert,
& Rosenstiel,1992), zero-suppressed BDDS
(ZBDD) (Minato, 1993), algebraic decision dia-
grams (ADD) (Bahar, Frohm, Gaona, Hachtel,
Macii, Pardo, & Somenzi, 1993), Kronecker FDDs
(Drechsler, Theobald, & Becker, 1994), binary
moment diagrams (BMD) (Bryant, & Chen, 1995),
free BDDs (Bern, Meinel, & Slobodova,1995),
multiterminal BDDs (MTBDD) and hybrid BDDs
(Clarke, Fujita, & Zhao, 1996), Fibonacci decision
diagrams (Stankovic R.S., Astola, Stankovic, M.,
& Egiazarian, 2002) etc. Overviews about dif-
ferent types of BDDs can be found for example
in (Sasao, & Fujita, 1996, Drechsler, & Becker,
1998; Karpovsky, Stanković, & Astola, 2008).
For more detail about the many ways to represent
logic functions, we refer for example to (Sasao,
1999; Astola, & Stankovic, 2006).
Most of these listed different types of BDDs,
however, suffer from the memory explosion, which
limits their usability for large designs. Regarding
test and fault simulation, the listed types of BDDs
derived from logic functions don’t represent in-
herently the structural details of the circuits they
represent, which are related to signal paths and
structural location of faults. For the lack of direct
representation of such structural information in the
model, indirect and less efficient methods should
be applied when the structure related tasks are
being solved like test generation, fault simulation
and fault diagnosis.
Structurally synthesized BDDs (SSBDD)
were proposed and developed with the goal to
represent, simulate and analyze structural features
of circuits (Ubar, 1996; Jutman, Raik & Ubar,
2002). The most significant difference between
the function-based BDDs and SSBDDs is in the
method how they are generated. While BDDs
are generated on the functional basis by Shan-
non expansion, or other types of expansions (like
Davio or Reed-Muller expansions), which handle
only the Boolean function of the logic circuit, the
SSBDD models are generated by superposition of
BDDs that extracts both, functions and data about
structural signal paths of the circuit. The linear
complexity of the SSBDD model results from the
fact that a digital circuit is represented as a system
of SSBDDs, where for each fanout-free region
(FFR) a separate SSBDD is generated. SSBDDs
have found efficient applications in simulation of
structural aspects and faults for test generation
(Raik & Ubar, 1998), hazard and timing analysis
with multi-valued simulation (Ubar, 1998; Jut-
man, Ubar & Peng, 2001), fault simulation and
fault coverage analysis (Ubar, Devadze, Raik &
Jutman, 2008 & 2010), design error diagnosis
(Ubar & Borrione, 1999).
Recent research has shown that generalization
of BDDs for higher levels provides a uniform
model for both gate and RT or behavioral level
simulation (Ubar, Moraviec & Raik, 1999), mixed-
level fault simulation (Ubar, Raik, Ivask & Brik,
2002), hierarchical test generation (Raik & Ubar,
2000), and high-level design verification (Jenih-
hin, Raik, Chepurov & Ubar, 2009).
409
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
The disadvantage of the traditional hierarchical
approaches to test is the conventional application
of gate-level stuck-at fault (SAF) model. It has
been shown that high SAF coverage cannot guar-
antee high quality of testing (Huisman, 1993). The
types of faults that can be observed in a real gate
depend not only on the logic function of the gate,
but also on its physical design. Good possibilities
to combine logical and physical level fault model-
ing are provided by the extensions of the SAF-
based fault model like pattern fault (Blanton &
Hayes, 2003) or conditional fault models (Holst &
Wunderlich, 2008). A similar pattern related fault
modeling approach called functional fault model
was proposed in (Ubar, 1980) for the module level
fault diagnosis in combinational circuits. Based
on this extended SAF fault model, a deterministic
defect-oriented test pattern generator DOT was
developed in (Raik, Ubar, Sudbrock, Kuzmicz &
Pleskacz, 2001), which allows to prove the logic
redundancy of not detected physical defects from
the given fault model.
Whereas the methods of synthesis of different
types of BDDs are well described in the literature,
the formal methods of generation of high-level
DDs (HLDD) have not been discussed in the
literature so far. This gap will be filled up in this
chapter. We present first, a short overview about
SSBDDs including the definition, synthesis, and
main algorithms of diagnostic modeling. In the
next Section we show possibilities to improve
the model by reducing its complexity while
retaining all the information about the structure
of the original circuits. Next, we generalize the
graph model of logic level SSBDDs for higher
levels of abstraction with introducing high level
DDs (HLDD). We compare them with BDDs
and present main algorithms of diagnostic mod-
eling based on HLDDs as the generalizations
of similar algorithms for BDDs. Then, three
methods for synthesis of HLDDs for represent-
ing digital systems at higher levels are proposed:
iterative superposition of HLDDs for high-level
structural representations of digital systems, sym-
410
bolic execution of procedural descriptions of the
functionality of systems, and creation of vector
HLDDs (VHLDD) on the basis of using shared
HLDDs for compact representing of a given set
of high level functions. The last section is devoted
to functional fault modeling which helps to map
faults from lower to higher levels in multi level
diagnostic modeling of systems with DDs.
DECISION DIAGRAMS
AND DIGITAL TEST
Diagnostic Modeling of Digital
Circuits with SSBDDs
In the following we present a definition of
SSBDDs, method of synthesis of SSBDDs, and
describe briefly fault simulation and test genera-
tion algorithms for SSBDDs with the purpose to
generalize all this topics for HLDDs.
Consider first, the following graph theoreti-
cal definitions of the BDD. We use the graph-
theoretical definitions instead of traditional ite
expressions (Bryant, 1986; Drechsler & Becker,
1998) because all the procedures defined further
for SSBDDs are based on the topological reasoning
rather than on graph symbolic manipulations as
traditionally in the case of BDDs. The topological
approach allows to generalize all the algorithms
related to low logic level BDDs (SSBDDs) for
high functional or behavioral level DDs.
Definition 1: A BDD that represents a Boolean
function y=f(X), X = (x1,x2, …, xn), is a
directed acyclic graph Gy = (M,Γ,X) with
a set of nodes M and a mapping Γ from M
to M. M = MN ∪ MT consists of two types
of nodes: nonterminal MN and terminal MT
nodes. A terminal node mT ∈ MT = {mT,,0,
mT,1} is labeled by a constant e∈{0,1} and is
called leaf, while all nonterminal nodes m ∈
MN are labeled by variables x ∈ X, and have
exactly two successors. Let us denote the as-
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
sociated with node m variable as x(m), then
m0 is the successor of m for the value x(m) =
0 and m1 is the successor of m for x(m) = 1.
Γ(m) ⊂ M denotes the set of all successors
of m ∈ M, and Γ -1(m) ⊂ M denotes the set
of all predecessors of m ∈ M. For terminal
nodes mT ∈ MT we have Γ (mT) = ∅. There
is a single node m0 ∈ M where Γ -1(m) = ∅
called root node.
Definition 2: For the assigned value of x(m) = e,
e ∈ {0,1}, we say the edge between nodes
m ∈ M and me ∈ M is activated. Consider a
situation where all the variables x ∈ X are
assigned by a Boolean vector Xt ∈ {0,1}n
to some values. The activated by Xt edges
form an activated path through nodes l(m0,
mT) ⊆ M from the root node m0 to one of the
terminal nodes mT∈ MT.
Definition 3: We say that a BDD Gy = (M,Γ,X)
represents a Boolean function y=f(X), iff
for all the possible vectors Xt ∈ {0,1}n a
path l(m0, mT) ⊆ M is activated so that y =
f(Xt) = x(mT).
Definition 4: Consider a BDD Gy=(M,Γ,X) which
models a gate-level tree-like combinational
circuit represented by a Boolean formula
(equivalent parenthesis form [21]) y = P(X)
where X is the vector of literals, the nodes m
∈ MN are labeled by x(m) where x ∈ X, and
|M | = |X |. Each literal x ∈ X represents an
Figure 1. A circuit and a SSBDD which correspond to the Boolean Formula (1)
input and a signal path of the given tree-like
circuit. The BDD is called a structurally
synthesized BDD (SSBDD) iff there exists
one-to-one correspondence between the
literals x ∈ X and the nodes m ∈ MN given
by the set of labels { x(m) | x ∈ X, m ∈ MN},
and iff for all the possible vectors Xt ∈ {0,1}
n
a path l(m0, mT) is activated, so that y =
f(Xt) = x(mT).
An example of such a SSBDD for the Boolean
formula
y = (x 11x 12 ) ∨ x 12x 31x 4 )(x 13x 22x 32 ) (1)
which corresponds to the gate-level circuit in
Figure 1a is represented in Figure 1b. We call
this graph structurally synthesized because it is
derived from and represents the structure of the
Formula (1) or the corresponding circuit in Figure
1a. There is a one-to-one mapping between the
nodes in graph and the literals in the Formula (1)
or in the corresponding signal paths in the circuit
in Figure 1a.
Synthesis of SSBDDs
For synthesis of SSBDDs for a given gate network,
the graph superposition procedure is used. If the
411
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
label x(m) of a node m in the SSBDD Gy is an
output of a subnetwork which is represented by
another SSBDD Gx(m) then the node m in Gy can
be substituted by the graph Gx(m). In this graph
superposition procedure the following changes
in Gy and Gx(m) are made.
Algorithm 1: Graph superposition procedure
1) The node m will be removed from Gy.
2) All the edges in Gx(m) that were con-
nected to terminal nodes mT,e in Gx(m)
will be cut and then connected, corre-
spondingly, to the successors me of the
node m in Gy. Here, mT,e is the terminal
node labeled by constant e∈ {0,1}
3) All the incoming edges of m in Gy will
be now incoming edges for the initial
node m0 in Gx(m).
We assume that the BDDs for elementary
components (gates) of the circuit are given as a
library. Let us call these library BDDs as elemen-
tary SSBDDs. Starting from the SSBDD of the
output gate, and using iteratively Algorithm 1, we
compress the initial model of the gate-network (by
each substitution we reduce the model by one node
and by one graph). To avoid the explosion of the
SSBDD model, we generate the SSBDDs only
for tree-like subnetworks. As the result we get a
macro-level logic network where each macro (a
tree-like subcircuit) is represented by a SSBDD.
An example of the graph superposition pro-
cedure for a combinational circuit represented
by a Boolean formula (equivalent parenthesis
Figure 2. Example of superposition of SSBDDs
412
form) y = (x1 ∨ x21) (x22 ∨ x3) is shown in Figure 2.
For simplicity, the values of variables on edges
of the SSBDD are omitted (by convention, the
right-hand edge corresponds to 1 and the down-
wards directed edge to 0). Likewise, the terminal
nodes with constants 0 and 1 are omitted: leaving
the graph to the right corresponds to y = 1, and
down, to y = 0.
We start with the output AND gate and its BDD
Gy which consists of two nodes a and b. The input
a of the AND gate is simultaneously the output
of the OR gate represented by the BDD Ga which
consists of the nodes x1 and x21. First, we substitute
the node a in Gy by the graph Ga. Thereafter the
node b in Gy is substituted by the graph Gb which
consists of the nodes x22 and x3. The final graph
which represents the whole circuit except the
fanout stem x2 consists of the nodes x1, x21, x22,
and x3. The full SSBDD model for the given
circuit consists of two graphs: the graph created
(in Figure 2), and the graph consisting of a single
node x2.
In general case, the SSBDD model for a given
combinational circuit consists of SSBDDs for all
tree-like fanout-free regions (FFR) and of 1-node
SSBDDs for all primary inputs which have fan-
out branches.
Unlike the “traditional” BDDs, SSBDDs
directly support test generation and fault simula-
tion for gate-level structural SAF faults without
representing these faults explicitly. The advantage
of the SSBDD based approach is that the library
of components is not needed for structural path
activation, as it is in the case of traditional gate-
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
level path activation. This is the reason why
SSBDD based test generation procedures do not
depend on whether the circuit is represented on
the logic gate level or on the logic macro-level
where the logic macro means an arbitrary single-
output gate-level subcircuit of the whole gate
network. Moreover, the test generation procedures
developed for SSBDDs can be easily generalized
for higher level DDs to handle digital systems
represented at higher levels.
From the above-described synthesis of
SSBDDs, automatic fault collapsing results for the
given gate-level circuits which can be formulated
as the corollary resulting from Algorithm 1.
Corollary 1: Since all the SAF faults on the inputs
of a FFR according to the approach of fault
folding (To, 1973) form the collapsed fault
set of the FFR, and since all these faults are
represented by the faults at the nodes of the
corresponding SSBDD, then the creation of
the SSBDD is equivalent to the fault col-
lapsing procedure similar to fault folding.
Assume a node m with label x(m) represents
a signal path p(m) in a circuit. Suppose the path
p(m) goes through n gates. Then, instead of 2n
faults of the path p(m) in the circuit, only 2 faults
related to the node variable x(m) should be tested
when using the SSBDD model.
Theorem 1: Let G(C) be the SSBDD model
generated for the combinational circuit C
by Algorithm 1. Then, any set of tests that
checks all the SAF faults at the nodes of
G(C) checks all the SAF faults in C.
The proof follows from Corollary 1 and Theo-
rem 5 in (To, 1973).
Differently from the traditional gate level
approaches to test generation and fault simula-
tion which use the collapsed fault list apart from
the simulation model, the SSBDD based test
generation and fault simulation are carried out
on the macro-level (FFRs as macros) with direct
representation of the faults in the model.
Test Generation with SSBDDs
Consider a combinational circuit as a network of
gates, which is partitioned into interconnected
tree-like subcircuits (macros). This is a higher
level (macro-level) representation of the same
circuit. Each macro is represented by a SSBDD
where each node corresponds to a signal path from
an input of the macro to its output. In the tree-
like subcircuits only the stuck-at faults at inputs
should be tested. This corresponds to testing all
the nodes in each SSBDD.
Test generation for a node m in SSBDD, which
represents a function y = f(X) of a tree-like sub-
circuit (macro), is carried out by the following
procedure.
Algorithm 2: Test generation for a node m in the
SSBDD Gy, y = f(X)
1) A path lm = l(m0, m) from the root node
of SSBDD to the node m is activated.
2) Two paths lm,e = l(m,mT,e) consistent with
lm, where e∈{0,1}, from the neighbors
me of m to the corresponding terminal
nodes mT,e should be activated. If the
node m is directly connected via e-edge
to mT,e, no path lm,e should be activated
for this particular value of e.
3) For generating a test for a particular
stuck-at-e fault x(m) ≡ e, e∈{0,1}, the
opposite assignment is needed:
x (m ) = e.
4) All the values assigned to node vari-
ables (to variables of X) form the local
test pattern Xt (input pattern of the
macro) for testing the node m in Gy
(for testing the corresponding signal
path p(m) on the output y of the given
tree-like circuit).
413
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
As an example, consider test generation for
the fault x11 ≡ 0 at the node x11 in Figure 1b. To
test this fault we have to assign to x1 the opposite
value x1 = 1. Since x11 is the root node, the path lm
is empty and there is no need to activate it. The
path lm,1 can be activated by assignment x2 = 1,
and the path lm,0 can be activated by assignment
x3 = 0. The generated test pattern is: Xt = (x1, x2,
x3, x4)t = (110-). The pattern detects all the faults
(x11 ≡ 0, x21 ≡ 0, y ≡ 0) along the signal path from
x11 to y in the circuit in Figure 1a.
This example demonstrated test generation
in a single SSBDD. In general case, the test gen-
eration for embedded macro-components of the
circuit the local test patterns for macros should
be extended to the final test patterns in terms of
primary inputs of the circuit, similarly as for the
gate-level circuits where fault propagation and
line justification procedures are needed. The
difference, however, is that these procedures are
carried out on the higher macro level (instead of
the gate level) whereas the macros of the circuit
are represented by SSBDDs. The fault propagation
through a macro from the input x to its output y is
carried out similarly to the test generation for the
node m labeled by x in the corresponding SSBDD
Gy as explained in Algorithm 2. Line justification
for the task y = e is carried out by activating a
path in the graph Gy from the root node to the
terminal node mT,e.
Fault Simulation with SSBDDs
Fault simulation of a test pattern Xt on the SSBDD
Gy, which represents a function y = f(X) of a
tree-like subcircuit (macro), is carried out by the
following procedure.
Algorithm 3: Fault simulation of a test pattern
Xt in Gy, y = f(X)
1) The path l = l(m0,mT,e) activated by the
pattern Xt in Gy is determined.
2) For each node m ∈ l, its successor m*
is determined so that m* ∉ l, and the
414
path lm* = l(m*,mT,e*) activated by the
pattern Xt is determined; if mT,e* ≠ mT,e
then the fault of the node m is detected
by Xt otherwise not.
As an example, consider the test pattern Xt =
(x1, x2, x3, x4)t = 1011 which activates the path l =
(x11, x21, x12, x31, x4, #1) in Figure 1b. According
to Step 2 of Algorithm 3 we have: lx11*= lx12 = (x12,
x31, x4, #1); lx21* = (#1); lx12* = lx13(x13, #0), lx31* =
lx13(x13, #0); lx4* = lx13(x13, #0). From this it follows
that by the given test pattern the faults x12 ≡ 0, x31
≡ 0, x4 ≡ 0 are detected.
Fault Diagnosis with SSBDDs
The fault simulation procedure described by
Algorithm 3 can be used during fault diagnosis
based on the effect-cause fault location concept.
Suppose there has been an error detected on the
output y of the circuit in Figure 1a by the pattern
Xt = 1011. According to the example, the fault
candidates should be suspected first along the path
l activated by the pattern. After fault simulation
the set of candidate faults has been pruned to the
subset { x12 ≡ 0, x31 ≡ 0, x4 ≡ 0}
The experiments of logic-level simulation with
SSBDD model were carried out on ISCAS’85
benchmarks. They were run for each circuit on
two levels of abstraction: gate-level and SSBDD.
This allows to directly measure the effect of the
chosen model on simulation speed. Figure 3 il-
lustrates the average speed-up brought by usage
of the SSBDD model compared to the gate-level
simulation speed. The fault simulation (fat dashed
line) shows the most noticeable acceleration.
Other simulation algorithms vary in decreasing
the runtime by 2,5 up to almost 4 times compared
to algorithms working on the gate-level netlist
model. This effect is possible due to shift from
lower gate level to a higher macro level when
working with SSBDD model.
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Figure 3. Logic level simulation speed-up for different algorithms
MODELING DIGITAL CIRCUITS WITH
MULTIPLE INPUT SSBDDS
Introduce now a new type of SSBDD model called
Structurally Synthesized Multiple Input BDDs
(SSMIBDD) by extending the superposition
procedure beyond the fanout nodes of the given
circuit. The goal is to further compress the SSBDD
model by exploiting the effect of the superposi-
tion procedure where by each superposition step
a node and a graph from the whole BDD model of
the gate-level network are removed. To avoid the
negative effects of fanout stems xi with branches
xi,1, xi,2,... xi,k, the superposition in the graph G
under construction is allowed only for a single
branch xi,j, 1 ≤ j ≤ k. For the remaining branches
of the same fanout stem a link will be established
from other graphs via corresponding additional
inputs inserted into the graph G.
In Figure 4, a hypothetical initial circuit is
represented with three outputs 1, 2, 3 and four
fanout stems 4, 5, 6, 7. The circuit is modeled
by two SSMIBDDs G1 and G2 with root nodes,
respectively, 1 and 2, and by a SSBDD G3 with
the root node 3. The links from the nodes 4, 5, 6
in G2, and from the nodes 6, 7 in G3 to the graph
G1 are available via additional inputs to G1. In
other words, when calculating the output values
2 and 3, respectively, in graphs G2 and G3, the
values of variables in nodes 4 and 5 (6 and 7)
can be calculated by entering the graph G1 (G2)
at additional inputs 4 and 5 (6 and 7).
Algorithm 4: Generation of SSMIBDDs
The general procedure of constructing the
SSMIBDD is as follows. Let us have a circuit
C with a set of outputs Y, a set of fanout stems
FS, and a library of BDDs for all the gates in C.
For all outputs of Y we start a generation of SS-
MIBDD by Algorithm 1. Let us have two graphs
to be superpositioned: G2 for the gate g2 into G1
for the gate g1. We have now three possibilities:
1) If for all input variables x of the gate g1 , x ∉
FS, then Algorithm 1 is carried out as it is.
2) If there is a single x: x ∈ FS, then before the
superposition the nodes in G1 are reordered
in such a way that the node m labeled with
x ∈ FS will be the last node in G1, i.e. Γ(m)
= ∅. The node m will get an additional input
x as the link to other branch variables (in
other graphs) of this fanout stem. Note that
for all other nodes m*∈ Γ -1(m), no additional
inputs can be inserted.
3) If there are more than one input variables x: x
∈ FS, then only one of them can be selected
for further superpositioning according to the
case 2, and for all other variables we have to
start the construction process of new graphs.
415
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams

Figure 4. Concept of SSMIBDDs

An example of a combinational circuit with
three outputs and 9 internal fan-outs is depicted
in Figure 5a, and the corresponding SSMIBDD
model consisting of only 3 graphs is shown in
Figure 5b. SSBDD model for this circuit would
consist of 12 graphs. The number of faults to be
processed in the gate level circuit in Figure 5a is
before fault collapsing 84. Using the FFR based
SSBDD model with 27 nodes we can collapse
the fault set up to 54 faults (corresponding to the
representative faults at 9 inputs and 18 internal
fan-out branches of the circuit). Using the SS-
MIBDD model with 18 nodes in Figure 5b allows
collapsing the fault set up to 36 faults.
The SSMIBDD model can be used for test
generation in a similar way as the SSBDD
model was used for test generation and for fault
simulation. Consider as an example test generation
for the fault SAF-13/1 (stuck-at-1 on the upper
branch of the node 13) in Figure 5a with SS-
MIBDDs in Figure 5b. Let find by Algorithm 2
the input pattern which propagates the faulty
signal from 13 to the output 20 to make the fault
observable. To activate the paths l13 = (17,16,13)
and l13,0 = (13,5,#0) (l13,1 is empty) in G20 we as-
sign: x17 = 0, x16 = 1, x13 = 0, and x5 = 0. In G19 we
see that x17 = 0 can be justified by x15 = 0. How-
ever, in G20 we see that x15 = 0 is already justified.

Figure 5. Circuit with high number of fan-outs and its SSMIBDDs

416
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
In G21 we justify x16 = 1 by x8 = 1, and x13 = 0 by
x6 = 0. Hence, the test pattern for detecting the
fault SAF-13/1 is: Xt = (x5, x6, x8) = 001. The fault
is propagated to the output 20 via the highlighted
path in Figure 5a. To solve the same task on the
gate-level would be more complex because of the
higher complexity of the model.
The method of generating SSMIBDDs de-
scribed above starts from the outputs of the
given circuits. Using this procedure, in the best
case, a circuit with n outputs can be represented
by n SSMIBDDs. However, it is easy to see that
the number of graphs can be further reduced by
sharing the same subgraphs by different output
graphs, and correspondingly, the number of nodes
in the whole model can be further reduced, which
results in further fault collapsing.
An example of the minimal SSMIBDD for the
circuit in Figure 6a is presented in Figure 6b. The
new SSMIBDD model with 7 nodes contains only
14 collapsed faults as targets for test generation.
The nodes x1, and x5 represent signal paths in the
circuit from x1 to y1 and from x5 to y2, respectively.
The node x3 represents the path from the lower
input of g1 to y1. The node x2 represents the path
from the upper input of g3 to y1. The nodes ¬x3 and
¬x4 represents the paths from the inputs of g2 to
both outputs y1 and y2. And, finally, the node ¬3
represents the path from the upper input of g6 to
y2. More detailed discussion of using SSMIBDDs
where the graphs for different output functions
Figure 6. Minimized SSMIBDD for the ISCAS’85 benchmarks family circuit c17
are merged into the same graph can be found in
(Ubar, Mironov, Raik & Jutman, 2010).
A comparison of the complexities of gate-
level circuits, SSBDDs and SSMIBDDs in terms
of fault collapsing was carried out using ISCAS’85
and ISCAS’89 circuits. The differences in the
number of nodes between the gate level, SSBDD
and SSMIBDD models are shown in Figure 7.
The average minimization gained for SSMIBDDs
in the number of nodes (and also in the size of
collapsed fault sets) for ISCAS’85 and ISCAS’89
circuits is up to 2,4 times compared to the gate
level, and up to 1,4 times compared to the SSBDD
model.
The results prove that the SSMIBDD model
is more compact than the previously discussed
SSBDD or gate level models, and as the result
allows better fault collapsing which in its turn has
the influence to the efficiency and speed of test
generation and fault simulation.
MODELING DIGITAL SYSTEMS WITH
HIGH LEVEL DECISION DIAGRAMS
Test generation methods developed for SSBDDs
have an advantage compared to other logic level
methods, namely that they can be easily gen-
eralized to handle the test generation and fault
simulation problems at higher levels of systems.
The possibility of generalization results from the
417
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Figure 7. Comparison of the complexities of gate-level circuits, SSBDDs and SSMIBDDs
topological similarity of representing systems with
DDs at lower and higher levels.
Consider a digital system S as a network of
components (or subsystems) where each compo-
nent is represented by a function z = f (z1, z2,..., zn)
= f (Z) where Z is the set of variables (Boolean,
Boolean vectors or integers), and V(z) is the set
of possible values for z ∈ Z which are finite. In
the case of representing next state functions z ∈ Z.
Definition 5: A decision diagram (DD) which
represents a digital function z=f(Z) is a
directed acyclic graph Gz = (M,Γ,Z) with
a set of nodes M and a mapping Γ from M
to M. M = MN ∪ MT consists of two types
of nodes: nonterminal MN and terminal MT
nodes. The terminal nodes m ∈ MT may be
labeled either by variables z(m) ∈ Z, digital
functions z(m) = fm(Zm), Zm ⊆ Z, or constants.
The nonterminal nodes m ∈ MN are labeled
by variables z(m) ∈ Z, and have successors
whose number may 2 ≤ |Γ(m)| ≤ |V(z(m))|.
A mapping exists between the values of
z(m) and the successors in Γ(m). Denote the
successor of m for a given value of z(m) as
mz(m) ∈ Γ(m). Γ(m) ⊂ M denotes the set of
all successors of m ∈ M, and Γ -1(m) ⊂ M
denotes the set of all predecessors of m ∈ M.
For terminal nodes mT ∈ MT we have Γ (mT)
= ∅. There is a single node m0 ∈ M where Γ
-1
(m) = ∅ called root node.
418
Definition 6: For the assigned value of z(m) = e,
e ∈ V(m), we say the edge between nodes
m ∈ M and me ∈ M is activated. Consider
a situation where all the variables z ∈ Z are
assigned by a vector Zt to some values from
the domains of definition V(z), z∈ Z. The
activated by Zt edges form an activated path
through nodes l(m0, mT) ⊆ M from the root
node m0 to one of the terminal nodes mT∈ MT.
Definition 7: We say that a decision diagram Gz
represents a function z = f (z1, z2, …, zn) = f
(Z), iff for each value V(Z) = V(z1) × V(z2)
×... × V(zn), a path in Gz is activated from
the root node z0 to a terminal node mT ∈MT,
so that z = z(mT) is valid.
It is easy to see that SSBDDs represents a
special case of DDs where for all z ∈ Z, V(z) =
{0,1}, and two terminal nodes are labeled by the
Boolean constants 0 and 1.
Depending on the class of the system (or its
representation level), we may have various DDs,
where nodes have different interpretations and
relationships to the system structure. In RTL
descriptions, we usually partition the system
into control and data paths. Nonterminal nodes
in DDs may correspond then to the control path,
and they are labeled by state or output variables
of the control part serving as addresses or control
words. Terminal nodes in DDs correspond in this
case to the data path, and they are labeled by the
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
data words or functions of data words, which cor-
respond to buses, registers, or data manipulation
blocks. When using DDs for describing complex
digital systems, we have to first, represent the
system by a suitable set of interconnected com-
ponents (combinational or sequential subcircuits).
Then, we have to describe these components by
their corresponding functions which can be rep-
resented by DDs.
In Figure 8, a RTL data-path and its compressed
DD is presented. The variables R1 and R2 represent
registers, IN denotes the input bus, the integer
variables y1, y2, y3, y4 represent control signals,
M1, M2, M3 are multiplexers, and the functions
R1+R2 and R1*R2 represent the adder and multiplier,
correspondingly. Each node in DD represents a
subcircuit of the system (e.g. the nodes y1, y2, y3, y4
represent multiplexers and decoders,). The whole
DD describes the behavior of the input logic of the
register R2. To test a node means to test the cor-
responding to the node component or subcircuit.
Test pattern generation and fault simulation
algorithms can be derived easily by generalization
of algorithms 2 and 3, respectively. The test gen-
eration for terminal and non-terminal nodes
slightly differs, as it can be seen from the follow-
ing algorithms 5 and 6.
Figure 8. Representing a data path by a decision diagram
Algorithm 5: High-level test generation for a
non-terminal node m∈ MN in the DD Gz
1) A path lm = l(m0, m) from the root node
of SSBDD to the node m is activated.
2) For all e ∈ V(m), non-overlapping
paths lm,e consistent with lm from me up
to terminal nodes mT,e are activated. If
the node m is directly connected via
e-edge to mT,e, no path lm,e should be
activated.
3) A proper set of data (the values of the
variables in Z) is found, so that the in-
equality z(mT,,1) ≠ z(mT,2) ≠ … ≠ z(mT,n)
holds, where n = |Γ(m)|.
4) The test data consist of a static part TS
(the values generated in Steps 1-3), and
of a dynamic part TD (all the values
from V(m)). To implement the test, a
loop is created where TS is repeated
|V(m)| times for all the values of z(m)
∈ V(m) in TD. Before each loop all the
values in TS are restored.
The paths in the SSBDD activated by the
described procedure are illustrated in Figure 9.
Note, the test generation according to Algo-
rithm 5 is carried out entirely on the high-level,
based on the high-level fault model, according to
419
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Figure 9. Test generation for the node m with low level and high-level DDs
which the behavior of the node m is tested ex-
haustively. It is reasonable, since the domains of
definitions of control variables are not large.
Satisfaction of the inequality in Step 3 of Algorithm
5 is the guarantee that the node m under testing
behaves correctly.
As an example, consider test generation for
testing the multiplexer M3 represented by the node
y3 in DD in Figure 8b. We activate 4 paths lm,e for
each value e = 0,1,2,3 of y3. Two of them, lm,1,
lm,2, for values y3 = 1 and y3 = 2, respectively, are
“automatically” activated since the successors of
y3 for these values are terminal nodes. The control
values for the test are found by activating the path
lm with assigning y4 = 2, and by activating two
paths lm,0 and lm,3 with assigning y1 = 0 and y2 =
0, respectively. The test data R1 = D1, R2 = D2, IN
= D3 are found to satisfy the inequality R1 + R2 ≠
IN ≠ R1 ≠ R1 * R2.
According to Step 4 of Algorithm 5, the follow-
ing test program for the control variable y3 results:
Test program for control part:
For e = 1,2,3,4
BEGIN
Load the data registers R1 = D1, R2 =
D2
Carry out the tested working mode at
y3 = e, y1 = 0, y2 = 0, y4 = 2 and IN
420
= D3
Read the value of R2,e
END.
Algorithm 6: Test generation for a terminal node
m ∈ MT in the DD Gz
1) A path lm = l(m0, m) from the root node
of SSBDD to the node m ∈ MT is ac-
tivated. This will be the static part TS
of the test.
2) The proper sets of values TD = {Ztm},
t =1,2,..., n, for testing the function
fm(Zm) represented by the terminal node
m is selected. This operation should
be carried out at the lower (e.g. gate)
level if the implementation details for
fm(Zm) are given, which directly leads
to hierarchical approach to test genera-
tion. Otherwise high level fault models
for testing fm(Zm) should be used.
3) To implement the test, a loop is created
where TS is repeated n times for all the
patterns in TD. Before each loop the
values in TS are restored.
As an example, consider test generation for test-
ing the multiplexer M3 represented by the node y3
in DD in Figure 8b. By activating the path to this
node (shown in bold in Figure 8b) we generate a
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams

control word (y2, y3, y4) = (0, 3, 2). To find the proper
values of R1 and R2 we need to descend to the lower
level (e.g. gate level) and generate test patterns by
a low level ATPG for the low level implementation
of the multiplier. Let us have a test set of n test
patterns (D1,1,D2,1; D1,2,D2,2; … D1,n,D2,n) generated
for the multiplier with input registers R1 and R2.
From above, according to Step 3 of Algorithm 6,
the following test program results:
Test program for data part:
For all the values of t = 1,2, …, n
BEGIN
Load the data registers R1 = D1,t, R2
= D2,t
Carry out the tested working mode
at the control values (y2,y3 y4) =
(0,3,2)
Read the value of R2,t
END.
In test pattern simulation, a path is traced in
the graph, guided by the values of input variables
in Z until a terminal node is reached, similarly as
in the case of SSBDDs. In Figure 8, the result of
simulating the vector Zt = (y1, y2, y3, y4, R1, R2, IN)
= 0,0,3,2,10,6,12 is R2 = R1*R2 = 60 (bold arrows
mark the activated path). Instead of simulating by
a traditional approach all the components in the
circuit, in the DD only 3 control variables are vis-
ited during simulation (y4, y3, y2,), and only a single
data manipulation R2 = R1*R2∈ MT is carried out.
The feasibility and advantages of using HLDDs
in diagnostic modeling of digital systems was dem-
onstrated by using the test generator DECIDER
(Raik & Ubar, 2000). The results regarding the
speed of test generation and fault coverage (Table
1) were compared with other known test generators
for complex sequential digital systems HITEC
(Niermann, Patel, 1991) and GATEST (Rudnick,
Patel, Greenstein, Niermann, 1994). The experi-
mental results show the high speed of the ATPG
DECIDER which is explained by the DD based
hierarchical approach used in test generation
The advantage of using DDs for test generation
and fault simulation is the very close similarity of
handling low logic level and higher behavior or
register transfer level representations of digital
systems. This helps to develop uniform approach-
es and generalize the concepts of solving different
test problems like test generation, fault simulation,
fault diagnosis, testability analysis etc. from logic
level to higher levels. To do this, we need formal
methods for HLDD synthesis from common high-
level representations of digital systems.
SYNTHESIS OF HIGH-LEVEL
DECISION DIAGRAMS FROM THE
NETWORK OF COMPONENTS
In this chapter we present two methods for syn-
thesis of high level DDs for representing digital
systems. The first method is based on iterative

Table 1. Comparison of test generators

HITEC GATEST DECIDER

Circuit # Faults
FC% T, s FC% T,s FC% T, s
gcd 454 81.1 170 91.0 75 89.9 14
sosq 1938 77.3 728 79.9 739 80.0 79
mult 2036 65.9 1243 69.2 822 74.1 50
ellipf 5388 87.9 2090 94.7 6229 95.0 1198
risc 6434 52.8 49020 96.0 2459 96.5 151
diffeq 10008 96.2 13320 96.4 3000 96.5 296
Average FC% 76.9 87.9 88.6

421
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
superposition of HLDDs, and can be regarded as
a generalization of Algorithm 1 used for synthesis
of SSBDDs in Section 2. This method can be used
in cases when the system is given structurally by
a network of components (subsystems), and for
each component its HLDD is given. An example
of such presentation of a system is depicted
in Figure 10a. The second method is based on
symbolic execution of procedural descriptions,
which correspond to the functional representa-
tions of systems at higher behavioral levels. This
method can be used in cases when the system is
given functionally as a procedure in a hardware
description language or in the form of a flow-chart.
An example of such presentation of a system is
depicted in Figure 10b. Consider in this Section
the first method.
The system in Figure 10a consists of the con-
trol and data parts. The finite state machine (FSM)
of the control part is given by the output function
y = λ (q’, x) and the next-state function q = δ (q’,
x), where y is an integer output vector variable,
which represents a microinstruction with four
control fields y = (yM, yz, yz,1, yz,2), x = (xA, xC) is
a Boolean input vector variable, and q is the in-
teger state variable. The value j of the state vari-
Figure 10. Digital system as a network of components and a flowchart of its behavior
422
able corresponds to the state sj of the FSM. The
apostrophe refers to the value from the previous
clock cycle. Assume that the functions λ and δ are
represented by HLDDs y = Gy (q’, x), and q = Gq
(q’, x), respectively.
The data path consists of the memory block M
with three registers A, B, C together with the ad-
dressing block ADR, represented by three DDs: A
= GA (yM, z), B = GB (yM, z), C = GC (yM, z); of the
data manipulation block CC where z = Gz (yz, z1,
z2); and of two multiplexers z1 = Gz,1 (yz,1, M) and
z2 = Gz,2 (yz,2, M). The block COND performs the
calculation of the condition function x = Gx (A, C).
The component level model of the system
consists of the following set of DDs: N={Gy,Gq,
GA,GB,GC,Gz,Gz,1,Gz,2,Gx}. Using now the follow-
ing chain of superpositions of DDs:
A = GA (yM, z) = GA (yM, Gz (yz, z1, z2))= GA (yM,
Gz (yz, Gz,1 (yz,1, M), f4 (yz,2, M))) = GA (yM, yz, yz,1,
yz,2, M) = GA (y, M) = GA (Gy (q’, x), M) = G’A
(q’, x, A, B, C),
and repeating a similar chain of superposition
steps for the graphs B = GB (yM, z) and C = GC
(yM, z), we create a new compact DD model of the
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Figure 11. High Level Decision Diagrams for the system represented as a network in Figure 10a
system N’= {Gq, G’A, G’B, G’C} which is presented
in Figure 11. The part of the model related to the
data path is represented by three DDs G’A, G’B,
G’C where, for simplicity, the terminals nodes for
the cases where the value of the function variable
does not change, are omitted.
For synthesis of the HLDD model for a given
network of component HLDDs, the iterative graph
superposition of two HLDDs is used, which can
be regarded as the generalization of Algorithm 1.
If the label z(m) of a node m in the HLDD Gz is
an output of a component in the network, which
is represented by another HLDD Gz(m), then the
node m in Gz can be substituted by the graph Gz(m).
In this graph superposition the following chang-
es in Gy and Gx(m) are made.
Algorithm 7
1) The node m labeled by z(m) is removed
from Gy.
2) All the edges in Gz(m) that were con-
nected to terminal nodes mT,e in Gz(m)
will be cut and then connected, to the
successors me of the node m in Gy,
respectively.
3) All the incoming edges of m in Gy will
be now the incoming edges for the root
node m0 in Gz(m).
Note, this procedure corresponds exactly (!)
to Algorithm 1 for synthesis of SSBDDs with
the only difference in the ranges of values for e
(binary vs. integers in V(m)). If a terminal node
m in a graph Gy is labeled by a data variable z(m)
which is represented by another graph Gz(m) then
the procedure is trivial: the node m in Gz can be
simply substituted by the graph Gz(m). Superposi-
tion of a non-terminal node m labeled by z(m) in
Gz by the graph Gz(m) is only then possible when
all the values z(m) in V(z(m)) have one-to-one
mapping to the terminal nodes of Gz(m) labeled by
constant values from the same set V(z(m)).
Consider as an example the data path as a
network of components in Figure 8a and the
descriptions of the working modes of these com-
ponents in Table 2.
Superposition of two graphs R2 = GR2 (y4, R2,
e) for the register input logic and the multiplexer
M3 with internal variable e, e = Ge (y3, R1, IN, d)
is illustrated in Figure 12. The whole fully com-
pressed HLDD for the data-path in Figure 8a is
depicted in Figure 8b.
SYNTHESIS OF HIGH-LEVEL
DECISION DIAGRAMS FROM
PROCEDURAL DESCRIPTIONS
Consider a procedure representing the behavior
level description of a digital system. The proce-
dure can be represented by a flow chart which is
a directed graph where a path Pi can be defined
as usual in the graph theory.
423
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams

Table 2. Working modes of the components in

path condition PCi = ∧ C ij is extracted as a
Figure 8a j :cij ∈ci

set of constraints on the input variables (condi-

y1 M1/a y2 M2/b y3 M3/e y4 R2 tions), such that the path is feasible iff the PCi is
0 R1 0 R1 0 M1+R2 0 0 satisfied. The path Pi is not feasible iff PCi ≡ 0.
1 IN 1 IN 1 IN 1 R2 Denote the set of all feasible paths as P.
2 R1 2 M3 In this section, a procedure similar to symbolic
3 M2*R2 execution is used for HLDD synthesis. We assume
that all the assignment statements correspond to
updating of the related data-path registers during
Denote by Y = {y1, y2,... ym} a set of system the current cycle. Based on that we introduce a
variables which values will be changed when “simplified” cycle based symbolic execution pro-
the procedure is executed. These changes are cedure. The procedure of HLDD synthesis based
determined by the assignment statements along on the procedural description of a digital system
the paths of the flow chart. A path Pi represents is carried out as follows.
a sequence of assignment statements Si = {sik}
and a sequence of conditional expressions Ci = Algorithm 8
{cil}. Here k = 1,2,...,m, and each sik ∈ Si denotes 1) Insertion of states into the flow-chart
an expression whose value will be assigned to yk. of the digital system
If the value of yk will not change in the current 2) Cycle based symbolic execution of
path Pi then formally sik ≡ y’k where the apostro- the procedure given by the flow-chart.
phe means that the value of y’k is taken from the Generation of the table of paths by
previous cycle. determining for each feasible path Pi
A path Pi is executable (or feasible) if there the sets of assignment statements Si
are input data such that the program can be exe- and the sets of constraints Ci.
cuted along that path. Otherwise, it is un-execut- 3) Generation of the conditional expres-
able (or infeasible). For deciding the feasibility sions yk = ∨i:P ∈PPC i ⋅ si,k for each
of program paths, symbolic execution of programs i

system variable yk ∈ Y.
and constraint solving techniques are used (Zhang,
2004). By symbolic execution for each path Pi, a

Figure 12. Superposition of two HLDDs

424
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams

4) Synthesis of HLDDs for each system of system variables is 4 (y1 = A, y2 = B, y3 = C,

variable yk by factorization of its con-
ditional expression.
As an example, consider the procedure in
Figure 10b which represents the system in Figure
10a. First, according to Step 1 of Algorithm 8,
the states are inserted into the flow chart of the
procedure under analysis in a similar way as the
algorithmic state machines are marked by states
during synthesis of FSM [46]. There are six states
s0, s1,..., s5, inserted into the procedure in such a
way that during the state transfer each data variable
can be changed only once. A global state variable
q is introduced to map the states to values of q.
For example, q = i for the state si.
y4 = q). Only these assignment statements sik, k =
1,2,3,4, are noted in the last column of Table 3,
where the system variable may change its value
during the transition.
Next, for all the left-hand side system variables
A, B, C and q in the last column of Table 3 we
create HLDDs which describe the cycle-based
behavior of these variables during execution of
the procedure in Figure 10b.
Steps 3 and 4 of Algorithm 8 are illustrated in
the following for the system variable A. In Table
4, all the paths Pi are depicted where the value of
A may be changed according to the expression A
= y1= si,1 if the path constraint PCi = ∧ j :c ∈c C ij
ij i

is satisfied.
During cycle based symbolic execution (Step
According to Step 3, from Table 4 the follow-
2 of Algorithm 8), the path trees are created only
ing expression for the behavior of the system
for the restricted regions between neighboring
variable A can be derived:
states, avoiding in this way the explosion of
paths. As a result of tracing all the transitions
A = (q=0)(B+C) ∨ (q=1)(xA=0) (¬A + 1) ∨
between the states in the flow graph, a table is
constructed where each row corresponds to a path
(q=3)(xC=1)(¬C+B) ∨
between neighboring states. The result of cycle
based symbolic execution of the procedure in (q=4)(xA=0)(xC=0)(A+ ¬B + C + 1)
Figure 10b is presented in Table 3. The number
According to Step 4, by factorization of this
Table 3. Results of symbolic execution of the expression, a HLDD for the variable A can be
procedure in Figure 10b derived in a similar way as BDDs are derived by
using Shannon factorization [16]. The only differ-
Path Constraints Ci Assignment statements ence is that instead of Boolean factorization we
Pi Si = {sik}
q xA xB xC use multi-valued factorization, depending on the
1 0 A = B + C; q = 1 possible number of values |V(z)| of the constraint
2 1 0 A = ¬A + 1; q = 4 variable z.
3 1 1 B = B + C; q = 2
4 2 0 C = A + B; q = 5
Table 4. Results of symbolic execution for the
5 2 1 C = ¬C; q = 3
variable A
6 3 0 C = A + B; q = 5
7 3 1 A = ¬C + B; q = 5 q xA xB xC A
8 4 0 B = ¬B 0 B+C
9 4 0 0 A = A + ¬B + C; q = 5 1 0 ¬A + 1
10 4 0 1 q=5 3 1 ¬C + B
11 4 1 C = ¬C; q = 5 4 0 0 A + ¬B + C

425
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams

The HLDDs created by factorization of the to calculate by a single run the values of all the
expressions yk = ∨i:P ∈PPC i ⋅ si,k , k = 1,2,3,4, for components of the vector variable Y = (y1, y2,... yn).
i
An example of merging the DDs GA, GB, GC
all the four variables A, B, C and q are shown in
and Gq in Figure 11 into a single VHLDD GM =
Figure 11. To underline that the values of variables
Gq, A, B, C to calculate the value of the vector M =
are taken from the previous cycle we use the
A.B.C.q is shown in Figure 13. For calculating and
apostrophes at the data and state variables.
assigning the new values to different components
In this model, for simplicity, all the edges in the
of the vector variable M, we introduce a new type
graph Gz for the function z = f(Z), which lead into
of a node in VDD called addressing node labeled
the terminal node labeled by the graph variable z,
by an addressing variable i. The vector decision
are dropped. This case corresponds to the assign-
diagrams offer the capability to efficiently repre-
ment where the value of z remains unchanged.
sent the array variables (corresponding to register
In other words, if we reach a non-terminal node
blocks and memories) for calculating and updating
m where the successor me for the value z(m) = e
their values. VHLDDs are particularly efficient for
is missing, the graph variable z holds the same
representing functional memories with complex
value as in the previous cycle. For example, in
input logic – with shared and dedicated parts for
the graph A in Figure 11, if q’ = 4 and xA ∨ xC =
different memory locations. In general case, all
0, the variable A will hold its value, i.e. A = A’.
the registers of the data path can be combined in
the VHLDD model as a single memory block.
Using VHLDDs allows significally to increase
VECTOR HIGH-LEVEL
the speed of simulation. Consider, as an example,
DECISION DIAGRAMS
the input pattern Zt = (q’,xA, xB, xC) = (1,1,0,0). As
the result of simulation of this pattern on all the
To further reduce the complexity of HLDDs
4 graphs in Figure 13 (the paths traversed are
we introduce the concept of vectorizing a set of
shown by bold lines and grey nodes) we get: A =
HLDDs. A set of HLDDs {Gy1, Gy2,... Gyn} can be
A’, B = B’ + C’, C = C’ and q = #2. The simulation
merged into a single vector HLDD (VHLDD) Gy1,
needs 11 visits to the decision nodes (note that
,... yn with less complexity as a result of sharing
y2
for A and C we stop simulation in the node where
similar parts in separate DDs. Such a graph allows

Figure 13. Representing a set of 4 HLDDs by a single vector HLDD

426
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
the successors for xA=1 and q’=1, respectively,
are missing, which means that the variables hold
their values). When simulating the same input
pattern in the joint VHLDD in Figure 13, we get
the same result by visiting only 5 decision nodes
whereas the variables A and C need no simulation
at all. The addressing node i in the joint graph
points by the values on the output edges at the
vector components which should be assigned by
the new value. These values are calculated by the
expressions in the terminal nodes.
FUNCTIONAL FAULT MODELING
IN DIGITAL SYSTEMS
New failure mechanisms in today’s deep-sub-
micron electronic devices cannot be modeled by
traditional stuck-at faults (SAF) which in case of
DDs are directly associated with the nodes of the
graphs. As the result, new advanced fault models
are continuously being developed to improve the
confidence of test quality measures and to increase
the accuracy of fault diagnosis. The types of faults
that can be observed in a real gate depend not only
on the logic function of the gate, but also on its
physical design. Good possibilities to combine
logical and physical level fault modeling provide
pattern fault model (Blanton & Hayes, 2003) or
conditional fault models (Mahlstedt, Alt & Hol-
lenbeck, 1995; Holst & Wunderlich, 2008). A
similar pattern related fault modeling approach
called functional fault model was proposed in
(Ubar, 1980) for the module level fault diagnosis
in combinational circuits.
Consider a parametric model of a component
(e.g. a complex gate) in a combinational circuit
with a correct function y = fy (x1,x2,...xn), and in-
cluding a Boolean fault variable Δ to represent an
arbitrary physical defect (Δ = 0 when the defect
is missing, and Δ = 1 when the defect is present)
as a generic function
y * = fy * (x 1, x 2 ,..., x n , ∆) = ∆fy ∨ ∆fy ∆
where fyΔ represents the faulty function of the
component because of the defect Δ. The solution
Wy(Δ) of the Boolean differential equation
∂fy *
= 1
∂∆
describes a condition which activates the defect Δ
to produce an error on the output y of the compo-
nent. The parametric modeling of a given defect Δ
by the condition Wy(Δ) = 1 allows to use it either
for defect-oriented fault simulation (to check
whether the condition Wy(Δ) = 1 is fulfilled), or
for defect-oriented test generation under the con-
straint Wy(Δ) = 1 when a test pattern is searched
for detecting the defect Δ. A similar approach is
described in (Devadze & Ubar, 2010) for simula-
tion of extended class of faults.
If the components of the circuit represent
standard library (complex) gates, the described
analysis for finding conditions should be made
once for all library components, and the sets of
calculated conditions will be included into the
library of components in the form of fault tables.
The defect characterization may be computation-
ally expensive, but it is performed only once for
each library cell. An example of the functional
fault table for the complex gate AND2,2/NOR2
is presented in Table 5 (Ubar, Kuzmicz, Pleskacz
& Raik, 2001). The defect lists WFy of library
components embedded in the circuit can be ex-
tended by additional physical defect lists WSy for
the interconnect structure in the neighboring of
the component to take into account also different
defects (bridging faults, crosstalks etc.) outside
the components. For these defects additional
characterization should be carried out by a similar
way as for the library cells.
The functional fault model Wy(Δ) allows to use
for test generation and fault simulation for any
427
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams

Table 5. Library defect table for a complex gate AND2,2/NOR2

i Fault di Erroneous function fdi Input patterns tj

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 B/C not((B*C)*(A+D)) 1 1 1 1
2 B/D not((B*D)*(A+C)) 1 1 1 1
3 B/N9 B*(not(A)) 1 1 1 1 1 1 1
4 B/Q B*(not(C*D)) 1 1 1 1 1 1 1 1 1
5 B/VDD not(A+(C*D)) 1 1 1
6 B/VSS not (C*D) 1 1 1
7 A/C not((A*C)*(B+D)) 1 1 1 1
8 A/D not((A*D)*(B+C)) 1 1 1 1
9 A/N9 A*(not(B)) 1 1 1 1 1 1 1
10 A/Q A*(not(C*D)) 1 1 1 1 1 1 1 1 1
11 A/VDD not(B+(C*D)) 1 1 1
12 C/N9 not(A+B+D)+(C*(not((A*B)+D))) 1 1 1 1 1
13 C/Q C*(not(A*B)) 1 1 1 1 1 1 1 1 1
14 C/VSS not(A*B) 1 1 1
15 D/N9 not(A+B+C)+(D*(not((A*B)+C))) 1 1 1 1 1
16 D/Q D*(not(A*B)) 1 1 1 1 1 1 1 1 1
17 N9/Q not((A*B)+(B*C*D)+(A*C*D)) 1
18 N9/VDD not((C*D)+(A*B*D)+(A*B*C)) 1
19 Q/VDD SA1 at Q 1 1 1 1 1 1 1 1
20 Q/VSS SA0 at Q 1 1 1 1 1 1 1 1 1

physical defect Δ traditional stuck-at fault test
generators or fault simulators developed for the
logic level. Denote by yW(Δ) the value of a node y
determined by the functional fault condition Wy(Δ)
= 1. To generate a test for a defect Δ, a test pattern
should be generated for the stuck-at fault y ≡ (yW(Δ)
⊕1) under the additional constraint Wy(Δ) = 1.
Otherwise, if a test pattern detects a SAF fault at
the node y, then all the defects Δ are detected as
well if the test pattern does not contradict with
Wy(Δ) = 1.The described approach has been used
for defect oriented test generation (Raik, Ubar,
Sudbrock, Kuzmicz & Pleskacz, 2005) and fault
simulation (Ubar, Devadze, Raik & Jutman, 2010)
based on the SSBDD model.
Table 7 gives the results of test generation
for ISCAS’85 circuits using the deterministic
defect-oriented test generator DOT generation
(Raik, Ubar, Sudbrock, Kuzmicz & Pleskacz,
2005). Column 2 shows the total number of de-
fects simulated in the circuits as described in this
Section. Column 3 reflects the number of gate
redundant defects (GRD). These are defects that
cannot be covered by any input pattern of the gate.
For example, AND-short between the two inputs
of the AND gate is a GRD. In column 4, circuit
redundant defects (CRD) are counted. CRDs
are defects that cannot be covered by any input
pattern of the circuit. Next, three fault coverages
to characterize the capability of SAF model for
testing physical defects are provided. The defect
coverages are measured for the test sets which
cover 100% stuck-at faults. The first coverage
(column 5) shows the ratio of detected defects
versus the total set of defects. In column 6, gate
redundancy has been excluded (i.e. all GRDs are

428
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams

Table 7. Results of defect oriented test generation

Redundant defects Defect coverage

Object Defects Gate level System level SAF test with different redundancies
New method
GL SL With GL&SL With SL Without
c432 1519 226 0 78.6 99.05 99.05 100
c880 3380 499 5 75.0 99.50 99.66 100
c2670 6090 703 61 79.1 97.97 99.44 100
c3540 7660 985 74 80.1 98.52 99.76 99.97
c5315 14794 1546 260 82.4 97.53 100 100
c6288 24433 4005 41 77.0 99.81 100 100

excluded from the list of possible defects). The
third coverage measure (column 7) shows the
test efficiency (or test coverage). In this column,
both, gate redundancy and circuit redundancy
have been excluded. The last column shows the
defect coverage achieved by the defect oriented
test generator DOT.
When using HLDDs for test generation pur-
poses, the idea of functional fault model was used
in generating tests for terminal nodes in DDs.
Remind that the main concept of the fault model
used in HLDDs was to test exhaustively each
node. For non-terminal nodes which model the
control variables such a concept is meaningful
because of the low number of possible values for
these variables. The situation is different with
terminal nodes which model the units of data
paths. In this case, hierarchical approach is advis-
able. Assume the terminal node mT in a graph Gy
is labeled by a multiplication expression A * B.
Test pattern for activation of the working mode
y = A * B is generated at the higher level using
the HLDD Gy. The set of local test patterns {At,
Bt} to be applied to the inputs A and B of the
multiplier are generated at the lower gate-level
(possibly, even in the defect oriented way). The
set of patterns {At, Bt} can be regarded as a set of
constraints (as the functional fault model) when
testing the terminal node mT. The set of local test
patterns {At, Bt} can be regarded also as an inter-
face between two levels in hierarchical test gen-
eration: they are generated at lower level and used
in a test plan which is generated at higher level.
CONCLUSION
In this chapter a unified method for diagnostic
modeling of digital systems at lower logic and
higher functional levels was described. The
method is based on using decision diagrams
where the well known BDDs can be regarded as
a special case of (high level) DDs. A special class
of BDDs called structurally synthesized BDDs
was introduced and analyzed. SSBDDs allow to
represent the internal structure of the circuit in
the model which is important regarding the fault
modeling. It was also shown that a byproduct
of the SSBDD synthesis is fault collapsing. A
method for diagnostic modeling of complex digital
systems with high level DDs was described and
compared with using BDDs. We described three
methods for synthesis of HLDDs for representing
digital systems at higher behavior, functional or
register-transfer levels. The first method is based
on symbolic execution of procedural descriptions,
which corresponds to functional representation
of systems at the behavioral level. An approach
called cycle-based symbolic execution was dis-
cussed as a tool for formalized HLDD synthesis,
which prevents the explosion of the number
of paths to be executed. The second method is

429
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
based on iterative superposition of HLDDs, and
the created model corresponds to the high-level
structural representation of the system. The sec-
ond method can be regarded as a generalization
of the superposition of BDDs developed for the
creation of SSBDDs. Surprisingly, the procedure
for superposition of HLDDs corresponds exactly
(!) to the superposition procedure of SSBDDs. The
third method aims at creation of vector HLDDs
(VHLDD) on the basis of using shared HLDDs
for compact representing of a given set of high
level functions. A functional fault model known
also as pattern fault model or conditional fault
model was described and connected to the concept
of diagnostic modeling of digital systems with
DDs. It was shown that this fault model serves as
a natural interface between two levels in hierarchi-
cal or multi-level diagnostic modeling.
REFERENCES
Akers, S. B. (1978, Oct). Functional Testing with
Binary Decision Diagrams. J. of Design Automa-
tion and Fault-Tolerant Computing, 2, 311–331.
Al-Yamani, A., & McCluskey, E. J. (2004). Test
quality for high level structural test. In Proc. of
9th IEEE High-Level Design Validation and Test
Workshop, (pp.109-114).
Alizadeh, B., & Fujita, M. (2010). Guided Gate-
level ATPG for Sequential Circuits using a High-
level Test Generation Approach (pp. 425–430).
ASP-DAC.
Astola, J. T., & Stanković, R. S. (2006). Funda-
mentals of Switching Theory and Logic Design.
Berlin: Springer.
Bahar, R. I., Frohm, E. A., Gaona, C. M., Hachtel,
G. D., Macii, E., Pardo, A., & Somenzi, F. (1993).
Algebraic decision diagrams and their applica-
tions. In Int. Conf. on CAD, (pp. 188-191).
430
Baranov, S. (2008). Logic and System Design of
Digital Systems. Tallinn, Estonia: Tallinn Univer-
sity of Technology Press.
Bern, J., Meinel, C., & Slobodova, A. (1995).
Efficient OBDD-based manipulation in CAD
beyond current limits. In 32nd Conference on
Design Automation, (pp. 408-413).
Blanton, R. D., & Hayes, J. P. (2003). On the Prop-
erties of the Input Pattern Fault Model. ACM Trans-
actions on Design Automation of Electronic Sys-
tems, 8(1), 108–124. doi:10.1145/606603.606609
Bryant, R. E. (1986). Graph-based algorithms
for Boolean function manipulation. IEEE
Transactions on Computers, C-35(8), 667–690.
doi:10.1109/TC.1986.1676819
Bryant, R. E., & Chen, Y.-A. (1995). Verifica-
tion of arithmetic functions with binary moment
diagrams. In Proc. 32nd ACM/IEEE DAC.
Clarke, E. M., Fujita, M., & Zhao, X. (1996).
Multi-terminal binary decision diagrams and
hybrid decision diagrams. In Sasao, T., & Fujita,
M. (Eds.), Representations of Discrete Functions
(pp. 93–108). Amsterdam: Kluwer Academic
Publishers.
Drechsler, R., & Becker, B. (1998). Binary Deci-
sion Diagrams. Amsterdam: Kluwer Academic
Publishers.
Drechsler, R., Theobald, M., & Becker, B. (1994).
Fast FDD based minimization of generalized
Reed-Muller forms. In European Design Automa-
tion Conf., (pp. 2-7).
Fummi, F., & Sciuto, D. (2000). A hierarchical test
generation approach for large controllers. IEEE
Transactions on Computers, 49(4), 289–302.
doi:10.1109/12.844343
Holst, S., & Wunderlich, H. J. (2008). Adaptive
Debug and Diagnosis Without Fault Dictionaries.
In Proc. 13th European Test Symposium, (pp.
199-204).
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Huisman, L. M. (1993). Fault Coverage and
Yield Predictions: Do We Need More than 100%
Coverage? In Proc. of European Test Conference,
(pp. 180-187).
Ichihara, H., Okamoto, N., Inoue, T., Hosokawa,
T., & Fujiwara, H. (2005). An Effective Design
for Hierarchical Test Generation Based on Strong
Testability. In Proc. of Asian Test Symposium –
ATS’2005, (pp. 288-293).
Jenihhin, M., Raik, J., Chepurov, A., & Ubar, R.
(2009). PSL Assertion Checking Using Tempo-
rally Extended High-Level Decision Diagrams.
Journal of Electronic Testing: Theory and Ap-
plications – JETTA, 25(6), 1-12.
Jutman, A., Raik, J., & Ubar, R. (2002, May).
On Efficient Logic-Level Simulation of Digital
Circuits Represented by the SSBDD Model. In
23rd Int. Conf. on Microelectronics, (Vol. 2, pp.
621-624).
Jutman, A., Ubar, R., & Peng, Z. (2001). Algo-
rithms for Speeding-Up Timing Simulation of
Digital Circuits. In IEEE Proc. of Design Auto-
mation and Test in Europe (pp. 460–465). DATE.
Karpovsky, M. G., Stanković, R. S., & Astola,
J. T. (2008). Spectral Logic and Its Applications
for the Design of Digital Devices. New York:
Wiley-Interscience. doi:10.1002/9780470289228
Kebschull, U., Schubert, E. & Rosenstiel, W.
(1992). Multilevel logic synthesis based on func-
tional decision diagrams. IEEE EDAC’92.
Kundu, S. (2004). Pitfalls of Hierarchical
Fault Simulation. IEEE Trans. on CAD of IC
and Systems, 23(2), 312–314. doi:10.1109/
TCAD.2003.822099
Lee, C. Y. (1959, July). Representation of Switch-
ing Circuits by Binary Decision Programs. The
Bell System Technical Journal, 985–999.
Lee, J., & Patel, J. H. (1997). Hierarchical test
generation under architectural level functional
constraints. IEEE Trans. Computer Aided Design,
15, 1144–1151.
Mahlstedt, U., Alt, J., & Hollenbeck, I. (1995).
Deterministic Test Generation for Non-Classical
Faults on the Gate Level. In 4th Asian Test Sym-
posium, (pp. 244-251).
Makris, Y., Collins, J., Orailoglu, A., & Vishakan-
taiah, P. (2000, May). Transparency-Based Hierar-
chical Test Generation for Modular RTL Designs.
In Proc. of ISCAS’2000, (pp. II-689 - II-692).
Minato, S. (1995). Zero-suppressed BDDs for set
manipulation in combinational problems. In Proc.
30th ACM/IEEE DAC, (pp. 272-277).
Minato, S. (1996). BDDs and Applications for
VLSI CAD. Amsterdam: Kluwer Academic Pub-
lishers.
Minato, S, Ishiura, N., & Yajima. (1990). Shared
binary decision diagrams with attributed edges for
efficient Boolean function manipulation. In Proc.
27th IEEE/ACM ICCAD’90, (pp.52-57).
Mirkhani, S., Lavasani, M., & Navabi, Z. (2002).
Hierarchical Fault Simulation Using Behavioral
and Gate Level Hardware Models. In Proc. of
Asian Test Symposium – ATS’2002, (pp.374-379).
Misera, S., & Urban, R. (2010). Fault simulation
and fault injection technology based on SystemC.
In Ubar, R., Raik, J., & Vierhaus, H. T. (Eds.),
Design and Test Technology for Dependable
Systems-on-Chip. Hershey, PA: IGI Global.
Misera, S., Vierhaus, H.-T., Breitenfeld, L., &
Sieber, A. (2006). A Mixed Language Fault
Simulation of VHDL and SystemC. In Proc. of
9th EUROMICRO Conf on Digital System Design
(DSD’06).
431
 Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Niermann, T., & Patel, J. H. (1991). HITEC: A
test generation package for sequential circuits.
In Proc. European Conf. Design Automation
(EDAC), (pp.214-218).
Plakk, M., & Ubar, R. (1980). Digital Circuit Test
Design using the Alternative Graph Model. Auto-
mation and Remote Control, 41(5, Part 2), 714-
722. New York: Plenum Publishing Corporation.
Raik, J., & Ubar, R. (1998). Feasibility of Structur-
ally Synthesized BDD Models for Test Generation.
In Proc. of the IEEE European Test Workshop,
(pp.145-146).
Raik, J., & Ubar, R. (2000). Fast Test Pattern
Generation for Sequential Circuits Using Decision
Diagram Representations. Journal of Electronic
Testing: Theory and Applications – JETTA, 16(3),
213–226.
Raik, J., Ubar, R., Sudbrock, J., Kuzmicz, W., &
Pleskacz, W. (2005). DOT: New Deterministic
Defect-Oriented ATPG Tool. In Proc. of 10th
IEEE European Test Symposium, (pp.96-101).
Ravi, S., & Jha, N. K. (2001). Fast test generation
with RTL and gate-level views. In International
Test Conference, (pp.1068-1077).
Rudnick, E. M., Patel, J. H., Greenstein, G. S.,
& Niermann, T. M. (1994). Sequential circuit test
generation in a genetic algorithm framework. In
Proc. Design Automation Conference, (pp. 698-
704).
Sasao, T. (1999). Switching Theory for Logic Syn-
thesis. Amsterdam: Kluwer Academic Publishers.
Sasao, T., & Fujita, M. (Eds.). (1996). Representa-
tions of Discrete Functions. Amsterdam: Kluwer
Academic Publishers.
Srinivasan, A., Kam, T., Malik, Sh., & Bryant.
(1990). Algorithms for discrete function manipu-
lation. In Proc. of Informations Conference on
CAD, ICCAD-90, (pp.92-95).
432
Stanković, R. S., Astola, J., Stanković, M., &
Egiazarjan, K. (2002). Circuit synthesis from
Fibonacci decision diagrams. VLSI Design. Spe-
cial Issue on Spectral Techniques and Decision
Diagrams, 14, 23–34.
To, K. (1973). Fault Folding for Irredundant and
Redundant Combinational Circuits. IEEE Trans-
actions on Computers, C-22(11), 1008–1015.
doi:10.1109/T-C.1973.223637
Ubar, R. (1976). Test Generation for Digital Cir-
cuits with Alternative Graphs. In [in Russian)].
Proceedings of Tallinn Technical University,
409, 75–81.
Ubar, R. (1980). Detection of Suspected Faults
in Combinational Circuits by Solving Boolean
Differential Equations. Automation and Remote
Control, 40(11, Part 2), 1693–1703.
Ubar, R. (1996). Test Synthesis with Alternative
Graphs. IEEE Design & Test of Computers,
(Spring): 48–57. doi:10.1109/54.485782
Ubar, R. (1998). Multi-Valued Simulation of Digi-
tal Circuits with Structurally Synthesized Binary
Decision Diagrams. In Multiple Valued Logic,
(Vol.4, pp.141-157). New York: OPA (Overseas
Publ. Ass.) N.V. Gordon and Breach Publishers
Ubar, R., & Borrione, D. (1999). Single Gate De-
sign Error Diagnosis in Combinational Circuits.
Proceedings of the Estonian Acad. of Sci. Engng,
5(1), 3-21.
Ubar, R., Devadze, S., Raik, J., & Jutman, A.
(2007). Ultra Fast Parallel Fault Analysis on
Structural BDDs. IEEE Proc. of 12th European
Test Symposium – ETS’2007, (pp.131-136).
Ubar, R., Devadze, S., Raik, J., & Jutman, A.
(2008). Fast Fault Simulation in Digital Circuits
with Scan Path. In IEEE Proc. of 13th Asia and
South Pacific Design Automation Conference –
ASP-DAC’2008, (pp. 667-672).
Diagnostic Modeling of Digital Systems with Multi-Level Decision Diagrams
Ubar, R., Devadze, S., Raik, J., & Jutman, A.
(2010a). Parallel X-Fault Simulation with Critical
Path Tracing Technique. IEEE Proc. of Design
Automation and Test in Europe - DATE.
Ubar, R., Mironov, D., Raik, J., & Jutman, A.
(2010b). Structural Fault Collapsing by Super-
position of BDDs for Test Generation in Digital
Circuits. In 11th Int. Symp. on Quality Electronic
Design – ISQED’2010, (pp. 1-8).
Ubar, R., Moraviec, A., & Raik, J. (1999). Cycle-
based Simulation with Decision Diagrams. In
IEEE Proc. of Design Automation and Test in
Europe (pp. 454–458). DATE.
Ubar, R., Raik, J., Ivask, E., & Brik, M. (2002).
Multi-Level Fault Simulation of Digital Systems
on Decision Diagrams. In IEEE Workshop on Elec-
tronic Design, Test and Applications – DELTA’02,
(pp. 86-91).
Vedula, V. M., & Abraham, J. A. (2002). Program
Slicing for Hierarchical Test Generation. In Proc.
of VTS, (pp. 237-243).
Yi, J., & Hayes, J. (2006). High-Level Delay
Test Generation for Modular Circuits. IEEE
Transactions on Computer-Aided Design of In-
tegrated Circuits and Systems, 25(3), 576–590.
doi:10.1109/TCAD.2005.853697
Zhang, J. (2004). Symbolic Execution of Program
Paths Involving Pointer and Structure variables.
In 4th Int. Conf. on Quality Software – QSIC’04,
(pp. 1-6).
This work was previously published in gn and Test Technology for Dependable Systems-on-Chip, edited by Raimund Ubar, Jaan
Raik and Heinrich Theodor Vierhaus, pp. 92-118, copyright 2011 by Information Science Reference (an imprint of IGI Global).
KEY TERMS AND DEFINITIONS
Binary decision diagrams (BDD): A directed
acyclic graph based data structure that is used to
represent a Boolean function.
Fan-out free region (FFR): A subcircuit in
the given network of logic gates, which does not
include reconverging in this region fan-out stems.
Functional fault model: A model which rep-
resents any arbitrary change in the logic function
of a subcircuit or a circuit component because of
the fault, by determining (1) a signal line where
the erroneous value will appear, and (2) a logic
condition which is needed for activating the fault.
High-level DDs (HLDD): A general case of de-
cision diagrams which represent digital systems at
higher than Boolean level such as register transfer,
behaviour or transaction levels using word level
variables and expressions as node labels in DDs.
Shannon expansion: A method by which a
Boolean function can be represented by the sum
of two sub-functions of the original function.
Structurally synthesized BDDs (SSBDD):
A BDD which represents both the function and
the structure of a single FFR of a digital circuit
through one-to-one mapping between the nodes
in the BDD and the signal paths in the circuit.
Structurally synthesized multi-input BDDs
(SSMIBDD): A general case of SSBDD which
represents two or more FFRs in a single BDD
with separate root nodes for each FFR.
Superposition of BDDs: A method to construct
a composite BDD from two given simpler BDDs.
Vector high-level DDs (VHLDD): A com-
posite HLDD constructed on the basis of sharing
similar sub-HLDDs to represent concurrently and
in a compact way the high-level functions of more
than two modules of a digital system.
433
434

Chapter 26
Semantic Annotation of
Geospatial RESTful Services
Using External Resources
Victor Saquicela
Universidad Politécnica de Madrid, Spain

Luis. M. Vilches-Blázquez
Universidad Politécnica de Madrid, Spain

Oscar Corcho
Universidad Politécnica de Madrid, Spain

ABSTRACT
RESTful services are increasingly gaining traction over Web Services (WS-*). As with WS-* services,
their semantic annotation can provide benefits in tasks related to their discovery, composition, and media-
tion. In this chapter, the authors present an approach to automate the semantic annotation of geospatial
RESTful services using a cross-domain ontology like DBpedia, domain ontologies like GeoNames, and
additional external resources (suggestion and synonym services). They use combinations of these re-
sources to discover meanings for each of the parameters of the geospatial RESTful services and perform
semantic annotations of them.

INTRODUCTION Machine-oriented Web applications and APIs that

In recent years, since the advent of Web 2.0 ap-
plications and given some of the limitations of
“classical” Web services (WS-*) based on SOAP,
WSDL, and Open Geospatial Consortium (OGC)
services, Representational State Transfer (REST)
services have become an increasing phenomenon.
DOI: 10.4018/978-1-4666-2038-4.ch026
are conformant to the REST architectural style
(Fielding, 2000), normally referred to as REST-
ful Web services, have started appearing mainly
due to their relative simplicity and their natural
suitability for the Web. Nowadays the largest
online repository of information about Web 2.0
mashups and APIs is ProgrammableWeb.com.
This aggregator site provides information on 5,465
mashups and 2,647 APIs that were registered be-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Semantic Annotation of Geospatial RESTful Services Using External Resources
tween September 2005 and December 2010, as of
the time of the writing of this chapter. Mashups
tagged as “mapping” represent a 34% mashups
(2,411 mashups) of the listed ones, what represents
the importance of geospatial information in the
generation of these applications. With respect to
APIs, GoogleMaps is the most used with a 42%,
that is, this API is used on 2,149 mashups. These
data show the importance of geospatial informa-
tion in the context of the REST world.
However, using geospatial RESTful services
still requires much human intervention since the
majority of their description pages are given in the
form of unstructured text in a Web page (HTML),
which contains a list of the available operations,
their Uniform Resource Identifiers (URIs) and
parameters (also called attributes), expected
output, error messages, and a set of examples
of their execution. This hampers the discovery,
composition and mediation between services,
which may be required in the development of
applications, and which are also important tasks
in WS-* approaches.
An approach in order to solve this difficulty
is to think of semantic annotation activities of
these services.
Besides, it can also play an important role in
the Semantic Web by providing data to semantic
software agents, as can be seen in (Ferreira, et
al.,2009; Alarcon, et al., 2010).
Semantic annotation refers to the task of at-
taching meaningful descriptions to the service and
the served geospatial data or processes (Maué,
et al., 2009). Traditionally, semantic annotation
approaches for services have focused on defining
formalisms to describe services, and have been
normally applied to WS-* service description
formalisms and middleware. More recently, these
(usually heavyweight) approaches have started
to be adapted in a more lightweight manner for
the semantic description of RESTful services
(Maleshkova, et al., 2009a; Kopecký, et al., 2008;
Lathem, et al., 2007). However, most of the
processes related to the annotation of RESTful
services (Maleshkova, et al., 2009b; Alowisheq,
et al., 2009) still require a large amount of human
intervention. First, humans have to understand the
informal descriptions provided in the RESTful
service description pages, and then the semantic
annotation of RESTful services has to be auto-
mated as much as possible.
In this chapter, we address these two main
challenges by: (1) providing syntactic descrip-
tions of geospatial RESTful services that allow
their automatic registration and invocation, and
(2) interpreting and enriching the geospatial
RESTful services’ parameters, by means of their
semantic annotation.
The main contribution of our work is an ap-
proach to perform an automatic lightweight se-
mantic annotation process of RESTful services.
This process is implemented in a system, which
takes into account a cross-domain and a geospatial
ontology (DBpedia ontology combined with Geo-
Names ontology), its SPARQL Endpoint, as well
as different external resources such as synonyms
and suggestion services. We use combinations of
these resources to discover meanings for each of
the parameter of the RESTful services and perform
semantic annotations of them. These semantic an-
notations might provide benefits in tasks related
to discovery, composition, and mediation of these
services. Furthermore, semantic annotations of
RESTful services might help to developers and
users in the creation of mashups.
The remainder of this paper is structured as fol-
lows: Next, we present background in the context
of semantic annotation of WS-* and RESTful ser-
vices (from a general and geographical viewpoint).
Later, we introduce our approach for automating
the annotation of RESTful services, including
explanations on how we derive their syntactic
description and semantic annotation. Finally,
we present some conclusions of this chapter and
identify future lines of work.
435
 Semantic Annotation of Geospatial RESTful Services Using External Resources
BACKGROUND AND
RELATED WORK
In this section, we describe briefly main topics
related to our work such as: RDF, SPARQL and
REST. Likewise, we present existing approaches
associated with semantic annotations and their
relationships with our proposal.
Resource Description
Framework (RDF)
The Resource Description Framework (RDF)
(Klyne, et al., 2004) is a metadata model and
language recommended by the W3C for building
an infrastructure of machine-readable semantics
for the data on the Web, a long-term vision known
as Semantic Web. The motivation behind the
development of RDF by the W3C was, as Tim
Berners-Lee pointed out for the Semantic Web,
to have a common and minimal language to en-
able to map large quantities of existing data onto
it so that the data can be analyzed in ways never
dreamed of by its creators (Berners-Lee, 1998).
In the RDF model (Gutierrez, et al., 2005),
the universe to be modeled is a set of resources,
essentially anything that can have a universal re-
source identifier, URI. The language to describe
them is a set of properties, technically binary
predicates. Descriptions are statements very much
in the subject-predicate-object structure. In ad-
dition, the RDF specification includes a built-in
vocabulary with a normative semantics (RDFS).
This vocabulary deals with inheritance of classes
and properties, as well as typing, among other
features allowing the descriptions of concepts
and relationships that can exist for a community
of people and software agents, enabling knowl-
edge sharing and reuse. That is, using this simple
model, it allows structured and semi-structured
data to be mixed, exposed, and shared across dif-
ferent applications. More details about RDF can
be found in this website1.
436
SPARQL Query Language for RDF
SPARQL Query Language for RDF (SPARQL)
(Prud’hommeaux, et al., 2008) is a query language
for RDF. This language can be used to express
queries across diverse data sources, whether the
data is stored natively as RDF or viewed as RDF
via middleware. SPARQL contains capabilities
for querying required and optional graph patterns
along with their conjunctions and disjunctions.
SPARQL also supports extensible value testing
and constraining queries by source RDF graph.
More details about SPARQL can be found in the
W3C Specification2.
A SPARQL endpoint is a conformant SPARQL
protocol service as defined in the SPROT (stands
for SPARQL Protocol for RDF) specification.
A SPARQL endpoint enables users (human or
other) to query a knowledge base via the SPARQL
language. Results are typically returned in one or
more machine-processable formats. Therefore,
a SPARQL endpoint is mostly conceived as a
machine-friendly interface towards a knowledge
base. Both the formulation of the queries and the
human-readable presentation of the results should
typically be implemented by the calling software,
and not be done manually by human users.
Representation State
Transfer (REST)
Representation State Transfer (REST) is a style
of software architecture for distributed systems
such as the World Wide Web. This term was in-
troduced by Roy Thomas Fielding in his doctor
dissertation “Architectural Styles and the Design
of Network-based Software Architecture.” REST
emphasizes scalability of component interactions,
generality of interfaces, independent deployment
of components, and intermediary components to
reduce interaction latency, enforce security, and
encapsulate legacy systems (Fielding, 2000).
A RESTful web service is based on the fol-
lowing principles (Fielding, 2000):
Semantic Annotation of Geospatial RESTful Services Using External Resources
• Resources and representations: A re-
source is a single piece of web applica-
tion that you make accessible to clients.
Because the resources you have are not
transportable over the network, “providing
them” means providing representations of
their state.
• Addressability and connectedness:
Resources have their representations. In
REST, every resource must have at least
one address, that is, one URI. To address
the resources, you simply specify the URI.
This concept is called “addressability.”
• Uniform interface: You need a communi-
cation protocol/interface to use. In a REST
architecture, such an interface must be uni-
form which means that whatever URI you
acceded, the interface should be the same.
• Statelessness: Statelessness means that
a web application is not responsible for
keeping any information about the state
of its clients. The client is responsible for
tracking its own actions. The service main-
tains its resources and provides a uniform
interface to the clients.
The following semantics are defined (Field-
ing, 2000):
• HTTP GET is used for obtaining a rep-
resentation of a resource. A consumer
uses it to retrieve a representation from a
URI. Services provided through this inter-
face must not incur any obligation from
consumers.
• HTTP DELETE is used for removing rep-
resentations of a resource.
• HTTP POST is used for updating or creat-
ing the representations of a resource.
• HTTP PUT is used for creating representa-
tions of a resource.
Related Work
Most research in the semantic annotation of
RESTful services has focused on the definition of
formal description languages for creating semantic
annotations. The main proposed formalisms for
describing these services are: the Web Application
Description Language (WADL) which describes
RESTful services syntactically, MicroWSMO
(Maleshkova, et al., 2009c) which uses hREST
(HTML for RESTful services) (Maleshkova, et
al., 2009c; Kopecký, et al., 2008), and SA-REST
(Maleshkova, et al., 2009b; Lathem, et al., 2007)
which uses SAWSDL (Maleshkova, et al., 2009a)
and RDFa to describe service properties.
In Battle et al. (2008), the authors introduced
an approach to annotate WADL documents linking
them to ontologies. Among these approaches, some
authors propose rather heavyweight approaches
for semantic description, which are normally
derived from Web Service (WS-*) semantic de-
scription frameworks like WSMO or OWL-S.
An example is proposed in (Ferreira et al., 2009),
which makes use of a specific selection of existing
languages and protocols, reinforcing its feasibility.
Firstly, OWL-S is used as the base ontology for
services, whereas WADL is used for syntactically
describing them. Secondly, the HTTP protocol is
used for transferring messages, defining the action
to be executed, and also defining the executing
scope. Finally, URI identifiers are responsible
for specifying the service interface. Neverthe-
less, these languages are strongly influenced by
existing traditional Web Services.
Other approaches are more lightweight, for in-
stance, the proposals of Maleshkova et al. (2009a,
2009b). The authors advocate an integrated light-
weight approach for formally describing semantic
RESTful services. This approach is based on use
of the hREST and MicroWSMO microformats,
which enable the creation of machine-readable
service descriptions and the addition of semantic
annotations. Furthermore, the authors present
SWEET, a tool that effectively supports users in
437
 Semantic Annotation of Geospatial RESTful Services Using External Resources
creating semantic descriptions of RESTful ser-
vices based on the aforementioned technologies.
Our work can be considered as an extension
of the work presented in Lerman et al. (2007) and
Ambite et al. (2009), in which the development
of domain-independent approaches to semanti-
cally label Web services is described. The authors
propose to automatically learn the semantics of
information sources labelling the input and output
parameters used by the source with semantic types
of the user’s domain model. In our approach we
have dealt with RESTful services and have used
semantic repositories to try to semantically label
these services.
Another approach is presented in Heß et al.
(2003). This approach classifies data type us-
ing HTML treated Web form files as the Web
service’s parameters. They use Naïve Bayes to
classify assigned semantic types to the input and
output parameters. However, this approach does
not consider values (instances) associated with
different parameters in the annotation process of
a web service.
Likewise, our work is also similar to the ap-
proaches related to schema matching or integration
(Rahm, et al., 2001; Doan, et al., 2003). In these
proposals, the main goal is to establish semantic
mappings between two different schemas. In our
approach, the developed system sets matchings
between different parameters of a RESTful service
and the DBpedia ontology.
Regarding the geospatial approaches, in Gang
(2009) the authors illustrate using REST semantic
methods to construct geospatial web services. They
present two methods of manual mapping geospa-
tial resource to semantic world according to the
elements of REST. Moreover, in Sousa (2009), the
authors propose an approach based on Semantic
annotations that use geographic metadata and on-
tologies to describe heterogeneous geospatial data.
It uses RDF/XML files and Federal Geographic
Data Committee (FGDC) metadata schema to link
ontology terms and metadata. In our approach, we
438
only considered RESTful services without Open
Geospatial Consortium (OGC) standards.
In the same way, ESRI (2010) has developed the
GeoServices REST Specification that provides a
standard way for Web clients to communicate with
Geographic Information System (GIS) servers
through Representational State Transfer (REST)
technology. Clients issue requests to the server
through structured URLs. The server responds with
map images, text-based geographic information,
or other resources that satisfy the request.
AN APPROACH FOR AUTOMATIC
SEMANTIC ANNOTATION OF
GEOSPATIAL RESTFUL SERVICES
In this section, we present our approach visual-
ized in Figure 1 for automating the syntactic
and semantic annotation of geospatial RESTful
services. Our prototypical system consists of
three main components, including invocation
and registration, repository, and semantic annota-
tion components, which are enriched by diverse
external resources. Next, we briefly describe the
different components, illustrating the descriptions
with some sample services.
A Sample Set of RESTful Services in
the Geospatial Domain
In this section, we present two different REST-
ful services related to the geospatial domain to
guide the explanation of the proposed semantic
annotation process. These services are taken from
programmableweb.com:
• Service 1. http://ws.geonames.org/
countryInfo?country=ES
This service retrieves information related to a
‘country.’ More specifically, it returns informa-
tion about the following parameters: ‘capital,’
‘population,’ ‘area’ (km2), and ‘bounding box
Semantic Annotation of Geospatial RESTful Services Using External Resources
Figure 1. RESTful service semantic annotation system
of mainland’ (excluding offshore islands). In the
specified URL, we retrieve information about
Spain.
• Service 2. http://nominatim.openstreetmap.
org/reverse?format=xml&lat=40.4167&l
on=-3.7003& zoom=18&addressdetails=1
This service retrieves information about
places. More specifically, it returns parameters
like: ‘latitude,’ ‘longitude,’ ‘road,’ etc. In the
specified URL, we retrieve information about a
theatre (Teatro Reina Victoria, Madrid, Spain).
In these examples, each service has a URL
with different input parameters. In the case of the
‘Service 1,’ input parameter is ‘Country’ and its
associated value is ‘ES.’ The inputs parameters of
‘Service 2’ are constituted by ‘format,’ ‘lat,’ ‘lon,’
‘zoom’ and ‘addressdetails.’ These parameters are
used in our approach to automate the semantic
annotation of geospatial RESTful services.
Syntactic Description:
Invocation and Registration
Our system takes as input Web applications and
APIs, which are known by users, or users can add
manually a URL of an available RESTful service.
In this case, we add manually different URLs of
services and obtain automatically information
related to each of the aforementioned RESTful
service. Once URLs have been added, our system
invokes the RESTful service with a sample of
parameters and analyzes the response to obtain
a basic syntactic description of a parameter set,
which is used like inputs and outputs.
In this process, our system uses the Service
Data Object (SDO) API to perform the invocation
of the RESTful service and determine whether it
is available or not. SDO is a specification for a
programming model that unifies data program-
ming across data source types and provides robust
support for common application patterns in a
disconnected way (Resende, 2007). The invoca-
tion process is performed as follows: first, it takes
the input parameters and their values, which are
given to the service as part of a URL. Then, the
system invokes the service, which translates our
“RESTful service call” into a query to the specific
service, including the URL and related parameters.
The service invocation of a specific REST-
ful service may return diverse formats, such as
HTML, JSON, XML, etc. In our work, we use
only the XML response for describing the service.
439
 Semantic Annotation of Geospatial RESTful Services Using External Resources

Figure 2. XML response of two sample RESTful services

The results of invoking both services are shown
in Figure 2.
These XML responses are processed using
SDO, which enables to navigate through the XML
and extract output parameters of each service.
The result of this invocation process is a syntac-
tic definition of RESTful services in XML, which
can be expressed in description languages like
WADL or stored into a relational model. In this
work, we use a relational model as data model as
a consequence of the simplicity of WADL for
showing concepts. Figure 3shows the different
output parameters of each service. In this case,
our approach only takes into account parameters
with an associated value. Therefore, parameters
as ‘addresparts’ (see Figure 3) are not considered.
The output parameters are registered and stored
in a repository. This repository is a database spe-
cifically designed to store syntactic descriptions
of RESTful services. We selected this storage to
increase efficiency in the recovery of the REST-
ful services.
Finally, both input and output parameters are
considered as inputs in our semantic annotation
process.

Figure 3. Syntactic description of RESTful service

440
Semantic Annotation of Geospatial RESTful Services Using External Resources
Semantic Annotation
Without proper descriptions, the use of such a
RESTful service is limited to a small user group.
Therefore, they have to be annotated with descrip-
tive metadata to make it usable to a broad audience.
Otherwise people will neither be able to find the
resource using search engines nor will they be
able to evaluate if the discovered resource satis-
fies their current information need (Maué, 2009).
Thus, in our approach once the RESTful service
is syntactically described with all its identified
input and output parameters, we proceed into its
semantic annotation. We understand the semantic
annotation as a reference which establishes a Link
Annotation (Bechhofer, et al., 2002) between the
application specific metadata (a parameter of a
RESTful service) and a shared external vocabulary
(an ontology). We follow a heuristic approach
that combines a number of external services and
semantic resources to propose annotations for the
Figure 4. Semantic annotation process
parameters as show in Figure 4. Next, we describe
the main components of the semantic annotation.
Ontological Resources
A growing amount of geo-data, being produced
and made available using various representa-
tion formats, need to be used in the context of
domain-specific use cases where the semantic
accuracy of the data descriptions is of the essence.
Such accuracy is achieved by using controlled
vocabularies (ontologies) defined and used by
authoritative entities in each of these domains. In
our approach, we use two widespread ontologies,
such as the DBpedia and GeoNames.
DBpedia is a community effort to extract struc-
tured information from Wikipedia and to make
this information available on the Web. DBpedia
allows you to ask sophisticated queries against
Wikipedia, and to link other data sets on the Web to
Wikipedia data. We hope this will make it easier for
441
 Semantic Annotation of Geospatial RESTful Services Using External Resources
the amazing amount of information in Wikipedia
to be used in new and interesting ways, and that
it might inspire new mechanisms for navigating,
linking, and improving the encyclopaedia itself.
With respect to its ontology3, it is a shallow,
cross-domain ontology, which has been manually
created based on the most commonly used info-
boxes within Wikipedia. The ontology currently
covers over 259 classes, which form a subsumption
hierarchy and are described by 1,200 different
properties. Moreover, this ontology currently
contains about 1,478,000 instances (413,000
instances belong to “Place” class).
GeoNames is integrating geographical data
such as names of places in various languages,
elevation, population and others from various
sources. All lat/long coordinates are in WGS84
(World Geodetic System 1984). Users may manu-
ally edit, correct, and add new names using a user
friendly wiki interface. Likewise, the GeoNames
geographical database is available for download
free of charge under a creative commons attribu-
tion license.
Regarding its ontology4, it makes it possible to
add geospatial semantic information to the World
Wide Web. All over 6.2 million GeoNames top-
onyms have a unique URL with a corresponding
RDF web service.
Using Semantic Sources in
the Annotation Process
At this stage, the list of syntactic parameters
obtained previously is used to query the DB-
pedia and GeoNames SPARQL Endpoints and
retrieve associated results for each parameter.
A SPARQL endpoint enables users (human or
other) to query a knowledge base via the SPARQL
language. Results are typically returned in one or
more machine-processable formats. Therefore,
a SPARQL endpoint is mostly conceived as a
machine-friendly interface towards a knowledge
base. Both the formulation of the queries and the
human-readable presentation of the results should
442
typically be implemented by the calling software,
and not be done manually by human users. Next,
we describe the process:
• First, the system retrieves all the classes
from the DBpedia ontology whose names
have an exact match with each parameter
of the RESTful service. If the system ob-
tains correspondences from the matching
process, it uses these DBpedia concepts
individually to retrieve samples (concept
instances) from the DBpedia SPARQL
Endpoint. Likewise, when a parameter
matches an ontology class related to some
geospatial information; such as latitude,
longitude, or bounding box, our system
retrieves samples from the GeoNames
SPARQL Endpoint. The resulting informa-
tion (RDF) is suggested automatically to
the system and registered as a possible val-
ue for the corresponding parameter. When
a parameter matches more than once in the
DBpedia ontology, our system only con-
siders concepts that have information (in-
stances), and automatically discards those
ontology concepts without instances.
In order to retrieve information about identi-
fied parameters of RESTful services the system
has registered the DBpedia and/or GeoNames
SPARQL Endpoint as a service. This service
enables automatically invocation of SPARQL
queries over DBpedia and GeoNames Endpoints.
Next, we present queries used by the system for
retrieving DBpedia/GeoNames information.
This SPARQL query (see Figure 5) enables
to retrieve classes of the DBpedia or GeoNames
ontology. The results of this query are compared
to the concepts with each parameter of a service.
• Next, the system tries to find correspon-
dences between parameters of the RESTful
service and DBpedia properties. If the sys-
tem obtains some correspondences, it uses
Semantic Annotation of Geospatial RESTful Services Using External Resources

Figure 5. SPARQL query for retrieving classes of the ontologies (filter omitted)

these DBpedia properties individually to
retrieve information of the DBpedia or
GeoNames SPARQL Endpoint, as de-
scribed above. Furthermore, this informa-
tion is registered as a possible correct value
for the corresponding parameter.
This SPARQL query (see Figure 7) enables to
retrieve possible values for a certain class of the
previous ontologies.
This SPARQL query (see Figure 8) enables to
retrieve possible values for a certain property of
the aforementioned ontologies.

This SPARQL query (see Figure 6) enables to
retrieve properties of the DBpedia or GeoNames
ontology. The system uses results to compare them
with each parameter identified in the syntactic
description.
• Finally, with the obtained classes and prop-
erties, the system calls the DBpedia and
GeoNames SPARQL Endpoints to retrieve
values (instances) for those classes and
properties, so that now we have possible
values for them. Next we show some query
examples:
Enriching the Semantic Annotations
Since we request exact matches with DBpedia
(and GeoNames) classes and properties, our
system does not establish correspondences with
ontology classes or properties for all parameters of
the RESTful service, since there are many lexical
and syntactic variations that the parameter names
may have, and because in some cases the informa-
tion that is being requested may not be available
in any of the external sources that are consulted.
In order to annotate semantically the parameters
that did not match any DBpedia resource, we use
additional external services to enrich the results.

Figure 6. SPARQL query for retrieving properties of the ontologies (filter omitted)

Figure 7. SPARQL query for retrieving possible values for a class

443
 Semantic Annotation of Geospatial RESTful Services Using External Resources
Figure 8. SPARQL query for retrieving possible values for a property
Below we describe the main characteristics of the
external services that are considered in the system.
Spelling Suggestion Services: Web search
engines (e.g. Google, Yahoo, and Microsoft)
usually try to detect and solve users’ writing
mistakes. Suggestions services, also called “Did
You Mean,” are spelling algorithms, which aim
at solving these spelling mistakes. For example,
when a user writes ‘countryName’these algorithms
suggest ‘country’ and ‘name’ separately.
In our system, we use the Yahoo Boss service
to retrieve suggestions about the parameters that
we have obtained in the previous steps and for
which we have not obtained any candidate in our
semantic resources. Thus, for each parameter that
the system did not find a correspondence with
classes or properties in DBpedia (nor GeoNames),
this service is invocated for obtaining a list of
suggestions to query DBpedia (and GeoNames)
again. The output is registered and stored into
the repository. Following the previous example,
the parameter ‘countryName’ is not found in the
DBpedia ontology. Nevertheless, the added service
allows separating this parameter in ‘country’ and
‘name,’ and then it calls to the DBpedia SPARQL
Endpoint for obtaining results.
Use of Synonyms: This external service is
incorporated into the system to retrieve possible
synonyms for a certain parameter. This service
tries to improve the semantic annotation process
when our system does not offer results for the
previous steps, that is, when we still have param-
eters in a RESTful service without any potential
annotations.
As an example, we may have a parameter
called ‘address.’ The invocation process uses the
synonyms service to retrieve a set of synonyms
444
of ‘address’ such as extension, reference, men-
tion, citation, denotation, destination, source, cite,
acknowledgment, and so on. These outputs are
registered and stored into the repository, and then,
the service calls to the DBpedia (and GeoNames)
SPARQL Endpoints for results.
Checking the Semantic
Annotation of RESTful Services
In order to check the collected sample individuals
and initial semantic annotations of the previous
process, our system invocates the RESTful service
that was already registered in the repository and
validates the input and output parameters.
For the validation of the input parameters, our
system selects, for each parameter, a random subset
of the example instances (of classes and/or proper-
ties) coming from the DBpedia (and GeoNames)
ontology that we have obtained and registered
before. Next, it makes several invocations of the
RESTful service iterating over these registered
values. The system does not check this with all the
possible combination of collected instances for all
parameters for two reasons: first, because of the
combinatorial explosion that may be produced in
such a case, and second because many RESTful
services have invocation limitations.
If the service returns results from the invoca-
tion, then the service is deemed as executable,
and the corresponding annotations are marked as
valid. If a service cannot be invoked successfully,
the service is classified as non-executable and is
automatically discarded from the list of services
that can be automatically annotated.
For the validation of the output parameters,
our system only takes into account executions
Semantic Annotation of Geospatial RESTful Services Using External Resources
with the correct inputs from the input sets that
have been considered before. Next, the system
compares the outputs obtained after execution with
the information already stored in the repository
due to the initial retrieval processes done before
with DBpedia (and GeoNames), and external
utility services. If the output can be matched, our
system considers the output annotation as valid.
Finally, the correspondences that have been
established between the different parameters
of the RESTful service and the DBpedia (and
GeoNames) ontology are registered and stored
in the repository, so that they can be used later.
In such a way, the RESTful service is annotated
semantically and it will allow generating semantic
descriptions or annotations of any of the types that
were identified in the background section (WADL,
hREST, SA-REST, MicroWSMO, etc.). Figure 9
provides an abbreviated form of this description
for our exemplar “Service 2.” In this description,
inputs/outputs parameters are associated with
classes and properties of the aforementioned on-
tologies (DBpedia or GeoNames). For instance, the
“Country” parameter is related to www.dbpedia.
org/resource/Country.
Figure 9. Semantic annotation of a RESTful service
EXPERIMENTAL RESULTS
In order to evaluate our approach we use 12 dif-
ferent RESTful services founded in http://www.
programmableweb.com/, which are characterized
to contain geospatial information. The list of
RESTful services can be seen in this website5.
This analysis follows the three steps described in
the semantic annotation. First, our system identi-
fies correctly 16 parameters calling directly the
DBpedia and GeoNames ontologies, but it fails to
recognize 161 parameters. Second, the system uses
the suggestion service and calls the used ontolo-
gies. In this case, it identifies 41 correspondences,
but it fails to recognize 120 parameters. Third, the
system uses the synonyms service and calls the
ontologies. It identifies 19 correspondences, but
fails to recognize 101. A detailed view of these
results is shown in Figure 10.
We cannot guarantee the success of the system
in all the cases, because in some cases the system
has not found any correspondence between REST-
ful service parameters and the concepts or prop-
erties of the DBpedia and GeoNames ontologies.
After having analysed our findings, we have
seen that some parameters are useless, because
they refer to a navigation process through RESTful
445
 Semantic Annotation of Geospatial RESTful Services Using External Resources
Figure 10. Results of the service test
service results, for example: page, total, hits, etc.
These parameters make it difficult to carry out the
right annotation semantic process. We are planning
to discard these types of parameters in the future.
To the best of our knowledge, there are no
available results from existing research works
to compare our results against. Likewise, these
preliminary results prove the feasibility of our
system and highlight that is possible to carry out
an automatic semantic annotation of RESTful
services.
CONCLUSION
In this chapter, we have proposed an approach to
perform an automatic semantic annotation process
of RESTful services. This process is implemented
in a system that takes into account the DBpedia
ontology and its SPARQL Endpoint, for gen-
eral annotation, and GeoNames and its SPARQL
Endpoint for geospatial specific results, as well
as different external resources such as synonyms
and suggestion services. We use combinations
of these resources to discover meanings for each
of the parameters of the RESTful services that
a user may select and perform semantic annota-
tions of them.
446
To illustrate our work and guide the explana-
tions of the proposed semantic annotation process
we have used two exemplary RESTful services
related to the geospatial domain. Besides, we
have presented some preliminary experimental
results that prove the feasibility of our approach,
at least in the geospatial domain, and show that
it is possible to carry out a semantic annotation
of RESTful services automatically, again at least
in this domain.
Future work will focus on the development of
a GUI that will ease the introduction of existing
services by users for their semantic annotation,
probably incorporated in any existing RESTful
semantic annotation tool/utility suite. Further-
more, we also plan to make improvements to the
proposed system, related to the matching process
and the use of similarity metrics, so as to improve
the results that have been demonstrated in our
evaluation. In the same sense, we also aim at
improving in the SPARQL queries to DBpedia
and other semantic resources that may be associ-
ated or not to a specific domain to better explore
the knowledge of this resource in the annotation
process, and optimize the use of suggestion and
synonyms services. Additionally, we will adapt the
described process to add semantic annotation to the
Open Geospatial Consortium services; concretely
Semantic Annotation of Geospatial RESTful Services Using External Resources
we will focus on Web Feature Services (WFS).
Moreover, we will incorporate ontology domains
in the semantic process for taking advantage of
specific domain characteristics. Finally, we are
planning to generate outputs of semantic annota-
tion process based on different formats, such as:
hRESTS, SA-REST and MicroWSMO. These
files (outputs) will be published in a SPARQL
Endpoint and/or in a public repository.
REFERENCES
Alarcon, R., & Wilde, E. (2010). Linking data from
RESTful services. Linked Data on the Web (LDOW
2010). Retrieved from http://www.linkeddata.org.
Alowisheq, A., Millard, D. E., & Tiropanis, T.
(2009). EXPRESS: Expressing restful semantic
services using domain ontologies. In Proceedings
of the International Semantic Web Conference
2009, (pp. 941-948). ACM.
Ambite, J. L., Darbha, S., Goel, A., Knoblock,
C. A., Lerman, K., Parundekar, R., & Russ, T. A.
(2009). Automatically constructing semantic web
services from online sources. In Proceedings of
the International Semantic Web Conference, (pp.
17-32). ACM.
Battle, R., & Benson, E. (2008). Brinding the
semantic web and web 2.0 with representational
state tranfer (REST). Web Semantics, 6, 61–69.
doi:10.1016/j.websem.2007.11.002
Bechhofer, S., Carr, L., Goble, C. A., Kampa, S.,
& Board, T. M. (2002). The semantics of semantic
annotation. In Proceedings of On the Move to
Meaningful Internet Systems, 2002 - DOA/Coo-
pIS/ODBASE 2002 Confederated International
Conferences DOA, CoopIS and ODBASE 2002
(pp. 1152–1167). London, UK: Springer-Verlag.
doi:10.1007/3-540-36124-3_73
Berners-Lee, T. (1998). Principles of design.
Retrieved from http://www.w3.org /DesignIssues/
Principle s.html.
Doan, A., Domingos, P., & Halevy, A. Y. (2003).
Learning to match the schemas of data sources:
A multistrategy approach. Machine Learning,
50(3), 279–301. doi:10.1023/A:1021765902788
ESRI. (2010). GeoServices REST specifica-
tion version 1.0. White Paper. Retrieved from
http://www.esri.com /library/whitepapers/pdfs/
geoservices-rest-spec.pdf.
Ferreira Filho, O. F., & Grigas Varella Ferreira,
M. A. (2009). Semantic web services: A RESTful
approach. Paper presented at the IADIS Inter-
national Conference WWW/INTERNET 2009.
Rome, Italy.
Fielding, R. (2000). Architectural styles and the
design of network-based software architectures.
PhD Dissertation. Irvine, CA: University of
California.
Gang, Y. (2009). A research on semantic geospa-
tial web service based REST. In Proceedings of
the International Forum on Computer Science-
Technology and Applications, 2009, (vol 1), (pp.
208-210). IFCSTA.
Gutierrez, C., Hurtado, C., & Vaisman, A. (2005).
Lecture Notes in Computer Science: Vol. 3532.
Temporal RDF (pp. 93–107). Berlin, Germany:
Springer.
Heß, A., & Kushmerick, N. (2003). Learning
to attach semantic metadata to web services. In
Proceedings of the International Semantic Web
Conference 2003. ACM.
Klyne, G., & Carroll, J. J. (Eds.). (2004). RDF
concepts and abstract syntax. W3C Recommenda-
tion. Retrieved from http://www.w3c.org.
447
 Semantic Annotation of Geospatial RESTful Services Using External Resources
Kopecký, J., Gomadam, K., & Vitvar, T. (2008).
hRESTS: An HTML microformat for describing
RESTful web services. In Proceedings of Web
Intelligence, (pp. 619-625). Web Inelligence.
Lathem, J., Gomadam, K., & Sheth, A. P. (2007).
SA-REST and (s)mashups: Adding semantics to
RESTful services. In [ICSC.]. Proceedings of
ICSC, 2007, 469–476.
Lerman, K., Plangprasopchok, A., & Knoblock, C.
A. (2007). Semantic labeling of online informa-
tion sources. International Journal on Semantic
Web and Information Systems, 3(3), 36–56.
doi:10.4018/jswis.2007070102
Maleshkova, M., Gridinoc, L., Pedrinaci, C.,
& Domingue, J. (2009c). Supporting the semi-
automatic acquisition of semantic RESTful service
descriptions. Paper presented at European Seman-
tic Web Conference 2009. Geneva, Switzerland.
Maleshkova, M., Kopecky, J., & Pedrinaci, C.
(2009a). Adapting SAWSDL for semantic annota-
tions of RESTful services. Paper presented at the
Workshop: Beyond SAWSDL at OnTheMove
Federated Conferences & Workshops. Vilamoura,
Portugal.
Maleshkova, M., Pedrinaci, C., & Domingue,
J. (2009b). Semantically annotating RESTful
services with SWEET. Paper presented at the 8th
International Semantic Web Conference. Wash-
ington, DC.
Maué, P. (Ed.). (2009). Semantic annotations in
OGC standards. OGC Discussion Paper 08-167.
Retrieved from http://portal.opengeospatial.org /
files/?artifact_id=34916.
This work was previously published in Discovery of Geospatial Resources: Methodologies, Technologies, and Emergent Ap-
plications, edited by Laura Díaz, Carlos Granell and Joaquín Huerta, pp. 156-171, copyright 2012 by Information Science
Reference (an imprint of IGI Global).
448
Prud’hommeaux, E., & Seaborne, A. (2008).
SPARQL query language for RDF. Retrieved from
http://www.w3.org /TR/2008/REC-rdf-sparql-
query-20080115/.
Rahm, E., & Bernstein, P. (2001). On matching
schemas automatically. The VLDB Journal, 10(4).
doi:10.1007/s007780100057
Resende, L. (2007). Handling heterogeneous data
sources in a SOA environment with service data
objects (SDO). In Proceedings of the ACM SIG-
MOD International Conference on Management
of Data, (pp. 895-897). ACM.
Sousa, S. R. (2009). A semantic approach to
describe geospatial resources. In Proceedings of
the ER 2009 Workshops (CoMoL, ETheCoM, FP-
UML, MOST-ONISW, QoIS, RIGiM, SeCoGIS) on
Advances in Conceptual Modeling—Challenging
Perspectives, (pp. 327-336). Berlin, Germany:
Springer.
ENDNOTES
1
http://www.w3.org/RDF/
2
http://www.w3.org/TR/rdf-sparql-query/
3
http://wiki.dbpedia.org/Ontology
4
http://www.geonames.org/ontology/ontol-
ogy_v2.2.1.rdf
5
http://www.oeg-upm.net/files/RESTFulAn-
notationWeb/RESTfulservices.xlsx
 449

Chapter 27
Framework for Graphical
User Interfaces of Geospatial
Early Warning Systems
Martin Hammitzsch
Helmholtz Centre Potsdam, GFZ German Research Centre for Geosciences, Germany

ABSTRACT
An important component of Early Warning Systems (EWS) for man-made and natural hazards is the
command and control unit’s Graphical User Interface (GUI). All relevant information of an EWS is
concentrated in this GUI and offered to human operators. However, when designing the GUI, not only
the user experience and the GUI’s screens are relevant, but also the frameworks and technologies that
the GUI is built on and the implementation of the GUI itself are of great importance. Implementations
differ based on their applications in different domains but the design and approaches to implement the
GUIs of different EWS often show analogies. The design and development of such GUIs are performed
repeatedly on some parts of the system for each EWS. Thus, the generic GUI framework of a geospatial
EWS for tsunamis is introduced to enable possible synergistic effects on the development of other new
related technology. The results presented here could be adopted and reused in other EWS for man-made
and natural hazards.

INTRODUCTION disaster management related activities regarding

As part of an Early Warning and Mitigation System
(EWMS), the Decision Support System (DSS)
provides processing, assessment, visualization,
decision support, analysis, warning and man-
agement functions for the purpose of supporting
DOI: 10.4018/978-1-4666-2038-4.ch027
threats (Raape et al., 2010) of natural or man-
made hazards such as tsunamis. Thus, the DSS
is intended to help the officer on duty to become
aware of a current situation, to assess incoming
information, to exploit synergies of information
fusion and analysis, to assess the impact and con-
sequences and to make informed decisions (Raape

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems
et al., 2010). As a result, the DSS and especially its
user interface are taken into account for a tsunami
EWMS initially designed and implemented in the
project Distant Early Warning System (DEWS).
Hammitzsch et al. (2010) outline that the DEWS
project has the objective of creating a new gen-
eration of interoperable early warning systems
based on an open sensor platform. This platform
integrates Open Geospatial Consortium (OGC)
Sensor Web Enablement (SWE) compliant sen-
sor systems for the rapid detection of hazardous
events, such as earthquakes, sea level anomalies,
ocean floor occurrences and ground displacements
for the case of tsunami early warnings. Based on
the upstream information flow, DEWS focuses
on the improvement of downstream capacities
for warning centers, especially by improving
information logistics for effective and targeted
warning message aggregation for a multilingual
environment. Even though DEWS was primarily
focused on tsunami early warning in the Indian
Ocean region, Lendholt and Hammitzsch (2011)
explain that the system offers a modular design
that serves as a reference architecture for early
warning systems independent of the hazard type
and region. A generic reference architecture for
early warning systems should have components
designed for serving in new deployments and
new sites without re-programming or compila-
tion procedures. Instead, deployment-specific
add-ons or plug-ins should be easy to add in
a configurable system based on the needs of a
specific scenario. This means, also, that the user
interface has to be not bound to specific hazard
characteristics and must be developed and imple-
mented independently from specific infrastructure
characteristics. Wächter et al. (2011) report that
these developments are resumed and continued by
the project Collaborative, Complex, and Critical
Decision-Support in Evolving Crises (TRIDEC,
TRIDEC 2010) focusing on real-time intelligent
information management and including the design
and implementation of a robust and scalable ser-
vice infrastructure supporting the integration and
450
utilization of existing resources with accelerated
generation of large volumes of data.
METHODOLOGY
After providing the motivation for the presented
results and the application of free and open source
software (FOSS), the technological framework for
a command and control unit’s GUI to be applied
in each EWS for man-made and natural hazards
is introduced with a simplified architecture for
the Command and Control User Interface (CCUI)
used in natural crisis management (NCM) for
tsunamis developed in DEWS and TRIDEC. In
this context, the selected GUI frameworks and geo
frameworks are introduced together with an ap-
plication oriented Geographic Information System
(GIS) and the relevant standards and technologies.
MOTIVATION AND PRECONDITIONS
In ‘The Global Survey Of Early Warning Systems,’
Annan (2006) concludes that considerable short-
comings and gaps remain, especially in developing
countries, where basic capacities, equipment and
resources are often not available. The UN (2006)
reports that systems for some hazards, such as
tsunamis, are often absent. In this regard, Löwe et
al. (2011) summarize that the Boxing Day Tsunami
of 2004 killed over 240,000 people in 14 coun-
tries and inundated the affected shorelines with
waves reaching heights of up to 30 m. Whereas
tsunami early warning capabilities have improved
in the meantime by continuing the development
of modular Tsunami Early Warning Systems
(TEWS), recent tsunami events, such as the Chile
2010 and the Honshu 2011 tsunami, demonstrate
that there are still shortcomings in the systems.
Furthermore, most of the disaster management
systems currently in place are proprietary systems
(Chen et al., 2010). These proprietary systems are,
however, challenged by limitations such as high
Framework for Graphical User Interfaces of Geospatial Early Warning Systems
cost, low flexibility, and constrained compatibil-
ity (Ashcroft et al., 2002; Burghardt, 2006). For
example, Chen et al. (2010) exemplify adopting
the FOSS concept due to following reasons: to
evade many of the existing challenges, such as a
lack of funds to generate proprietary software; to
endorse greater flexibility, wealth of knowledge,
and improved compatibility; and to make wide-
spread deployment and customization a reality.
In this regard, Steiniger and Bocher (2009)
argue that the development of FOSS has experi-
enced a boost over the last few years. The variety
of FOSS that can be found on desktop computers
ranges from word processors and web browsers
to drawing and scientific applications. In the GIS
domain, the widespread use of FOSS is apparent
as well. This rise in popularity of free GIS tools
can be measured using four indicators. The first
indicator is the number of projects started in the last
few years. The second indicator is the increasing
financial support by governmental organizations
for the foundation of FOSS GIS projects. The third
indicator is the download rates of free desktop
GIS software. Finally, a fourth indicator is an
increasing number of cases of using open source
GIS software. Along with this trend towards the
application of FOSS goes the number of research
publications that mentions the use of open source
software tools and libraries (i.e., Mitasova & Ne-
teler, 2004). Moreover, software and algorithms
developed in research projects are increasingly
published under open source licenses (e.g., Badard
& Braun, 2003, Pebesma, 2004, Burghardt et al.,
2005, Buliung & Remmel, 2008). As such, it is
important to note that the FOSS movement that
postulates freedom of use and modification for
software is not restricted to software only. Rather,
one regards free software as ‘the foundation of a
learning society where we share our knowledge
in a way that others can build upon’ (Lee, 2011).
Hence, this movement also includes the free avail-
ability of data that forms a basis for our knowledge.
Reid and Martin (2001) discussed the potential of
open source software for implementing spatial data
infrastructures (SDIs). They conclude that ‘open-
source software provides great potential to make
available components for SDI implementations
that are affordable by resource poor organizations’
and add that ‘many building blocks for SDIs are
already available.’ Ramsey (2007) remarks that
‘existing [FOSS GIS] products are now entering a
phase of rapid refinement and enhancement’. He
further emphasizes that ‘open source [GIS] soft-
ware can provide a feature-complete alternative
to proprietary software in most system designs.’
Thus, a FOSS concept is followed to increase
the utility of the CCUI and to leverage benefits such
as low cost, flexibility, and improved compatibil-
ity. Furthermore, the presented framework based
on FOSS tries to fill the gaps outlined partially
by providing an approach to copy findings and
development to implement a CCUI for natural
and man-made hazards.
COMMAND AND CONTROL
USER INTERFACE
Based on profound FOSS community activities
and results described hereinafter, the CCUI will be
based on the Eclipse Rich Client Platform (RCP)
and the User-friendly Desktop Internet GIS (uDig)
Software Development Kit (SDK), with their
behind frameworks and standards. Hammitzsch
et al. (2010) summarize that existing standards
have been integrated, wherever possible. The
CCUI integrates various services standardized
by OGC. Using the Web Map Service (WMS)
and Web Feature Service (WFS) specifications,
spatial data are utilized to present the situation
picture and to integrate a simulation system via
the Web Processing Service (WPS), to identify the
affected areas. Warning messages are compiled and
transmitted using the Common Alerting Protocol
(CAP) for the content together with addressing
information using the Emergency Data Exchange
Language – Distribution Element (EDXL-DE).
Both standards are specified by the Organization
451
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems
for the Advancement of Structured Informa-
tion Standards (OASIS). Internal interfaces are
realized with SOAP web services. Standards
and functionalities are implemented mainly via
plug-ins, a central concept of the Eclipse RCP
that enables the building of platform applications
with user interfaces. Figure 1 depicts the stack of
frameworks, libraries, technologies and plug-ins
adopted and realized in the CCUI.
Atop of various library and functionality plug-
ins, the CCUI reveals its functionalities to the
operator via a GUI assembled by so-called per-
spectives (Figure 1, Element 1). Each perspective
determines visible actions and provides views
within a window that regard the dedicated major
tasks of the operator. The composition of the GUI’s
graphical part (i.e., the perspectives) is described
in Hammitzsch et al. (2012) and forms the basis
for the technological approach presented in this
paper. Hammitzsch et al. (2012) explain that the
following main perspectives support operators
doing their tasks with EWSs for man-made and
natural hazards. The Monitoring Perspective
Figure 1. CCUI framework stack
452
(Figure 1, Element 2), exemplarily shown in
Figure 2, provides a survey of a specific area and
contributes an overall situation picture to the
operator, with geo-spatial information to track
running events. The Forecasting Perspective
(Figure 1, Element 3), exemplarily shown in
Figure 3, supports the operator in analyzing dif-
ferent probable forecasts provided by the simula-
tion system. The Message Composition Perspec-
tive (Figure1, Element 4), exemplarily shown in
Figure 4, facilitates the operator to prepare and
send warning messages or system messages. The
Dissemination Perspective (Figure 1, Element 5)
provides a comprehensive overview of the status
of disseminated messages sent through the dif-
ferent dissemination channels and allows observa-
tion of all disseminations initiated for specific
user groups. The Situation Picture Perspective
(Figure 1, Element 6) also known as the Map
Perspective is included via uDig and provides a
set of preconfigured and selected thematic map
layers, also allowing the incorporation of dy-
namic additional information on the map to ana-
Framework for Graphical User Interfaces of Geospatial Early Warning Systems

Figure 2. CCUI monitoring perspective with exemplary setup for Thailand developed in DEWS

lyze the specifics of a given situation. The Main-
tenance Perspective (Figure 1, Element 7) enables
the operator to maintain sensors and sensor net-
works by means of requesting sensor observations
and planning services as well as rebinding and
disconnecting them. The Administration Perspec-
tive (Figure 1, Element 8) allows the administra-
tion of configurations and settings needed for
information logistics. Additionally, each of the
perspectives ships its own libraries, employed

Figure 3. CCUI forecasting perspective with evaluation setup for Turkey developed in TRIDEC (Ozel
et al., forthcoming)

453
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems
Figure 4. CCUI message composition perspective with validation setup for Indonesia developed in
continued DEWS activities (Hammitzsch et al., 2012)
solely by the relevant perspective. This distribu-
tion allows the deployment of each perspective
as a standalone application or, as outlined here,
in a compound application together with other
perspectives.
CCUI Application Framework
The CCUI application framework (Figure 1,
Element 9) forms the basis of the introduced
framework stack and is composed of the Java
Platform Enterprise Edition (Java EE; Figure 1,
Element 10) together with components built upon
GeoTools including a GeoAPI implementation
(Figure 1, Element 11), Eclipse RCP (Figure 1,
Element 12) and uDIG (Figure 1, Element 13).
These components form the basis for the imple-
mentation of CCUI specific components. The
resulting platform that is employed to build the
CCUI on top is introduced briefly in this section.
454
Eclipse and Eclipse Rich
Client Platform
Aniszczyk (2010) describes Eclipse as an open
source community focused on providing an
extensible development platform for building
software and developing a universal platform
for application frameworks and exemplary tools
that make it easy and cost-effective to build and
deploy software. There is a large consortium of
major software vendors, solution providers, cor-
porations, educational and research institutions,
and individuals working cohesively to create an
ecosystem that enhances and cultivates the Eclipse
platform with complementary products, capabili-
ties, and services.
While the Eclipse platform is designed to
serve as an open tools platform, it is architected
so that its components could be used to build
almost any client application. The minimal set of
Framework for Graphical User Interfaces of Geospatial Early Warning Systems
plug-ins needed to build a rich client application is
collectively known as the Eclipse RCP. Applica-
tions can be built using a subset of this platform.
This rich set of applications is still based on the
dynamic plug-in model, and the user interface is
built using the same toolkits and extension points.
The layout and function of the workbench is under
the fine-grained control of the plug-in developer,
in this case (Irawan, 2009).
With regard to Eclipse RCP, Irawan et al. (2010)
comments that many people that have built, or
are building RCP applications state that the main
value they get from using the RCP is that it allows
them to build a professional-looking application
quickly, with a native look-and-feel, on multiple
platforms, allowing them to focus on their value
added. They appreciate that the components that
form the RCP are of high quality, are actively
maintained, and are open source. They often
discover, after the initial adoption of RCP, that
there are many other Eclipse components avail-
able. Several individuals have also discovered
that the inherent extensibility of Eclipse allows
them to build not only a closed-form product but
also an open-ended platform in their own domain.
User-Friendly Desktop Internet GIS
‘uDig,’ its Software Development Kit
and the Eclipse RCP Target Platform
uDig is an open source desktop application frame-
work, built with RCP technology. uDig can be used
as a stand-alone application, can be extended with
RCP plug-ins and can be used as a plug-in for an
existing RCP application (Refractions, 2010a).
Thus, uDig is used as an off-the-shelf product,
and it also provides an environment to imple-
ment solutions for domain specific applications
(WhereGroup, 2010). Refractions (2010a) reports
that the goal of uDig is to provide a complete Java
solution for Desktop GIS data access, editing, and
viewing. uDig aims to be the following: user-
friendly, providing a familiar graphical environ-
ment for GIS users; desktop located, running as
a thick client, natively on Windows, Mac OS/X
and Linux; Internet oriented, adhering to standards
such as OGC WMS, WFS, Web Coverage Service
(WCS) and de facto geospatial web services such
as GeoRSS, Keyhole Markup Language (KML)
and tiles; and GIS ready, providing the framework
on which complex analytical capabilities can be
built and gradually subsuming those capabilities
into the main application.
Thus, the integration of uDig as an open source
GIS framework in Eclipse RCP based applica-
tions saves efforts on developing necessary GIS
functionality and leverages the flexibility of the
underlying Eclipse framework. However, Steini-
ger and Bocher (2009) also point out that there are
two perceived disadvantages that result from the
use of Eclipse RCP as the core component: The
first disadvantage is the size of the application,
and the second disadvantage is that the GUI is
quite similar to the development environment for
programming and thus, may be too complex for
end users. On the positive side, the user and the
developer documentation are very good, and the
project management and software development
processes are outlined and documented.
Using Eclipse RCP, the development of uDig
started between 2004 and 2005 and has been
initiated by Refractions Research Inc. in Canada,
which also develops the spatial extension PostGIS
for the PostgreSQL database (Steiniger & Bocher,
2009). Based on Eclipse RCP, uDig is developed
in Java and was initially focused on the editing
of vector data. However, since 2007 the uDig
team has been joined by the JGrass team, which
works on raster analysis functionality (Antonello
& Francheschi, 2007). Besides JGrass, further
projects are involved in the development. Because
of the joint work of these projects and the possi-
bility of using uDig as a frontend for the PostGIS
database (which has already a very wide user
base), the user and developer community grows
consistently (Steiniger & Bocher, 2009). uDig is
further developed and maintained at the present
time under the Refraction Research and uDig user
455
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems
community to make it a better and more powerful
application. This activity follows the vision to fill
in functional gaps in the two technology com-
munities: the open source geospatial community
and the open standards geospatial community, as
represented by OGC (Tress & Moestofa, 2009).
To perform standardized operations for core GIS
functionality such as data reading, coordinate
re-projection, and rendering, uDig uses the well-
established GeoTools library (Refractions, 2010b).
Open Source Java GIS
Toolkit ‘GeoTools’
GeoTools is a free, open source Java geospatial
toolkit for working with both vector and raster
data (GeoTools, 2010) and is a library that pro-
vides classes, objects, and methods required to
implement standards conformant to geospatial
applications (Santokhee, 2008). The library con-
sists of a large number of various modules that
can be combined according to programmer needs
(Santokhee, 2008) and that allow you to do the fol-
lowing: access GIS data in many file formats and
spatial databases; work with an extensive range of
map projections; filter and analyze data in terms
of spatial and non-spatial attributes; compose and
display maps with complex styling; and create and
analyze graphs and networks (GeoTools, 2010).
Santokhee (2008) reports that GeoTools can be
readily extended by adding new modules, either
for custom applications or as contributions to the
library. GeoTools supports additional formats
through the use of plug-ins. A developer can
control the formats supported by the application
by only including the plug-ins required. Thus,
GeoTools can be used to build applications of
many different types covering most processes of
GISs, including network server applications and
desktop clients. GeoTools leverages a modern,
object oriented structure and uses common, well
documented computational design patterns. The
library consists of loosely dependent modules to
present third party programmers with a coherent,
456
easily usable code base. GeoTools implements the
core global geospatial standards of the Interna-
tional Organization for Standardization (ISO) and
of the OGC to ensure maximal interoperability
with other geospatial software. GeoTools is a
project that is growing continuously and breaking
new grounds by implementing untested standards.
The library is therefore still evolving to address
weaknesses in certain areas, to implement new
functionality, to work around defects in existing
standards and to embrace newer standards when
proposed. GeoTools provides almost no graphical
user interface functionality. Thus developers plan-
ning applications that will depend on the library
should carefully assess the relevant strengths and
weaknesses of GeoTools.
Importantly, not only is GeoTools an open
source Java GIS toolkit providing implementations
of many OGC specifications as they are developed,
but it also is associated with the GeoAPI project
that creates geospatial Java interfaces (OSGEO,
2010).
GeoAPI’s Java Language Programming
Interfaces for Geospatial Applications
In 2003 the GeoTools project started an outreach
program in the form of the GeoAPI project as an
effort to collaborate with other projects working in
the same domain and as a line of communication
to the standards community (GeoTools, 2011). The
GeoAPI provides a set of Java language program-
ming interfaces which include many of the data
structures and manipulation methods needed for
GIS applications (OGC, 2011). It provides neutral,
interface-only APIs derived from OGC and ISO
standards to reduce duplication and to increase
interoperability (Desruisseaux & Garnett, 2010).
The need for a standardized set of programming
interfaces that remains constant across different
toolkits is reflected in GeoAPI’s history, which
includes an attempt for GeoAPI to become a
standard. Desruisseaux et al. (2010) recollects
that the GeoAPI project was initially an initiative
Framework for Graphical User Interfaces of Geospatial Early Warning Systems
of various open source communities wanting to
reduce work duplication. The goal was to make
it easier to exchange pieces of software between
independent projects, so that a project does not
need to reinvent functions already provided by
another project. In 2004, GeoAPI merged with the
GO-1 initiative from OGC. In September 2004, the
creation of a GeoAPI working group has been ap-
proved by OGC voting members. In May 2005, the
GO-1 final specification, which includes GeoAPI
interfaces, finally, has been accepted as an OGC
standard. Discussing why a standardized set of
programming interfaces is required instead of the
standardized OGC web services, Desruisseaux et
al. (2010) proposes that web services are efficient
ways to publish geographic information using
existing software. However, some users need to
build their own solution, for example as a wrap-
per on top of their own numerical model. Many
existing software packages provide sophisticated
developer toolkits, but each toolkit has its own
learning curve, and one cannot easily switch from
one toolkit to another or mix components from
different toolkits. Using standardized interfaces,
a significant part of the API can remain constant
across different toolkits, thus reducing both the
learning curve and the interoperability challenges.
OGC (2011) summarizes that the GeoAPI
interfaces closely follow the abstract model and
concrete specifications published collaboratively
by the ISO in its 19100 series of documents
and the OGC in its abstract and implementation
specifications. GeoAPI provides an interpretation
and adaptation of these standards to match the ex-
pectations of Java programmers. These standards
provide GeoAPI with the richness which comes
from the expertise of the specification writers.
Clients benefit from the potential for interoper-
ability which comes from using a well-defined,
standardized data model. Implementors benefit
from having a pre-defined set of well-considered,
formal boundaries to modularize their develop-
ment work. So the GeoAPI interfaces provide a
layer which separates client code, which would call
the API, from library code, which implements the
API. These interfaces are not an implementation.
Clients can use the API without concern for the
particular implementation which they will use. One
of various groups, that have implemented differ-
ent subsets of GeoAPI, is GeoTools introduced.
Java Programming Language
and Computing Platform
Based on the use of available frameworks and
libraries, the characteristics of the underlying
technology used (i.e., Java) will be summarized,
to outline the potentials of this technology for
the future adoption of the CCUI in related EWS
application domains.
Jendrock et al. (2006) argue that developers
today increasingly recognize the need for dis-
tributed, transactional, and portable applications
that leverage the speed, security, and reliability
of server-side technology. In the world of infor-
mation technology, enterprise applications must
be designed, built, and produced for less money,
with greater speed, and with fewer resources. With
the Java EE platform, development of enterprise
applications is easy and fast. The aim of the Java
EE platform is to provide developers with a pow-
erful set of APIs while reducing the development
time, reducing the application complexity, and
improving the application performance. Further-
more, Jendrock et al. (2010) adds that the Java
EE platform is developed through the Java Com-
munity Process (JCP), which is responsible for all
Java technologies. Expert groups, composed of
interested parties, have created Java Specification
Requests (JSRs) to define the various Java EE
technologies. The work of the Java community
under the JCP program helps to ensure Java tech-
nology’s standard of stability and cross-platform
compatibility.
Importantly, Java EE is platform independent
and supports applications on various computing
platforms ranging from super computers and
enterprise servers on the one end to mobile and
457
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems
embedded devices on the other end. It is a de-facto
industry standard for developing applications. Java
EE is one of the most widely used technologies
in its field. Finally, Java EE is free and supported
by a large worldwide community and with lots
of vendors in the information technology (IT)
domain.
Generic CCUI Plug-Ins
The CCUI application framework introduced in
Figure 1 is supplemented with required plug-ins
(Figure 1, Element 14). The core plug-in (Figure
1, Element 15) provides core functionality for
graphical parts of the application such as perspec-
tives, views, and internal events. The configura-
tions plug-in (Figure 1, Element 16) contains the
central configurations and settings used to store
all of the plug-in preferences, e.g., region and
hazard dependent configurations for the differ-
ent setups depicted in Figures 2 through 4. The
libraries plug-in (Figure 1, Element 17) contains
the libraries accessed by other plug-ins, e.g., to
work with standards such as CAP, EDXL-DE
and SOAP. Further plug-ins provide function-
alities to be used optionally (Figure 1, Element
18) by other plug-ins built on top. For example,
the AspenCloud widgets (Figure 1, Element 19)
extend the available Standard Widget Toolkit
(SWT) widgets integrated in Eclipse RCP and thus
form a new widget with an entirely new capacity,
meeting the specific requirements of the CCUI,
which are related to time management. Another
optional plug-in named gazetteer (Figure 1, Ele-
ment 20) contributes to the main application by
providing access to a gazetteer service to assist
the operator in the task of navigating around the
map by typing a location name and centering the
map on its coordinates. Other optional plug-ins
might be integrated if desired.
458
EWS Specific CCUI Plug-Ins
The application plug-in (Figure 1, Element 21) is
based on the introduced application framework, the
required plug-ins and the optional plug-ins. The
application plug-in is the plug-in from which the
application is started and which forms the basis
for specific EWS implementation. On the basis of
the described layers of the CCUI framework, EWS
specifics are implemented by further plug-ins.
Special purpose plug-ins (Figure 1, Element 22)
extend functionality, for example, with extensions
to the GeoTools library (Figure 1, Element 23) and
required messaging functionality for communica-
tion (Figure 1, Element 24). Standards defined by
the OGC are covered by OGC specific plug-ins
(Figure 1, Element 25). uDig already provides
the required functionality to communicate via
WMS and WFS (Figure 1, Element 26). However,
other functionalities to access sensors via SWE
(Figure 1, Element 27) and to instruct geospatial
processing via WPS (Figure 1, Element 28) are
added by further plug-ins.
Hazard Specific CCUI Plug-Ins
Additional sensor specific plug-ins (Figure 1,
Element 29) such as the Tsunami Service Bus
(TSB) plug-in (Figure 1, Element 30) and the
earthquake (Figure 1, Element 31) and sea level
(Figure 1, Element 32) plug-ins extend the generic
and standardized access covering system and haz-
ard specifics (Figure 1, Element 33), in this case
for earthquake and sea level monitoring. Further
hazard specifics and convenience functionality
might be provided by other plug-ins such as the
tsunami functionalities plug-in (Figure 1, Element
34), which provides specific functionality to deal
with aspects related to tsunami risks in a geospatial
EWS for tsunami warning and mitigation. Also,
the already introduced perspectives that reveal the
functionality of one or more operators on duty are
hazard specific plug-ins that expose graphically
the required and implemented functionality.
Framework for Graphical User Interfaces of Geospatial Early Warning Systems
Realization as Desktop GIS
preferred to WebGIS
The CCUI is developed as a Desktop GIS utilizing
an open standards based Service Oriented Archi-
tecture (SOA) with the adoption of free and open
source technologies, software and frameworks,
as discussed earlier. This section outlines briefly
why the Desktop GIS approach has been applied.
Nalani (2009) explains that the rapid growth of
the Internet has led to an increasing concern over
the WebGIS approach, with the World Wide Web
(WWW) offering new application areas. This ap-
proach provides access to geographic information
via the WWW and is used to distribute informa-
tion to potentially numerous users. Because of
the considerable interest in geospatial data visu-
alization through web services that employ GIS
the widespread availability of internet and related
services such as WWW technology have been
widely applied in geospatial data visualization
and in providing the infrastructure to allow wide
access to geo-referenced data or GIS services.
Further, these technologies provide an advanced
means to assist with visual data exploration and
spatial decision making systems in the context
of landslides, earthquakes, flood modeling and
debris flow.
Additionally, Siddiquee et al. (2009) state that
the WebGIS is comparatively new but is a very
fast growing sub-set of GIS. WebGIS is achieving
particular significance for spatial data handling
over the web. In other words, WebGIS are used for
distributing and processing geographic informa-
tion via the Internet and the WWW. It is attain-
ing increasing momentum and acceptability for
different levels of users, such as geospatial data
handlers and producers as well as governmental
and non-governmental agencies. Because the
WebGIS is platform independent, this capabil-
ity reduces the necessity for purchasing costly
Desktop GIS software. WebGIS then refers to
GIS that uses web technologies as a method for
communication between the elements of a GIS
(Avraam, 2009).
However, the WhereGroup (2010) argues that
Desktop GIS enables a wide range of users to
process geospatial data and to utilize a wealth of
geospatial functionality based on comparatively
cheap and readily available Personal Computers
(PCs). Over time, Desktop GIS has become a typi-
cal example that represents a tendency to move
from powerful centralized computing hardware
to decentralized networks of PCs for geo-data
processing. With the introduction of distributed
architectures and web based processing services,
the use of geospatial data has moved from its lim-
ited traditional application areas to a mainstream
source of information. At the same time, the
relevance of stand-alone Desktop GIS products
is steadily declining. While the total number of
installed seats is more or less stable, most Desktop
GIS now plug into service oriented SDIs as data
sources and thus combine the advantages of local
processing capabilities with distributed architec-
tures. In this regard, Avraam (2009) summarizes
that the interaction between the components is
usually very direct in Desktop GIS. Data, maps
and analysis happen on the same computer, so
communication is done internally. Enterprise
Desktop GIS often allows the user to commu-
nicate with data remotely, and sometimes even
analysis is performed remotely. WebGIS enables
the communication of all components to happen
through the web, enabling diverse data, analysis
algorithms, users and visualization techniques
that may be hosted at any location on the web.
Summarizing the WebGIS approach, there is
a major focus on cloud computing (Armbrust et
al., 2009) in which the Desktop GIS approach
changes dramatically from a traditional mono-
lithic desktop application, accessing a few re-
mote services only, to a desktop application that
provides access to both integrated functionality
and distributed services on servers and the cloud.
With the increasing performance of accessing
network nodes in the WWW and the growing
459
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems
availability of widespread services in the cloud,
a modern GIS has to support the access and usage
of distributed services to be suitable for the future.
However, cloud computing is still an emerging
approach in its current state, and does not cover
professional requirements that need to be met
to be implemented in (enterprise) Desktop GIS.
Apparently, WebGIS solutions do not provide the
large amount of professional functionalities today
provided by Desktop GIS solutions. Also, Web-
GIS solutions are often specialized applications
with single components in what is termed GIS,
providing limited functionality tailored to specific
and not extensive application fields. Moreover,
limited web mapping applications accessing the
so-called GeoWeb (Leclerc et al., 2001) are often
misleadingly described as WebGIS. Finally, a ma-
jor drawback of the cloud dependent WebGIS is
that the response time can be long, depending on
a number of factors, such as connection capacity,
data volume, network traffic, and processor power
(Bonnici, 2005). Thus, an application specific GIS
such as an EWS for tsunamis should implement a
Desktop GIS approach based on a geo framework
that provides both access to distributed services
and access to integrated functionality, allowing
utilization of the appropriate functionality based
on the required needs.
DEVELOPMENT AND RESEARCH
CONTINUED AS FOSS INITIATIVE
Löwe et al. (2011) summarize that all develop-
ments of the CCUI are based to the largest extent
on FOSS components and industry standards. They
announce that emphasis has been and will be made
on leveraging open source technologies that sup-
port mature system architecture models wherever
appropriate. All open source software produced is
foreseen to be published on FOSSLAB, a publicly
available software repository provided by the Cen-
tre for GeoInformation Technology (CeGIT) of the
GFZ German Research Centre for Geosciences.
460
FOSSLAB is a platform for the development of
FOSS projects in a geospatial context, allowing
users to save, advance and reuse results achieved
in previous and on-going project activities and
enabling further development and collaboration
with a wide community including scientists, de-
velopers, users and stakeholders. Furthermore,
FOSSLAB constitutes an umbrella encompass-
ing a number of related geoinformatic activities,
such as the documentation of best practices for
working experiences and results with SDIs, GIS,
spatial processing on clusters and in clouds, and
geomatics related topics.
APPLICATIONS
In this regard, readers interested in the ideas cov-
ered in this paper and readers who want to make
use of those ideas in constructive ways are enabled
to act on or apply those ideas by firstly visiting the
related FOSSLAB sub-projects at www.fosslab.
org. Secondly, using the products and results of-
fered by the FOSSLAB sub-projects, asking and
answering questions, reporting bugs and making
feature requests are critical parts of the projects’
communities. Still only a few key persons repre-
sent the communities in small, gradually evolv-
ing groups and provide instructions successively
and on request. It is the user feedback which can
strongly drive the projects and the technologies.
Beyond the user activities, it would be greatly
appreciated getting help with the development
of the projects. This can mean getting involved
with the discussions on the development mailing
lists, answering user questions, providing patches
for bug-fixes or features, and helping to improve
the documentation.
In this context it is important to mention that
behind the work presented here various software
components form the overall system by provid-
ing the required data and services, which finally
expose their information and functionality to an
operator on duty (OOD) with the CCUI. These
Framework for Graphical User Interfaces of Geospatial Early Warning Systems
required software components have to be built,
deployed and configured appropriately on an op-
erating system (OS) with other suitably installed
FOSS before the CCUI can be used. Based on
activities in the DEWS successor TRIDEC stable
versions of these components, used in the problem
domain of natural crisis management for tsunamis,
are bundled irregularly together with the pre-
installed OS on a raw disk image to be used with
many virtualization software packages. Alterna-
tively, there is the opportunity to download these
components, publicly shared on the FOSSLAB
sub-projects one by one, and to install them with
the help of available instructions.
CONCLUSION
This paper covers the investigation and results
of selecting the appropriate GUI and geospatial
frameworks for the implementation of the CCUI
and exposes the generic framework designed and
developed in conjunction with the realization of the
CCUI. Specifically, the user interface framework
presented for early warning systems is an attempt
to provide an appropriate and useful concept for the
business of public safety, emergency management
and homeland security. The concept presented is
independent of a specific hazard domain, thus
provides a generic solution for specific applica-
tions in the area of EWS and might be adopted in
other systems for managing natural and man-made
hazards. Within DEWS and TRIDEC, the designed
concepts have been realized and implemented
as generic software components validated and
approved by different interest groups in several
project-internal and public live-demonstrations,
demonstrating the complete early warning system
prototype including the CCUI as a concept for an
adaptable and extensible user interface for early
warning systems (Lendholt & Hammitzsch, 2011).
In this regard, Hammitzsch et al. (2012) assert that
the overall system, including the CCUI and its
framework, has received preliminary evaluation
by the stakeholders and domain experts and is cur-
rently under further development. More extensive
evaluation and revision of this system is expected
to be completed in the future with multiple experts
from research and government agencies offering
their opinions on the system. Through open-ended
survey questions, the requirements of the system
have to be controlled and successively refined.
In addition, the unique benefits of the system in
providing a FOSS multi hazard framework concept
implementation, and the relevance of the approach
followed here should be affirmed by the respective
stakeholders. Finally, decision-makers in local
and regional authorities already have expressed
interest in the system (ICT, 2010).
ACKNOWLEDGMENT
The work presented in this paper is part of the
research and development in the DEWS project
(contract no. 045453), partially funded by the 6th
Framework Program of the European Commis-
sion. DEWS is based on the collective creativity,
idealism and optimism of a global community
sharing achievements of performed work and find-
ings with free and open source licenses. Thus, it is
not possible to acknowledge all of the individuals
and institutions that have contributed. The author
would like to express his sincere thanks to all of
those who contributed towards the development
of DEWS and its CCUI by providing the vital
foundations with their efforts and collectively
shared results. Further findings and develop-
ments are conducted within the TRIDEC project
(contract no. 258723), partially supported by the
7th Framework Program of the European Com-
mission. The author is grateful to the anonymous
reviewers for their comments, which have helped
to improve the earlier version of this paper.
461
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems
REFERENCES
Aniszczyk, C. (2010). Eclipse IDE project re-
sources. Retrieved October 31, 2010, from http://
www.ibm.com/developerworks/opensource/top-
projects/eclipse/eclipse-starthere.html
Annan, K. A. (2006). Foreword to the global
survey of early warning systems. In Proceedings
of the Third International Conference on Global
Survey of Early Warning Systems, Bonn, Germany.
Armbrust, M., Fox, A., Griffith, R., Joseph, A.
D., Katz, R. H., & Konwinski, A. …Zaharia,
M. (2009). Above the clouds: A Berkeley view
of cloud computing. Retrieved December 27,
2011, from http://www.eecs.berkeley.edu/Pubs/
TechRpts/2009/EECS-2009-28.html
Ashcroft, J., Daniels, D. J., & Hart, S. V. (2002).
Crisis Information Management Software (CIMS)
feature comparison report. Washington, DC:
National Institute of Justice.
Avraam, M. (2009). Geoweb: Web mapping and
web GIS. Retrieved November 3, 2010, from
http://michalisavraam.org/2009/03/geoweb-web-
mapping-and-web-gis/
Badard, T., & Braun, A. (2003). OXYGENE:
An open framework for the deployment of geo-
graphic web services. In Proceedings of the 21st
International Cartographic Conference, Durban,
South Africa.
Bonnici, A. M. (2005). Web GIS software com-
parison framework. Retrieved November 8, 2010,
from http://www.webgisdev.com/webgis_frame-
work.pdf
Buliung, R. N., & Remmel, T. K. (2008). Open
source, spatial analysis, and activity-travel behav-
iour research: Capabilities of the aspace package.
Journal of Geographical Systems, 10(2), 191–216.
doi:10.1007/s10109-008-0063-7
462
Burghardt, D., Neun, M., & Weibel, R. (2005).
Generalization services on the web – classification
and an initial prototype implementation. Cartog-
raphy and Geographic Information Science, 32(4),
257–268. doi:10.1559/152304005775194665
Chen, R., Thiyagarajan, T., Rao, R. H., & Lee,
J. K. (2010). Design of a FOSS system for flood
disaster management. In Proceedings of the 7th
International Conference on Information Systems
for Crisis Response and Management, Seattle, WA.
Desruisseaux, M., & Garnett, J. (2010). Home.
Retrieved October 31, 2010, from http://docs.
codehaus.org/display/GEO/Home
Desruisseaux, M., Garnett, J., et al. (2010). FAQ.
Retrieved October 31, 2010, from http://docs.
codehaus.org/display/GEO/FAQ
DEWS. (2010). Distant early warning system.
Retrieved December 17, 2011, from http://www.
dews-online.org
GeoTools. (2010). GeoTools user guide frequently
asked questions. Retrieved October 29, 2010,
from http://docs.geotools.org/stable/userguide/
faq.html
GeoTools. (2011). End of GeoAPI involvement.
Retrieved December 18, 2011, from http://geoto-
olsnews.blogspot.com/2011/01/end-of-geoapi-
involvement.html
Hammitzsch, M., Lendholt, M., & Esbrí, M. A.
(2012). User interface prototype for geospatial
early warning systems – a tsunami showcase.
Natural Hazards and Earth System Sciences, 12,
555–573. doi:10.5194/nhess-12-555-2012
Hammitzsch, M., Lendholt, M., & Wächter, J.
(2010). Distant early warning system for tsuna-
mis – A wide-area and multi-hazard approach.
In Proceedings of the Conference of the Euro-
pean Geosciences Union: Vol. 12. Geophysical
Research Abstracts. Retrieved August 17, 2010,
from http://meetingorganizer.copernicus.org/
EGU2010/EGU2010-4496.pdf
Framework for Graphical User Interfaces of Geospatial Early Warning Systems
Irawan, H., McCrary, T., Barbareau, L., et
al. (2010). RCP FAQ. Retrieved October 31,
2010, from http://wiki.eclipse.org/index.php/
RCP_FAQ#Why_should_I_build_my_applica-
tion_on_the_Eclipse_Rich_Client_Platform.3F
Irawan, H., Vogel, L., Ebert, R., et al. (2009).
Rich client platform. Retrieved October 31, 2010,
from http://wiki.eclipse.org/index.php/Rich_Cli-
ent_Platform
Jendrock, E., Ball, J., Carson, D., Evans, I., Fordin,
S., & Haase, K. (2006). The Java EE 5 tutorial:
For Sun Java System Application Server Platform
Edition 9 (Java Series). Upper Saddle River, NJ:
Prentice Hall.
Jendrock, E., Evans, I., Gollapudi, D., Haase, K.,
& Srivathsa, C. (2010). The Java EE 6 tutorial
1: Basic concepts (Java Series). Upper Saddle
River, NJ: Prentice Hall.
Leclerc, Y., Reddy, M., Iverson, L., & Heller, A.
(2001). The GeoWeb - A new paradigm for finding
data on the Web. In Proceedings of the Interna-
tional Cartographic Conference, Beijing, China.
Lee, M. (2011). What is free software and why
is it so important for society? Retrieved June 13,
2011, from http://www.fsf.org/about/what-is-
free-software
Lendholt, M., & Hammitzsch, M. (2011). Generic
information logistics for early warning systems. In
Proceedings of the 8th International Conference
on Information Systems for Crisis Response and
Management, Lisbon, Portugal.
Löwe, P., Hammitzsch, M., & Lendholt, M.
(2011, May 19-20). Significance of FOSSGIS
in the TRIDEC Project. Paper presented at the
Conference of Geoinformatics, Prague, Czech
Republic. Retrieved June 11, 2011, from http://
geoinformatics.fsv.cvut.cz/gwiki/Significance_
of_FOSSGIS_in_the_TRIDEC_Project
Mitasova, H., & Neteler, M. (2004). GRASS as
an open source free software GIS: Accomplish-
ments and perspectives. Transactions in GIS, 8(2),
145–154. doi:10.1111/j.1467-9671.2004.00172.x
Nalani, H. A. (2009). WebGIS based 3D visu-
alization of geospatial data. In Proceedings of
the Applied Geoinformatics for Society and En-
vironment Conference, Stuttgart, Germany (pp.
212-215). Retrieved November 3, 2010, from
http://www.gis-news.de/papers/AGSE_Proceed-
ings_2009_07_01_1025.pdf
OGC. (2011). GeoAPI. Retrieved December 18,
2011, from http://www.geoapi.org
OSGEO. (2010). GeoTools InfoSheet. Retrieved
October 31, 2010, from http://www.osgeo.org/
geotools
Ozel, N. M., Necmioglu, O., Yalciner, A. C.,
Kalafat, D., Yilmazer, M., & Comoglu, M. …
Erdik, M. (forthcoming, June 17-22). Tsunami
early warning in the Eastern Mediterranean, Ae-
gean and Black Sea. Paper to be presented at the
22nd International Ocean and Polar Engineering
Conference, Rhodes (Rodos), Greece.
Pebesma, E. J. (2004). Multivariable geosta-
tistics in S: The gstat package. Computers &
Geosciences, 30(7), 683–691. doi:10.1016/j.
cageo.2004.03.012
Raape, U., Teßmann, S., Wytzisk, A., Steinmetz,
T., Wnuk, M., & Hunold, M. …Jirka, S. (2010).
Decision support for tsunami early warning in
Indonesia: The role of OGC standards. In M.
Konecny, T. L. Bandrova, & S. Zlatanova (Eds.),
Geographic information and cartography for risk
and crisis management (Vol. 2, pp. 233-247).
Berlin, Germany: Springer-Verlag.
Ramsey, P. (2007). The state of open source GIS.
Paper presented at the annual Free and Open
Source Software for Geospatial Conference,
Vancouver, BC, Canada.
463
 Framework for Graphical User Interfaces of Geospatial Early Warning Systems

Refractions Research. (2010a). A GIS framework
for eclipse. Retrieved October 31, 2010, from
http://udig.refractions.net/
Refractions Research. (2010b). Developing with
uDig. Retrieved October 31, 2010, from http://
udig.refractions.net/developers/
Reid, J., & Martin, F. (2001). The open source
movement and its potential in implementing
spatial data infrastructures. Paper presented at
the International Symposium on Spatial Data
Infrastructure, Melbourne, Australia.
Results, I. C. T. (2010). New tsunami early warn-
ing system stands guard. Retrieved August 17,
2010, from http://cordis.europa.eu/ictresults/
index.cfm?tpl=article&ID=91371
Santokhee, A. (2008). GeoTools. Retrieved June
13, 2011, from http://www.resc.rdg.ac.uk/twiki/
bin/view/Resc/GeoTools
Siddiquee, Z. H., Strzalka, A., & Eicker, U.
(2009). Publication of energy consumption data of
Scharnhauser Park via Web GIS. In Proceedings
of the Conference of the Applied Geoinformatics
for Society and Environment, Stuttgart, Germany
(pp. 186-190). Retrieved November 3, 2010, from
http://www.gis-news.de/papers/AGSE_Proceed-
ings_2009_07_01_1025.pdf
Steiniger, S., & Bocher, E. (2009). An overview
on current free and open source desktop GIS de-
velopments. International Journal of Geographi-
cal Information Science, 23(10), 1345–1370.
doi:10.1080/13658810802634956
Tress, S., & Moestofa, A. (2009). uDig – An
overview of open source desktop GIS application.
In Proceedings of the Conference of the Applied
Geoinformatics for Society and Environment,
Stuttgart, Germany.
TRIDEC. (2010). Collaborative, complex, and
critical decision processes in evolving crises.
Retrieved December 17, 2011, from http://www.
tridec-online.eu
UN. (2006). Global survey of early warning sys-
tems. In Proceedings of the Third International
Conference on Early Warning, Bonn, Germany.
Wächter, J., Fleischer, J., Häner, R., Küppers,
A., Lendholt, M., & Hammitzsch, M. (2011).
Development of tsunami early warning systems
and future challenges. In Proceedings of the Con-
ference of the European Geosciences Union: Vol.
13. Geophysical Research Abstracts. Retrieved
December 17, 2011, from http://meetingorganizer.
copernicus.org/EGU2011/EGU2011-11015-2.pdf
WhereGroup. (2010). Desktop GIS. Retrieved
November 3, 2010, from http://www.wheregroup.
com/en/desktop_gis

This work was previously published in the International Journal of Open Source Software and Processes (IJOSSP), Volume 3,
Issue 4, edited by Stefan Koch, pp. 49-63, copyright 2011 by IGI Publishing (an imprint of IGI Global).

464
 465

Chapter 28
User-Friendly Geoportal
Interfaces for Geospatial
Resource Discovery
Victor Pascual Ayats
Spatial Data Infrastructure of Catalonia (IDEC), Spain

ABSTRACT
The Spatial Data Infrastructure of Catalonia (IDEC) was launched in 2002. From the beginning, the
Metadata Catalog (MC) service has been considered as one of the main pieces of the infrastructure.
Building a metadata catalog is important for any Spatial Data Infrastructure to foster resource interop-
erability and integration. In addition to organizing, classifying, and sorting metadata records, one of
the hardest parts in the IDEC was to design web applications that allow users to easily discover and
access such geospatial resources. This chapter reviews the different trends in building friendly user
interfaces of web applications to search and discovery metadata records through the evolution of user
interface of the IDEC Geoportal.

INTRODUCTION focuses more on the design of web applications

This chapter attempts to explain how to create
a web application to search metadata records
focusing on the evolution experimented by the
IDEC (Spatial Data Infrastructure of Catalonia)
metadata search client application over the past
8 years. The first part provides a general intro-
duction to the metadata catalog technologies
and the standardized process. The second part
DOI: 10.4018/978-1-4666-2038-4.ch028
and how the knowledge and observation of the
“user experience” has permitted the creation of
more useful and effective applications in terms
of user interfaces and functionalities.
To build a metadata catalog from a technologi-
cal point of view means implementing a complex
system based on ISO1 standards (International
Standards Office) for the metadata description
and Open GeoSpatial Consortium2 (OGC) speci-
fications for the catalog service implementation,
which allows geodata and geoservices providers

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
to register resources, and final users—consum-
ers—to locate and use these resources.
Nevertheless, a Spatial Data Infrastructure
(SDI) is not only a technological implementa-
tion, but also is a set of institutional agreements
and policies to identify and encourage local and
regional institutions, research centers, universi-
ties, and companies to describe—creating meta-
data—and to share its geographic information.
We might say that the IDEC has achieved some
significant results as its catalog had more than
37,000 metadata records from over 200 providers
at the end of 2010.
All of this work may go unnoticed if an SDI
is not able to build good front-end applications—
such as a shop window—to discover, find, or bind
the metadata records. We are talking about web
services and a Web Catalog Client Application
(WCCA).
This chapter explains and justifies the changes
applied during the last 8 years to the IDEC WCCA.
With every change and update, there is always
the same aim of trying to facilitate access to the
information in the most accessible and easy way
possible. All of these changes have been follow-
ing the same trend, decreasing the complexity and
search options in the design of the web interface,
while the server processes—hidden from the end
user—have been increasing exponentially.
This trend in minimal user interfaces has had
nothing to do with successive versions released by
the specification producers such as OGC or ISO.
The key factors have been: (1) improvements re-
lated to the design and web programming; (2) web
usability and a commitment to focus on respond-
ing to user demands; and (3) a perspective change
from data centric offer to user centric demand.
Obviously, the emergence of new web-pro-
gramming techniques such as AJAX (Asynchro-
nous JavaScript and XML) and Javascript API
(Application Programming Interface) frameworks
like jQuery3 and OpenLayers4 have impacted web
applications’ design and performance; however,
the biggest impact has probably been the existence
466
of a search engine like Google5. It has changed
the users’ behavior when using web search appli-
cations. This impact has affected all web search
applications, even the metadata catalogs, and
has been called “googlization” (Battelle, 2003;
Salkever, 2003). Google offers a single search
box. Usually what we are looking for appears well
positioned in the first page and it is also very fast.
This does not happen in most metadata search
engines, despite having a smaller and very well
described volume of information.
Finally, the future work section will introduce
the latest improvements developed in the meta-
data search using natural language programming
techniques.
OGC CATALOG SERVICES
SPECIFICATION
Among the applications, which lie at the core of
all SDIs, perhaps the most important ones are the
metadata catalogs. Implementing a catalog may
be a tedious and think-less task but it is vital to
the success of an SDI. The job basically consists
of locating all possible producers of cartographic
data (departments, agencies, companies, or even
universities) and convincing them to make an
inventory of all their geographic information and
to describe it following a standardized procedure.
This is what we call “creating metadata.”
This metadata, which is stored in eXtensible
Markup Language6 (XML) encoded files accord-
ing to the International Standards Office (ISO)
standards, 19115, are loaded into the metadata
catalog so that they can be interrogated by any
user. It has to be said that the creation of metadata
is not an easy task and involves describing fields
such as map creator, creation date, revisions data,
accessibility, information formats, quality, distri-
bution, and the coordinate box corresponding to
the information described.
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
Context
In an optimal scenario, all SDI users should be
able to find out what geographic information ex-
ists in a particular area of the territory (catalog)
in question, view this information (viewer), and
finally, if it is of interest, download it (process).
To achieve this goal any SDI development should,
adopt geospatial interoperability standards in order
to facilitate communication between components.
There is a consortium called the Open Geospa-
tial Consortium (OGC) whose job is to decide on
the specifications to be used. OGC’s role in the
geospatial world is not that different from W3C’s
role in the standardization of the World Wide Web.
OGC is a nonprofit consortium, created in 1994
and made up of over 400 enterprises, governmental
agencies, and universities, which work together
towards developing open specifications.
The purpose of these specifications is to pro-
mote geospatial interoperability. In any SDI the
interoperability between components is based on
the OGC’s specifications. The success of these
specifications lies in the fact that they are pro-
moted and created by the same companies that
will later implement them in their commercial
products. In particular, an OGC specification is a
technical document describing the communication
interfaces between servers and how they are to be
implemented. These specifications make no men-
tion of the architecture, platform, or programming
languages to be used. Before being considered as
an OpenGIS Implementation Specification, each
document must be developed and tested by various
working groups within the OGC and then voted on.
One of the most commonly used OGC
specifications is Web Map Service (WMS). The
purpose of WMS is to enable the superimposed
visualization of complex geographic information
distributed over various servers. A client may re-
quest other servers, which are also based on this
specification in order to search for geographic
information. Once the data is found, the client
may make use of it simultaneously and visualize
different geographic data from different servers in
the same environment. Each request contains a set
of specific parameters defined by the WMS speci-
fication which are understood by all map servers
that meet the specification. Thus, when we say a
map server is standard and is WMS compliant, it
means that the server is capable of responding to
such requests, while the already mentioned WMS
or others such as WCS (Web Coverage Service)
and WFS (Web Feature Service) are specifications
related to data visualization.
The OGC specification related to the discovery
of data is OGC Catalogue Services Specification.
In the late 90’s the OpenGIS consortium—nowa-
days Open GeoSpatial consortium—needed to
define a specification that allowed interoper-
ability among metadata catalogs focused on
geospatial information. The motivation to define
a specification was very clear due to vast amounts
of geospatial information resources that were
distributed among multiple databases, stored in
many formats, and accessed by many different
interfaces (Nerbert, 1999).
Within the range of OpenGIS documents and
specifications, the Abstract Specification docu-
ment (OGC, 1999) describes, documents and cre-
ates a conceptual model sufficient enough to allow
for the creation of Implementation Specifications.
Among them, the topic number 13 focuses on
Catalog Services.
The term “Catalog” is described as a “set of
service interfaces which support organization,
discovery, and access of geospatial information.
Catalog services help users or application soft-
ware to find information that exists anywhere in
a distributed computing environment” (Kottman,
1999).
Kottman (1999) defines catalogs as having
three essential purposes:
• To assist in the organization and manage-
ment of diverse Geospatial data and ser-
vices for discovery and access,
467
 User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
• To discover resource information from
diverse sources and gather it into a single
searchable location, and
• To provide a means of locating, retrieving
and storing the resources indexed by the
catalog.
The OGC Catalog Services Specification- ver-
sion 1.0.0 was released in 1999. It establishes a
general framework to access metadata and speci-
fies a design pattern to define interfaces to publish
and search collections of descriptive information
(metadata). It also defines both an information
model and a service model. Figure 1 shows the
conceptual architecture proposed by OGC.
From version 1.0.0 until present day, version
2.0.2, OGC has defined several protocol bindings
and application profiles of the OGC Catalog
Services to meet the needs of stakeholders in a
wide variety of application domains. Table 1 shows
the different versions and key issues in the
specification.
Figure1. Catalog service architecture
468
Protocol Bindings
Catalog Service for the Web (CS-W) is prob-
ably the most used protocol binding in metadata
catalogs OGC implementations. It is based on
Hypertext Transfer Protocol (HTTP) a protocol
that is widely used to exchange information on
the web. CS-W has been available since version
2.0 and permits the interaction between a client
and a server using a standard request-response
model of the HTTP protocol, that is, a client sends
a request to a server using HTTP and expects to
receive a response to the request or an exception
message. Request and response messages are
encoded as keyword-value pairs within a request
URI or using an XML entity-body. Requests may
also be embedded in a messaging framework
such as SOAP.
Application Profiles
There are several profiles of the current OGC CS-W
binding. On the one hand, the ISO 19115/19139
metadata profile7 specifies an application profile
for ISO metadata with support for XML encoding
per ISO 191398. This application profile is widely
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
Table 1. OGC catalog services specification history
Year Version
1999 1.0
2001 1.1
2002 1.1.1
2003 2.0.0
2004 2.0.0
2005 2.0.1
2005 2.0.1
2007 2.0.2
2009 2.0.2
201? 3.0.0
implemented in Europe, and has been chosen by
INSPIRE9 as part of the discovery services.
Om the other hand, the ebRIM (ISO / TS
15000-3) profile10 applies the CS-W interfaces to
the OASIS11 ebXML registry information model
(ebRIM 3.0) so as to provide a general and flex-
ible web-based registry service that enables users,
software agents or human-to locate, access, and
make use of resources in an open, distributed
system. It provides mechanism to retrieve, store,
and manage many kinds of resource descriptions
beyond ISO metadata.
The application profiles constrain CS-W re-
quests and therefore the interoperability among
catalogs. A CS-W request needs to include: details
of how search criteria will be encoded; what XML
application profile will be used to encode meta-
data; and how specific metadata content will be
encoded in the chosen application profile.
The future version 3 will be based on Open-
Search12 with Geo and Time extensions as the
“baseline” query operation for all profiles, which
means the complete CSW 2.0.2 implementation
without any additional profiles or extension
packages.
Key Issues
Implementation for the OLEDB, CORBA, and ANSI
Z39.50 (ISO 23950) environments.
HTTP Protocol Binding. CS-W (Catalog Service for the
Web)
ISO19115/ISO19119 Application Profile for CS-W 2.0
Ebrim (ISO / TS 15000-3) Application profile for CS-W
OWL Application profile for CS-W
OGC OpenSearh protocol binding
WEB CLIENT APPLICATION
A web client is an application that accesses a remote
service on another computer system, known as a
server and is reliant on a common web browser
to render the application executable.
The client presents a Graphical User Interface
(GUI) that allows users to interact with the server;
a GUI represents the information and actions avail-
able to a user through graphical icons, buttons,
visual indicators and typed text box navigation.
The actions are usually performed through direct
manipulation of the graphical elements.
The GUI design has been evolving based on
trends in internet. In a web catalog client appli-
cation the evolution has clearly been influenced
by the Google search engine, both in its behavior
and design.
First Web Catalog Client
Applications
In the early 2000s, the web client applications
focused on searching metadata that showed a web
interface with many search options. Many of these
user interfaces were like a one by one mapping
process between the web search options and the
core queryable elements defined by the OGC
Catalogue Service Specification (see Table 2).
469
 User-Friendly Geoportal Interfaces for Geospatial Resource Discovery

Table 2. OGC catalog service common queryable elements

Name Definition
Subject The topic of the content of the resource b
Title A name given to the resource
Abstract A summary of the content of the resource
AnyText A target for full-text search of character data types in a catalogue
Format The physical or digital manifestation of the resource
Identifier An unique reference to the record within the catalogue
Modified Date on which the record was created or updated within the catalogue
Type The nature or genre of the content of the resource. Type can include general categories, genres or
aggregation levels of content.
BoundingBox A bounding box for identifying a geographic area of interest
CRS Geographic Coordinate Reference System (Authority and ID) for the BoundingBox
Association Complete statement of a one-to-one relationship

Figure 2 shows the web catalog client of IDEC
in 2003. Notice within the GUI you can see up to
18 boxes text that the user could fill out before to
launch a query.
The web application was structured and de-
signed following the three classic search op-
tions—WHAT, WHERE and WHEN—that are
heritage from applications used mainly in librar-
ies (Baldonado, 1997). Looking at this web user
interface we may notice that, obviously, these
kind of application were focused on a small set
of expert users, who might build a request with
up to eight concatenated different filters. So users
were asking what exist about this place? Instead
of What is in this place?
We can separate a catalog system in two sides:
the server side, which contains the database, the
metadata records, and the client side, which is the
web client application. It is possible to observe that
these client applications delegate the complexity
to the client side and therefore to the end user.
The server side, does not have to interpret the
request, it just has to send it directly to the catalog
as a mapping translation by just replacing values.

Figure 2. IDEC catalog web client: year 2003

470
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
This way of proceeding provides a few options to
potential users, not experts, and most importantly,
it provides high quality information, since the user
only finds what has been accurately described in
the metadata.
This can work in an environment such as
libraries, for example, where a small group of
people are responsible for describing informa-
tion. But how does it work in an SDI distributed
environment? And when do you have more than
200 metadata producers?
First Attempt at Simplification
At the end of 2005, the IDEC catalog systems
implemented the version 2 of the OGC speci-
fication with the Ebrim application profile and
the CS-W HTTP protocol binding. The CS-W
protocol facilitates the possibility to promote the
catalog as a web service, which means to offer the
possibility to query the catalog without needing
to enter into the IDEC website.
At the client application level, this is a first
attempt to “simplify” the user interface, by adding
a new search option called “quick search.” In fact,
this option simultaneously searches several fields
of metadata: title, description, purpose, scale, and
Figure 3. IDEC catalog web client: year 2005
keywords. The “WHEN” option was removed
from the user interface because it was rarely used
by users. The map is presented as a complex tool
with many options, and with the unique purpose
of populating the coordinates text boxes within
the WHERE section.
Figure 3 shows a first evolution of the previ-
ous design (see Figure 2). This reduces the text
boxes 18 to 11 and it begins to show the map as
important piece of the GUI.
These changes were made, more or less, in an
intuitive way, just attending to feedback received
from some well-known users, but a year after the
change (2006) we note that visits to the catalog
had just grown a little. We decided to develop a
study about user behavior.
A WEB SEARCHING STUDY
Web searching studies fall into three categories: (1)
those that primarily use transaction-log analysis,
(2) those that incorporate users in a laboratory
survey or other experimental setting, and (3) those
that examine issues related to or affecting Web
searching (Bernard, 2003).
471
 User-Friendly Geoportal Interfaces for Geospatial Resource Discovery

In 2007, an internal study was performed,
based on use transaction-log analysis—all user
actions are stored in log files—and focusing on
the following research questions:
• What are the search options used by the
user?
• Does the user combine them?
• What is the number result pages queried?
• What are the trends in search topics?
• Does the user find what he is looking for?
We defined 11 variables (see Table 3), using
them to classify 5,000 sessions.
Trying to determine how the user searches for
information: If he uses a single text box or com-
bines search options and especially if he uses the
map as a spatial interface. Another variable referred
to categorizing the results and whether the user
found what he was looking for; to determine if
the user go forward to see more search results
(pagination) or used the print option. The final
purpose was to determine if the catalog was a
useful tool, and the conclusion was a negative
response.
Notice that the study revealed some very in-
teresting issues. First of all, 32% of the requests
return zero results. Most of these requests were
queries made by combining more than two search
options. A query example that would be made
with zero results found:
FREETEXT {Enology} + KeyWord {Base
Mapping} + ISOThemes {Ecology}
Indeed, there was not any metadata record that
met these conditions, but there were other metadata
related to Ecology and Enology that could be of
interest to the user and were never found.
Second, there was an extensive use of the
“quick search” option and more than 66% of the
inputs referred to place names. This meant, for
example that although there are metadata from a
particular place, if the place name did not appear
explicitly written in the metadata, than there was
no way to find it, unless a direct coordinate—
spatial—search was used. This only happened
in 11% of the requests. Notice that probably the
most powerful searches supported by an OGC
catalog are the spatial ones, such as those based
on a bounding box.
Third, 42% of the searches contain the search
option “by organization,” and fourth, 84% of
users only view one result page. This trend will
inform us how persistent searchers are when
locating information or services on the catalog.
Table 4 shows percentages in the search options
used by the users.

Table 3. Variables defined in the web study

Variable
Single search option
Spatial search option
Combined search options
Combined+ spatial search option
Due Pagination
Query metadata details
Search metadata Services
Search metadata Datasets
Search category
User IP address
User Type
Definition
Use only “quick search option”
Use only spatial search using the Map
Combine several search options
Combine several search option but using the Map.
Perform pagination or not
Query the candidates or not
Interested in services
Interested in Datasets.
Evaluating search option try to classify the search in topics
User origin (University, Government,...)
Classify the user depends precedence (Public sector, research,..

472
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery

Table 4. Interoperability levels in the literature

pened in many catalog client implementations.
Search options % used request The unity of the interface is important because
parameters it has the potential to link concepts and objects
Quick search 69.63% together (Bernard, 2003).
By Organizations 42.01% A second step was to reduce the number of
By Map (coordinates) 11.28% search options to 4 and permit a direct search us-
By Keywords 23.34% ing one of them. These options were in some way
By IsoTopics 35.12% “controlled”—using suggestions list—in order to
avoid possible no result found.
The four search options were:
The final conclusions of the study were the
following: • By place (WHERE): this option offers
a suggestion list while the user is typing
• Most people only use the simple text box a place name, retrieved from a gazetteer
like Google. service. The purpose was to translate the
• Most users will not pass the first page (no place name into concrete coordinate posi-
pagination). tions and request the catalog in a spatial
• Many requests contain references to a way, so that the metadata does not contain
place name. specific place names, but the geographical
• Expert users who combine more than two extent belonging to that place.
or three search options obtain unsatisfac- • By keyword (WHAT): as by place, this
tory results. option offers a suggestion list from an in-
• The information contained in some meta- dexed table that indexes not only keywords
data records were wrong or, in some cases, but all the terms contained in the metadata.
limited. This allows searching for any word, not
• Many of these do not contain explicit refer- only keywords.
ences to their geographical location. • By Organization: this option shows a bul-
leted list with a hyperlink based text for
each organization (provider) and the num-
ACTIONS FOR FRIENDLY ber of metadata from datasets, series and
USER INTERFACE services available in the catalog. A user
just need one click to access the metadata
Looking at the conclusions and adding some other provider. As mentioned by Bernard (2003)
inputs received from end users and metadata pro- a bulleted list of links is useful, for instance
viders, we decided to create guidelines in order to in a drop-down list.
improve the catalog system performance. • By Map: The map is considered as a search
option too, not as before when the map was
Make the Search More only a “tool” used to obtain bounding box
Intuitive, Useful, and Promote coordinates. With just one click on the map
the Spatial Search or by drawing a box, a spatial request is
sent to the catalog.
The first step was trying to maintain the unity of
the interface and, for instance, not showing the Figure 4 shows how these four search op-
result candidates in a separate window as hap- tions were localized among the web catalog cli-

473
 User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
Figure 4. IDEC catalog web client with 4 search options
ent application. As you see “by place” and “by
keyword” options are top-right side, while “by
organization” is at the bottom-right side and “by
map” is in left side.
Making the Candidates’ Response
More Useful
The candidates’ response found in a request are
shown as a link list with summaries and intuitive
icons, as Bernard (2003) mentions, “the summary
condition was perceived as being the easiest in
finding information, being visually pleasing, pro-
moting comprehension, participants’ satisfaction
with the site, and looking professional.”
The results may be filtered by useful topics
such as: scale, format, time, provider, or just free
text. Notice that the values of these filters—less
free text option—are populated depending on re-
474
sponse content information, avoiding again the “no
results found” message. The results were drawn
in the map as rectangles, showing the extent that
was described in the metadata record. The map
rectangles and candidates are both connected, so
it is possible to select a rectangle and to highlight
the candidate and vice versa.
Another important point was when the results
contains metadata from services—for instant
OGC Web Map Service (WMS)—allowing the
user to connect the service and more importantly
to facilitate information about the availability of
the service. To do this a service monitoring tool
called Service Status Checker (SSC) was used.
The SSC is a tool for monitoring services,
which was designed by the U.S. Federal Geo-
graphic Data Committee13 (FGDC) to test the
geospatial Web services such as WMS, WFS,
ArcIMS services, and Z39.50 (FGDC, 2010).
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
The SSC provides real-time tests (accepts XML
requests, perform testing, and immediately returns
test results in XML) and file tests (tests performed
daily services previously recorded and stored the
results in a database).
To access this second service, servers need to
be previously registered. The service can offer
two important facts regarding the WMS server:
• Availability: Indicates whether the service
is working or not.
• Reliability: Measures the percentage of
successful responses. An unsatisfactory re-
sponse is any slow or incorrect response.
Figure 5 shows a candidate response and the
linkage existing between candidates and the map
Figure 5. IDEC catalog web candidates and filter options
Minimize the Response Time
To achieve this objective, the use of the CS-W
protocol binding as a message request-response
between client and server has been discarded.
This does not affect the interoperability of the
catalog, which remains accessible to third parties
via CS-W as a web service.
The CS-W implies the use of XML meta-
language to build requests and responses. In par-
ticular, a web client send request parameters to a
middle ware service that builds a CS-W-compliant
XML request and sends it to the server. Then, the
server receives the XML request, parses it to build
a Structured Query Language (SQL) sentence to
the database, and translates the responses database
475
 User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
Figure 6. New conceptual architecture
to CS-W-compliant XML responses and sends it
back to the client.
The best way to eliminate steps and minimize
the response time would be: (1) to use Structured
Query Language14 (SQL) sentences directly to
request the catalog’s database: (2) to use JavaS-
cript Object Notation15 (JSON) message format to
interchange messages between client and server
(see Figure 6). JSON is lighter and easier to ma-
nipulable than CS-W xml.
Review and Improve the Quality and
Identification of Important Details of
Metadata
A review quality protocol has been setup. Basi-
cally, its function is to review whether the meta-
data has been described properly and assure that
there is consistency in the information provided.
For instance, some of the controls established by
the protocol that look at “title” and “description”
metadata items are self-explanatory and discour-
age the use of codes or abbreviations in these
items. Another control is related to bounding box
definition. A municipality metadata cannot have
regional bounding box coordinates, for instance.
476
Extending the Concept and
Implementation of the Catalog
They are a small set of actions aimed to promote
the use of the catalog:
• Provide additional channels to disseminate
the catalog results such as syndication RSS
and visualization with Google Earth.
• Provide an OpenSearch geo16 wrapper to in-
tegrate the catalog in the third application.
• Facilitate searching several online catalogs
simultaneously. This is why we developed
the software CatalogConnector17, which
is a Java servlet application that sends re-
quests to several OGC CSW catalogs, pro-
cesses the responses and shows them in a
unitary way. It supports different CSW ver-
sions (2.0.0, 2.0.1, 2.0.2) and application
profile implementations (ebRIM and ISO)
and it also implements OpenSearch-Geo
interfaces.
• Use the catalog to search other resources
related to geospatial data. Thanks to the
ebRIM application profile it is possible to
catalog other kinds of resources such as
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
sensors, images data models or schema. So
it is feasible to set up specialized catalogs.
SERVER SIDE VS. CLIENT
SIDE COMPLEXITY
Reducing the search options to four (see Figure
4) in the web client side has meant a significant
increase in the server process.
There is a clear relationship between usability
and simplifying the search in the client side and
a growing complexity and concatenation of pro-
cesses in the server. The tendency to create web
applications that are easier to use and focused
on the “web-mass” implies building a complex
system; on the server side to understand and to
answer the user properly.
One way to quantify this trend is to subtract
the number of processes on the server side and the
number of search options on the client side. A posi-
tive result means that there are more processes on
the server and more complexity, whereas a negative
result means more complexity at the client side
of the application. Figure 7 shows the evolution
between server processes and web client search
options, from 2002 to 2010. The IDEC catalog
has had clear tendencies to increase the server
side processes and decrease the search options
Figure 7. Server side and client side complexity
on the client side. So it went from a value of -8 in
2002 to a value of + 4 in 2010. In other words, the
value +4 means that there are four search options
in web application and 8 server processes to deal
and setup these 4 search options.
For the search by place option, it has been
necessary to implement a gazetteer server and a
process to build spatial queries from a point (x,
y coordinates). This means that when a user re-
quests for the place called “X,” the request that
will be sent to the catalog server will contain the
coordinates of “X” place not the string “X.”
For the search by keyword indexing processes
have been created to relieve significant informa-
tion from metadata and eliminate duplicated
keywords. For the organization list and filters to
candidates, statistical processes based on unique
values categorizations have been created. In this
way, it is possible to offer the user a link to a text
list with relevant information about the number
of records provided by providers. All of this is
available without any click by the user. Finally,
statistical processes based on unique values cat-
egorizations have been created to start automati-
cally processes each time the catalog has been
updated with new metadata records in order to
keep the catalog synchronized.
477
 User-Friendly Geoportal Interfaces for Geospatial Resource Discovery
FUTURE WORK: A WEB
SEMANTIC APPROACH
Nowadays there are two objectives that must be
filled, simplifying the client web application,
joining the search boxes “by place” and “by
keyword” in a single one, and providing mecha-
nisms to understanding what the user is looking
for. To achieve this, it is mandatory to provide
the catalog with some semantic capabilities and
more complexity on the server side.
Basically, we are testing several approaches
using semantic techniques dealing with text
strings, such as:
• The stemming process reducing each word
to its stem or root (not exactly their mot-
to lexicon that would be a pure linguistic
process, but rather an algorithmic known
approach).
• Removing stop words (stop words) such
as prepositions, adverbs, conjunctive, and
disjunctive clauses and ultimately all those
words that do not provide information in
the context of the problem.
• Calculating the distance between what us-
ers search for and the content indexed.
• Automatic place names recognition—
geoparsing—and connection with gazet-
teer service—translating a places name
into coordinates.
• Decomposing the request into bigrams
(two words) and providing a scoring sys-
tem for matching. For instance, the fol-
lowing search “ Barcelona’s water quality”
will be searched as:
◦◦ Exacts string “Barcelona’s water
quality.”
◦◦ By each bigram “water quality” and
“water Barcelona.”
◦◦ By every word “Barcelona,” “water,”
“quality.”
◦◦ By coordinates of Barcelona and
bigrams.
478
The use of these techniques should provide
more reliable results and permit the non-expert
user to interact normally with a metadata catalog.
Another key issue will be the new versions
of the OGC Catalog Specification. The fact that
this future version will be based on OpenSearch
application profile, must allow a better interoper-
ability among catalogs and the most important, the
efforts made implementing semantic techniques,
will not remain in only a web application domain
and may be re-used as a OGC standard catalog
service.
REFERENCES
Baldonado, M., Chen-Chuan, K., Gravano,
L., & Paepcke, A. (1997). The stanford digital
library metadata architecture. International
Journal on Digital Libraries, 1(2). doi:10.1007/
s007990050008
Battelle, J. (2003). The creeping googlization
meme. BusinessWeek Online. Retrieved May
10, 2011, from http://battellemedia.com /ar-
chives/000145.php.
Bernard, J., & Spink, A. (2005). How are we
searching the world wide web? A comparison of
nine search engine transaction logs. Retrieved
May 10, 2011, from http://citeseerx.ist.psu.edu
/viewdoc/download?doi=10.1.1.90.4299&rep=r
ep1&type=pdf.
Bernard, M. (2003). Optimal web design. Re-
trieved May 10, 2011 from http://uwf.edu /ddaw-
son/d3net/documents/ web_usability/optimal%20
web%20design.pdf.
ISO 19115. (2003). Geographic information -
Metadata. ISO 19115:2003. Geneva, Switzerland:
International Organization for Standardization
(ISO).
User-Friendly Geoportal Interfaces for Geospatial Resource Discovery

Kottman, C. (1999). The OpenGIS abstract Voges, U., & Senkler, K. (2004). ISO19115/
specification: Topic13: Catalog services, ver 4.0. ISO19119 application profile for CSW 2.0 (CAT2
OpenGIS Project Document 99-113. New York, AP ISO19115/19). OpenGIS Project Document
NY: OpenGIS Consortium Inc. 04-038r2. New York, NY: OpenGeoSpatial Con-
sortium Inc.
Martell, R. (2005). OpenGIS catalogue services
- ebRIM (ISO/TS 15000-3) profile of CSW. Open-
GIS Project Document 05025r3. New York, NY:
ENDNOTES
OpenGeoSpatial Consortium Inc.
Nebert, D. (2002). OpenGIS catalog services
1
International Standards Office, http://www.
specification, ver 1.1.1. OpenGIS project docu- iso.org
ment 02-087r3. New York, NY: Open GIS Con-
2
GeoSpatial Consortium, http://www.open-
sortium Inc. geospatial.org/
3
Jquery, http://jquery.com/
Nebert, D. (2004). OpenGIS - Catalogue services 4
OpenLayers, http://openlayers.org
specification, ver 2.0. OpenGIS Project Document 5
Google Search, http://www.google.com
04-021. New York, NY: Open GIS Consortium Inc. 6
eXtensible Markup Language, http://www.
Nebert, D. (2004). Developing spatial data in- w3.org/XML/
frastructures: The SDI cookbook. Retrieved May
7
ISO 19115/19139 metadata profile,
10, 2011, from http://www.gsdi.org /docs2004/ http://www.iso.org/iso/catalogue_detail.
Cookbook/ cookbookV2.0.pdf. htm?csnumber=32557
8
ISO 19139, http://www.iso.org/iso/cata-
Nebert, D., & Whiteside, A. (2005). OpenGIS logue_detail.htm?csnumber=32557
- Catalogue services specification, ver 2.0.1. 9
Infrastructure for Spatial Information in the
OpenGIS Project Document 04-021r3. New York, European Community, http://www.inspire-
NY: OpenGeoSpatial Consortium Inc. geoportal.eu/
Nebert, D., Whiteside, A., & Vretanos, P. (2007).
10
Ebrim (ISO / TS 15000-3) profile, http://
OpenGIS - Catalogue services specification, ver portal.opengeospatial.org/files?artifact_
2.0.2. OpenGIS Project Document 07-006r1. New id=31137
York, NY: OpenGeoSpatial Consortium Inc.
11
Organization for the Advancement of Struc-
tured Information Standards, http://www.
Nogueras-Iso, J., Zarazaga-Soria, F. J., Béjar, R., oasis-open.org
Álvarez, P. J., & Muro-Medrano, P. (2005). OGC 12
OpenSearch interface, http://www.open-
catalog services: A key element for the develop- search.org/Home
ment of spatial data infrastructures. Computers 13
Federal Geographic Data Committee, http://
& Geosciences, 31(2), 199–209. doi:10.1016/j. www.fgdc.gov
cageo.2004.05.015 14
Structured query language, http://
Salkever, A. (2003). Google here, there, and ev- en.wikipedia.org/wiki/SQL
erywhere. BusinessWeek Online. Retrieved May
15
JavaScript Object Notation, http://www.json.
10, 2011, from http://www.businessweek.com / org/
technology/content/dec2003/ tc20031216_9018_
16
OpenSearch geo, http://www.opensearch.
tc047.htm. org/Specifications/OpenSearch/Extensions/
Geo/1.0/Draft_1

This work was previously published in Discovery of Geospatial Resources: Methodologies, Technologies, and Emergent Ap-
plications, edited by Laura Díaz, Carlos Granell and Joaquín Huerta, pp. 261-276, copyright 2012 by Information Science
Reference (an imprint of IGI Global).
479
480

Chapter 29
Development and Application
of a Spreadsheet-Based Spatial
Decision Support System (SDSS)
Dossay Oryspayev
University of Northern Iowa, USA

Ramanathan Sugumaran
University of Northern Iowa, USA

John DeGroote
University of Northern Iowa, USA

ABSTRACT
Spatial decision support systems (SDSS) are decision support tools which have been used widely in ad-
dressing complicated issues involving a spatial component. The use of SDSS has increased greatly over
the last few decades especially in fields such as planning, natural resources management, and environ-
mental science. Traditionally, SDSS have been developed with Geographic Information Systems (GIS)
technology as a major component and used in application areas in which the use of GIS technology
has been common. GIS software is often expensive and requires significant expertise, which can lead
to under-utilization of GIS-based SDSS. In this paper, we describe the development of a freely avail-
able SDSS extension developed for Microsoft Excel, a very commonly used spreadsheet application.
The purpose of this SDSS is to expand potential SDSS use to a wider potential audience for research,
management, and teaching purposes.

INTRODUCTION computer systems that support decision makers in

Spatial decision support systems (SDSS) are
decision support systems (DSS) which utilize
geographical or spatial data. Sugumaran and De-
Groote (2010, p. 14) defined SDSS as “integrated
DOI: 10.4018/978-1-4666-2038-4.ch029
addressing semistructured or unstructured spatial
problems in an interactive and iterative way with
functionality for handling spatial and nonspatial
databases, analytical modeling capabilities, deci-
sion support utilities such as scenario analysis,
and effective data and information presentation
utilities.” Research on SDSS has been strongly

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
influenced by that of DSS (Malczewski, 1999) and
has followed advances in geospatial technologies;
especially Geographic Information Systems (GIS)
software advances (Sugumaran & DeGroote,
2010). The use of geospatial techniques and
technologies such as GIS, remote sensing, and
Global Positioning Systems (GPS) has grown
tremendously over the last several decades. As
Keenan (2006) pointed out, many phenomena have
a geographic component which can be captured,
described, and analyzed in GIS which has led to
its playing an important role in a wide range of
fields. The development and utilization of SDSS
technologies grew tremendously from the mid-
1990s based on the number of articles published
on the subject (Malczewski, 2006; Sugumaran &
DeGroote, 2010). This followed a similar trend in
the uptake of geospatial technologies. Drivers of
this increased use were more powerful desktop
computers, the development of GIS software
with more friendly graphical user interfaces (e.g.,
ArcView (ESRI, 1996)), as well as built-in GIS
development environments (e.g., Avenue which
was bundled with ArcView (ESRI, 1996)).
The vast majority of SDSS applications devel-
oped to date have had GIS as a major component
of the architecture. The architectures of SDSS
have varied with some of them being developed
entirely within the GIS software (embedded), some
with a tight coupling (common user-interface)
between GIS and other modeling or mutli-criteria
evaluation software, or a loose coupling between
GIS and other components through formatting
and exchange of data files. In a review of SDSS
composed of GIS and multi-criteria decision
analysis (GIS-MCDA) techniques, Malczewski
(2006) found that loose coupling of systems was
the most common method of integration. Thus,
the development and application of SDSS has
generally required possession of GIS software
and experience in using such software. While
open-source and free GIS software are becoming
more widely available, the majority of SDSS have
been developed using commercial GIS software
which can range in cost from several hundred to
many thousands of US dollars.
Due to the GIS-centric nature of SDSS, adop-
tion has naturally occurred in disciplines which
were also adopters of geospatial technologies. In
a review of SDSS publications, Sugumaran and
DeGroote (2010) listed the application of SDSS
from most to least frequent being natural resources
management, environmental, urban, agriculture,
emergency planning/hazard analysis, transporta-
tion, business, utility/communications/energy,
and public health. Malczewski (2006) found
that what he classified as environment/ecology
was the most common application area. Fields
such as natural resources management were early
adopters of GIS and remote sensing technologies.
Other fields such as business and public health
have been later adopters of technologies such as
GIS and thus have had slower uptake of SDSS.
Some of the reasons for lack of adoption are the
lack of spatial data knowledge of the users, lack
of training and education for these technologies
within their fields, and the expensiveness of GIS
systems. Uran and Janssen (2003) analyzed five
SDSS applications and found one of the major
drawbacks to effective use was that they required
a high level of GIS expertise in order to carry out
operations in the SDSS. They specifically indi-
cated that the outputs of the model (spatial data)
still required further spatial processing in order
to properly evaluate results. This type of situation
limits effective use of SDSS to GIS experts.
Although spatial data processing, analysis,
management, and visualization functionality are
often essential for SDSS, a given SDSS often uti-
lizes a mere fraction of the tools available in GIS
software. However, as the GIS software have been
developed to offer a wide-range of functionality
for use in many disciplines, the interfaces and tools
will not necessarily be intuitive for non-expert
users. Thus for many SDSS, unless the users are
experienced GIS users, or the developers invest
in the careful design of user interfaces and tools
for ease of use, the SDSS might be under-utilized.
481
 Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
There are two main data models utilized for
storing and displaying spatial data in GIS soft-
ware. These are vector and raster. In a vector
data model, entities in the real world are divided
into clearly defined features which are based on
point, line, and polygon geometry (Sugumaran
& DeGroote, 2010). The simplest vector feature
is a point which is defined by x, y (and possibly
z) coordinates. Line and polygon geometry are
defined by points and connecting line segments
with the polygon having the beginning and end
point at the same point. The raster data model on
the other hand divides space into an array of (usu-
ally square) cells (Longley, Goodchild, Maguire,
& Rhind, 2011) with each cell having a value that
represents some aspect of the real world (e.g., land
use, elevation, precipitation amount, etc.). Both
of these data models have advantages and disad-
vantages in terms of data storage, management,
and analysis. The vector data model is better for
topological or network analyses, provides explicit
geographic representation at all scales, allows
for production of high quality cartographic prod-
ucts, and allows sophisticated attribute handling.
However, the vector data model has complex
data structures, can require significant computer
resources for processing and display, and requires
expensive data collection techniques (Burrough &
McDonnell, 2000; Lo & Yeung, 2002). The raster
data model has the advantages of being easy to
understand, allows rapid reading and analysis, and
is easily displayed on devices that display pixels
(Clarke, 2011). The raster model is often used for
remotely sensed imagery collected from airplanes
and satellites. Image processing software allows
the production of derived data products from this
remotely sensed imagery which also is in a raster
data format. The raster data model is insufficient
for representing each individual entity from the real
world (e.g., signs, trees, fire hydrants). However,
the raster data model is useful for storing and dis-
playing phenomena which vary continuously over
space such as elevation or topography, rainfall, or
categorical representations of land use.
482
Most GIS software will have functionality and
mechanisms for handling both raster and vec-
tor data models often with vector functionality
making up the majority of tools available. The
major commercial GIS software applications
from companies such as ESRI (http://www.esri.
com), Intergraph (http://www.intergraph.com),
and Bentley (http://www.bentley.com/en-US/)
are vector-based although they usually have ex-
tensions for raster data handling. In his review
of 319 articles detailing GIS-MCDA (type of
SDSS) applications, Malczewski (2006) reported
approximately equal number of studies utilizing
raster and vector data models with some of these
using both data models. The simple nature of the
raster data model is similar to the structure of
other data concepts (e.g., a matrix) and the way
other computer applications (e.g., spreadsheets
and databases) store data. Although spreadsheet
software were not developed with the idea of
storing spatial data, they are capable of being
formatted so that the cells are square with each
cell holding a single value similar to the raster
data model in GIS software. One of the most
widely used spreadsheet applications, Microsoft
Excel, provides excellent tools for resizing cells
(square) and for changing the representation (e.g.,
color, pattern, texture) of each cell according to
the value stored. Microsoft Excel has a wealth
of mathematical, logical, and other functional-
ity that can be carried out on a cell-by-cell basis
and based on examination of neighboring cells.
Although these attributes/tools do not approach
the level of cartographic capabilities present in
GIS software, they do provide the possibility of
mapping both input and modeled values in a map.
The main advantages of using spreadsheet soft-
ware such as Microsoft Excel are that the software
is much less expensive and is much more widely
used throughout the world. There are a broad
range of potential users that are familiar with the
functionality of Microsoft Excel that also might
be interested in analysis of some data that has a
spatial component. For these reasons, spreadsheet
Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
software such as Microsoft Excel is conducive to
the development of SDSS which utilize raster data
models and which do not require the extensive
set of spatial data, management, and analysis
functionality present in expensive GIS software.
This article is not the first to propose the idea
of implementing spatially-explicit modeling or
the development of an SDSS using spreadsheets.
As early as 1991 (Ali, Ross, & Beveridge, 1991)
the idea was put forward in a paper describing the
implementation of a system for assessing spatial
parameters related to potential fish aquaculture in
Pakistan. The authors transformed hard-copy map
data for climatic, topography, transportation, and
market characteristics into a series of separate grids
which stored values representing the suitability
for each cell for aquaculture. Using spreadsheet
functionality, they weighted each separate grid
and combined them in order to come up with a
final grid representing potential suitability for
aquaculture in each square. The authors concluded
that their approach was a cheap and effective
alternative to using GIS.
In 1997, a paper describing an approach to
teaching raster GIS operations using spread-
sheets was published in the Journal of Geography
(Raubal, Gaupmann, & Kuhn, 1997). The authors
proposed and demonstrated the usage of spread-
sheets for spatial analysis (SDSS) using Microsoft
Excel. They proposed the use of Excel for execut-
ing common categories of raster operations used
in GIS software such as local, neighborhood, and
zonal functions. They demonstrated the use of
simple overlay in which two grids (precipitation
and elevation) were combined on a cell-by-cell
basis based on some constraints and an output grid
representing suitability for fir (type of tree) growth
was calculated. Their overall goal was to develop
tools that could be used in introductory courses for
teaching raster operations. They wanted the tools
to allow users to easily manipulate the raster data
and additionally let the user interactively analyze
the operations in a process. The advantages to us-
ing the spreadsheet software included the simple
user interface, familiarity to students, low cost,
and easy access to viewing code or macros used
to calculate outputs. The authors pointed out that
spreadsheet applications could be especially well
suited for SDSS using cellular automata models.
Indeed, Microsoft Excel had already been used in
a coupled SDSS system utilizing cellular automata
methods (Engelen, White, Uljee, & Drazan, 1995).
A custom application with a vegetation dy-
namics model combined with models of envi-
ronmental determinants was developed within
Microsoft Excel (Berardi, 2002). The purpose of
the application was to predict vegetation change
within the Astroni crater, a state nature reserve
near Naples, Italy. The application was developed
using the Visual Basic programming language and
included simulation models, a spatial data visual-
ization facility, and user friendly interfaces. The
choice of Microsoft Excel was made because it
was commonly-used software that was relatively
inexpensive and had user friendly interfaces both
for usage and programming through Visual Basic
(VB). The conclusion was that it is not necessary
to use expensive and complex software to develop
modeling systems for ecosystem management.
Spatially explicit modeling capabilities within
spreadsheet applications have continued to be de-
veloped within spreadsheet applications, especial-
ly using Microsoft Excel. Li, Shen, and Li (2004)
developed an SDSS which was based in Microsoft
Excel that embedded mapping functionality from
MapObjects. The natural analogy between raster-
based data structures and analysis techniques in
both GIS and spreadsheets is described in a paper
for the 2009 International Conference on System
Sciences (Keisler, Blake, & Wagner, 2009). This
paper presented two examples of analytical models
built within Microsoft Excel. The first example
was an Excel-based tool used to assist adminis-
trators in deciding on a good location for a new
parking facility at a university while the second
was a model for examining potential locations
for a retail store. While the authors acknowledge
limitation of spreadsheet technology for spatial
483
 Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
analysis in comparison to GIS software, they argue
that for non-GIS experts spreadsheet technologies
provide a less-complex and relatively inexpensive
alternative which allows for greater intra- and
inter-organizational collaboration. Specifically
they mention that millions of business and other
spreadsheet users would be exposed to spatial
processing functionality that they otherwise likely
would not be. They also point out that there are
great opportunities for these types of applications
to be used in teaching settings. The usefulness
of spreadsheet-based decision support tools with
spatially-explicit capabilities has been demon-
strated. However, all of the applications described
were developed for very specific purposes and
would not be applicable in other disciplines. To
this point, as known to the authors, there have
been no generic SDSS tools developed within a
spreadsheet environment. These types of tools
could be used by a wide variety of disciplines.
The goal of this paper is to describe the de-
velopment, implementation and application of a
Microsoft Excel extension that can be used for
spatial decision support. This extension is called
Spreadsheet-SDSS. The following sections will
describe the rationale behind the development of
Figure 1. Overall architecture and development environments used for Spreadsheet-SDSS
484
Spreadsheet-SDSS and the techniques used, give
an overview of the user interface, explain the com-
ponents of the extension, and finally use a case
study to demonstrate the functionality available.
DEVELOPMENT OF SPREADSHEET-
SDSS EXTENSION
The Spreadsheet-SDSS was developed using
the Visual C# programming language, using
tools from Microsoft such as Visual Studio (VS)
2008, Visual Studio Tools for Office (VSTO),
and Microsoft Office Excel 2007. Figure 1 shows
the general architecture and development tools
used for Spreadsheet-SDSS. When installed, the
extension can be added to Excel as a new ribbon.
The Spreadsheet-SDSS extension was originally
created as an add-in to the Microsoft Office 2007
environment.
In Microsoft Excel 2007, the process of adding
new tabs and groups to the ribbon was not avail-
able as an option from inside the Excel environ-
ment itself directly. Rather, the user had to go
through outside options to do this which were not
very efficient. In order to package the tools for
Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
easier installation, we utilized VS 2008 with VSTO
support to do this in a much more efficient way.
However, in Microsoft Office 2010 customization
of the ribbon is considerably easier. Users can
add custom tabs and groups of his/her own choice
and associate them with their own macros using
the Visual Basic for Applications development
environment. We have tested and verified that our
add-in properly works in a Microsoft Office 2010
environment also.
Microsoft Excel was chosen as it is one of
the most popular spreadsheet based applications
used around the world. It is used in a very wide
range of disciplines and has extensive math-
ematical, statistical, financial, logical and other
functionality. It also has capabilities that make it
amenable to the use of cell-based raster spatial
data. Specifically it has the capability for manu-
ally, or through programming, resizing the width
and height of the cells in order that square cells
can be utilized for raster representation. It also
has a zoom functionality that lets a user view a
small or large number of cells. This functionality
approximates zooming functionality in mapping
and GIS software.
There have been a significant number of
discussions concerning what functionality an
architecture must have before being considered
an SDSS (Densham & Goodchild, 1989; Keenan,
2003; Sugumaran & DeGroote, 2010). For the
purpose of this paper we are considering an SDSS
to consist of at least the following components:
• Database management utilities
• Modeling capability
• Dialog management
• Map display and report generation utilities
In the Spreadsheet-SDSS, database manage-
ment tools were created to allow the importation
of ESRI ASCII raster files (ESRI, 2011b). These
types of files are one of the most common distri-
bution formats for raster spatial data. Also ESRI
ASCII raster files (ESRI, 2011b) can be created
from many other raster data formats using open-
source and free software such as the Geospatial
Data Abstraction Library (GDAL, 2010), which
is a translator library for geospatial data formats.
In the future, functionality for importing other
raster data formats will be made available within
Spreadsheet-SDSS.
Several types of modeling operations which
have been used in numerous SDSS have been
implemented in Spreadsheet-SDSS. The first and
simplest method is Boolean Overlay including the
following Boolean operators: AND, OR, NOT, and
XOR. An in-depth discussion of Boolean opera-
tions is given by Malczewski (1999). Boolean
Overlays are often carried out using raster data in
GIS environments for combining multiple layers
for applications such as land suitability analysis.
An example SDSS application using Boolean
Overlay was carried out in locating a new hotel
project based on transportation features, land
use, elevation, existing tourist facilities, and the
coastline in Mauritius (Beedasy & Whyatt, 1999).
The second major class of modeling operation
implemented is Weighted Linear Combination
(WLC) which is a map overlay technique that at-
tempts to capture the value judgments and expert
opinions within the spatial analysis procedures by
the assigning of weights of relative importance
to each map layer (Malczewski 1999; Sugumaran
& DeGroote, 2010). In the hotel example given
above, higher weights would be given to layers
considered more important resulting in output that
isn’t expressed as suitable/unsuitable but which
has a range of scores indicating the level of suit-
ability. A WLC technique was used in an SDSS for
environmental planning and watershed manage-
ment in Missouri, USA (Sugumaran, Meyer, &
Davis, 2004). A more complicated method called
Ordered Weighted Averaging (OWA) has also been
implemented. The OWA method, as implemented
in Spreadsheet-SDSS, was originally described by
Yager (1988) and is explained for spatial contexts
in greater detail in other publications (Jiang & East-
man, 2000; Malczewski, 1999; Malczewski et al.,
485
 Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
2003). In addition to the above listed models, we
have implemented weight estimating methods. An
example is the Analytical Hierarchical Processing
(AHP) method, which is a pairwise comparison
method (Malczewski, 1999). This method has
the user set the relative importance of different
factors as compared to each other and then uses
an algorithm to set weights for the factors for
use in a model such as WLC. So for example, in
relation to the hotel example mentioned above,
the user might indicate distance to coastline is
more important than the distance to other tourist
facilities and this will be reflected in how the AHP
algorithm assigns the weight for the respective
layers. Numerous studies have used AHP and
WLC methods. For example these techniques
were used by Strager and Rosenberger (2006) for
identifying priority areas for land conservation.
The user interface or dialog management com-
ponent of Spreadsheet-SDSS was implemented
in VS 2008 and VSTO tools using the Visual C#
programming language and will be described in
more detail below. The map display capability
was realized in Microsoft Excel worksheets by
adding functionality for resizing the width and
height of the cells (to square cells). In addition
code was written that alters the color of cells based
on the values of the cells. Currently the report
generation functionality generates reports which
describe the distribution of values (count of cells)
and the area in acres, hectares, and square miles
Figure 2. The Spreadsheet-SDSS extension or plug-in in Microsoft Excel
486
for specific cell values. It is envisioned that the
report generation utility will be enhanced in later
versions of the extension.
USER INTERFACE OF THE
SPREADSHEET-SDSS EXTENSION
In Figure 2, the user interface of Spreadsheet-
SDSS is shown as a Microsoft Excel Add-In. The
extension is divided into five categories: Data,
Model, Map, Navigation, and Help.
The first tool (Add Layers) on the Data section
allows users to import ESRI ASCII raster files
(EARF, 2011) into Excel. The second tool (Change
Colors) allows users to manipulate the way the
layer is displayed (i.e., choose color schemes).
The Change Values tool allows the user to reclas-
sify values in the layer. In the Model section, the
user selects the model they would like to run.
They can choose from WLC Direct (without AHP),
OWA, WLC using AHP, or Boolean Overlay. The
next section, Map, provides some basic carto-
graphic capabilities. In this category there are
tools that allow users to insert a north arrow (North
Arrow) and text (Text), receive a report on a
layer (Report), change the legend of the layer
(Change Legend), and lastly view a print preview
(Print Preview) of the layer. The Navigation sec-
tion provides the user the ability to zoom in and
out for viewing data at different scales while the
Contents (has been changed to Info in latest imple-
Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)

mentation) section provides help and information
on the extension. The Spreadsheet-SDSS (for
Microsoft Excel 2007 and 2010), as well as instal-
lation instructions, can be freely downloaded at
http://www.geotree.uni.edu/SpreadsheetSDSS.
aspx (GeoTREE Center, 2011).
EXAMPLE APPLICATION
In order to demonstrate the utility of the Spread-
sheet-SDSS application, we provide an example
of a potential real-world application using the
WLC using AHP technique. In this example ap-
plication, the goal is to find a suitable location
for construction of a new elementary school in a
growing small city (~40,000 people). Four raster
layers (Table 1) are considered in this example as
important determinants of locating a new school.
In this example all layers had a spatial resolution of
30 m2 square and the layers were made up of 250
columns and 398 rows for a total of 99,500 cells.
The general flow for using the Spreadsheet-
SDSS is illustrated in Figure 4. In our case study
example, we first import (using Add Layer tool)
the four layers (Table 1 and Figure 3) into Micro-
soft Excel. Next the model is chosen (in this case
the ‘WLC using AHP’ button). After clicking on
this button a user interface is displayed in which
the user chooses which layers are to be included
in the modeling effort. The layers that can be
chosen are worksheets in the Excel workbook.
The next step in this example, as the WLC using
AHP model is being used, is to set the relative
importance (more detail below) of each layer in
relation to each other in the AHP dialog (Figure
5). After the user sets these values and clicks
Compute the model runs and a new raster output
(worksheet) is produced and added to the Excel
workbook. The user can then use the Map tools
to add cartographic elements, produce a descrip-
tive report, or use the navigation tools to zoom in
and out on the new layer.
In the example provided, the selection of
values in the AHP pairwise comparison is based
on the perceived level of importance of the layers
in relation to each other. The original AHP
method developed by Saaty (1980), and adopted
in Spreadsheet-SDSS, used a range of values from
1 (equal importance of two variables) to 9 (one
variable has extreme importance in relation to
other variable) in a pairwise comparison. In our
example, shown in Figure 5, the pairwise com-
parison indicates that population change is con-
sidered of strong importance (score of 5) in
comparison to schools, schools are considered
equally important as compared to parcel size, and
land use is considered of moderate importance in
relation to parcel size. These pairwise comparison
scores (and their reciprocals – e.g., reciprocal
score of 1/5 for schools in comparison to popula-
tion change) are used in Spreadsheet-SDSS, using

Table 1. Description of raster layers used in hypothetical real-world application of Spreadsheet-SDSS

Layer # Layer Name Description

1 Population Population change in the city from 1990-2000 derived from Census blocks in city limits. The two
change categories (0 and 1) represent whether the Census block gained (1) or lost (0) people over the 10 year
time period.
2 Schools Represents a 2 km buffered area around the six elementary schools that already exist in the city. The ‘0’
represents areas within 2 km of an elementary school. The ‘1’ represents areas that lie outside of this
area.
3 Land use Another binary category that represents land uses that are either (1) suitable for purchase and construc-
tion of a new school or (0) not suitable for building a new elementary school. This classification was
based on landcover datasets (e.g., agricultural land was suitable while wetlands were not).
4 Parcels A raster layer that portrays the size of land parcels in the city based on the assumption that a parcel
needs to be a minimum of 10 acres in size.

487
 Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)

Figure 3. The raster layers used in hypothetical Figure 4. Generalized workflow for running a
real-world application of Spreadsheet-SDSS model in Spreadsheet-SDSS

Figure 5. The AHP pairwise comparison interface in Spreadsheet-SDSS

488
Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)

algorithms described in more detail by Saaty est gray areas are determined by the model to be
(1980) and Malczewski (1999), to develop weights the most suitable locations for a new elementary
that are used in the WLC method. In a real ap- school. The darkest areas were determined by
plication of this type of spatial decision situation the model to be unsuitable for the construction
(i.e., locating a new school), a variety of stake- of a new school (the center of town). Note that
holders (e.g., school administrators, city planners, this analysis produces a range of suitability. The
parent teacher association members) would help output produced from the WLC Direct and OWA
decide on the important layers and also the pair- would look similar to that seen in Figure 6 in
wise comparison scores. In our example, popula- that the output would be a range of scores with
tion growth is determined to be the most important the lowest score being the least suitable and the
factor in selecting a site for the new school. The highest score being the most suitable. The Bool-
logic for this is that these areas of the city would ean Overlay model only accepts rasters that have
likely have the largest increase in the number of values of zero (bad or unsuitable) or one (good or
school age children entering the system. It is suitable) and the output would always be a raster
deemed to be preferred over the locations of exist- with cells of zero (unsuitable) and one (suitable)
ing elementary schools as these schools are also. All commercial raster GIS software (e.g.,
likely already at or near full capacity. Land use ArcGIS (ESRI, 2011a) and IDRISI (Clark Labs,
is slightly favored when compared to existing 2011)) can carry out Boolean Overlays. In recent
school locations as it is necessary to build a new versions, the ArcGIS (ESRI, 2011a) software has
school on land that can be easily converted for
other uses and zoned as such. Population growth
was the most important factor in locating a new Figure 6. The output raster data from the hypo-
school building, so it is slightly favored over both thetical real-world application Spreadsheet-SDSS
parcel size and land use. When using the AHP
using WLC model in the Spreadsheet-SDSS, the
decisions of how to score the pairwise comparisons
are very important and require careful consider-
ation by the stakeholders who are involved in the
spatial decision making process. Of course it is
possible for the users to alter the values and run
the model multiple times in order to see how the
outcome changes.
The output produced from running the AHP
using WLC model in Spreadsheet-SDSS for the
school location example is shown in Figure 6. The
output worksheet contains a score for each cell
representing its suitability for locating the school.
In this example, the scores range from zero (dark)
to one (light gray) as all of the input cells from
the layers were classified as ones (favorable) or
zeroes (unfavorable). The input rasters to this
model could have had any numeric values (i.e.,
not only zero and ones) and the resulting score
would have been scaled accordingly. The light-

489
 Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
added a simple WLC tool (called Weighted Sum)
added. IDRISI (Clark Labs, 2011) has WLC,
AHP and OWA tools (Clark Labs, 2011). For us-
ers familiar with the ArcGIS (ESRI, 2011a) and
IDRISI (Clark Labs, 2011) environments these
tools are relatively easy to use.
CONCLUSION AND FUTURE
DIRECTIONS
In this article we have described the development
and application of a new Spatial Decision Support
System (SDSS) which is available as an exten-
sion to Microsoft Excel. The Spreadsheet-SDSS
provides spatial decision support capabilities in a
widely used spreadsheet program and thus opens
up this technology to a wider audience than is
typical of SDSS. The vast majority of SDSS de-
velopment has traditionally included a commercial
GIS component. These GIS packages are often
expensive and require a high level of experience.
Thus, SDSS based on these GIS packages can be
under-utilized. Spreadsheet software packages are
generally less expensive and more widely used
than GIS software. The spreadsheet environment
is analogous to the raster data model often used
within GIS software and thus provides an envi-
ronment which allows spatial data to be modeled.
Indeed numerous examples of existing SDSS
applications developed in spreadsheet software
have been documented. However, these SDSS
were generally developed for specific applica-
tion areas or purposes. The Spreadsheet-SDSS
extension described in this article provides a
more generic framework which includes several
modeling techniques which have often been used
in GIS and SDSS applications. A hypothetical
real-world example was used to demonstrate the
functionality of the Spreadsheet-SDSS extension.
The Spreadsheet-SDSS is provided as an exten-
sion or plug-in to Microsoft Excel and provides
a rich set of functionality and a user-friendly
interface. There is functionality for importing
490
commonly used spatial raster data format, reclas-
sifying values, running models, and visualizing
modeling results cartographically. The modeling
tools include commonly used methods in SDSS
including Weighted Linear Combination, Ordered
Weighted Averaging, Analytical Hierarchical
Processing (with WLC), and Boolean Overlay.
Although, these modeling techniques have been
built into GIS software to provide SDSS capa-
bilities, the inclusion of these capabilities within
Microsoft Excel provides a means of exposing a
wider range of users and decision-makers to their
applicability in spatial decision making situations.
It is envisioned that disciplines such as business,
various agencies of government, economics, and
social sciences would benefit as there is a higher
rate of use of Microsoft Excel as compared to
GIS tools/software. Additionally, there are in-
stances in which these disciplines are involved
with issues or problems that at least have a partial
spatial component. The extension would provide
an inexpensive yet powerful tool to improve the
efficiency of many organization’s workflows.
Finally, the Spreadsheet-SDSS extension could
prove to be an extremely valuable aid in teaching
environments. The extension could be used in a
wide range of disciplines in order to demonstrate
real-world issues that have a spatial component.
Given the value of Spreadsheet-SDSS de-
scribe in this article, it must be pointed out that
the software cannot be considered as having the
breadth of functionality that many GIS software
offers for SDSS development. For example, the
Spreadsheet-SDSS, based on the Microsoft Excel
environment, has limited cartographic function-
ality for display purposes. Also, commercial
GIS software packages such as ArcGIS (ESRI,
2011a) and IDRISI (Clark Labs, 2011) are the
result of decades of development and thus there
are tremendous number of spatial analysis and
processing tools that could potentially be utilized
for SDSS development by the experienced user
or developer. Some of these tools or functional-
ity could be added to Spreadsheet-SDSS with
Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
further development efforts. An example of a
class of functions which are in commercial GIS
but not in Spreadsheet-SDSS are neighborhood
analysis functions, which consider the influence
of proximal cells.
Although this article demonstrated the potential
utility of the Spreadsheet-SDSS extension, there
are still improvements that could be made in the
future. Improvements to be made are inclusion
of more models, more flexible import/export
functionality for different raster formats, and
improved cartographic functionality.
REFERENCES
Ali, C. Q., Ross, L. G., & Beveridge, M. C. M.
(1991). Microcomputer spreadsheets for the imple-
mentation of geographic information systems in
aquaculture: a case study on carp in Pakistan.
Aquaculture (Amsterdam, Netherlands), 92,
199–205. doi:10.1016/0044-8486(91)90021-X
Beedasy, J., & Whyatt, D. (1999). Diverting the
tourists: A spatial decision support system for
tourism planning on a developing island. Inter-
national Journal of Applied Earth Observation
and Geoinformation, 1(3), 163–174. doi:10.1016/
S0303-2434(99)85009-0
Berardi, A. (2002). ASTROMOD: a computer
program integrating vegetation dynamics mod-
eling, environmental modeling and spatial data
visualization in Microsoft Excel. Environmental
Modelling & Software, 17, 403–412. doi:10.1016/
S1364-8152(02)00011-7
Burrough, P. A., & McDonnell, R. A. (2000).
Principles of geographical information systems.
Oxford, UK: Oxford University Press.
Clark Labs. (2011). IDRISI Taiga GIS and Image
Processing Software. Retrieved February 22,
2011, from http://www.clarklabs.org/products/
idrisi-taiga.cfm
Clarke, K. C. (2011). Getting started with geo-
graphic information systems. Upper Saddle River,
NJ: Pearson/Prentice Hall.
Densham, P. J., & Goodchild, M. F. (1989). Spatial
decision support systems: a research agenda. In
Proceedings of the Annual Conference on Geo-
graphic Information Systems and Land Informa-
tion Systems, Orlando, FL (pp. 707-716).
Engelen, G., White, R., Uljee, I., & Drazan, P.
(1995). Using cellular automata for integrated
modeling of socio-environmental systems. En-
vironmental Monitoring and Assessment, 34,
203–214. doi:10.1007/BF00546036
Environmental Systems Research Institute (ESRI).
(1996). ArcView GIS: The geographic information
system for everyone. Redlands, CA: ESRI.
Environmental Systems Research Institute (ESRI).
(2011a). ESRI Products: Desktop GIS. Retrieved
February 22, 2011, from http://www.esri.com/
products/index.html#desktop_gis_panel
Environmental Systems Research Institute (ESRI).
(2011b). ESRI ASCII Raster format. Retrieved
February 23, 2011, from http://resources.esri.
com/help/9.3/arcgisengine/java/GP_ToolRef/spa-
tial_analyst_tools/esri_ascii_raster_format.htm
GDAL. (2010). Geospatial Data Abstraction Li-
brary: Version 1.6.0 [Software]. Beaverton, OR:
Open Source Geospatial Foundation. Retrieved
from http://www.gdal.org
GeoInformatics Training, Research, Education,
and Extension (GeoTREE) Center. (2011) Spread-
sheet Spatial Decision Support System (SDSS):
Version 1.1 [Software]. Retrieved February 22,
2011, from http://www.geotree.uni.edu/Spread-
sheetSDSS.aspx
Jiang, H., & Eastman, J. R. (2000). Applica-
tion of fuzzy measures in multi-criteria evalu-
ation in GIS. International Journal of Geo-
graphical Information Science, 14(2), 173–184.
doi:10.1080/136588100240903
491
 Development and Application of a Spreadsheet-Based Spatial Decision Support System (SDSS)
Keenan, P. B. (2003). Spatial decision support
systems. In Mora, M., Forgionne, G., & Gupta,
J. N. D. (Eds.), Decision making support sys-
tems: Achievements and challenges for the new
decade (pp. 28–39). Hershey, PA: Idea Group.
doi:10.4018/978-1-59140-045-5.ch003
Keenan, P. B. (2006). Spatial decision support
systems: A coming of age. Control and Cyber-
netics, 35(1), 9–27.
Keisler, J., Blake, R., & Wagner, J. (2009). Imple-
menting Geographic Information Systems (GIS)
in spreadsheet models: What, why, and how. In
Proceedings of the 42nd Hawaii International
Conference on System Sciences (pp. 1-10).
Li, Y., Shen, Q., & Li, H. (2004). Design of spatial
decision support systems for property profession-
als using MapObjects and Excel. Automation
in Construction, 13, 565–573. doi:10.1016/j.
autcon.2004.04.004
Lo, C. P., & Yeung, A. K. W. (2002). Concepts
and techniques of geographic information systems.
Upper Saddle River, NJ: Prentice Hall.
Longley, P. A., Goodchild, M. F., Maguire, D. J.,
& Rhind, D. W. (2011). Geographic information
systems & science. Hoboken, NJ: John Wiley &
Sons.
Malczewski, J. (1999). GIS and multicriteria deci-
sion analysis. New York, NY: John Wiley & Sons.
Malczewski, J. (2006). GIS-based multi-
criteria decision analysis: a survey of the
literature. International Journal of Geo-
graphical Information Science, 20(7), 703–726.
doi:10.1080/13658810600661508
This work was previously published in the International Journal of Decision Support System Technology (IJDSST), Volume 3,
Issue 3, edited by Pascale Zaraté, pp. 1-14, copyright 2011 by IGI Publishing (an imprint of IGI Global).
492
Malczewski, J., Chapman, T., Flegel, C., Walters,
D., Shrubsole, D., & Healy, M. A. (2003). GIS
– multicriteria evaluation with ordered weighted
averaging (OWA): case study of developing wa-
tershed management strategies. Environment &
Planning A, 35, 1769–1784. doi:10.1068/a35156
Raubal, M., Gaupmann, B., & Kuhn, W. (1997).
Teaching Raster GIS operations with spreadsheets.
The Journal of Geography, 96(5), 258–263.
doi:10.1080/00221349708978800
Saaty, T. L. (1980). The analytical hierarchy pro-
cess. New York, NY: John Wiley & Sons.
Strager, M. P., & Rosenberger, R. S. (2006). In-
corporating stakeholder preferences for land con-
servation: Weights and measures in spatial MCA.
Ecological Economics, 58, 79–92. doi:10.1016/j.
ecolecon.2005.05.024
Sugumaran, R., & DeGroote, J. (2010). Spatial de-
cision support systems: Principles and practices.
Boca Raton, FL: CRC Press/Taylor & Francis
Group. doi:10.1201/b10322
Sugumaran, R., Meyer, J. C., & Davis, J. (2004).
A web-based environmental decision support
system for environment planning and watershed
management. Journal of Geographical Systems,
6, 1–16. doi:10.1007/s10109-004-0137-0
Uran, O., & Janssen, R. (2003). Why are spa-
tial decision support systems not used? Some
experiences from the Netherlands. Computers,
Environment and Urban Systems, 27, 511–526.
doi:10.1016/S0198-9715(02)00064-9
Yager, R. R. (1988). On ordered weighted av-
eraging aggregation operators in multicriteria
decisionmaking. IEEE Transactions on Sys-
tems, Man, and Cybernetics, 18(1), 183–190.
doi:10.1109/21.87068
 493

Chapter 30
A Package-Based Architecture
for Customized GIS
Andrés Pazos
Universitat Jaume I, Spain

José Poveda
University of Texas, USA

Michael Gould
Universitat Jaume I, Spain

ABSTRACT
In this chapter we present a package-based component architecture for the specific deployment and
maintenance of public sector applications, specifically corporate Geographical Information Systems
(GIS). The three-tier architecture defines a group of server-side packages, a kernel and external packages,
that are dynamically requested and assembled at execution time according to the needs of individual
users to produce on demand customized GIS applications. An intermediate middleware layer handles
user authentication and version control. In order to demonstrate the proposed architecture, a practical
prototype has been implemented using Java WebStart. This prototype demonstrates the creation of several
GIS client applications, with increasing levels of functionality, and based on free software packages.

INTRODUCTION and in general any service relying on location

Geographic Information Systems (GISs) have
migrated over the past decade, from stand-alone
scientific niche applications to key parts of corpo-
rate IT infrastructures in public sector (and com-
mercial) organizations (Longley et al, 2001). GISs
provide services such as vehicle fleet tracking,
spatial planning (optimal location of resources),
DOI: 10.4018/978-1-4666-2038-4.ch030
information as part of the context within which
other information is interpreted. Corporate GISs
are complex, multi-participant (cost-sharing) and
multi-user computing environments complicating
practical tasks such as data sharing among depart-
ments and routine system maintenance. Each of
potentially thousands of users may have specific
requirements based on their capacity to use the
GIS, and hardware and software configuration
requirements. Conventionally, system administra-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

tors install custom upgrades and extensions one
workstation at a time, or offer all users the same
monolithic package to be downloaded remotely.
Until recently, GIS software packages were largely
monolithic; although experience shows that many
users exercise only a fraction of the functionality
offered by a full-featured GIS, system admin-
istrators were required to install and support a
full, often costly, commercial license for each
and every user. In search of just the right set of
functionality for particular users, two primary op-
tions have emerged: end user development (EUD)
which assumes the user has software development
skills (Morch, 2004) or server-side configuration
of software modules, here termed packages based
on the terminology common in the Java, Perl, and
Python communities.
In organizations with an elevated number of
users, and especially in the public sector which
is extremely cost conscious, an extensible and
easily-customizable GIS application solution is
needed. This is possible assuming two levels of
application flexibility. On one hand at administra-
tion level the system should present scalability that
allows the use of existing software modules or
components, and on the other hand, at user level,
the user should be able to determine locally the
customization, installing or updating only the parts
from the kernel of the system needed to perform
the work (Morch, 1997).
The component-based software development
(CBSD) paradigm (Brown, 2000) is helping to
bring this extensibility and flexibility to reality.
In the CBSD approach new code development is
minimized and system upgrades become the task
of replacement of well-bounded functional units
of the system. In our proposed architecture, the
components are grouped into packages, defined as
a functional, replaceable set of services. A pack-
age is a higher-level aggregation of functionality
compared to an ActiveX component or a JavaBean.
Adherence to interface standards allows acqui-
sition of packages developed from third-party
developers (Commercial off-the-shelf (COTS)
494
A Package-Based Architecture for Customized GIS
or open source) and adapting them to the CBSD.
These applications may be composed of a light
kernel, which implements the basic functionality,
and a set of independent packages offering ex-
tended functionality (commonly called plug-ins).
In this case, each package implements a specific
part of the GIS—data conversion, visualization,
and so forth, is connected to the kernel through
published interfaces in order to compose the final
application desired by the user.
The architecture described here centralizes
software maintenance and distribution, exploiting
the existence of ubiquitous Internet infrastructure
throughout the public sector (in the developed
world). Conventional client/server architectures
do not cope well with certain authentication and
monitoring requirements. For this, the authors
moved to 3-tier software architecture. A central
server stores and dispatches the different pack-
ages that would compose the final application,
while the middleware provides a natural place to
locate the adaptive behaviour (McKinley, 2004).
The client interface is through a specific HTTP
browser that connects to the middleware services.
System administrator takes charge of controlling
the server side adapting and updating the pack-
ages to the CBDS.
The case study that motivated this work stems
from our interaction with the public administra-
tion, specifically the Infrastructure and Transport
Department of a regional government authority.
This organization has more than 400 computers
users, many of whom utilize GISs on a sporadic
basis. They commonly use proprietary desktop
applications with high annual license payments.
However, according to internal surveys, many of
these users require only a specific subset of GIS
functionality. Moreover, users are distributed
geographically, so it is important to centralize
the remote maintenance and the installation of
the software in order to save time and money.
The architecture proposed here is valid for
both commercial and free software, but additional
monetary benefit naturally accrues for the case
A Package-Based Architecture for Customized GIS
of GIS applications based on free software com-
ponents (Anderson, 2003). Today, it is possible
to reuse existing software libraries that cover the
basic functionality of a GIS application, and in
many cases source code is available so that these
libraries can be modified or extended by system
administrators. Examples of these libraries are
Geotools (www.geotools.org) and Terralib (Ca-
mara, 2000), which are distributed under the Lesser
General Public License (LGPL) (GNU, 1999), so
they are valid for open source developments as
well as for commercial purposes.
RELATED WORK
The structure of the package-based architecture is
directly related to the object-oriented frameworks
(Fontoura, 2001) concept. The main difference
is that instead of a collection of smaller compo-
nents, we implement a collection of higher-level
packages with predefined collaborations between
them and extension interfaces. In order to con-
nect each developed package to the kernel, it
must apply the extension interfaces, and connect
to the points of extension called hot spots (Pree,
1995). Our packages are pieces of software that
implement a specific functionality, deployable as
a unit of versioning and replacement. The package
contains immutable code, and metadata informa-
tion necessary for connection to the kernel. This
structure must allow a dynamic recomposition
of the software transparent with respect to the
application source code.
Software components are well defined by
Szyperski (1998) as binary units of independent
production, acquisition, and deployment that
interact to form a functioning system. The most
well-know component model is ActiveX/COM+
(Rogerson, 1997), an interesting model for
Windows programming integration but not very
interesting for dynamic composition. Moreover,
the option of programming custom solutions using
ActiveX components is viable for groups of users,
but not for individual customization.
In the field of GIS a widely-implemented
example of CBSD is the ESRI MapObjects (Hart-
man, 1997) platform: an ActiveX control and a
collection of dozens of automation objects that
provide basic GIS and mapping capabilities. These
components must be embedded in an application
development environment, such as Visual Basic,
beyond the capability of typical GIS users and
therefore the domain of end user composition.
We are interested in a higher aggregation of
components (self-contained packages) that can
be included into the main application without
the need to modify the source code of the base
system, and that can be maintained and updated
centrally by the system operator
ARCHITECTURE OVERVIEW
In search of a low cost GIS solutions with an
optimal level of functionality, we modify the tradi-
tional concept of a monolithic GIS application and
redistribute it to better fit a changing corporative
environment, one which is integrated in a high-
speed network (100 Mbps LANs). The proposed
architecture would not have been viable in most
organisations 10 years ago, as shown in previous
studies (Anitto, 1994) due to the inherent com-
munication capabilities required. The proposed
architecture is defined keeping in mind two key
system features: minimal system maintenance
costs due to centralization and scalability of the
system in a natural way, avoiding costly reengi-
neering processes.
In order to achieve the desired flexibility, the
system is separated into a kernel and a set of ex-
ternal packages. The kernel implements the main
thread of the system and is in charge of building
the client side GIS application. In this process, the
requested packages are loaded in order to build
the final application.
495

System maintenance is high-overhead cost,
and so we follow Isnard (2004), in assuring that
the system administration design also builds
maintenance tasks into the computing environ-
ment. The proposed architecture centralizes all
software maintenance and distribution, by means
of a central server that contains the packages for
future composition of custom GIS applications.
Corporate Needs in GIS
After studying the existing solutions for Corporate
Geographic Information Systems (CGIS), we have
identified the following key factors to take into
consideration:
• Customization. Following EUD prin-
ciples, the user must be able to select and
install only the specific packages needed to
perform his/her particular work and not all
the functionalities of the CGIS, assuming
access priveledges.
• Extensibility. The system must be scalable
according to the new needs of the organi-
zation, allowing the total re-use of the ex-
isting system.
• Software distribution. Organizations
operate in a decentralized fashion, and
conduct business in many geographical
regions, therefore they require distributed
computer support. The software must be
downloaded and installed across a net-
work, avoiding repetitive, manual local
installations.
• Maintenance. A centralized architecture
helps in the administration of the system,
because it allows updates of each indepen-
dent piece (package) with a minimal cost.
General Structure
Traditionally, the GIS architecture follows a client-
server schema, where the data processing takes
place on the server side and the results are sent to
496
A Package-Based Architecture for Customized GIS
the clients for the its representation. At the server
side, the system loses flexibility and introduces a
performance bottleneck at executing time. On the
other hand, at the client side the system overloads
communication between the server and the clients
due to redundant functionality sent to the different
clients. Taking into consideration these aspects,
the schema that has been adopted corresponds to
a three-tier software architecture (see Figure. 1) as
a result of adding to a simple server-client schema
a new tier located between the client and server
in charge of user authentication. If we take into
consideration the possible geographic informa-
tion databases of the system, the architecture (see
Figure. 1) can be naturally extended to a N-tier
software architecture.
The proposed architecture exploits the main
benefits of the web as transfer protocol (ubiquity,
portability, reliability and trust) to deliver the
different pieces that will conform to the GIS ap-
plication. The client layer is responsible for the
interaction with the user and for the generation
of the queries sent to the middleware with the
user profile validation and authentication ser-
vices. These requirements, in practical terms, are
achieved with a web-based system architecture
where the client is a standard web browser and the
middleware is implemented as a web server. The
server layer, as repository, contains the packages
that will be sent to the client after the processing
of the middle tier messages. The interchange of
messages between the different layers is imple-
mented as XML files (Aloisio, 1999).
The process to get the final GIS application at
the client side is as follows:
1. Connection to the middleware layer through
the client browser, the first time the user
composes the GIS application and each
time that the GIS application configuration
is updated.
2. Validation and authentication with user
name and password. According to the user
profile and package visibility privileges, the
A Package-Based Architecture for Customized GIS
middleware displays a list of the possible
packages (read here functionalities) that can
be added to the final GIS application.
3. Selection of packages (functionalities) that
the user requests to install.
4. XML configuration. After receiving the
result of the selection, the middleware dy-
namically creates a XML file that contains
the user profile and packages configuration
information. This configuration file is sent
to the server.
5. Processing at server side. The server pro-
cesses the XML configuration file and sends
the selected packages to the client. These
packages are automatically connected to the
client kernel thereby composing the updated
GIS application.
Application Composition
After downloading the requested packages, the
kernel builds and launches the application auto-
Figure 1. Architecture overview
matically. Figure 2 shows the general structure of
the packages composition.
The CGIS has been designed in such a way
that packages fit to the general structure of the
kernel. All the parts of the kernel to be accessed
from external packages must be defined according
to an open and well defined specification. In the
literature we find that these parts have been called
“hot spots” or “point of extension” (Anderson
and Moreno-Sanchez, 2003). After receiving the
requested packages, the kernel examines them
and extracts the information contained in a valid
XML configuration file. This configuration file
must be defined according to the Document Type
Definition (DTD) of the kernel and contains in-
formation about the points of extension to be
extended. For example, if for a particular GIS
application (an instantiation of the CGIS) a par-
ticular extension of the GIS application menu is
required, the XML configuration file must contain
a label for the package to be included in the menu
and for the associated actions.
Besides the hot spots, the kernel of the applica-
tion exposes an interface that allows the packages
to the access internal objects. For access to the
internal kernel objects, the packages must use this
public interface of the kernel.
FURTHER FEATURES
Once the interfaces for the interaction of the pieces
(kernel and other packages) and the hot spots of
the CGIS are defined, developers are able to build
up new packages with additional functionality and
to simply include them in the CGIS for possible
use by any client. The remainder of the CGIS
does not have to be modified, because of the
structure of the CGIS that allows by design the
direct inclusion or replacement of new packages.
The task of uploading new packages to the CGIS
repository corresponds to the administrator of the
system. The users (CGIS clients) connected to the
middleware tier, are able to add to the GIS client
497

Figure 2. Package composition
application new packages (functionalities) if they
are visible according to the user profile.
The user needs to connect to the middleware
only the first time that the application is built and
downloaded or when an updating of any installed
packages is need After the first time the GIS ap-
plication has been downloaded and installed at
the client side, the GIS application remains in
the local computer. However, the user always
may uninstall existing packages or the whole GIS
application and download it again.
The administrator of the CGIS or any user with
administrator privilege through the middleware
tier is able to list the available packages, to delete
any of them or to add new ones. The administrator
of the CGIS grants user privileges.
CONCLUSION
In this paper we examine some of the limitations
of GIS applications involved in public sector
corporate environments. Specifically, we con-
centrated on some crucial issues of integrating
component-based software development and
three-tier architectures for geographic information
systems. These ideas are the result of previous
investigation into the uses of GIS software in the
public administration and the possibilities of open
source in this field. We have attempted to find the
498
A Package-Based Architecture for Customized GIS
appropriate design for corporate environments in
an effort to minimize maintenance costs, while at
the same time promoting the interoperability and
flexibility of the resulting software framework.
In this article, we proposed an architecture,
named CGIS, that allows the optimum distri-
bution and installation of the GIS application,
where each user can customize his/her applica-
tion interactively. The software architecture has
been validated through the implementation of a
testing prototype. This development has been a
key aspect in order to improve and demonstrate
the design features. The architecture is universal,
not strictly related to any protocol, programming
language or platform, though some characteris-
tics of the selected implementation language, as
for example platform independence, makes the
system even more universal. In particular, in the
validation implementation Java WebStart was
utilized with success.
Finally, the architecture creates a simple yet
effective framework for the development of GIS
applications. With the inclusion of independent
packages in a central server, different needs of
the corporative environment can be supported.
As future work, we propose the implementation
of new packages to extend functionalities of the
current prototype. It is important to make an effort
to incorporate more open source libraries such as
Geotools 2 and Terralib that allow the creation
A Package-Based Architecture for Customized GIS
of new packages with a minimum effort. On the
other hand, the system can be extended in the
middleware layer. We propose to add functional-
ity for monitoring and controlling of downloads
of the different users in order to produce useful
statistics for administration purposes.
REFERENCES
Aloisio, G., Millilo, G., & Williams, R. D. (1999).
An XML architecture for high performance web-
based analysis of remote-sensing archives. Future
Generation Computer Systems, 16, 91–100.
Anderson, G., & Moreno-Sanchez, R. (2003).
Building Web-based spatial information solu-
tions around open specifications and open source
software. Transactions in GIS, 7, 447.
Anitto, R. N., & Patterson, B. L. (1994). A new
Paradigm for GIS data Communications. URISA
Journal, 64-67.
Brown, A. W. (2000). Component-Based Develop-
ment. Prentice Hall PTR.
Camara, G., et al. (2000). Terralib: Technology in
Support of GIS Innovation. II Workshop Brasileiro
de Geoinformática, Geoinfo2000. Sao Paulo.
Fontoura, M., Pree, W., & Rumpe, B. (2001). UML
Profile for Framework Architectures. Addison-
Wesley/ACM Press.
Gosling, J., & McGilton, H. (1996). White Paper:
The Java Language Environment. Java Articles:
http://java.sun.com
Guo, J. (2003). An Approach for Modeling and
Designing Software Architecture. 10th IEEE
International Conference and Workshop on
the Engineering of Computer-Based Systems
(ECBS’03). Alabama.
Hartman R. (1997). Focus on GIS Component
Software. OnWordPress.
Isnard E., Perez, E., & Galatescu, A. (2004).
MECASP – An Environment for Software Main-
tenance and Adaptation. ERCIM News, No.58,
45-46.
Longley, P. A., Goodchild, M. F., Maguire, D.
J., & Rhind, D. (2001). Geographic information
systems and science. Wiley and Sons.
McKinley, P., Masoud, S., Kasten, E., & Cheng, B.
(2004). Composing adaptive software, computer
IEEE. Computers & Society, 37(7), 56–64.
Monson-Haefel, R. (1999). Enterprise JavaBeans.
O’Reilly Press.
Morch, A. (1997). Three Levels of End-User Tai-
loring: Customization, Integration, and Extension.
M. Kyng & L. Mathiassen (Eds.), Computers and
Design in Context. The MIT Press, Cambridge,
MA, 51-76.
Morch, A., Stevens, G., & Won, M. (2004).
Component-based technologies for end-user
development. Communications of the ACM,
47(9), 59–62.
Open Geospatial Consortium Inc. (OGC). Web
Map Service 1.3. (2004). www.opengeospatial.org
Pissinou, N., Makki, K., & Park, E. K. (1993).
Towards the Design and Development of a New
Architecture of Geographic Information Systems,
ISBN:0-89791-626-3, 565-573, Washington.
Pree, W. (1995). Design Patterns for Object-
Oriented Software Development. Wokingham:
Addison-Wesley/ACM Press.
Rogerson, D. (1997). Inside COM: Microsoft’s
Component Object Model. Microsoft Press.
Schmidt, R. (2001). Java Networking Launching
Protocol & API Specification. Java Articles http://
java.sun.com.
499

KEY TERMS AND DEFINITIONS
Application Programming Interface (API):
The interface to a library of language-specific
subroutines, for instance a graphics library that
implement higher level graphics functions.
Binding: Language-dependent code that al-
lows a software library to be called from that
computer language.
Client-Server Architecture: The term client/
server was first used in the 1980s in reference to
personal computers (PCs) on a network. The actual
client/server model started gaining acceptance in
the late 1980s. The client/server software archi-
tecture is a versatile, message-based and modular
infrastructure that is intended to improve usabil-
ity, flexibility, interoperability, and scalability as
compared to centralized, mainframe, time sharing
computing. A client is defined as a requester of
services and a server is defined as the provider of
services. A single machine can be both a client and
a server depending on the software configuration.
Commercial Off-the-Shelf (COTS): A gen-
eral term for software products that are made and
available for sale or lease.
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 311-319,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
500
A Package-Based Architecture for Customized GIS
Component: In object-oriented programming
and distributed object technology, a component
is a reusable program building block that can be
combined with other components in the same or
other computers in a distributed network to form
an application.
Corporate Geographic Information System
(CGIS): A Corporate GIS is a GIS defined to be
used as a corporate resource by the members of
an enterprise or institution.
Extensible Markup Language (XML): Is
a flexible way to create common information
formats and share both the format and the data
on the World Wide Web, intranets, and elsewhere.
Graphical User Interface (GUI):The graphi-
cal user interface: or GUI, provides the user with
a method of interacting with the computer and its
special applications, usually via a mouse or an-
other selection device. The GUI usually includes
such things as windows, an intuitive method of
manipulating directories and files, and icons.
Middleware: Layer(s) of software between
client and server processes that deliver the extra
functionality behind a common set of APIs that
client and server processes can invoke.
 Section 3
Tools and Technologies

This section presents extensive coverage of various tools and technologies available in the field of Geographic
Information Systems that practitioners and academics alike can utilize to develop different techniques. These
chapters enlighten readers about fundamental research on the many tools facilitating the burgeoning field of
Geographic Information Systems. It is through these rigorously researched chapters that the reader is provided
with countless examples of the up-and-coming tools and technologies emerging from the field of Geographic In-
formation Systems. With 18 chapters, this section offers a broad treatment of some of the many tools and tech-
nologies within the Geographic Information Systems field.
502

Chapter 31
Algorithms for 3D Map
Segment Registration
Hao Men
Stevens Institute of Technology, USA

Kishore Pochiraju
Stevens Institute of Technology, USA

ABSTRACT
Many applications require dimensionally accurate and detailed maps of the environment. Mobile map-
ping devices with laser ranging devices can generate highly detailed and dimensionally accurate coor-
dinate data in the form of point clouds. Point clouds represent scenes with numerous discrete coordinate
samples obtained about a relative reference frame defined by the location and orientation of the sensor.
Color information from the environment obtained from cameras can be mapped to the coordinates to
generate color point clouds. Point clouds obtained from a single static vantage point are generally
incomplete because neither coordinate nor color information exists in occluded areas. Changing the
vantage point implies movement of the coordinate frame and the need for sensor position and orienta-
tion information. Merging multiple point cloud segments generated from different vantage points using
features of the scene enables construction of 3D maps of large areas and filling in gaps left from occlu-
sions. Map registration algorithms identify areas with common features in overlapping point clouds and
determine optimal coordinate transformations that can register or merge one point cloud into another
point cloud’s coordinate system. Algorithms can also match the attributes other than coordinates, such
as optical reflection intensity and color properties, for more efficient common point identification. The
extra attributes help resolve ambiguities, reduce the time, and increase precision for point cloud regis-
tration. This chapter describes a comprehensive parametric study on the performance of a specialized
Iterative Closest Point (ICP) algorithm that uses color information. This Hue-assisted ICP algorithm,
a variant developed by the authors, registers point clouds in a 4D (x, y, z, hue) space. A mobile robot
with integrated 3D sensor generated color point cloud used for verification and performance measure-
ment of various map registration techniques. The chapter also identifies various algorithms required to
accomplish complete map generation using mobile robots.

DOI: 10.4018/978-1-4666-2038-4.ch031

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Algorithms for 3D Map Segment Registration
INTRODUCTION
Complete and dimensionally accurate maps of
the environments are of interest to many domains
including surveying, search and rescue, security,
defense and construction. Laser based scanning
devices (Light Detection And Ranging-LIDAR)
are generally used to generate point clouds that
describe spatial information in the form of numer-
ous discrete point coordinate measurements. Point
data are acquired by measuring time of flight of
scattered light or phase shift between incident
and reflected light to find the distance between
the object surface and the scanning device (Blais,
2004). The speed of scanning discrete points can
be enhanced by pulse and phase based measure-
ment technologies (Blais, 2004). Precise rotation
mechanisms with high-resolution encoders spin a
2D LIDAR device to generate a 3D point cloud.
Point cloud scanners have been mounted on air-
planes (Browell et. al. 1990) and ground vehicles
(Gebre, et al. 2009) to create large area terrain
maps. When vision sensors are integrated with
the laser ranging systems, point clouds can also
contain the color information of the scene. Opti-
cal imagery from the camera is associated with
point coordinates to produce color point clouds
(Andresson, 2007).
A 3D point cloud obtained from a single
vantage point is seldom adequate to construct a
complete map. Generation of a complete map of
an environment requires merging or registration of
map segments taken from various vantage points.
The registration enables construction of large-scale
global 3D maps (Thrun, 2003). Registering the
map segments is trivial if precise position and ori-
entation of the sensor are accurately known about
a global reference frame. Position sensors such
as inertial measurement units or those relying on
global positioning systems are prone to errors and
can be highly inaccurate under certain conditions.
The map registration process determines the rigid
body translation and rotation of the sensor as its
output (Thrun, 1993, 2003). The map registration
quality varies depending upon the sensor resolu-
tion and the extent of overlap between the map
segments. Different techniques exist for merging
3D maps by exploiting geometric features and
measuring surfaces. The most popular registra-
tion algorithm for point cloud registration is the
iterative closest point (ICP) algorithm (Thrun,
2003). In ICP, the corresponding closest points in
different point clouds are associated and optimal
rigid transformation required to minimize a mean-
square error of separation between the associated
points (Bsel, 1992) is iteratively found. The color
attributes of the sampled point can be utilized in
ICP progress to increase computational speed and
provide higher accuracy. Anderson (2007) filtered
the point set data based on hue before conducting
traditional ICP. Houng et al., (2009) processed
images to extract corresponding visual features
that are used in registration process.
In this chapter, we examine the algorithms
required for a mobile robot to generate a dimen-
sionally accurate and complete map of an area
without prior information about the area. We focus
particularly on the techniques for registration of
map segments taken from various vantage points.
The chapter also describes a mobile robotic sys-
tem with a color point cloud scanner and various
algorithms required for accomplishing the mis-
sion of generating a complete and dimensionally
accurate map of an area.
MOBILE MAPPING WITH COLOR
POINT CLOUD SCANNERS
Color point clouds are created by synchronizing
range sensors such as the LIDAR with video/still
cameras. LIDAR devices discretely measure the
distance between a light source and a reflection
target at a high frequency. By changing the path
of the light through mirrors and actuators, a point
cloud of a 3D space is produced. A calibrated
vision sensor maps the color information to the
sampled points. Installing such a scanning sensor
503

on a mobile platform extends its range and enables
mapping of large areas.
3D Color Scanner
The 3D color scanner used in this effort consists
of a 2D LIDAR and two 1.3 megapixel high-frame
rate video cameras installed on the LIDAR scan-
ning plane. The LIDAR and the cameras move
such that the scan plane is rotated about an axis
within the plane, thus generating 3D color point
clouds. Figure 1 shows that the LIDAR consists
of a rotating mirror which is driven about Y axis
(degree of freedom: θ) and the scan plane is
rotated about Z axis (φ: degree of freedom). The
rotations are controlled by servomotors installed
on the axes. The cameras are calibrated to be on
the LIDAR scan plane and a forty-pixel wide
image stripe is extracted from the cameras. The
color information is then matched, in real-time,
to the points ranged by the LIDAR. The relative
distance between cameras and LIDAR is pre-
configured and images are pre-aligned. The 2D
LIDAR generates scans at a frequency of 38 Hz
and the cameras provide imagery at 60 frames per
second. Time synchronization establishes that the
pixel color is mapped to each ranged point. Use of
two cameras reduced occlusions due to the offset
between the LIDAR mirror and the camera lens.
All areas visible to the LIDAR are visible to one
of the two cameras. The 2D range measurement
along with the scanner rotation position (φ) is used
Figure 1. 3D scanning devices built with 2D commercial scanners
504
Algorithms for 3D Map Segment Registration
to generate the coordinate in a spherical coordinate
system, which is transformed to Cartesian system
as necessary. Figure 1 also shows a picture of a
compact version of the system.
The 3D color scanner is mounted on a mobile
vehicle for mapping large areas. This mobile
mapping system generates color point cloud data.
Figure 2 shows the mobile system with the scan-
ner installed on top of the vehicle. The vehicle
has no global positioning devices other than wheel
encoders. Cameras and short-range infrared sen-
sors enable observation of terrain conditions,
collision avoidance and allow a remote operator
to drive the vehicle. Map data and video feeds are
transmitted using an on-board wireless commu-
nication system. This mapping system performs
scans only when it is stopped. The vehicle can
localize itself from the map observations and
moves directly from one vantage point to the next
and acquires additional map information. This
system can generate color point cloud maps with
0.25° angular resolution in the vertical scanning
direction with a coverage angle of 100°. In the
rotation (φ) direction, the resolution is at 0.1° with
coverage angle 300°. The map segment from one
vantage point covers a maximum radius of 80
meters.
The data elements produced by the scanner
are shown in Figure 3. Figure 3(a) shows the
camera image taken from the vantage point de-
picting scene visible to the scanner. The 3D color
point cloud generated at that vantage point is
Algorithms for 3D Map Segment Registration

Figure 2. Mobility platform used for 3D color map construction in large area (Gebre et al., 2009)

shown in Figure 3(b). In this figure, the coordinate
(x,y,z) and the color (r,g,b) for all the pixels are
known. The point density (spatial resolution of
the point cloud) varies on the left and right sides
of the color scan scene depending upon the distance
between the scanned point and the scanner. The
closest area to the scanner has the highest den-
sity of points. The scanner also records the optical
reflection intensity of laser beam. The intensity
information is combined with range measurement
data and shown in Figure 3(c). The object surface
material, color and distance towards scanner cause
variations in intensity data. Similarities between
intensity point cloud and color point cloud can
be observed between Figure 3(b) and (c) on
edges, doors, and windows.

Figure 3. High dimensional point cloud map segment taken from a single vantage point: (a) Image of
an urban building (b) Color point cloud map (c) Laser reflection intensity map

505

Algorithms for Complete Mapping
An autonomous robot with the color point cloud
scanner can reduce the surveying and map building
cost and time. However, several methodologies
for robust self-localization, map completeness
evaluation, map based navigation and 3D map
registration must evolve before a high degree of
autonomy can be achieved.
A mapping robot deployed at initial start
position must go through the four phases of the
mapping processes as shown in Figure 4. The
robot must be able to localize itself so it can navi-
gate the scene. This can be accomplished by 2D
SLAM (Simultaneous Localization and Mapping)
techniques or other methods. Methodologies for
establishing the map completeness and detection
of occluded areas are necessary. Determination
of the optimal vantage point for filling in the
occluded areas and exploring unmapped areas is
also a critical step. As the navigation is based on
imprecise mapping and localization information,
the map segment registration based on 3D color
point clouds is the last but crucial step in building
the complete map of a given area. In this subsec-
Figure 4. Map completeness orientated robotic
mapping process
506
Algorithms for 3D Map Segment Registration
tion, we discuss the algorithms that address each
of these tasks.
Robot Self-Localization
The self-localization problem requires mobile
mapping robot to determine its location in an
unknown environment. Localization is critical
because robot cannot effectively navigate to the
next waypoint without the location information.
Map registrations require location and pose es-
timates. Usually robot is equipped with multiple
position and orientation sensors like GPS, Inertial
Measurement Unit (IMU), odometer, and wheel
encoders to measure real-time pose and position.
Multiple position and location sensors return robot
position information with certain level of error due
to reasons like sensor precision, GPS signal noise
and errors, sensor drift for IMU and inaccurate
measurements from other sensors.
The main challenge for robot localization is
to escape location sensor noise, drift errors, and
constantly provide accurate location and position
reference for the robot. Probabilistic self-local-
ization techniques based on maximum-likelihood
estimation have been applied to address this
problem. These techniques assume that the noise
of position sensor follows certain probabilistic dis-
tribution, which can be described mathematically.
They also assume that two subsequent map results
are highly comparable to each other and several
landmarks can be quickly identified. Therefore,
accurate relative position and location can be
solved by comparing current map with a previous
map in short time intervals, and probabilistically
maximizing similarity between two maps (Olson,
2000). Map could be generated by different sensors
like stereo cameras, sonar or laser range finders.
Landmarks extracted from maps are commonly
applied in the self-localization process to reduce
computation cost. Whyte and Bailey (2006) uti-
lize the relative localization results between two
neighbor vantage points to merge the two maps.
Algorithms for 3D Map Segment Registration
A two-step process, termed as Simultaneous
Localization and Mapping (SLAM), typically lo-
calizes the robot. The robot position is established
from multiple but imprecise sensor measurements
and comparison of landmarks in the scene. The
position sensor data is improved using sensor
fusion techniques by Spletzer (2003). Location
information is estimated based on previous loca-
tion, driving command information and current
sensor measurements. In SLAM, probabilistic
methods are applied to reduce sensor noise ef-
fects. Extended Kalman filter and particle filters
and noise models improve the location estimates
(Montemerlo, et al. 2003). The SLAM solution
has been expanded into 3D space with a six de-
gree of freedom (6DOF) SLAM which applies
sensor measurement and robot kinematics models
(Nücher, 2005). Landmark extraction and map
comparison entail the major computation effort
during the SLAM progress. Real-time SLAM
has been demonstrated with stereovision sensors
(Davison, 2003).
The SLAM technique simultaneously consid-
ers the localization and mapping mission (Thrun,
et al., 2000). The SLAM problem can be described
by a joint posterior:
P (x t , m | z 0:t , u 0:t , x 0 ) (1)
Where, xt is the state vector representing the
robot location and orientation, mi is the vector
representing the ith landmark location, zit is the
robot mapping measurement about ith landmark
at time t, and ut is the control vector applied at t-1
time to drive robot to state xt at time t.
The SLAM problem requires that Equation (1)
be solved for the time, t, and the latest robot state
vector xt be computed. Solving the joint posterior
from, 0-t requires an observation model and a mo-
tion model based on Bayes Theorem (Whyte and
Bailey, 2006). The observation model determines
the probabilistic distribution of observation zt with
known vehicle state and landmarks location as:
P (z t | x t , m ) (2)
The robot motion model describes probability
on state transition of robot state vector, xt with
known previous state xt-1 and control input ut
P (x t | x t −1, ut ) (3)
The transition of state vector is assumed as a
Markov process, implying that the next robot state
xt can only be determined on previous state xt-1
and latest control input ut and not the history of
states. The state of robot is independent of both
observations and landmarks. Equation (1) can be
recursively solved in a Prediction (time update)
and Correction (Measurement update) form.
Prediction:
P (x t , m | z 0:t −1, u 0:t , x 0 ) =
∫ P(x t
| x t −1, ut )P (x t −1, m | z 0:t −1, u 0:t −1, x 0 )dx t −1
(4)
Correction:
P (x t , m | z 0:k , u 0:t , x 0 )
P (z t | x t , m )P (x t , m | z 0:t −1, u 0:t , x 0 ) (5)
=
P (z t | z 0:t −1, u 0:t )
Equation (4) and Equation (5) recursively solve
latest robot state joint posterior. Robot state can be
predicted from the motion model P (x t | x t −1, ut )
and control input at time t. The observation model
P (z t | x t , m ) is applied to correct state prediction
with observation and mapping at time t.
In order to find solutions to the SLAM prob-
lem, proper practical descriptions about motion
and observation model in Equation (2) and Equa-
tion (3) should be provided with reliability and
efficiency. Extended Gaussian Filter (EKF) is
applied to represent these models on state-space
507

model with additive Gaussian noise (Welch and
Bishop, 1995). The EKF based SLAM simplifies
motion model as:
x t = f (x t −1, ut ) + wt (6)
f(xt-1,ut) is the robot kinematics model and wt
is the additive uncorrelated Gaussian disturbances
with zero mean and covariance Qt. The observation
model can be described as:
z t = h(x t , m ) + vt (7)
In which, h(xt,m) is the observation geometry
description and vt is the additive uncorrelated
Gaussian disturbance with zero mean and covari-
ance Rt. Equations (6) and (7) can be applied to the
SLAM prediction and correction. In EKF-SLAM
process, the mean and covariance of both motion
model and observed motion should be updated at
every time t. Other probabilistic methods such as
Particle Filter (PF) (Montemerlo et al., 2003) and
Graph Filter (GF) are used to solve the SLAM
problem. A typical SLAM method is implemented
on 2D space, however, SLAM in 3D space with 6
Degree of Freedom (6DOF) on robot kinematics
have been implemented by expanding landmarks
state, motion model and observation model into
3D space (Nücher, 2005).
Map Completeness Evaluation
The map completeness problem can be addressed
with several methodologies including grid oc-
cupancy, obstacle recognition and object view
completion detection. The completeness of map is
calculated by occupancy grid map (Thrun, 2003),
which entails projecting the acquired map on an
occupancy grid and calculating the occupancy
level. Possible mapping area is determined based
on the contour of the objects and separating the
map into areas that can be potentially mapped or
impossible to map (Oh et al., 2004). Terrains are
508
Algorithms for 3D Map Segment Registration
extracted from current incomplete map for pos-
sible paths for navigation. The map evaluation
also returns possible explorative area that is acces-
sible to the mobile robot but not mapped. If map
completeness is the most important factor for the
mission, algorithms that evaluate latest exploration
status after every scan may require assessment of
the complete map and not just the current map
segment. There are many techniques to evaluate
the completeness of mapping, namely, grid based
occupancy map (Thrun, 2003), network/graph,
cell based map (Zelinsky, 1994) and template
based completeness evaluation (Oh et al, 2004).
The occupancy grid map is one of the most
commonly used methods to determine map
completeness. Area of interest is gridded and
acquired maps from different vantage position
are transferred into or projected onto the grid.
Grid is marked as occupied when data exists on
this grid, every grid should be represented with
certain level of occupancy, which is computed by
density of point cloud map on this grid. Map can
be assumed as complete all the mapped objects
form self-closed contours or closed contours with
the boundaries of the mapped area.
A major challenge in map completeness evalu-
ation is deciding whether an area can be mapped.
For example, when mapping robot is performing
indoor exploration, space behind wall of the hall-
way may not be accessible. Contours extracted
from latest global map may be used to determine
possible navigation paths. Possible mapping area
exists for contours with gaps. Ascertaining that
the gaps in map contours are indeed traversable
paths requires discerning traversable pathways
in the map.
Map-Based Navigation
Determination of the next vantage point may
depend upon several criteria: best view, coverage
of unmapped areas, areas of overlap with current
map, localization, accessibility and traveling costs.
Two steps are required for determination of the
Algorithms for 3D Map Segment Registration
next vantage position. The first step is the genera-
tion of candidate positions and second step is the
selection of optimal vantage point from the list.
The candidate vantage positions can be created
based on frontier exploration algorithm (Basilico
& Amigoni, 2009) considering obstacles, position
and terrain conditions. The vantage position is
selected between candidate positions that have
the best view coverage and shortest traveling cost.
Next vantage point should be decided based on
the best view to fill occluded regions and cover
as much new area as possible. Frontier based
exploration algorithm provides vantage point
candidates for the best view point, these candidate
points are evaluated to determine best vantage
point for next mapping.
Computing vantage position for mapping based
on previous vantage positions and incomplete map
is known as the Next Best View (NBV) problem
(Yamauchi, 1997; Basilico & Amigoni, 2009).
NBV algorithms navigate robot to acquire maxi-
mum uncovered area. A certain level overlapped
area ensures that the robot has enough landmarks
to navigate between the current and the next best
view vantage point. Frontier based algorithm can
be applied to provide candidate positions for the
next best view point. Based on the regions on the
boundary between mapped and unmapped space,
the frontier can be extracted. Considering the range
for mapping sensor constraints, next mapping
position on the frontier can then be generated.
Current frontier should be evaluated in occupancy
grid map so that the frontier grid positions that
cover more unoccupied can be selected to accel-
erate the coverage of the area. These candidate
points can be evaluated based on the criteria for
the exploration and time and power requirements
for reaching the vantage point.
The map data acquired from various vantage
points must be registered into global map space
using various registration algorithms. Although
this section describes the various algorithms
required for complete map generation, the focus
of this chapter remains on the registration aspect
of the mapping exploration.
ALGORITHMS FOR REGISTERING
MAP SEGMENTS
Three-dimensional point cloud segments acquired
from different locations have to be combined to-
gether as complete large-scale map. Position and
orientation information required for registration
can be provided directly by mobile platform sen-
sors such as GPS and IMU (Thrun 1993). In most
cases, position information acquired from sensor
is reasonably accurate. However, the orientation
information is costly and relatively imprecise
because orientation sensor measurement can be
affected by external disturbances like magnetic
field variations and sensor integration drift with
time. Position and orientation information can also
be provided by indirect techniques based on both
rough position sensor measurement and common
geometric feature identification. Figure 5 shows
two maps generated from separate vantage points.
The left map on the top row shows map gener-
ated with robot facing towards one side of the
building, the right map shows the map generated
from the second vantage point. The bottom figure
shows the map data from the first vantage point
registered into the coordinate system of the second
location. Registering the two segments produces
the complete map of the façade of the building.
Comparing with the SLAM algorithm, map
registration techniques focus on generating ac-
curate map details rather than localization of the
robot in a global coordinate system (Arun, 1987;
Bsel, 1992; Lorusso, 1995; Rusinkiewucz, 2001).
Discrete range points received from color point
cloud sensor contain detailed spatial information
about the environment. Different techniques exist
for merging such point clouds together by exploit-
ing geometric features and measuring surfaces.
Map registration techniques such as Iterative
Closest Point (ICP) algorithm proposed by Bsel
509

Figure 5. Map segments generated from two vantage points (Top) and registered map (Bottom)
(1992) has been applied to stitch two neighbor
3D point cloud maps together into one map based
on their common coverage area. Upon conver-
gence, ICP algorithm terminates at a minimum.
Several algorithms are in existence for calculating
the minimum average distance between two point
clouds. Singular Value Decomposition (SVD)
method by Arun (1987), eigen-system methods
that exploit the orthonormal properties of the
rotation matrices, and unit and dual quaternion
techniques were adopted in ICP process. Quater-
nion based algorithms have been used in ICP for
map fusion by Bsel (1992), SVD based algorithms
are widely used in ICP and 6DOF SLAM (Arun
1987, Nucher, 2005, Joung et al., 2009) as they
are robust to reach local minimum and easy to
implement. Several variants of ICP are reported
by Rusinkiewucz (2001) to increase the speed
and precision. Corresponding points sampling,
matching, weighting and rejecting are some
methods used to accelerate the ICP algorithm. In
the ICP algorithm, associating corresponding
points in two point cloud data sets is the most
critical step. Nearest neighbor search in 2D or 3D
space is commonly used for associating the cor-
510
Algorithms for 3D Map Segment Registration
responding points. Parallel ICP algorithms have
been developed by Robertson (2002) to accelerate
computation speed. Point to plane registration
method (Lorusso, 1995, Rusinkiewucz, 2001,
Salvi et al., 2007) accelerates the ICP iteration
and convergence.
Other techniques include the point signature
method by Chua (1997), which uses signature
points to describe curvature of point cloud and
matches corresponding signature points during the
registration process. Spin image based methods
compute 2D spin image to represent surface char-
acterization and solve the registration problem by
finding best correspondence between two different
scan spin images (Johnson 1997). Other methods
like principle component analysis (Chung and Lee,
1998) and algebraic surface model (Tarel et al.,
1998) are based on the point cloud surface geo-
metrical features. The normal vector distribution
can be translated into an orientation histogram in
an Extended Gaussian Image (EGI) (Makadia &
Daniilidis, 2006). Rigid motion required to register
two point clouds is solved from the cross covari-
ance function (Chibunichev & Vilizhev, 2008) of
the two EGI images. Rigid motion could also be
Algorithms for 3D Map Segment Registration
solved in Fourier domain by computing Discrete
Fourier Transform on Rotation Group on SO(3)
(SOFT) (Joistekecm and Ricjnirem, 2008).
Registration of color point clouds has been
considered (Ferbabdez, et al., 2007; Druon,
2007; Newman et al., 2006; Anderson, 2006,
2007). By applying proper calibration on the
hybrid sensor system (Joung et al., 2009; New-
man, Cole, Ho, 2006), range measurement and
visual information can be integrated together to
construct a visually accurate representation of the
scene. Color mapped 3D data was used in map
registration by weighted red, green, blue data.
The corresponding point search during the ICP is
conducted on both the coordinate and color data
(Johnson, Kang, 1997). Hue filters were also used
to constrain the closest point search in every ICP
iteration (Druson, 2007). Color data can be used
to estimate initial alignment of pair wise scans
using Scale Invariant Feature Transform (SIFT)
techniques. Color attributes transferred in YIQ
color model can also be weighted to construct
new variant together with range information for
ICP fine registration. Depth-interpolated Image
Feature (DIFT) algorithm solves corresponding
points between two images and registers color
point clouds based on extracted correspondences
(Anderson, Lilienthal, 2010).
In this chapter, we introduce hue assisted ICP
algorithm for registration of color point clouds.
The criteria for association are defined on a 4D
space rather than 3D geometric space. The fourth
dimension selected is the hue, representing the
intrinsic color values of the pixel. While achieving
the effect of a hue-based filter, hue-association
reduces the nearest neighbor search burden con-
siderably (Men & Pochiraju, 2010). The remaining
sections of the paper describe the approach and
the performance of the algorithm under several
hue distributions in the scene.
HUE-ASSISTED ITERATIVE
CLOSEST POINT (H-ICP)
ALGORITHM
The primary hypothesis of this algorithm is that the
hue value can be applied to increase the accuracy
of point association and accelerate the registration
process. The major time and computation cost
during ICP is finding the correct points pairs.
Closest spatial distance is typically applied in 3D
ICP method. The distance value in 3D space can
be expanded into 4D space by adding weighted
hue value as the 4th dimension. By integrating hue
value into the closest point search, accuracy of
point association can then be improved.
Hue Invariance with Vantage Point
Hue value remains consistent about the same
point between images taken from two vantage
points, while the color values represented in red,
green and blue quantities usually differ because
of variation in light conditions. In order to apply
color to improve the association process, light-
ing effect should be removed. Color raw data
are transformed into representation of separate
chroma, lightness and brightness value. Figure 6
shows two camera images of different angles of
a color palette on a Rubik’s cube, four colors are
used on the same surface. Figure 6 also shows the
color pixels with the background and black frame
removed. Histograms showing the red, green and
blue value in RGB space for all the pixels are
shown in Figure 7. In the RGB histogram, R, G,
and B distributions of the image vary considerably
with the vantage point. When the RGB color space
is transformed into HSL space and histograms of
hue, lightness and saturation are plotted in Figure
8, the hue values remain relatively invariant with
the position of the camera. Therefore, hue value of
the pixel, taken from the Hue-Saturation-Lightness
(HSL) model, is used as the fourth dimension in
the point association process. In Figure 9, the
hue rendered point cloud of color point cloud in
511
 Algorithms for 3D Map Segment Registration

Figure 6. Rubik’s cube camera images take from two vantage points

Figure 3(b) is shown. Hue values are normalized
between 0 and 1. The hue distribution is typically
similar to the color distribution in Figure 3(b).
Construction of a Weighted 4D
Search Space
Both hue and range value have to be combined
together in the H-ICP variant as {xo, yo, zo, hw}
for point association. xo, yo, zo are the original
coordinate values with distance units and hw is the
weighted hue value. Hue values are normalized
to a 0-1 range and must be weighted during the
closest point search in the four-dimensional space.
In order to normalize the coordinates, we find the
bounding box for each point cloud segment and
the coordinate space is rescaled to a 0-1 range.
The normalized variant for point association is
{x, y. z, hw}, where x=xo/rx, y=yo/ry, z=zo/rz. rx, ry,
rz are the dimensions of the bounding box in x,
y, z directions.

Figure 7. RGB distribution varies with camera positions

512
Algorithms for 3D Map Segment Registration
Figure 8. HSL distribution: Hue remains invariant
Figure 9. Hue rendered point cloud of the scene shown in Figure 3
The weight value for the hue dimension should
be properly selected for point association. Since
both range and hue value are normalized from 0
to 1. Weight for hue represents its influence in
the nearest neighbor search process. Low weight
biases the point association towards the range
data and a high weight towards the hue values.
Small weight values for the hue correspond to the
traditional 3D-ICP. Hue weight should be selected
between 10% and 35% for accurate point asso-
ciation. Error in H-ICP will be evaluated by the
average mean square root distance of normalized
associated point pairs.
k-d Tree Based Point
Cloud Association
In 3D ICP algorithm, corresponding points are
searched according to the closest distance rule.
This may cause incorrect matching during single
iteration loop as Figure 10. Dashed line circle
illustrates range based nearest point association
results, in which all points in data set look for
nearest neighbor in 3D space. It takes more than
one iteration to pair correct nearest neighbor points
for given data points set. Grey circle denotes the
H-ICP nearest point search that also uses the
correct hue property in finding the best neighbor
in the model. Depending on the correct color
information, corresponding point can be locked
with less iteration.
513
 Algorithms for 3D Map Segment Registration

Figure 10. Point association based on nearest distance (dotted) and nearest distance and hue (solid)

The ICP computation speed and precision are
highly dependent on association process. Use of
a k-d tree for closest point search and association
or the Nearest Neighbor Search (NNS) problem
increases the speed and efficiency of the search.
The k-d tree is a spatial partitioning data structure
that stores and organizes data in a k dimensional
space. The k-d tree is a generalized type of bi-
nary tree, with every leaf node is a k-dimension-
al data point that splits the hyperspace into two
subspaces. Splitting is done sequentially from the
first dimension to the kth dimension. A typical k-d
tree in 2D space is shown in Figure 11(a). Each
point in the 2D space divides the space sequen-
tially into a left-right spaces (about x-axis) or into
a top-bottom spaces (about y-axis).
Nearest neighbor search can be done very ef-
ficiently on k-d trees. For a given point with known
coordinates in the data point cloud and a search
radius, the algorithm recursively moves down the
tree and follows the same procedure as insertion.
Search stops at a leaf node of the tree and the
points in the model tree within the search radius
are identified. The nearest point is obtained using
distance computation. Figure 11(b) shows the
nearest neighbor (red square) for the search point
at the center of the circle. The nearest point is
then regarded as the point associated with the
search point.
In 3D closest point search, the distance between
2 points between 2 point clouds is:
rij = (mix − d jx )2 + (miy − d jy )2 + (miz − d jz )2
(8)
in which, di{dix,diy,diz} and mj{mjx,mjy,mjz} are
point spatial coordinates in data and model point
cloud respectively.

Figure 11. k-d tree construction and nearest neighbor search in 2D space: (a) k-d tree construction in
2D space (b) 2D space nearest neighbor search in k-d tree

514
Algorithms for 3D Map Segment Registration

In 4D space, the 4th dimension for each point Use of hue as a fourth dimension is significant
should be weighed hue value dhw or mhw. The in those instances where the coordinate based
spatial value of points should be normalized by matching results in a non-unique registration. For
3D search radius rij as mentioned in section 4.1. example, if the points in the model and the data
In order to accomplish closest point search in point clouds belong to a plane, traditional coordi-
4D space, the distance between two normalized nate based ICP results in non-unique association
points di{dix,diy,diz,,hihw} and mj{mjx,mjy,mjz,mjhw } of points. In such cases, using the hue value may
should be: result in unique registration of the points. The
color assisted ICP algorithm in this paper can be
rij ' = (mix − d jx )2 + (miy − d jy )2 + (miz − d jz )2 + (mihw − d jhw )2 described as follows.
(9)
1. Estimate the initial transformation matrix R
or and T;
2. Construct k-d tree of model point cloud
M{m1,m2,m3…mM}, hue value has been
rij ' = rij2 + ∆hijw 2 (10)
weighted as the 4th dimension;
3. While merging error ε >preset error
In the ICP process, search radius effects the Use R and T to transfer data point cloud
computation time and final result. A constant D{d1,d2…dN}.
search radius is applied for all iteration loops.
 
Once the search radius is large, too many points D = RD + T
will be included as candidates and increases the
computational burden. Candidate points cloud will
4. For i=1 to length of data point cloud
be missed if search radius is too small. The search
Search closest point for point di {dix, diy, diz
radius is determined by the density of point cloud.
, dih} in model k-d tree
In 4D k-d tree search, the search radius comprises
If closest point mj exists in search range r
of two parts -- a distance part and weighted hue
Pair di and mj as {dk, mk};
part as seen in Equation (9). The search range for
k++;
3D distance is selected such that it ensures about
End If
50 candidate points within search radius. As hue
End For
value is not transformed at iteration, hue search
5. Acquire paired point cloud Dp and Mp, con-
is analogous to filtering. If the weight for hue is
tain N Points, calculate normalized mean
high, k-d tree search will bias toward hue dimen-
square root distance ε as error,
sion. Therefore, appropriate hue weighting ensures
that spatial search dominates over hue filtering. ε=
The ICP algorithm iteratively converges at 1 N

minimum error, which is described by mean square N
∑ (dix − mix )2 + (diy − miy )2 + (diz − miz )2
i =1
root of the spatial distance between paired points.
At each iteration, a rigid transformation matrix
is computed so that the distance error metric be- 6. Construct orthonormality matrix H
tween the associated points is minimized. Data (Equation14) and solve rigid rotation R
point cloud is transformed into the model space and translation T (Equation15, 16) for next
using the computed transformation matrix. This iteration;
iteration continues until error metric converges. End While

515
 Algorithms for 3D Map Segment Registration

Solving Rigid Transformation matrix H can be constructed based on m ′ {mi ′ ,

i=1… N} and d ′ {di ′ , i=1… N}.
ICP algorithm is an iteration process to calculate
rigid transformation matrix based on associated
S S xy S xz 
point clouds. Mi = {mix,miy,miz} represent the co-  xx
ordinates of the ith point in the model point cloud H = Syx Syy Syz 
and dj = {djx,djy,djz} is the jth point in data point  
S zx S zy S zz 

cloud. Rigid transformation (R) that minimizes
the error measure E(R,T) shown in Equation (11)
is determined. Where

N
N
1 S xx = ∑ m 'ix d 'ix
E (R,T ) =
N
∑m i
− (Rdi + T ) (11)
i =1
i =1 N
Syy = ∑ m 'iy d 'iy
A centroid for the associated points is calcu- i =1
N
(14)
lated as the first step (Equation12) and associ- S zz = ∑ m 'iz d 'iz
ated points are translated into centroid relative i =1
N
coordinates (Equation13). Orthonomal matrix of S xy = ∑ m 'ix d 'iy
associated points can then be constructed as shown i =1

in Equation14. Rotation R and translation T are

decoupled based on the gravity center of associ- Singular value decomposition is performed
ated points. Using Singular Value Decomposition on constructed H matrix for optimal rotation R
(SVD) methods, R can be determined as shown
in Equation15. Translation T is computed using H = U ΛV T (15)
Equation16.
where optimal rotation R = VU T .
N N
1 1 The translation T can be calculated as
m=
N
∑ m ,.d = N ∑ d
i i
(12)
i =1 i =1
T T
T = m − Rd (16)
In which, m = {m x , m y , m z } a n d
d = {d x , d y , d z } are the center points of associ- Convergence Criteria
ated points in model and data point clouds. N is
the amount of point pairs. The coordinated of Convergence and stopping criterion for the H-ICP
associated point in center point relative space iteration are discussed in this sub-section. An as-
should be sociation stability criteria is introduced as the one
of the convergence criteria. Association stability,
mi ′ = mi − m, di ′ = di − d (13) denoted as S, is defined as the number of points
which changed their paired point in any iteration.
If a point comes into association or changes its
In which, mi′ = {mix′ , miy′ , miz′ } and nearest neighbor, S is incremented. Large value of
di′ = {dix′ , diy′ , diz′ } are the ith associated point with S signifies that point association has not stabilized.
center relative coordinates. The orthonormality H-ICP iteration is terminated when S vanishes and
the distance error converges.

516
Algorithms for 3D Map Segment Registration
A pair wised color point H-ICP registration
was accomplished based on above criteria. Model
point cloud contained 122,409 points with color
attributes. Data point cloud is extracted from
model point cloud with a known rotation (θ z=5o).
The H-ICP registration process is compared with
3D ICP, error as shown in Figure 12(a). The as-
sociated point number reaches maximum after
the 5th iteration (Figure 12(b)), but error has not
converged. From Figure 12(c) the association
stability (S) reaches 0 after 15th and 26th iteration
for H-ICP and 3D ICP respectively. Error and
rigid transformation are shown in Figure 12(a)
and Figure 13. The known transformation (θ z=5o)
is recovered by the H-ICP and ICP algorithms.
Figure 12. Building color point cloud registration comparison between H-ICP and 3D ICP algorithm:
(a) Comparison of error convergence (b) Association number convergence (c) Association stability
convergence
MAP REGISTRATION WITH ICP AND
H-ICP
The hue distribution or the color of the model
is generally independent of the geometry. If the
entire body is painted with a color of a single hue,
H-ICP is as effective as the traditional ICP. In
this section, we describe the performance of the
algorithm under various hue distribution scenarios.
The Stanford bunny point cloud is considered as
the benchmark data set. In HSL color space, hue
value varies from 0- 360. The color correspon-
dence between RGB and hue is given in Table 1.
Environments with Fixed Hue
Distributions
For the first experiment, we textured the Stanford
bunny point cloud model as shown in Figure
14(a). In this model, the hue varies from 0 to 360
with from bottom to top at Z direction in seven
517
 Algorithms for 3D Map Segment Registration

Figure 13. Convergence of translation and rota-

Continuously Varied Hue along One
tion estimates during registration
Dimension

In the second simulation, a continuous hue distri-

bution is assigned to the bunny model. The hue
value is varied from 0 to 360, smoothly, along
the z (vertical) direction. The resultant model
and data clouds are shown in Figure 16 (a), (b).
Saturation and lightness value have been set as
constant at every point inside dataset. Hue value
can be calculated by Equation (17).

z j − z min
h = 360 (17)
z max − z min

h is the hue value at range point i, zi is the

coordinate distance for ith point at z direction, zmax
and zmin are maximum and minimum coordinate
Table 1. Hue and RGB values for several com- of the point cloud at z direction.
mon colors Continuous hue distribution on point cloud
data is registered together (Figure 16 (c)) and the
Color R G B Hue results are shown in Figure 17. A comparison of
Gray 128 128 128 0 model performance on discrete and continuous
Yellow 255 255 0 60 distribution of hue on the same model shows
Green 0 255 0 120 the expected acceleration in performance due to
Cyan 0 255 255 180 uniform distribution of hue on the model.
Blue 0 0 255 240
Magenta 255 0 255 300
Red 255 0 0 360
Figure 14. Registration of point clouds with
uniformly distributed hues: (a) Stanford Bunny
segments. Figure 14(b) also shows the initial point clouds with hue distributed as seven distinct
registration of the model and data point clouds stripes (b) Registered color point cloud
used for this simulation.
The H-ICP registration progress is shown in
Figure 15(a) and Figure 15(b). Figure 15(a) shows
the mean square error during the ICP process and
Figure15 (b) shows the number of points associ-
ated during iteration loops. Both data and model
point cloud after registration is shown in Figure
14(b). The hue-assisted ICP registers the point
and data clouds faster than the traditional coor-
dinate based ICP.

518
Algorithms for 3D Map Segment Registration

Figure 15. Registration comparisons between H-ICP and 3D ICP algorithm: (a) Mean square error
comparison (b) Associated point number comparison

Figure 16. Bunny model with continuous hue variation in one axis: (a) Data point cloud (b) Model point
cloud (c) Merged view

Figure 17. Registration comparisons between 7 segment hue model and continuous hue model: (a) Mean
square error comparison (b) Associated point number comparison

519

Randomized Hue on the Model
In this case, the model considered has a continu-
ously distributed hue but with a randomized and
noisy pattern. In this case, there is no geometric
pattern for the color on the object. The color
point clouds are rendered in Figure 18 (a, b).
The merged cloud point cloud after registration
is shown in Figure 18(c). Figure 19 shows the
error minimization iteration and comparison with
the seven-segment hue distribution model. In this
case, the hue confuses the nearest neighbor search.
The registration accuracy is also not as good as a
patterned hue case.
Figure 18. Bunny point cloud with randomized hue distribution: (a) Data point cloud (b) Model point
cloud (c) Merged view
Figure 19. Comparison between discrete and random hue distribution case: (a) Mean square error
comparison (b) Associated point number comparison
520
Algorithms for 3D Map Segment Registration
Effect of Imaging Noise
In the previous simulation, the imaging sensor is
assumed perfect. The hue on a point is assumed
to be recorded by the imaging sensor perfectly
in both model and data clouds. Some noise in
the color measurement can be expected when
the point clouds are generated from two vantage
points (Gebre et al., 2009). Considering this situa-
tion, we colorized the bunny model but with 50%
noise in the sensor. The points in the model and
data clouds differ in color by as much as 50%.
The resulting point clouds are shown in Figure
20(a, b). The merged color point cloud is shown
as Figure 20(c).
Algorithms for 3D Map Segment Registration
Figure 20. Hue mapped with noise: (a) Data point cloud (b) Model point cloud (c) Merged view
Hue assisted color ICP matching result in
camera noise color point cloud is compared with
3D ICP matching performance. From Figure 21,
noise in hue decreases the matching accuracy and
reduces the iteration efficiency. Two groups of
cloud point clouds are selected to evaluate the
performance of H- ICP algorithm compared with
typical 3D ICP. A known transformation point
cloud data pair was generated by transforming
model point cloud at 6DOF to compare the con-
vergence speed and registration accuracy as the
rigid transformation is already known. Outdoor
large scale area pair wised registration includes
8 pair wised data registration.
Registration with Six DOF Rotation
In this experiment, registration speed between 3D
ICP and H-ICP are compared using data and model
Figure 21. Comparison between H- ICP in and 3D ICP for noisy hue case: (a) Mean square error com-
parison (b) Associated point number comparison
point clouds with known (and exact) registration
transformation. Both H-ICP algorithm and 3D ICP
algorithm have been applied on a building data set
(Gebre et al., 2009). The data point cloud is taken
from a view position that is 10° off in Y and Z axis
from the model point cloud. Translation between
the point clouds is known to be 2.46, 2.612 and
0.347 along the X, Y, and Z respectively. Same
parameters for registrations are selected to be
the same as in the previous 1-DOF registration.
Error comparison and associated point number
comparison are shown in Figure 22(a) and (b).
Association stability is shown in Figure 22(c).
The evolution of rigid transformation during ICP
is shown in Figure 23. The H-ICP completes reg-
istration after 102th iteration and the traditional 3D
ICP after the 164th iteration, which demonstrates
the effectiveness of H-ICP for registering complex
521
 Algorithms for 3D Map Segment Registration

Figure 22. Registration comparisons between 3D ICP and H-ICP algorithm: (a) Mean square error
comparison (b) Associated point number comparison (c) Stability comparison

and realistic point clouds. The merged color point required to register the point clouds is shown in
cloud about building is shown in Figure 24. Table 2. H-ICP requires less number of iterations
than 3D ICP.
Sequential Registration of Multiple
Point Clouds Figure 23. Convergence of translation and rota-
tion estimates during registration
3D ICP and H ICP algorithms have been applied
on several outdoor map segments. Color point
clouds taken from eight different vantage points
have been registered together to construct a large
scale color point cloud map. Figure 25 shows the
top view of outdoor mapping area in aerial image.
This scene includes trees, road, electrical poles
and buildings. Figure 26 shows the registered map
and the vantage points from which map segments
are obtained. Pair-wise registration is applied to
construct a single map about the reference coor-
dinate of the first map segment. 3D search radius
in k-d tree was set as 1.5 and the 3D range data
was normalized based on this radius. Hue value
was normalized to a 0-1 range, hue search radius
was set to be 0.15, and hue weight was set to
5.0. The final error and the number of iterations

522
Algorithms for 3D Map Segment Registration

Figure 24. Registered data and model point clouds

Figure 25. Aerial image of outdoor mapping area and vantage positions

Figure 26. Top view of eight sequentially registered color point cloud maps

523
 Algorithms for 3D Map Segment Registration

Table 2. Sequential registration of multiple point

FUTURE RESEARCH DIRECTIONS
cloud maps

3D ICP H- ICP 3D ICP H- ICP Point clouds are inefficient representations of

Position Iterations Iterations Error Error geometry. Some of the future research directions
2 45 35 0.842 0.856 can include:
3 54 44 0.929 0.961
4 77 54 0.039 0.290 a. Efficient generation of higher order geo-
5 49 43 0.104 0.319 metric representations --- lines, surfaces and
6 66 59 0.165 0.179 solids from the point cloud data;
7 73 69 0.129 0.128 b. Map completeness measures that predict
8 99 95 0.068 0.070 the geometry missing in the occluded areas
based on a knowledge-base; and
c. Extra sensing modalities such as infrared
This experiment proves that faster registration or thermal imaging, acoustic/ultrasonic and
will be conducted by adding color value into radio frequency imaging to help determina-
registration progress. Position 3 and 4 acquired tion of materials in the scene.
point clouds have been registered together and
shown in Figure 27, Figure 27(a) describes two Architecture, surveying and engineering fields
different point clouds with two different colors; have considerable needs for automatic or semi-
point cloud at position 4 (black) has been registered automatic conversion of 3D point clouds into
into position 3 point cloud (blue). Combined point higher order line, surface and solid models that are
clouds with color are shown in Figure 27(b). compatible with commercial CAD software. This

Figure 27. Map registered from scans taken from two vantage points: (a) Registered position 4(black)
point cloud into position 3 (blue) point cloud (b) Color point cloud after registration

524
Algorithms for 3D Map Segment Registration
enables bringing the point cloud data into existing
business processes like generation of drawings for
code compliance, additions and modifications to
existing built areas and remodeling interior spaces.
CONCLUSION
This chapter describes an algorithm to introduce
color attribute into point cloud registration process
and fundamental algorithms for autonomous ro-
botic complete mapping. Normalization of range
data and hue value have been applied during the
registration process and quantitatively evaluate the
effect of hue search range and weight for the point
association process. Different hue distribution and
noise effect have been discussed with specific hue
rendered color point clouds. A building data set and
large-scale outdoor point cloud has been registered
using image data assisted algorithm. Use of the
hue value to assist the point association and error
minimization is shown to be effective during the
ICP iteration schemes. Higher dimensional point
association based on weighted hue and range data
leads more accurate point matching result, conduct
earlier convergence of ICP progress, and reduce
computation time. When rigid transformation is
been application in every iteration loop during the
ICP period, hue value does not change in space
transformation. However, in HSL data space,
Lightness should change according to the view
angle and light position. Corresponding point
search using additional lightness value could
be a further research field to increase Color ICP
algorithm.
ACKNOWLEDGMENT
The authors acknowledge the partial support
from US/ARMY ARDEC for the development
of ROAMS robot. We acknowledge the help of
Mr. Biruk Gebre for generating some of the scan
point cloud maps used in this chapter.
REFERENCES
Andreason, H., & Lilienthal, A. J. (2010). 6D scan
registration using depth-interpolated local image
features. Robotics andAutonomous Systems, 59,
157-165.
Andresson, H. (2007). Vision aided 3D laser
scanner based registration. Paper presented at
European Conference on Mobile Robots: ECMR.
Arun, K. S., Huang, T. S., & Blostein, S. D.
(1987). Least square fitting of two 3D-Point sets.
IEEE Transactions on Pattern Analysis and Ma-
chine Intelligence, 9(5), 698–700. doi:10.1109/
TPAMI.1987.4767965
Basilico, N., & Amigoni, F. (2009). Exploration
strategies based on multi-criteria decision mak-
ing for an autonomous mobile robot. Presented at
European Conference on Mobile Robots, Mlini/
Dubrovnik, Croatia.
Blais, F. (2004). Review of 20 years of range sen-
sor development. Journal of Electronic Imaging,
13, 231–243. doi:10.1117/1.1631921
Browell, E. V., Butler, C. F., & Ismail, S. (1990).
Airborne lidar observations in the wintertime
arctic stratosphere: Polar stratospheric clouds.
Geophysical Research Letters, 17, 385–388.
doi:10.1029/GL017i004p00385
Bsel, P. J. (1992). A method for registration of 3D
shapes. IEEE Transactions on Pattern Analysis
and Machine Intelligence, 14, 239–256.
Chibunichev, A. G., & Velizhev, A. B. (2008).
Automatic matching of terrestrial scan data
using orientation histograms. Presented at the
International Archives of the Photogrammetry,
Remote Sensing and Spatial Information Sciences,
Beijing, China.
Chua, C. J. R. (1997). Point signatures: A new
representation for 3D object recognition. Inter-
national Journal of Computer Vision, 25, 63–85.
doi:10.1023/A:1007981719186
525

Chung, D., & Lee, Y. D. S. (1998). Registration of
multiple range views using the reverse calibration
technique. Pattern Recognition, 31(4), 457–464.
doi:10.1016/S0031-3203(97)00063-0
Davison, A. J. (2003). Real-time simultaneous
localization and mapping with a single camera.
Presented at the 9th IEEE Computer Society
Conference on Computer Vision and Pattern
Recognition, Madison, Wisconsin.
Druon, S. (2007). Color constrained ICP for
registration of large unconstructed 3D/Color data
sets. Presented at IEEE International Conference
on Robotics and Automation, Roma, Italy.
Fernandez, J. C., & Singhania, A. (2007). An
overview of lidar point cloud processing software.
GEM Center Report, University of Florida.
Gebre, B., Men, H., & Pochiraju, K. (2009). Re-
motely operated and autonomous mapping system
(ROAMS). Paper presented at the 2nd Annual
IEEE International Conference on Technologies
for Practical Robot Applications, Woburn, MA.
Johnson, A. E., & Kang, S. B. (1997). Registration
and integration of textured 3D data. Presented
at International Conference on Recent Advances
in 3D Digital Imaging and Modeling, Ottawa,
Canada.
Joung, J. H., An, K. H., Kang, J. W., et al. (2009).
3D environment reconstruction using modified
color ICP algorithm by fusion of a camera and a
3D laser range finger. Presented at the 2009 IEEE
International Conferrnce on Intelligent Robos and
Systems, St. Louis, USA.
Johnson, A. (1997). Spin-images: A representation
for 3d surface matching. Doctoral dissertation,
Carnegie Mellon University, USA.
Kostelec, P. J., & Rockmore, D. N. (2008). FFT
on the rotation group. Journal of Fourier Analy-
sis and Application, 14, 145–179. doi:10.1007/
s00041-008-9013-5
526
Algorithms for 3D Map Segment Registration
Lorusso, A. (1995). A comparison of four algo-
rithms for estimating 3D rigid transformations.
Presented at British Machine Vision Conference.
Makadia, A., Iv, E. P., & Daniilidis, K. (2006).
Fully automatic registration of 3D point clouds.
Presented at the 2006 Computer Society Confer-
ence on Computer Vision and Pattern recognition,
New York, NY.
Men, H., & Pochiraju, K. (2010). Hue assisted reg-
istration of 3D point clouds. Presented at ASME
2010 Interaional Design Engineering Technical
Conference, Montreal, Canada.
Montemerlo, M., Thrun, S., et al. (2003). Fast-
SLAM 2.0: An improved particle filtering algo-
rithm for simultaneous localization and mapping
that provably converges. Presented at the Interna-
tional Joint Conference on Artificial Intelligence,
Acapulco, Mexico.
Newman, P., Cole, D., & Ho, K. (2006). Outdoor
SLAM using visual appearance and laser rang-
ing. Presented at IEEE International Conference
on Robotics and Automation, Orlando, Florida.
Nüchter, A. (2005). 6D SLAM with approximate
data association. Presented at IEEE International
Conference on Robotics and Automation, Barce-
lona, Spain.
Nuchter, A. (2007). Catched k-d tree search for
ICP algorithms. Presented at 6th International
Conference on 3-D Digital Imaging and Model-
ing, Montreal, Canada.
Oh, J. S., & Choi, Y. H. (2004). Complete
coverage navigation of cleaning robots using
triangular-cell-based map. IEEE Transactions on
Industrial Electronics, 51, 718–727. doi:10.1109/
TIE.2004.825197
Olson, C. F. (2007). Probabilistic self-local-
ization for mobile robot. IEEE Transactions
on Robotics and Automation, 16, 55–67.
doi:10.1109/70.833191
Algorithms for 3D Map Segment Registration
Robertson, C. (2002). Parallel evolutionary
registration of range data. Computer Vision and
Image Understanding, 87, 39–50. doi:10.1006/
cviu.2002.0981
Rusinkiewucz, S. (2001). Efficient variants of
the ICP algorithm. Presented at 3rd International
Conference on 3-D Digital Imaging and Modeling,
Quebec City, Canada.
Salvi, J., Matabosch, C., & Fofi, D. (2007).
A review of recent range image registration
methods with accuracy evaluation. Image and
Vision Computing, 25, 578–596. doi:10.1016/j.
imavis.2006.05.012
Spletzer, J. R. (2003). Sensor fusion techniques for
cooperative localization in robot teams. Doctoral
dissertation, University of Pennsylvania.
Tarel, J., Civi, H., & Cooper, D. (1998). Pose
estimation of free-form 3D objects without point
matching using algebraic surface models. Pre-
sented at IEEE Workshop on Model-based 3D.
Thrun, S. (1993). Exploration and model building
in mobile robot domains. Paper presented at IEEE
International Conference on Neural Networks,
San Francisco, USA.
Thrun, S. (2003). Robotic mapping: A survey. In
Lakemeyer, G. (Ed.), Exploring artificial intel-
ligence in the new millennium (pp. 1–35). Morgan
Kaufmann Publishers Inc.
Thrun, S., Fox, D., Burgard, W., & Dellaert,
F. (2000). Robust Monte Carlo localization for
mobile robots. Artificial Intelligence, 128(1-2),
99–141. doi:10.1016/S0004-3702(01)00069-8
Welch, G., & Bishop, G. (1995). An introduction
to the Kalman filter. Technical Report TR 95-041.
University of North Carolina, Department of
Computer Science.
Whyte, H. D., & Bailey, T. (2006). simultaneous
localization and mapping (slam): part i- the es-
sential algorithms, IEEE Robotics and Automation
Magazine.
Yamauchi, B. (1997). A frontier based approach
for autonomous exploration. Presented at IEEE
International Conference on Autorotation and
Robotics, Albuquerque, NM.
Zelinsky, A. (1994). Planning paths of com-
plete coverage of an unstructured environ-
ment by a mobile robot. The International
Journal of Robotics Research, 13(4), 315.
doi:10.1177/027836499401300403
ADDITIONAL READING
Blais, F. (2004). Review of 20 Years of Range Sen-
sor Development. Journal of Electronic Imaging,
13, 231–243. doi:10.1117/1.1631921
Druon, S. (2007), Color Constrained ICP for Reg-
istration of Large Unconstructed 3D/Color Data
Sets, Presented at IEEE International Conference
on Robotics and Automation, Roma, Italy.
Gebre, B., Men, H., & Pochiraju, K. (2009),
Remotely Operated and Autonomous Map-
ping System(ROAMS), Paper presented at the
2nd Annual IEEE International Conference on
Technologies for Practical Robot Applications,
Woburn, MA.
Fernandez, J. C., & Singhania, A. (2007). An
Overview of Lidar Point Cloud Processing Soft-
ware, GEM Center Report. University of Florida.
Lorusso, A. (1995), A Comparison of Four Algo-
rithms for Estimating 3D Rigid Transformations,
Presented at British Machine Vision Conference.
Makadia, A., Iv, E. P., & Daniilidis, K. (2006),
Fully Automatic Registration of 3D Point Clouds,
Presented at the 2006 Computer Society Confer-
ence on Computer Vision and Pattern recognition,
New York, NY.
Men, H., & Pochiraju, K. (2010), Hue Assisted
Registration of 3D Point Clouds, presented at
ASME 2010 Interaional Design Engineering
Technical Conference, Montreal, Canada.
527

Montemerlo, M., Thrun, S., et al. (2003), Fast-
SLAM 2.0, An Improved Particle Filtering
Algorithm for Simultaneous Localization and
Mapping that Provably Converges, Presented at
the International Joint Conference on Artificial
Intelligence, Acapulco, Mexico.
Nüchter, A. (2005), 6D SLAM with Approximate
Data Association. Presented at IEEE Interna-
tional Conference on Robotics and Automation,
Barcelona, Spain.
Nuchter, A. (2007), Catched k-d tree Search for
ICP Algorithms, Presented at 6th International
Conference on 3-D Digital Imaging and Model-
ing, Montreal, Canada.
Rusinkiewucz, S. (2001), Efficient Variants of
the ICP Algorithm, Presented at 3rd International
Conference on 3-D Digital Imaging and Modeling,
Quebec City, Canada.
Salvi, J., Matabosch, C., & Fofi, D. (2007). A
Review of Recent Range Image Registration
Methods with Accuracy Evaluation. Image and
Vision Computing, 25, 578–596. doi:10.1016/j.
imavis.2006.05.012
This work was previously published in Depth Map and 3D Imaging Applications: Algorithms and Technologies, edited by Aamir
Saeed Malik, Tae Sun Choi and Humaira Nisar, pp. 56-86, copyright 2012 by Information Science Reference (an imprint of
IGI Global).
528
Algorithms for 3D Map Segment Registration
Thrun, S. (2003). Robotic Mapping: A Survey,
Exploring artificial intelligence in the new mil-
lennium (pp. 1–35). Morgan Kaufmann Publish-
ers Inc.
KEY TERMS AND DEFINITIONS
3D Map: 3D map represents certain level of
spatial information about the geometry features in
specific area. The map is dimensionally accurate
and may have a relative or absolute reference.
3D Mapping: 3D mapping is the process of
applying measurement devices to construct 3D
map about specified environment.
Color Point Cloud: Discrete points group
with both dimensiaonl accurate measurement
and texture property, normally generated by both
ranging device and color camera.
Map Registration: A process to accurately
stitch pair or multiple point clouds together into
single point cloud.
Point Cloud: Discrete points group with ac-
curate 3D coordinates describing object surface
dimensional measuremnts, usually contructed by
laser ranging devices.
 529

Chapter 32
TVAL-Farm:
A Qualitative Enhancement
of the LESA Model

Art Rex
Appalachian State University, USA

Leah Greden Mathews

University of North Carolina at Asheville, USA

Anne Lancaster
Independent Researcher, USA

ABSTRACT
The Total Value Assessment Tool for Farmland (TVAL-Farm) is a tool which incorporates scenic qual-
ity and cultural heritage elements to create an enhanced Land Evaluation and Site Assessment (LESA)
model. The enhancement of the LESA model provides insight and a framework on how to collect and
incorporate qualitative public values within the quantitative environment of a Geographic Information
Systems (GIS). Inclusion of these public values is essential for holistically valuing land parcels and us-
ing LESA to make land protection decisions.

INTRODUCTION has been used across the country as a decision

This research presents the Farmland Values Proj-
ect1’s TVAL-Farm tool which incorporates scenic
quality and cultural heritage elements to create an
enhanced Land Evaluation and Site Assessment
(LESA) model. The LESA model, developed
by the United States Department of Agriculture
Natural Resource Conservation Service (NRCS),
DOI: 10.4018/978-1-4666-2038-4.ch032
making resource particularly in the context of
farmland preservation programs. This project
develops a method for enhancing LESA’s Site
Assessment Factor 3 (SA-3), factors measuring
“Other Public Values.” The enhancement of the
LESA model provides insight and a framework on
how to collect and incorporate qualitative public
values data within the quantitative environment
of a Geographic Information System (GIS). This
information is not typically included in farmland

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

preservation benefit estimates and thus represents
a significant bridge between the qualitative and
quantitative studies of the importance of farmland
to rural areas.
LESA is a numeric rating system created by the
NRCS to evaluate a parcel’s relative agricultural
importance. The land evaluation (LE) component
of a LESA system measures soil quality (Pease
& Coughlin, 1996). It is usually based on land
capability classes, important farmland classes, soil
productivity ratings and/or soil potential ratings
(Pease & Coughlin, 1996). The site assessment
(SA) component evaluates other factors that
contribute to the site’s agricultural importance
such as parcel size and on-farm investments. SA
factors may also consider indicators of develop-
ment pressure such as distance to sewer and water
infrastructure and public amenities like wildlife
habitat or scenic views (Pease & Coughlin, 1996).
LESA systems assign points and a relative weight
to each of the LE and SA factors. The sum of the
weighted ratings is the LESA score; the higher
the LESA score, the more significant the site for
agriculture. States and localities often adapt the
federal LESA system to meet their needs. Once a
local system is approved by NRCS, it supersedes
the federal system: NRCS is required to use the
local version to review federal projects (Pease et
al., 1994, p. 62).
Our enhancements focused on two benefits,
scenic quality and cultural heritage significance,
derived from survey information gathered in a
community mapping activity conducted as part
of the Farmland Values Project. Following the
methodology of LESA, we assigned weights to
each benefit category in order to derive the holistic
benefit valuation score. Alternative weighting
measures may also be calculated based on public
input from various stakeholder groups. The pri-
mary product of this research is an enhanced geo-
spatial database and assessment tool that includes
qualitative layers for several types of farmland
benefits; we call it the Total Value Assessment
Tool for Farmland (TVAL-Farm).
530
TVAL-Farm
While LESA has the potential for being a
useful tool for farmland preservation, it has not
been used in Western North Carolina. The need
for linking LESA and GIS has been stressed
(Soil and Water Conservation Society, 2003).
Once site and benefit measures are tied to GIS
data, future land use scenarios can be applied to
identify which strategies maximize all the factors
determined best for farmland preservation. The
enhanced GIS-LESA model we have developed,
TVAL-Farm, is a tool to help identify farmland
most suitable for protection efforts.
THE STUDY AREA
The study area used to test our methodology
consists of four contiguous counties in the North
Carolina Department of Agriculture and Consumer
Service’s (NCDACS) Western Mountain District:
Buncombe, Henderson, Madison, and Haywood
(Figure 1). These counties provide an excellent
test region for our methodologies because parts of
the region are fairly urbanized while others have
great potential for growth. Buncombe is the most
populated county in our study area with 314 people
per square mile and 6,454 non-farm establish-
ments. Henderson is less populated but growing
at a faster rate with 238 people per square mile
and 2,302 non-farm establishments. Haywood’s
population density is 97.6 people per square mile;
the county has 1,411 non-farm establishments.
Madison is the least populated with 44 people
per square mile and 309 non-farm establishments
(U.S. Census Bureau, 2004). Madison County is
perceived to be under an urgent threat of urban-
ization since the recent completion of Interstate
26 through the county now makes it more acces-
sible to commuters and tourists, thus raising the
likelihood that property values will increase and
create additional stressors on farmland.
According to the NCDACS, Buncombe and
Henderson counties have the highest cash receipts
from farm goods in Western North Carolina (NC-
TVAL-Farm
Figure 1. The study area
DACS, 2004). In the most recent Census of Ag-
riculture, 2002-2007, Buncombe County led the
region in loss of farmland acreage with a 24%
loss. Henderson and Madison saw similar losses
(22% and 21%, respectively) while Haywood lost
13% of farmland acreage (USDA NASS, 2009).
The area is popular with tourists largely
because of its scenic quality (Brothers & Chen,
1997; Kask et al., 2002; Mathews, Stewart, &
Kask, 2003), and regularly attracts individuals
who choose to retire or purchase a second home
in the region because of its high quality of life.
USDA’s National Resources Inventory shows
a rapid decrease in farmland in Western North
Carolina over the past 20 years (USDA NRCS,
2000); if this data is projected into the future much
of the remaining farmland will be lost in the next
20 years. Land use change is expected to continue,
in part, because the region has relatively lax land
use regulation. An exception is the steep slope
ordinance that Haywood and Buncombe counties
have recently adopted to regulate development
on mountainsides. These regulations have the
consequence of shifting development pressure
to mid-slope and valley lands which are also the
most desirable lands for agriculture.
AN ENHANCED LESA MODEL
Figure 2 illustrates the TVAL-Farm model input,
the selected ModelBuilder (ESRI ™) tool, and the
modeled results. All data are raster format with a
spatial resolution of 90 x 90 feet. The traditional
531

Figure 2. TVAL-Farm, a qualitative enhancement of the LESA model
LESA has four input data sets: population, land
value per acre, agricultural soils potential, and
land use/land cover. The TVAL-Farm enhance-
ment adds two more--scenic value and cultural
value--bringing the total to six model input layers.
Model Input, Data Collection, and
Processing
Land Use/Land Cover
2001 Land Use/Land Cover data was collected
from the U.S. Geological Survey’s National Land
Cover Dataset (Homer et al., 2004). Fourteen land
cover categories identified by the National Land
Cover Dataset (NLCD) are found in our study
area; these were regrouped into five classes for
the TVAL-Farm model (Table 1). A value of 5
indicates the land cover type most likely under
pressure to be developed; in our model, this in-
532
TVAL-Farm
cludes cultivated crops and pasture, areas which
are often considered as prime development lands.
Lands that were already developed or classified
as water were valued at the lowest end of the
scale (1) as these areas would be near impossible
to develop further without large capital outlays.
Population
Population data was acquired from Landscan’s 1
km population data base available from the Oak
Ridge National Laboratory (ORNL, 2003). Data
was re-sampled and recalculated to reflect the
population per square mile. Class ranges were
then assigned as indicated in Table 1. Areas with
moderate population density bring both threats
(development including roads, infrastructure,
housing, and commercial development) and op-
portunities (proximity to markets, opportunities for
roadside stands, and walk-in or drive by business
TVAL-Farm
Table 1. TVAL-Farm model values
Model Land Use/ Population (people/
Value Land Cover square mile)
Cultivated Crops,
5 41-400
Pasture/Hay
Forest (Deciduous,
4 4-40
Evergreen and Mixed)
Shrub/Scrub, Grass-
3 <4
land
Barren Land, Wet-
2 401-600
lands
Open Water, Devel-
1 oped > 1600
(all levels)
that are not economically viable in more remote
locations). As a result, in our model the Wildland
Urban Interface (WUI)/Mixed classification
(population density of 41-400 people per square
mile) was allocated the most important value of
5, as this would indicate areas that are under the
most pressure to be lost to development based on
population density. Values of 4 and 3 are classified
rural and wildland, respectively. These areas are
also susceptible to development but are further
from the densely populated urban and suburban
classes and as such are less likely to be developed
than WUI/Mixed category.
Agricultural Soils Potential
Data for TVAL-Farm’s Agricultural Soils Potential
layer consisted of two parts: each county’s digital
soils data published by Natural Resources and Con-
servation Service (USDA NRCS, 2009) and slope
data. The model’s classification of agricultural
soils potential is based on the NRCS definition
of prime agricultural soils. Prime agricultural soil
is comprised of the best combination of physical,
chemical and organic characteristics that is avail-
able for agricultural uses. Prime agricultural soils
are not excessively erodible or saturated with
water for long periods and infrequently flood or
Agricultural Soils Potential Land Value ($/acre)
0-5% slope with deeper loamy soils often in
< $10,000
flood prone areas that may be waterlogged
5-12% slope with lowland soils that are
fairly deep and well drained, transitioning to $10,001-25,000
foothills
12-20% slope with well drained mid-slope
$25,001-75,000
soils of varying depth and quality
20-30% slope, well drained soil with varied
$75,001-250,000
depth but primarily shallow
> 30% slope with very shallow, very well
> $250,000
drained, and often acidic soils
are protected from flooding. Prime soils are deep
loamy soils with gentle slopes (less than 5%);
these are classified as excellent in the TVAL-
Farm model with a value of 5. The poorest soils,
assigned a model value of 1, are on the steepest
slopes (greater than 30%), are very shallow, are
very well drained and are often acidic.
Land Value per Acre
Land value per acre is based on the tax value of
land parcels in each county. Cadastral data from
each county was combined into a single dataset
and polygon features were converted to point
features with the full parcel attribution attached.
Large parcels were given multiple points with the
same attribution as to cover those parcels with an
even distribution of values (approximately one
point for every 100 acres). From this feature point
dataset, a raster surface was generated based on
an interpolation of land values using a nearest-
neighbor algorithm. The least expensive land (less
than 10,000 dollars per acre) was assigned the
highest value of 5; because it is most affordable,
the land is the most vulnerable to development
while also being the most cost-effective to retain
in agriculture. The most expensive land (greater
than 250,000 dollars per acre) was assigned the
533

lowest value of 1 because it is less cost-effective
for agricultural use and is less vulnerable to de-
velopment.
The Enhancement Layers: Scenic
Quality and Cultural Heritage
To gather data on scenic quality and cultural heri-
tage value, The Farmland Values Project conducted
a community mapping activity in 2008 with 33
residents of the four-county study region. Com-
munity members were recruited to participate via
the Farmland Values Project survey and flyers,
radio and print media invitations. Participants
were asked to identify and locate the places that
were important to them using Google Earth ™,
and assign a scenic and cultural value rating to
each place they identified. Participants identified
236 places of interest in the study region. Ratings
ranged from 1 through 5 based on each individual’s
perception of value for that specific location. These
values were converted to raster format using the
nearest-neighbor method, resulting in a “scenic
value surface” and a “cultural value surface,”
where each cell in the interpolated grid was as-
signed a value from 1 to 5.
Modeling and Output
Modeling and the Weighted
Overlay Technique
Weighted overlay is a technique for applying a
common scale of values to diverse and dissimilar
input to create an integrated analysis. The TVAL-
Farm model classifies 6 different raster layers on
a standardized 1-5 scale: land value in dollars,
agricultural soils potential, population per square
mile, land cover type, and scenic and cultural pref-
erences. Each input raster is weighted based on its
relative influence and layers are added together
to result in a model output that varies according
to the weighting scheme applied. While many
different weighting schemes can be applied, we
534
TVAL-Farm
present here two variations derived from survey
data for illustration purposes: the rank importance
and choice models.
Rank Importance Model
The Rank Importance Model weightings were
derived directly from responses to a question on
the Farmland Values Project survey. The question
(Farmland Values Project, n.d.) asked respondents
to rank a set of statements in order of importance;
the community preferences identified in this
question were used to determine the weighting
scheme for the Rank Importance Model. Partici-
pants identified local food as the most important
benefit provided by farmland followed by scenic
value and cultural value; these preferences led to
heavier weights for these layers in the model. Since
we didn’t have data on local food production, we
used Agricultural Soils Potential, an indicator of
soil productivity and an essential factor in food
production, as a proxy for local food produc-
tion. Because local food was ranked highest by
respondents, in this model the Agricultural Soils
Potential layer was weighted heaviest (37%). The
cultural and scenic layers were assigned lower but
equal weights (18% each). The remaining layers
(Land Value, Land Cover Type, Population) were
weighted the least in this model (9% each). The
model influence of each layer is summarized in
Table 2; results appear in Figure 3.
Table 2. Percent influence of input layers in rank
importance and choice models
Rank Impor- Choice
Input Layer
tance Model Model
Land Cover 9% 11%
Population 9% 11%
Agricultural Soils Potential 37% 23%
Land Value 9% 11%
Scenic Value 18% 22%
Cultural Value 18% 22%
TVAL-Farm
Figure 3. Rank importance results
Choice Model
Choice models are used to learn how consum-
ers make decisions and to estimate nonmarket
environmental benefits or costs. We used a
choice model in the Farmland Values Project to
determine respondent preferences for farmland in
western North Carolina. Specifically, we asked
respondents to choose between different scenarios
representing the following bundles of character-
istics: the likelihood of farmland conversion, the
scenic quality of the rural landscape, the presence
of indicators of agricultural heritage are present,
the extent to which the land contributes to local
food production, the amount of natural services
that are provided by farmland, and the financial
contribution required for that scenario (Farmland
Values Project, n.d.).
One of the strengths of the choice model is
that it enables implicit prices to be estimated for
attributes, which in our case allows us to weight
the importance or influence of the various raster
layers in TVAL-Farm Choice Model. The Choice
Model weightings were estimated using data from
the choice model responses to the Farmland Values
Project survey. The Choice Model results indi-
cated that scenic quality, cultural heritage, and the
value of access to local food were approximately
equally valued by respondents (Mathews, 2009).
Thus in the TVAL-Farm Model, these layers are
weighted equally (Table 2). Once again, we used
the Agricultural Soils Potential layer as a proxy for
535

Figure 4. Choice model results
local food production. To complete the modeling
scenario the remaining three layers (Land Value,
Land Cover Type, Population) were weighted
equally but at only half the value of the other
three. Table 2 summarizes each layer’s model
influence; results appear in Figure 4.
Comparison of the Rank Importance and
Choice Model results yield some overlap of lands
valued the same in each model. Figure 5 illustrates
all lands from both the Rank Importance and
Choice Model results with the highest model
value of 5. This indicates the clusters of areas
where land is considered the most valuable
within our modeled parameters.
536
TVAL-Farm
SUMMARY
The Farmland Values Project developed the
TVAL-Farm Model to identify and quantify the
values that rural residents place on farmland. Our
methodology asked local residents to define and
rank important attributes of farmland that are
important in their communities. Two of these at-
tributes, scenic quality and cultural heritage, are
characteristics that are not typically included in
land valuations. The TVAL-Farm Model results
reflect local community preferences in a way that
the traditional LESA does not.
A related innovative feature of the spatial
database is its potential to identify clusters of
tracts that are most appropriate for preservation.
TVAL-Farm
Figure 5. Rank importance and choice models of concentrated benefits
Research by Brabec and Smith (2002) suggests
that while the number of preserved agricultural
acres is important, whether the preserved land re-
mains in active farming is also dependent on other
factors including the adjacency and contiguity of
protected parcels. Being able to readily identify
currently preserved parcels in a spatial database
which also incorporates information about the
other benefits of the tract will be useful to policy
makers, land trusts, and individual farmers. The
high and rising cost of farmland protection efforts
increases the need for efficient targeting of scarce
preservation dollars. Our spatial database can be
used to identify the most valuable farmland tracts
to preserve in order to maximize the efficiency
of farmland preservation expenditures by govern-
ment and nonprofit groups.
Quantifying data that is ordinarily qualita-
tive, such as scenic quality and cultural heritage,
is challenging for community mapping activity
respondents. They aren’t acclimated to thinking
about scenic quality or cultural heritage quanti-
tatively. They may interpret qualitative attributes
differently from researchers, and from other
respondents, which is a limitation of the model.
Additionally, the limited sample size represents
a small population and may not accurately reflect
the views of the residents within the study region.
Increasing the sample size could alter the surfaces
used in this research and the model outcomes.
Regardless of sample size, the surfaces generated
from our community mapping activity participants
represent the preferences of this select group.
Accordingly, as sample size increases, individual
537

responses are often generalized or lessened in the
generated data surface.
It is important to incorporate qualitative data
sources in available quantitative tools like GIS
in order to better understand the areas we study;
TVAL-Farm provides a method for doing so. The
result is a GIS mapping tool that will help our
communities better understand the relationship
between people, values, threats, opportunities
and farmland in our region.
REFERENCES
Brabec, E., & Smith, C. (2002). Agricultural
land fragmentation: The spatial effects of three
land protection strategies in the eastern United
States. Landscape and Urban Planning, 58(2-4),
255–268. doi:10.1016/S0169-2046(01)00225-0
Brothers, G., & Chen, R. J. C. (1997). 1995-96
Economic impact of travel to the Blue Ridge
Parkway: Virginia and North Carolina. Asheville,
NC and Roanoke. VA: The Coalition for the Blue
Ridge Parkway and the National Park Service.
Environmental Systems Research Institute (ESRI).
(2009). Online documentation. Retrieved No-
vember 10, 2010, from http://webhelp.esri.com/
arcgisdesktop/9.2
Farmland Values Project. (n.d.). Final report. Re-
trieved from http://www2.unca.edu/farmlandval-
ues/pdfs/Farmland_Values_Project_Quick_Facts.
pdf
Farmland Values Project. (n.d.). Results. Retrieved
from http://www.unca.edu/farmlandvalues/re-
sults.htm
Homer, C., Huang, C., Yang, L., Wylie, B., &
Coan, M. (2004). Development of a 2001 National
Landcover Database for the United States. Pho-
togrammetric Engineering and Remote Sensing,
70(7), 829–840.
538
TVAL-Farm
Kask, S. B., Mathews, L. G., Stewart, S., & Ro-
tegard, L. (2002). Blue Ridge Parkway Scenic
Experience Project Final Report. Report submit-
ted in fulfillment of obligations for Cooperative
Agreement #CA5143990137, National Park
Service. Retrieved August 26, 2009, from http://
www.nps.gov/blri
Mathews, L. G., Stewart, S., & Kask, S. B. (2003).
Blue Ridge Parkway Scenic Experience Project
Phase 2 Final Report. Retrieved August 26, 2009,
from http://www.nps.gov/blri
North Carolina Department of Agriculture and
Consumer Services. (2004). North Carolina agri-
cultural statistics. Retrieved March 14, 2005, from
http://www.ncagr.com/stats/general/general1.htm
Oak Ridge National Laboratories (ORNL). (2003).
LandScan (2003)TM High Resolution Global Popu-
lation Data Set. Retrieved April 10, 2005, from
http://www.ornl.gov/sci/landscan/index.shtml
Pease, J. R., & Coughlin, R. E. (1996). Land
evaluation and site assessment: A guidebook for
rating agricultural lands, second edition. Ankeny,
IO: Soil and Water Conservation Society.
Pease, J. R., Coughlin, R. E., Steiner, F. R., Suss-
man, A. P., Papazian, L., Pressley, J. A., & Leach,
J. C. (1994). State and local LESA systems: Status
and evaluation. In F. R. Steiner, J. R. Pease, &
R. E. Coughlin (Eds.), A decade with LESA: The
evolution of land evaluation and site assessment
(pp. 56-75). Ankeny, IO: Soil and Water Conser-
vation Society.
Soil and Water Conservation Society. (2003, June
3-4). Enhancing LESA: Ideas for improving the
use and capabilities of the land evaluation and
site assessment system. Report from the Farmland
Information Center Workshop, Nebraska City, NE.
United States Census Bureau. (2004). State and
county quick facts. Retrieved March 14, 2005,
from http://quickfacts.census.gov/qfd/maps/
north_carolina_map.html
TVAL-Farm
United States Department of Agricultural Natural
Resources Conservation Service (NRCS). (2009).
Web soil survey. Retrieved November 10, 2010,
from http://websoilsurvey.nrcs.usda.gov
United States Department of Agriculture Na-
tional Agricultural Statistics Service. (2009). 2007
census of agriculture county profile. Retrieved
August 26, 2009, from http://www.agcensus.
usda.gov/Publications/2007/Online_Highlights/
County_Profiles/North_Carolina/index.asp
United States Department of Agriculture Natural
Resources Conservation Service. (2000). 1997
national resources inventory (Revised December
2000). Retrieved March 14, 2005, from http://
www.nrcs.usda.gov/technical/NRI
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 3,
edited by Donald Patrick Albert, pp. 58-69, copyright 2012 by IGI Publishing (an imprint of IGI Global).
ENDNOTES
1
The Farmland Values Project was designed to
inventory, analyze, and display the multiple
nonmarket values associated with farmland
in four counties in Western North Carolina.
The Farmland Values Project was supported
by the National Research Initiative of the
U.S. Department of Agriculture’s Coopera-
tive State Research, Education, and Exten-
sion Service, Grant # 2005-35618-15647.
This manuscript is adapted from the Farm-
land Values Project final report. Complete
information about the project, results, data
and maps are available at http://www.unca.
edu/farmlandvalues.
539
540

Chapter 33
OpenStreetMap
Kevin Curran
University of Ulster, UK

John Crumlish
University of Ulster, UK

Gavin Fisher
University of Ulster, UK

ABSTRACT
OpenStreetMap is a collaborative web-mapping project that collects geospatial data to create and distrib-
ute online maps, freely available to anyone with an Internet connection. Once accessed, OpenStreetMap
allows Internet users to contribute and edit geospatial data, effectively making it the mapping equivalent
of Wikipedia. OpenStreetMap is maintained by volunteer cartographers from around the world who
use GPS devices, portable cameras, and laptops for field mapping. Collected data are complemented
with digitised open source aerial photography and free maps from the governmental and commercial
sources. This report provides a summary of OpenStreetMap as a remarkable example of participatory
geographic information systems (GIS).

INTRODUCTION attention to the locations of particular interest to

OpenStreetMap (OSM) is a collaborative web
mapping project—an example of participatory
GIS—which makes free crowd-sourced geospa-
tial data available to anyone with access to the
Internet. OpenStreetMap was started in 2004 by
Steve Coast, a programmer from the United King-
dom residing in London. Using his laptop and a
global positioning system (GPS) receiver, Steve
Coast began mapping his neighborhood, paying
DOI: 10.4018/978-1-4666-2038-4.ch033
him, and including much more detail than any
available online map service was providing at that
time. He correctly envisioned that making such
data augmentable, editable and freely available
online, where anyone could access or contribute
information, might create a snowball effect, which
would eventually lead to the development of,
in his own words, a “jigsaw map of the world”
(Musgrove, 2010).
Since its inception, OpenStreetMap has at-
tracted countless Internet users throughout the
world who have contributed to the project in differ-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
OpenStreetMap
ent ways. At the time of the first State of the Map
(SoTM) conference, set up by OpenStreetMap
Foundation in July 2007 in Manchester, there were
only 9,000 registered users. Starting from 2007,
Yahoo! allowed OSM to tap into its vertical aerial
photography database, which greatly facilitated the
progress of the OSM project (in 2010 Microsoft’s
Bing Maps also released its aerial imagery for the
OSM). By August 2008, the number of registered
users rose to 50,000; by March 2009, the number
rose to 100,000 members; and by the beginning
of 2010 the project had 200,000 members. In
November 2011 the total number of volunteers
reached 500,000.
Among other milestones, one should mention:
(a) valuable functionality was added in January
2008 making OSM available for cyclists who
started downloading map data to use it on trips;
(b) funding of 2.4 million euro became avail-
able from CloudMade, a company founded by
Steve Coast, that uses OpenStreetMap data for
providing commercial geospacial services; (c)
in September 2009 Flickr announced that it be-
gan supporting OpenStreetMap, which allowed
geotagged photos of landmarks to be uploaded
to the OpenStreetMap server (Lardinois, 2009).
Currently, netizens around the world utilize and
update data available at the project’s website
OpenStreetMap.org. Citizen cartographers range
from volunteers providing humanitarian relief
work in disaster regions—where up-to-date maps
are essential and could be a matter of life and
death for the victims—to residents of cities, small
towns and villages; who simply want to map their
neighborhoods or highlight certain landmarks that
might be of interest to others. At the time of this
writing, OpenStreetMap community continues to
grow with contributors across the globe submit-
ting novel data or editing existing information.
Probably the most significant use of Open-
StreetMap has come in the aftermath of the massive
earthquake that struck Haiti on January 12, 2010
with the epicenter located 25 km to the west of the
country’s capital Port-au-Prince. Pre-earthquake
maps of Haiti contained little more than a few
highways and roads, with Port-au-Prince being
signified by a shaded outline. Haiti is one of the
poorest countries in the Western Hemisphere, and
as very few of the population could afford GPS,
there was no point for the commercial mapping
services in creating accurate digital maps. The
lack of an accessible online cartographic reference
had serious impeding implications for the relief
efforts once the catastrophic earthquake killed
hundreds of thousands of people, and made close
to one million homeless. International rescue and
aid workers who were trying to provide much
needed relief encountered many problems iden-
tifying what roads and routes were available for
the rescue and delivery of the necessary supplies.
According to the website CrisisCommons.org,
there were a number of stages in the develop-
ment of the inclusive Haiti maps to allow the aid
workers to get help to the victims. Firstly, there
was an immediate search for all available period
and contemporary maps. These included various
historic maps and even maps unclassified by
the CIA. Due to critical circumstances, the New
York Public Library, which has a specialized Map
Rectifier facility, provided volunteer cartographers
with NYPL Map Warper, a software application
that allowed for digitizing of the historic maps of
Haiti and aligning them with contemporary maps.
At this point, OpenStreetMap volunteers began to
get deeply involved with the project. Before the
earthquake of 2010, they had already collected
some mapping data of the disaster areas, which
were used in conjunction with historic maps.
Starting from this point, the OpenStreetMap
community began collecting and editing the field
data and mapping collapsed bridges and impass-
able roads, locations of hospitals and refugee tent
camps, as well as accurately marking the posi-
tion of damaged buildings where victims could
remained trapped. The next vital step involved the
United Nations releasing satellite imagery of the
region for public use, a move that was followed
by commercial satellite overhead imagery provid-
541

ers such as GeoEye and Digital Globe. All these
combined efforts helped to collect a vast base of
cartographic knowledge, which has permitted the
OpenStreetMap community to build what has
become the most comprehensive map of Haiti.
The Haiti mapping effort may be the most
significant example of how OpenStreetMap has
had a major social impact, but there are many
other examples of OSM performing valuable
social service for society. A large part of the suc-
cess of OpenStreetMap was due to its diverse
user base around the world. The OpenStreetMap
official blog, user diaries and OSM forum threads
contain references to a number of local or even
global mapping events such, as Night of the Living
Maps, a global OSM map editing marathon. The
OpenStreetMap Wiki also highlights similar and
smaller events. These events could be anything
from a casual get-together for novice users to
explore neighborhoods using on-the-ground GPS
surveys or walking papers, with collected data
to be later imported and processed in the OSM
editing environment; to mapping parties, where
cartography enthusiasts meet with the aim of
inclusive mapping of a particular area; or events
such as armchair mapping—editing cartographic
information collected in the field, working with
available sets of data or digitizing aerial photog-
raphy. While armchair mapping provides only a
virtual camaraderie, the mapping parties are made
up of a number of people who then usually split
into groups of two. Using GPS units, volunteers
typically start with plotting the locations of ob-
scure roads and footpaths, local landmarks, etc.
Coordinates can be recorded in handwritten form
by novice users, or saved on GPS receivers by
the more experienced. These coordinates are then
brought to a person experienced with the OSM
editing environment such as JOSM, who uploads
the information into existing OpenStreetMap
dataset. The main objective of these and similar
volunteer efforts is to chart exhaustively by adding
layer after layer of detail to a map of a designated
area, and to keep it continuously up-to-date.
542
OpenStreetMap
OPENSTREETMAP
PRODUCTION TECHNIQUES
The data collection for OpenStreetMap is increas-
ing rapidly. Basically, participatory digital map
production techniques stem from on-the-ground
GPS or paper surveys, intimate knowledge of a
certain geographical community plus available
aerial imagery. Volunteers who travel by foot,
bicycle, car or boat study the landscape using
handheld GPS, a paper log, or voice recorder, and
create the initial OSM data file. The project can
greatly benefit from access to aerial photography
datasets from government agencies or commercial
services, and volunteer cartographers in different
countries work on securing them. Technical teams
are set up to manage the conversion and import
the data when large datasets are available to use in
the OSM. In some cases, GPS would not be used
if satellite imagery had been previously uploaded
to the OSM server. After the upload of data a map
editor is used to edit the data.
Another important form of contribution to the
OpenStreetMap project is based on an editor’s own
local knowledge, which allows features such as
schools, libraries, various community places and
landmarks, roundabouts and busy roads bypasses
to be placed on the map. A number of enthusiasts
who are extremely committed to the project will
methodically map whole towns or rural districts
over a period of time by systematically organising
mapping parties to intensively map a particular
area in the evenings or over weekends. A score of
other volunteers correct errors and add features
by performing a large number of smaller edits.
This is done via OpenStreetMaps built-in and
stand-alone map editors.
Several governments, including the United
States, have released official cartographic data
to OpenStreetMap, including Landsat 7 Satellite
Imagery, Prototype Global Shorelines and Tiger
Data. Other countries have also made available
the use of detailed local aerial photography
through the Open Aerial Map initiative. Some
OpenStreetMap
commercial companies have also donated data to
the OpenStreetMap project on suitable licences.
For instance, the Automotive Navigation Data, a
well-known provider of the digital maps for GPS-
based navigation devices, donated a complete
road dataset for the Netherlands and partial road
datasets for China and India. The use of Yahoo!’s
aerial photography and Bing Maps is an example
of a partnership with commercial map providers.
There is a long list of software that can be used
in the creation of online editable maps such as
OSM. The main purposes are map rendering and
representation, map data collection, and manage-
ment. Map editing software used in the production
and presentation of OpenStreetMap came from
different sources, each with its own licensing. Map
rendering and presentation software comes in the
form of interactive online web-based software,
desktop software, and mobile phone applica-
tions. OpenLayers is an open source JavaScript
library for displaying map data in web browsers.
It provides an API for building rich web-based
geographic applications similar to Google Maps
and Bing Maps (Surrey councils open access,
2010). Slippy map interface is used to display
map tiles rendering by other software such as
Figure 1. Kosmos OpenStreetMap rendering platform
Map OS Matic, which can generate city maps
by using OSM data with the map split in squares
allowing to easily look for streets and an index of
the streets with references to the squares on the
map. Other examples of such software include
Mapnik, Osmarender and Cartagen.
Kosmos, which is currently being replaced
by Maperitive, is an example of a lightweight
OpenStreetMap rendering platform that provides
many tasks, including the ability to render Open-
StreetMaps interactively, print the maps, set up a
local tile map server, create separate map render-
ing rules or share rules stored in OpenStreetMap
Wiki pages, view maps on a Java enabled mobile
phone, and to perform tasks from the command
line (Figure 1).
Other map rendering and presentation software
applications created for use on mobile phones and
GPS devices include AndNav2 and Navit for
Android smartphones; OpenTouchMap, which is
a is a slippy map optimised for touch screen de-
vices such as iPhone; and finally, GpsMid which
is a Java ME application that is a free, fully of-
fline, vector-based map application for mobile
phones. It displays one’s current position on a
543

zoomable map and can be used to search for and
navigate to roads or points of interest.
A prime example of mapping and data col-
lection software that is most commonly used
for OpenStreetMaps is GPSBabel, which many
contributors use to convert GPS track data from
various proprietary formats to the GPX format
that OpenStreetMap requires. GPSBabel is a
cross platform, free software to transfer routes,
tracks and waypoint data between GPS units and
mapping programs. It can convert over a hundred
different types of GPS data formats. GPSBabel
grants users the ability to freely move their own
waypoint data between the programs and the
hardware they choose to use.
In OSM there is no option of categorising
what is uploaded (e.g., whether it is a road, or
a river) therefore, the data needs to be edited.
The volunteers perform this task with the use of
map editing software. The most commonly used
editing application for OpenStreetMap is JOSM,
the original Java OpenStreetMap desktop editing
software application (Figure 2). JOSM is a feature-
rich editor with an interface that can seem complex
at first. It also requires certain installation and
configuration efforts to get some plug-in features
when working particularly with Yahoo! imagery.
The user can utilize JOSM software to edit
place names, build locations, add photo imagery,
Figure 2. JOSM using Yahoo WMS plug-in
544
OpenStreetMap
and much more. It is an essential application for
users of OpenStreetMap (as shown in Figure 3).
Other map editing applications available to
users of OpenStreetMap include Potlatch, an
online flash editor; Merkaartor, a desktop editor
similar to JOSM; OSM2Go, a mobile/desktop
editor; and Quantum GIS, another desktop editor
which includes an OpenStreetMap plug-in. The
OpenStreetMap interface is easy to navigate and
it is visually appealing. It includes a built-in search
to find places. Overall, OpenStreetMap offers
many functions and features allowing users to
edit maps, review recent edits, export map data,
view and upload GPS data, and make comments
and suggestions.
INTERACTING WITH
OPENSTREETMAP
There are various methods to augment and edit the
data in OpenStreetMap. Each individual method
is geared toward the individual technical skills
of users, and also depends on how much time
and effort the user would like to contribute. A
range of online and standalone editing software
applications support the OSM project, with the
most frequently used including Potlatch, JOSM
and Merkaator.
OpenStreetMap
Figure 3. JOSM displaying OpenStreetMap data and GPS track logs
Potlatch
Potlatch 2 is the OSM default online editor, which
opens when an Internet user goes to the edit tab on
the OSM interface—utilizing any major browser
and requiring only a Flash plug-in. It allows the
user, for example, to work with the satellite imag-
ery provided by Yahoo! and Bing for the purposes
of creating a map of a neighborhood. By double-
clicking on the image, the user creates what is
called a node. A node can be used to highlight
any single point of interest on the map, such as
a building, and can be used in conjunction with
other nodes. To complete the map, the nodes are
connected using waypoints. In the latest version
of OpenStreetMap, many public amenities such
as telephone boxes, bus stops, etc, have their own
highlighting icons, which the user simply needs to
drag from the toolbar and drop onto the required
place on the map to mark the location. If a par-
ticular point of interest does not already have a
designated icon, the user can simply double click
the correct location on the image, creating a node.
The user can then manually add the description
of that particular point of interest.
Roads, tracks and pathways can be added to
OpenStreetMap in a very similar way to places.
The user simply has to identify the road, etc,
on the Yahoo! or Bing overhead image. Single
clicking at the beginning of a road will produce
a node. The difference from the “point of inter-
est” entry is that once the user moves the mouse
pointer, the movement is traced with waypoints.
Another single click will produce another node,
and the tracing will continue from this point, al-
lowing the user to negotiate the tracing of bends
in a motorway, road, track etc. Once the end of a
particular road has been reached, a double click
will enter a final node that will signify the end of
that particular entry. The user is then free to enter
the necessary details, such as what type of road
it is, motorway, connecting road, cycle path, or
footpath, and any names that the entry may have,
such as M1, Dunfield Terrace, etc.
Once all the information that the user wants
to include has been added, they are provided with
the option to save the changes that they have
made for editing later, or to save the changes
and upload them to the OpenStreetMap server.
After the changes have been uploaded to the
OpenStreetMap server, other members of the
OpenStreetMap community can then review and
edit them if necessary.
545

Java OpenStreetMap
Java OpenStreetMap (JOSM) is an individual
stand-alone software application that allows
OpenStreetMap users to render GPS data. This
information can then be uploaded onto the Open-
StreetMap server. The user can either download
existing data associated with the area on the map
that they would like to edit or augment, or they
can edit and upload their own raw GPS data to
JOSM. This would most likely be decided when
establishing exactly how much data already is
available for a particular area. For example, the
accessible data, such as satellite imagery and exist-
ing OpenStreetMap entries, for major cities and
their surrounding areas is usually quite detailed.
With this in mind, it may be helpful for the user
to reference the existing data when making their
own entries, as it will enable them to establish
how accurate their GPS readings are—if need
be—as well as establish whether the data they
have compiled about particular items (streets,
landmarks, points of interest, etc.) already exists
within the server. The user may find that the cur-
rent data contained within the server may need to
be updated. On the flip side, as some urban areas
in the United Kingdom are having quite detailed
cartographic information available, some others,
for example, Derry City in the Northern Ireland,
has very little. The satellite imagery available
for the Derry City is poor and the information
exhibited on the OpenStreetMap is quite minimal
for anywhere other than the city center. With this
in mind, users wanting to update data associated
with an undermapped urban area such as Derry
City may need to collect and upload the raw data
from their GPS to JOSM.
When using JOSM, the user will initially gather
as much GPS data as possible using a GPS unit.
This information is then imported into JOSM as a
GPX file. The coordinates of the GPX file can then
be read by JOSM, and this in turn allows the user
to retrieve any existing data on that particular area
from the OpenStreetMap server. The information
546
OpenStreetMap
that is added is actually a series of coordinates.
These coordinates can then be traced over by the
user, creating nodes. These nodes can then be used
to signify an individual object, such as a street
postbox, or as part of a series of nodes, or way-
points making up a road or a track, etc. Photos or
images such as a statue, street sign, building, etc.,
can also be uploaded, giving the user additional
information on a particular location. Various types
of information may be included about each of
the nodes and waypoints. For example, if a node
signifies a street postbox, information should be
entered to detail that it represents an amenity, and
that it is indeed a postbox. If a series of nodes and
waypoints signifies the location of a road, then the
user should enter details of the type of road it is,
e.g., motorway, residential, cycle path, footpath
etc., the name if applicable, and whether it is one
way or not. JOSM uses a uniform tagging system,
which encourages users to tag all features on a
map in a consistent way.
Merkaator
Merkaator, like JOSM, is a stand-alone map editing
software application that allows users to access,
add and edit the datasets stored on the OpenStreet-
Map server. Once the data has been retrieved,
Merkaators’ mapping functions are very similar
to that of JOSM, in that it deals with nodes which
are used individually to signify points of interest,
or as a series to create waypoints signifying roads
and paths etc., which can in turn be linked to Ya-
hoo! or Bing satellite imagery. It has an intuitive
user interface, providing a transparent display of
map features. However, Merkaator places more
emphasis on the aesthetics of mapping, since it
allows mapping more precisely curved roads. On
the other side, the OpenStreetMap Wiki warns that
this application, still being in the phase of devel-
opment, is “very memory hungry and can freeze
the machine while page-swapping” (“Comparison
of editors,” 2012).
OpenStreetMap
DISCUSSION AND CONCLUSION
Given that a lot of mapping information with a
high level of detail is now available online via
MapQuest, Google, Yahoo! or Bing maps, it is
worth asking the question, what benefit does
access to free, editable maps provide to regular
Internet users? It is important to recognize that the
OpenStreetMap project has had major influence
on both the attitudes of citizens and governments
towards the issue of cartographic data availabil-
ity and accessibility. For example, over the past
years, The Guardian has spearheaded a campaign
in the UK to make cartographic data collected by
government-funded organizations freely available
to the public. The argument is that, if taxpayers’
money paid for the data to be collected, why
should taxpayers have to pay any more to access
the data? This in turn has resulted in Ordnance
Survey, the UK national mapping agency, making
maps in several formats available online for free
downloads from April 2010 (Sung, 2010). The
Independent newspaper (OS Maps, 2010) wrote
that Sir Tim Berners-Lee, the founder of the World
Wide Web, advised Prime Minister Gordon Brown
on allowing publicly funded map data to be freely
available in an effort to boost entrepreneurship and
kick-start cutting-edge technology enterprises in
the United Kingdom. According to Dr. Christopher
Board, chairman of the Charles Close Society for
the Study of Ordnance Survey Maps, having free
Ordnance Survey maps at the user’s disposal is
very important. Board claims that if it was left
up to the providers to map what they felt was
important, then the “most popular tourist areas
would be mapped regularly and kept up to date,
but there would be huge areas of agricultural land,
moor or croft that would be left untouched” (OS
Maps, 2010). Ordnance Survey maps can only be
implemented as a point of reference for the user,
however if any drastic changes may occur within
the landscape of the area mapped, this effectively
renders the map out of date and it is up to the Ord-
nance Survey cartographers to provide updates. A
major advantage that OpenStreetMap would have
over any Ordnance Survey maps would be that it
has the potential to be updated in real time. On the
other side, under the direct influence of the OSM,
Ordnance Survey has started to provide more free
services, such as OS OpenSpace, which allows
embedding maps of Great Britain into web pages.
The practices of commercial online map and
image providers have also been influenced by the
OpenStreetMap project. As mentioned, first Yahoo!
and then Bing Maps permitted volunteer mappers
to use their proprietary aerial photography as a
backdrop. Yahoo! Flickr also allows tagging photo
images with OSM tags to be easily searched and
transferred on OSM maps if needed. At the 4th
annual international State of the Map 2010 confer-
ence, MapQuest, owned by AOL, announced that
it decided to become “the first major mapping site
to embrace and encourage open source mapping
at scale” (MapQuest Opens Up, 2010). As part of
this initiative, MapQuest founded an open source
project – open.mapquest.co.uk mapping site – based
on the OpenStreetMap data. Additionally, in 2008,
Google started its own OSM-type collaborative
project called Google Map Maker. According to
the Times newspaper (Blakely, 2008), this service
is aimed mainly at amateur cartographers based
in “hyper-growth countries” like India, where the
infrastructure has a tendency to change at an alarm-
ing rate in comparison to more developed countries.
Google Map Maker creator Lalitesh Katragadda felt
that Google’s “enormous online reach” would be
the deciding factor on its popularity in comparison
to similar non-commercial applications such as
OpenStreetMap. Four years later, Google Map
Maker is available in 188 countries and its com-
munity has reached 25,000 members. Google Map
Maker invites Internet users to correct and update
Google Maps and provides step-by-step instructions
and technical support. However, as appealing and
functional as Google’s mapping applications may
look, one should remember that they serve as a
means of generating corporate revenue—not for the
altruistic purposes of mapping the world. On the
547

other hand, the OpenStreetMap project allows for
uninterrupted programming access to its datasets
and provides a foundation for similar crowdsourced
initiatives such as Information Freeway, Open-
CycleMap, OpenBusMap, OpenRouteService,
OpenOrienteeringMap, and other projects.
The OpenStreetMap initiative is growing
and expanding with new volunteers joining the
effort to build and maintain a free-to-use, com-
prehensive map of the world. With continuing
advances in GIS technology and new approaches
to gathering and editing digital cartographic data,
the seemingly utopian project steadily moves
towards fruition as time passes. New features
for OpenStreetMap are being added, more areas
in developed countries are getting covered, and
more and more developing countries are getting
involved with the OpenStreetMap project. Citizen
cartographers are enthusiastically exploring un-
orthodox ways of using OSM, from highlighting
the location of a communal event, to keeping a
crisis map up-to-date during a humanitarian or
relief mission. Haiti’s catastrophic earthquake in
2010 has shown how powerful participatory and
activist cartography can be when properly chan-
neled and implemented. A major contribution to
the success of OpenStreetMap work during the
Haiti crisis is that OSM gave remote volunteer
cartographers and onsite relief workers the ability
to work together and update geographical data on
the OSM in real time. Despite the fact that much
of the work is still needed to complete the project,
OpenStreetMap already serves as a remarkable
example of public participation GIS, and one of
the largest collective mapping endeavours in the
history of human civilisation.
REFERENCES
Blakely, R. (2008, October 4). Google enlists pub-
lic to map fast-changing world. The Times Online.
Retrieved from http://technology.timesonline.
co.uk/tol/news/tech_and_web/article4878039.ece
548
OpenStreetMap
Comparison of editors. (2012, January 3). Open-
StreetMap Wiki. Retrieved from http://wiki.
openstreetmap.org/wiki/Comparison_of_editors
Google experiments with hotel prices on Google
maps. (2010, March 24). The Independent. Re-
trieved from: http://www.independent.co.uk/
travel/news-and-advice/google-experiments-
with-hotel-prices-on-google-maps-1927650.html
Haiti OpenStreetMap. (2010). CrisisCommons.
Retrieved from http://www.crisiscommons.org/
Haiti-Open-Street-Map
Lardinois, F. (2009, September 28). Flickr now
supports OpenStreetMap tags. ReadWriteWeb.
Retrieved from http://www.readwriteweb.com/ar-
chives/flickr_now_supports_openstreetmap_tags.
php
Mapping party. (2009, May 8). Weait.com. Re-
trieved from http://weait.com/content/mapping-
party
MapQuest Opens Up. (2010, July 9). Map-
quest. Retrieved from http://blog.mapquest.
com/2010/07/09/mapquest-opens-up/
Musgrove, M. (2010, January 31). Citizen car-
tographers’ map the microcosms of the world.
The Washington Post. Retrieved from http://
www.washingtonpost.com/wp-dyn/content/ar-
ticle/2010/01/30/AR2010013000033.html
OpenStreetMap. (2012). Retrieved from http://
www.openstreetmap.org
OpenStreetMap Wiki. (2012). Retrieved
from http://wiki.openstreetmap.org/w/index.
php?title=Main_Page&oldid=726612
OS maps: No.1 in the charts since 1747. (2010,
April 2). The Independent. Retrieved from http://
www.independent.co.uk/news/uk/this-britain/
os-maps-no1-in-the-charts-since-1747-1934059.
html
OpenStreetMap

Sung, D. (2010, April 1). Ordnance Survey fi-
nally gives up the goods. Pocket-lint. Retrieved
from http://www.pocket-lint.com/news/32400/
ordnance-survey-free-up-data
Surrey councils open access to aerial photography.
(2010, April 6). Guardian Professional. Retrieved
from http://www.guardian.co.uk/government-
computing-network/2010/apr/06/surrey-heath-
aerial-photography-openstreetmap-gis-06apr10

This work was previously published in the International Journal of Interactive Communication Systems and Technologies (IJI-
CST), Volume 2, Issue 1, edited by Tatyana Dumova, pp. 69-78, copyright 2012 by IGI Publishing (an imprint of IGI Global).

549
550

Chapter 34
GeoCache:
A Cache for GML Geographical Data

Lionel Savary
PRiSM Laboratory, France

Georges Gardarin
PRiSM Laboratory, France

Karine Zeitouni
PRiSM Laboratory, France

ABSTRACT
GML is a promising model for integrating geodata within data warehouses. The resulting databases are
generally large and require spatial operators to be handled. Depending on the size of the target geo-
graphical data and the number and complexity of operators in a query, the processing time may quickly
become prohibitive. To optimize spatial queries over GML encoded data, this paper introduces a novel
cache-based architecture. A new cache replacement policy is then proposed. It takes into account the
containment properties of geographical data and predicates, and allows evicting the most irrelevant
values from the cache. Experiences with the GeoCache prototype show the effectiveness of the proposed
architecture with the associated replacement policy, compared to existing works.

INTRODUCTION 1999) are based on XML, which becomes a stan-

The increasing accumulation of geographical data
and the heterogeneity of Geographical Informa-
tion Systems (GISs) make difficult efficient query
processing in distributed GIS. Novel architectures
(Zhang, 2001) (Gupta, 1999) (Leclercq, 1999)
(Chen, 2000) (Paolucci, 2001) (Corocoles, 2003)
(Boucelma, 2002) (Stoimenov, 2000) (Voisard,
DOI: 10.4018/978-1-4666-2038-4.ch034
dard for exchanging data between heterogeneous
sources. Proposed by OpenGIS (OpenGIS, 2003),
GML is an XML encoding for the modeling,
transport, and storage of geographical informa-
tion including both the spatial and non-spatial
fragments of geographical data (called features).
As stressed in (Savary, 2003), we believe that
GML is a promising model for geographical data
mediating and warehousing purpose.

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
GeoCache
By their nature, geographical data are large.
Thus GML documents are often of important size.
The processing time of geographical queries over
such documents in a data warehouse can become
too large for several reasons:
1. The query evaluator needs to parse entire
documents to find and extract query relevant
data.
2. Spatial operators are not cost effective, es-
pecially if the query contains complex selec-
tions and joins on large GML documents.
Moreover, computational costs of spatial op-
erators are generally more expensive than those of
standard relational operators. Thus, geographical
queries on GML documents raise the problem of
memory and CPU consumption. To solve this
problem, we propose to exploit the specificities
of a semantic cache (Dar, 1996) with an optimized
data structure. The proposed structure aims at
considerably reducing memory space by avoid-
ing storing redundant values. Furthermore, a new
cache replacement policy is proposed. It keeps in
cache the most relevant data for better efficiency.
Related works generally focus on spatial data
stored in object-relational databases (Beckmann,
1990). The proposed cache organizations are
better suitable for tuple-oriented data structures
(Brinkhoff, 2002). Most cache replacement poli-
cies are based on Least Recently Used (LRU) and
its variants. Other cache replacement policies
proposed in the literature (Lorenzetti, 1996) (Cao,
1997) (Arlitt, 1999) deal with relational or XML
databases, but have not yet investigated the area
of XML spatial databases.
The rest of the paper is organized as follows:
Section 2 gives an overview of related works. Sec-
tion 3 presents our cache architecture adapted for
GML geographical data. Section 4 discusses about
the inference rules of spatial operators and presents
an efficient replacement policy for geographi-
cal data considering inference between spatial
operators. Section 5 shows some results of the
proposed cache implementation and replacement
policy. Finally, the conclusion summarizes our
contributions and points out the main advantages
of the proposed GML cache-based architecture.
RELATED WORKS
Cache Replacement Policy
In the literature, several approaches have been
proposed for cache replacement policy. The most
well-known is the Least Recently Used (LRU,
Tannenbaum, 1992). This algorithm replaces the
document requested the least recently. Rather at
the opposite, the Least Frequently Used (LFU)
algorithm evicts the document accessed the least
frequently. A lot of extensions or variations have
been proposed in the context of WWW proxy
caching algorithms. We review some in the sequel.
The LRU-Threshold (Chou, 1985) is a simple
extension of LRU in which documents larger
than a given threshold size are never cached. The
LRU-K (O’Neil, 1993) considers the time of the
last K references to a page and uses such infor-
mation to make page-replacement decisions. The
page to be dropped is the one with a maximum
backward K-distance for all pages in the buffer.
The Log(size)+LRU (Abrams, 1995) evicts the
document with the largest log(size), and apply
LRU in case of equality. The Size algorithm evicts
the largest document. The Hybrid algorithm aims
at reducing the total latency time by computing
a function that estimates the value of keeping a
page in cache. This function takes into account
the time to connect with a server, the network
bandwidth, the use frequency of the cache result,
and the size of the document. The document with
the smallest function value is then evicted. The
Lowest Relative Value (LRV) algorithm includes
the cost and the size of a document in estimat-
ing the utility of keeping it in cache (Lorenzetti,
1996). LRV evicts the document with the lowest
utility value.
551

One of the most successful algorithms is the
Greedy Dual-Size (GD-size) introduced by Cao
and Irani (Cao, 1997). It takes into account the
cost and the size of a new object. When a new
object arrives, the algorithm increases the rank-
ing of the new object by the cost of the removed
object. In the same spirit, the Greedy Dual-Size
Frequency (GDSF) algorithm proposed by Arlitt
et al. (Arlitt, 1999) takes into account not only
the size and the cost, but also the frequency of
accesses to objects. As an enhancement of GDSF,
Yang et al. introduce the time factor (Yang, 2003).
Combined to the Taylor series, it allows predicting
the time of the next access to an object. Thus, it
provides a more accurate prediction on future ac-
cess trends when the access patterns vary greatly.
But the main bottleneck of this approach is the
time consumption to recalculate the priority of
each object.
Spatial Cache Replacement Policy
Most proposed spatial cache replacement policies
are based on variants of LRU and are developed in
the context of relational databases. In the area of
spatial database systems, the effect of other page-
replacement strategies has not been investigated
except in (Brinkhoff, 2002).
Considering a page managed by a spatial data-
base system, one can distinguish three categories
of pages (Brinkhoff, 1993): directory pages (de-
scriptors), data pages (classical information), and
object pages (storing the exact representation of
spatial objects). Using the type-based LRU (LRU-
T), first the object pages are evicted, followed by
the data pages, and finally by the directory pages.
Using primitive based LRU (LRU-P), pages are
removed from buffer according to their respective
priorities. If a tree-based spatial access method
is used, the highest priority is accorded from the
root to the index directory pages, followed by the
data pages, and finally the object pages. Thus, the
priority of a page depends on its height in the tree.
552
GeoCache
Let us recall that in GIS jargon, the MBR of
an object is the minimum-bounding rectangle
of this object. The area of a page of objects is
the minimum rectangle including all MBRs of
that page. The margin is the border of an area.
Beckmann (1990) and Brinkhoff (2002) define
five spatial pages-replacement algorithms based
on spatial criteria:
1. Maximizing the area of a page (A): a page
with a large area should stay in the buffer
as long as possible. This result from the ob-
servation that the larger is the area, the more
frequently the page should be requested.
2. Maximizing the area of the entries of a page
(EA): instead of the area of a page, the sum
of the area of its entries (spatial objects) is
maximized.
3. Maximizing the margin of a page (M): the
margin of a page p is defined as the margin
of the MBR containing all entries of p. The
larger a page margin is, the longer it will
stay in the buffer.
4. Maximizing the margin of the entries of a
page (EM): instead of the margin of a page p,
that of the composing MBRs are considered.
5. Maximizing the overlaps between the entries
of a page (EO): this algorithm tries to maxi-
mize the sum of the intersection areas of all
pairs of entries with overlapping MBRs.
As a synthesis, Brinkhoff (2002) proposes
a combination of LRU-based and spatial page-
replacement algorithms. To evict a document, a
set of victim candidates is determined using the
LRU strategy. Then, the page to be dropped out
of the buffer is selected from the candidate set
using a spatial page replacement algorithm. The
page dropped by this selection is placed in an
overflow buffer, where a victim is evicted using
the FIFO strategy. Depending on its spatial and
LRU criteria, a requested page found in the over-
flow buffer is moved to the standard part of the
buffer, influencing the size of the candidate set.
GeoCache
Buffer cache techniques are mainly used in spa-
tial database systems in order to optimize queries
response time. The work conducted by Brinkhoff
uses a spatial index for better management of the
buffer cache. However, there is no spatial index
for GML documents, as they are encoded in XML.
Hence, the spatial criteria mentioned above could
not be applied. Other criteria must be considered to
handle geographical queries. Moreover, semantic
cache gives better performances than page or tuple
replacement strategies (Dar, 1996), but until now
it has not been really studied for geographical
queries where data are stored in XML.
CACHE STRUCTURE FOR
GEOGRAPHICAL QUERIES IN GML
Generally spatial data consume a lot of memory
space. Hence, caching spatial objects has a ten-
dency to flood the available space in cache. For
example in a spatial join query, a spatial object A
can match with several objects B1, B2, etc. Thence,
a same object A can be replicated many times in
spatial query results. This may considerably reduce
the available space in cache, especially when a
large amount of spatial fragments must be stored.
To avoid spatial object replication in cache, we
propose a simple data structure which facilitates
object identification and isolation. This structure
is divided into two parts. The first is devoted to
the non-spatial elements of the geographical data.
The second one contains non-redundant spatial
fragments of geographical data. i.e., only distinct
spatial objects are stored in cache. In semantic
cache, the semantic region is divided into two
parts (Chidlovskii, 1999): the region descriptor
describing each query result stored in cache, and
the region content where the data are stored. In
the case of geographical queries, we introduce two
kinds of region content: the non-spatial region
content, and the spatial region content.
The spatial region content contains non-
redundant spatial data of geographical query
results, whereas the non-spatial region content
contains non-spatial data of geographical query
results. These region contents are associated with
the geographical region descriptor. It contains
information about each geographical query stored
in cache. The cache is then divided into two parts
(see Figure 1): (i) the non-spatial part of the cache
composed of the non-spatial region content and it
associated description contained in the geographi-
cal region descriptor (section 3.1); (ii) the spatial
part of the cache composed of the spatial region
contents and it associated description contained in
the geographical region descriptor (section 3.2).
Non-Spatial Part of the Cache
All data are encoded in XML. For simplicity and
standard enforcement, we encode XML data as
DOM trees. More compact structures are possible,
but it would not change the relative results and
would require specific developments. Thence,
query results are stored in cache as DOM trees
that compose the non-spatial region content (see
Figure 2). To separate non-spatial and spatial data,
a generated identifier replaces each GML sub-
tree (XML fragment with “gml:” as namespace).
The identifier value is used to uniquely retrieve
the corresponding spatial fragment stored in the
Figure 1. General cache structure
553

Figure 2. Cache organization
spatial part of the cache structure. For non-spatial
data, the geographical region descriptor contains
classical information about each query (region
formula, replacement values, and pointers)
(Chidlovskii, 1999).
Spatial Part of the Cache
The second part of the cache contains a specific
structure designed, from one hand, to store only
distinct spatial objects, and from the other hand,
to quickly retrieve spatial fragments. Each spatial
fragment is stored as a DOM tree, since spatial
data are in GML. The spatial fragments compose
the spatial region content. Each spatial fragment
is referenced in a hash table. The identifier is
used as a key of the hash table. It is composed of
a counter value and a reference to the root tree
of the spatial fragment (see Figure 3). Thus, each
DOM tree corresponding to a spatial fragment
can be quickly retrieved in the spatial part of the
cache. The counter value indicates the number of
references to the spatial fragment from the non-
spatial part of the cache. Thence, the entry can
be deleted when the counter is 0. The hash table
composes the geographical region descriptor for
spatial data. It contains: identifiers ID, counters
and references to spatial fragments (see Figure 3).
554
GeoCache
Figure 3. Spatial part of the cache
Cache Insertion Algorithm
When a geographical query result must be stored
in the cache, the spatial and non-spatial fragments
have to be identified and stored in the correspond-
ing part of the cache. The insertion algorithm is
described in Figure 4. A key issue is the genera-
tion of the GML fragment identifier. We simply
use the centroïde of the object defined as a point
standing for the geometry center. More sophisti-
cated identifiers could be used, e.g., a numerical
encoding of the shapes to avoid confusion between
geographical objects of same centroïde.
A new DOM tree is created (line 00) for the
non-spatial fragment. The XML query result is
read homogeneously fragment by fragment. If a
non-spatial fragment is read (line 02), then it is
placed in the non-spatial part of the cache (line
03). Else, if a spatial fragment is read (line 05),
an identifier is automatically generated (line 06).
If the identifier already exists in the hash table
(line 07), the counter is incremented (line 08).
Otherwise, a Dom tree for the spatial fragment is
generated (line 11) and inserted into the hash
table with the new identifier (line 12). The gener-
ated identifier is added in the non-spatial DOM
tree (line 14).
GeoCache

Figure 4. Cache insertion algorithm

in the tree (line 02). The counter value related to
the spatial fragment is decremented (line 03). If
this value is equal to 0 (line 04), the DOM tree
corresponding to the spatial fragment is removed
from the hash table (line 05). Once all identifiers
have been scanned, if there is no identifier with
associated counter strictly greater than 0 (line 11),
the non-spatial tree corresponding to the query
result is removed from the cache (line 12).

GEOGRAPHICAL QUERY
PROCESSING

Spatial Operators with

Inference Rules

In this section, we introduce the spatial operators

we use to compose queries. We further detail the

Figure 5. Cache deletion algorithm

Cache Deletion Algorithm

When a geographical query result has to be evicted

from the cache, the DOM tree corresponding to the
non-spatial part of the result is logically removed.
Physically, only the referenced spatial fragments
with counter value 0 (no reference) are removed.
The algorithm is sketched in Figure 5 which
depicts the cache deletion algorithm. A Boolean
value defined at line 00 indicates if all counters
of a query result have no reference. The DOM
tree corresponding to the non-spatial part of the
query result is retrieved using the root node (line
01). Then the tree is scanned to get the spatial
fragments corresponding to each identifier stored

555
 GeoCache

inference rules between geographical predicates, • Contains (Path1, Path2). Determines if the
which are useful for determining if a query result spatial object specified by the path Path1
is included in another one. This is important for ‘spatially contains’ the spatial object of
semantic cache management. path Path2.

Spatial Operators and Predicates Some spatial predicates are based on topologi-
cal relationships and can be deduced from other
For geographical queries, not only conventional spatial predicates. This means the results of spatial
comparison operators θ = {≤, <, =, ≠, >, ≥} must queries can be contained in results of other queries,
be taken into account, but also spatial operators. with different operators. If inference rules between
To query the database, we use XQuery extended spatial predicates can be established, then it could
with spatial operators as proposed in GML-QL be possible to determine whether the result of a
(Vatsavai, 2002). For illustration purpose, we geographical query is contained in another one.
select the following spatial operators (OpenGIS, This feature may allow using the cached results
2003): instead of accessing the original document, and
thus, optimizing the query.
• Overlaps (Path1, Path2). Determines if For example, consider the two GML docu-
two spatial objects specified by the paths ments: Feature1.xml in Figure 6, which contains
Path1 and Path2 are overlapping. the non-spatial elements T1, T2, T3 and a spatial
• Distance (Path1, Path2) θ d. Determines fragment of type linearGeometry (OpenGIS,
if the distance between two spatial objects 2003) standing for a polyline; and Feature2.xml
specified by the paths Paht1 and Path2 sat- in Figure 7, which contains the non-spatial frag-
isfies the relation θ d, where d is a given ments N1, N2, and a spatial fragment also of type
distance. linearGeometry.
• Touch (Path1, Path2). Determines if two
spatial objects specified by the paths Path1 Figure 6. The Feature1.xml data
and Path2 are touching.
• Intersects (Paht1, Path2). Determines if
two spatial objects specified by the paths
Path1 and Path2 are intersecting.
• Within (Path1, Path2). Determines if the
spatial object specified by Path1 is within
the spatial object of path Path2.
• Equals (Path1, Path2). Determines if the
spatial object specified by the path Path1
is ‘spatially equal’ to the spatial object of
path Path2.
• Disjoint (Path1, Path2). Determines if the
spatial object specified by the path Path1 is
‘spatially disjoint’ to the spatial object of
path Path2.
• Crosses (Path1, Path2). Determines if the
spatial object specified by the path Path1
crosses the spatial object of path Path2.

556
GeoCache
Figure 7. The Feature2.xml data
Let R1 and R2 be two GML-QL queries:
R1:
For $b in doc(“Feature1.xml”)/Fea-
ture1/Object
For $c in doc(“Feature2.xml”)/Fea-
ture2/Object
Where Distance ($b/gml:LineString/
gml:coordinates,$c/gml:LineString/
gml:coordinates) < 1200
Return
<Result> { $b/T1 } { $b/T3 }
{$b/gml:LineString/gml:coordinates}
{ $c/N2 }
{$c/gml:LineString/gml:coordinates}
</Result>
R2:
For $b in doc(“Feature1.xml”)/Fea-
ture1/Object
For $c in doc(“Feature2.xml”)/Fea-
ture2/Object
Where Intersect ($b/gml:LineString/
gml:coordinates, $c/gml:LineString/
gml:coordinates)
Return
<Result>
{ $b/T1 }
{$b/gml:LineString/gml:coordinates}
{ $c/N2 }
{$c/gml:LineString/gml:coordinates}
</Result>
R1 determines if two objects Feature1 and
Feature2 are within a distance closer than 1200. In
this case, the query returns the geometry of the two
objects and the specified values. R2 determines if
two objects Feature1 and Feature2 intersects. In
this case, the query returns their geometry with
the other specified elements.
Two objects are intersecting if their distance is
equals to 0. Hence, the result of R2 is contained
in the results of R1 as all the values returned by
R2 are included in the values returned by R1, and
1200 > 0. Thus the result of R2 can be extracted
from that of R1 without scanning the database,
provided that the result of R2 is kept in cache. As
semantic cache replacement policies are based
on the semantic of queries, the query-processing
algorithm shall determine if a geographical query
result is contained in cache. Thus, an originality
of semantic caching in geographical databases is
the ability to compare geographical predicates.
Spatial Predicate Inference Rules
Let us denote a ⇒ b the implication and a ⇔ b
the equivalence inference rule. Let O1 and O2 be
two geometric objects. A simple inference rule is,
for all real d positive or null: Intersect (O1, O2) ⇒
(Distance (O1, O2) < d).
Using the nine operators defined in section
3.1, more inference rules are given in Figure
8. Notice that, for a couple of objects (O1, O2)
verifying a predicate P at level i, if the predicate
P implies the predicate Q at level i+1 and if the
predicate Q implies the predicate S at level i+2,
then the operator P at level i, infers the operator S
at level i+2. That is the implication of predicates
557

Figure 8. Spatial predicates inference rules
is transitive. It is important to take into account
all rules for better cache management.
In our proposed architecture (Savary, 2003),
spatial queries are computed using JTS, is a spa-
tial Java API proposed by the OGC (Open Geo-
spatial Consortium) and based on the OpenGIS
Simple Feature Specification (OpenGIS, 1999).
It implements the spatial operators introduced in
section 3.1. Hence, in this practical context, the
use of spatial predicate inference rules makes it
possible to determine if a new query is contained
or partially contained in cache.
Query Processing
Suppose an input query Q and a set C = {C1,
C2…, CK} of K cache queries. Then, according
to (Lee, 1999) there are five general cases of
query matching: exact match, containing match,
contained match, overlapping match and disjoint
match. Figure 9 illustrates the various cases. The
grey box in shop outline stands for the queries
Figure 9. Query match types
558
GeoCache
results C stored in cache. The white box with full
outline represents the new query Q. For an exact
match, the query Q already exists in cache: ∃ Ci
⊆ C \ Ci = Q. For a containing match, the query
Q is totally contained in the cache: ∃ D ⊆ C \ (Q
⊂ D ∧ D ⊄ Q). For a contained match, the query
Q contains some cache queries: ∃ D ⊆ C \ (Q
⊄ D ∧ D ⊂ Q). For an overlapping match, the
query Q is partially contained in cache: ∃ D ⊆
C \ (Q ⊄ D ∧ D ⊄ Q ∧ Q ∩ D ≠ ∅). Finally, for
a disjoint match, the query Q is not contained in
cache: Q ∩ C =∅.
For exact and containing match, Q can be
totally answered from the cache; but for contained
and overlapping match, an access to the database
is required in order to answer the remainder
query (Dar, 96). For geographical queries and
especially for join queries, the processing time
can be prohibitive. To palliate this disadvantage,
we propose to exploit the geographical region
descriptor.
GeoCache
General Query Processing Scheme
The geographical query processing is depicted
in Figure 10. When a new query Q must be pro-
cessed, the Query Analyzer determines if the query
can be answered from the cache, the database or
both. Then the Query Rewriter rewrites Q into
the remainder and probe queries. We introduce
two kinds of probe query:
• The elementary probe query allows re-
trieving totally or partially the data stored
in cache to answer the query Q.
• The optimized probe query aims at retriev-
ing the temporary spatial data for a further
Figure 10. Geographical query processing
optimization. This will be detailed in sec-
tion 4.2.2 bellow.
The elementary probe query is performed by
the cache Query Executor, with the non-spatial
and spatial data stored in cache. In the cases of
contained match, overlapping match, and disjoint
match, if an optimization is possible using the spa-
tial data stored in cache (as shown in the example
of section 4.2.2), then the required temporary
spatial data are determined using the optimized
probe query and forwarded to the Spatial Query
Executor. The temporary spatial data is composed
of n documents corresponding to the n features
implied in the remainder query. Each document
559

only contains the geometries of a specific feature.
The remainder query is split by the Query De-
composer into the non-spatial remainder query,
and the spatial remainder query. The non-spatial
remainder query is free of spatial predicates and
allows interrogating the database. The database
temporary results (non-spatial remainder query
results) are then forwarded to the Spatial Query
Executor. These database temporary results are
composed of n documents corresponding to the
n features implied in the remainder query. The
Spatial Query Executor performs the remainder
spatial query which contains the spatial predicates.
If the temporary spatial data is not empty, an
optimization is possible which avoids computing
spatial predicates. The result (remainder query
result) is then sent to the Query Combiner, which
merges the remainder and elementary probe
query results into a single GML document R.
The query result R of Q is then sent to the cache
Replacement Manager. The Cache Replacement
Manager determines then if the result has to be
placed in cache or not, and which cache queries
must be evicted.
Cache-Based Query Processing
and Optimizations
In the cases of contained match, overlapping
match, and disjoint match, data which are not
contained in cache are retrieved from the database.
To reduce the I/O and CPU costs, we propose to
exploit the spatial part of the cache. The idea is
to determine the set of spatial geometries implied
in the remainder query. If this set is known, then
it is possible to directly compare the geometries
contained in this set with the geometries contained
in the documents of the database temporary re-
sults. This set of spatial geometries results from
the optimized probe query. The optimized probe
query results (the temporary spatial data) contain n
documents corresponding to the n features implied
in the remainder query. Only spatial geometries
are stored in these documents. This allows to
560
GeoCache
avoid the process of spatial predicates for the
remainder query result. Only comparisons with
geometry features of database temporary result
and geometry features of their corresponding
temporary spatial data are required. Moreover,
it makes a better cache management, since the
size of those temporary spatial data documents
is reduced. Thus, performances are improved
especially for join queries, where multiple scans
are often necessary. For example, if the query Q
is the following:
For $b in doc(“Feature1.xml”)/Fea-
ture1/Object
For $c in doc(“Feature2.xml”)/Fea-
ture2/Object
Where Intersect ($b/gml:LineString/
gml:coordinates, $c/gml:LineString/
gml:coordinates)
Return <Result> { $b/a } { $c/b } {
$b/c } </Result>
Let Ci be the closest cache query of Q contained
in cache, i.e., the query containing all the required
entities to answer Q with a minimal number of
attributes not contained in Q (not mentioned in the
returned result of Q). Ci is found using the infer-
ence rules which have been previously presented
(figure 6). For instance, Ci is found to be the query:
For $b in doc(“Feature1.xml”)/Fea-
ture1/Object
For $c in doc(“Feature2.xml”)/Fea-
ture2/Object
Where Distance ($b/gml:LineString/
gml:coordinates, $c/gml:LineString/
gml:coordinates) < 9
Return
<Result> { $b/e } { $c/f } </Result>
Here, the optimized probe query contains the
intersect predicate. It will be applied to the cache
query result Ci. The resulting temporary spatial
data will be composed of two documents contain-
GeoCache

ing only geometries of Feature1 and geometries
of Feature2 that match. The geometries of these
documents will be compared, using their cen-
troïdes, with geometries of those of the database
temporary results. The required data a, b and c
of respectively Feature1 and Feature2 are then
retrieved to answer the query Q.
Cache Replacement Policy: B&BGDSF
When a new query result must be stored in a
saturated cache, the most irrelevant queries have
to be evicted. It is important to take into account
the constraints on size but also the cost and fre-
quency of access to spatial objects. Thus, we base
our replacement policy on the Greedy Dual-Size
Frequency (GDSF) algorithm proposed by Arlitt
et al. (Arlitt, 99). This strategy replaces the object
with the smallest key value for a certain cost func-
tion. When an object i is requested, it is given a
priority key Ki computed as follows:
Ci
K i = Fi * +L.
Si
where: Fi is the frequency usage of the object I;
Ci is the cost associated with bringing the object i
into cache.; Si is the size of the object I; and, L is
a running age factor that starts at 0 and is updated
cache and the minimal cost out of cache if it must
be recomputed, as formalized by equation (i).
Secondly, the total size of the set must be equal
or greater than the size of the new query result,
as formalized by equation (ii) bellow.
n
(i )MIN (∑ X i * CCacheCPU + (1 − X i ) * C DiskCPU )
i =1
n
(ii )∑ X i * S i ≥ T
i =1
In equation (i), Xi is set to 1 if object i is kept
in cache, and 0 otherwise, CCacheCPU stands for the
cost in CPU of the cached query result and CDiskCPU
stands for the cost in CPU to re-compute the query
result I from the database. The second constraint
concerned the size of the set of cached queries to
remove. It is given by equation (ii), where T stands
for the size of the new query result, and Si, the
size of a cached query result i. This is equivalent
to the following expression, which can be resolved
using the Branch and Bound algorithm:Where
n
(i )MAX (∑Yi * (CCacheCPU − C DiskCPU )).
i =1
Yi = 1 − X i
n n
(ii )∑Yi * S i ≤ ∑ Si − T
i =1 i =1

for each replaced object o to the priority key of

this object in the priority queue, i.e., L = Ko. PROOF:
Our cache replacement policy performs in two By replacing Xi by (1-Yi) in equation (i), we
steps:. First, the cost of each query result contained obtain:
in cache is computed using the GDSF policy. Ci
n
stands for the cost of a geographical query. The (i ) ⇔ MIN (∑ (CCacheCPU −Yi * (CCacheCPU − C DiskCPU ))
computation of Ci for the experimentations is dis- i =1

cussed in the experimentation section. Secondly,

The Branch and Bound algorithm (Pisinger, 04) is n
used to determine the most irrelevant query result Since ∑C CacheCPU
is a constant, the problem
in cache. In the case of spatial cache replacement i =1

policy as in the general case, the problem is to is reduced to the second expression:
determine the optimal set of candidates to be
removed. This set must have the minimal cost in

561
 GeoCache

n the performances of this algorithm compared to

(i ) ⇔ MIN (−∑Yi * (CCacheCPU − C DiskCPU ))
i =1
existing ones.
n
⇔ MAX (∑Yi * (CCacheCPU − C DiskCPU ))
n
i =1 EXPERIMENTATIONS
(ii ) ⇔ ∑ (1 −Y ) * S i i
≥T
i =1
n n The Geographical Data
⇔ ∑Y i
* Si ≤ ∑ Si − T Warehouse Architecture
i =1 i =1

This formalization is generally used to resolve
the knapsack problem, in operational research
(Martello, 90). To solve this operational research
problem, we choose the Branch and Bound algo-
rithm. The Branch and Bound algorithm take as
parameter: the GDSF cost of each cache query
result, the size of each query result, and the size
of the new query result.
Compared to GDSF, the proposed B&BGDSF
algorithm determines the optimal set of candidates
to be replaced by the new query result. Thus,
better cache management is done. The results
depicted in the experimentations below, show
Figure 11. Geographical data warehouse archi-
tecture
The cache has been implemented in a geographi-
cal data warehouse prototype. The architecture of
this prototype is depicted in Figure 11.
The geographical data warehouse is entirely
built using open sources components based on
Java and XML. The data extract from heteroge-
neous sources are encoded in SOAP (Simple
Object Access Protocol) - a standard protocol
defined by the W3C - messages and sent to the
DSA (Data Staging Area) to be processed. In the
DSA, the data are integrated using spatial ETL
process before it feeds the database of the data
warehouse. The XML database eXist has been
chosen as the data warehouse manager, since it
is an open source and allows querying the non-
spatial data using the XQuery language. The
geographical queries are computed using the JTS
(Java Topology Suite) packages which is OpenGIS
compliant. Users queries are sent to the data
warehouse embedded in a SOAP message.
The implemented architecture is composed of
four levels: sources, the DSA, the data warehouse
and the clients. The communication between these
four levels is based on SOAP protocol using the
apache tomcat Web application. At the source
level, Hibernate 2.0 is used for extracting data
from heterogeneous database. Hibernate has
been chosen since it allows interrogating any
database with the singular HQL language. The
data extracted from these sources are sent into a
SOAP message to the DSA in order to be cleaned
and integrated by the spatial ETL transformation
process. Once the data have been prepared, they
are sent to the data warehouse in SOAP messages
to feed the XML eXist database. By the same

562
GeoCache

Figure 12. Cache hit ratio

way, the user queries are sent into a SOAP mes-
sage to the data warehouse. The query analyzer
then extracts and processes the queries (section
4.2.1). The user level is composed of a graphi-
cal interface which has been implemented using
SWING Java technology. This interface allows
users to specify a request and display the queries
results. The queries results are encoded in SVG
(Scalable Vector Graphic) – the XML format for
vector graphic specified by the W3C - embedded
in SOAP messages and sent to the users to be
graphically displayed.

The Experimentation
Parameters and Results Figure 13. Cache coverage ratio

The experimentations were done on a Pentium IV

2.0 GHz with 512 MB of memory. The used dataset
is composed of 10000 and 1000 features stored in
two GML documents in the eXist (Meier, 2002)
XML database. They describe the road networks
of Lille (a French town) (Savary, 2005). Their
geometry is of type linearGeometry (OpenGIS,
2003). They have been generated by our ETL
process from the source, one was in format shape
(the exchange format of ArcGIS software), and
the other was stored in the Oracle 9i DBMS which
allows spatial data management.
Our cache replacement policy has been com-
pared with GDSF (without Branch and Bound
Figure 14. Cache hit ratio
selection), LFU and LRU, which is the most
popular in spatial database cache replacement
policy. The cache size has been moving from 10
to 100 Mega Bytes. To compare those policies,
we have simulated a flow of approximately 70
queries (point, range and join queries), possibly
made by different users. We have performed two
series of experimentations, consisting in two sets
of about 70 GML XQueries. The one (random.
gxq file) uses an arbitrary order of queries, while
in the second (freq.gxq) the most recent queries
are repeated. These two files contain different
queries. The evaluation of the first query set is
reported in Figures 12, 13 and 16. Whereas the

563
 GeoCache

Figure 15. Cache coverage ratio

results obtained using freq.gxq file are illustrated
in Figures 14, 15 and 17.
Figure 12 shows the hit ratio performances
realized with a random dataset queries. Here we
can see that in general, GDSF and B&BGDSF,
outperform LRU and LFU. When the cache size
becomes larger, GDSF roughly scale like B&BGDSF.
But with smaller cache size, our B&BGDSF algo-
rithm performs better than the other cache replace-
ment policies. This is because the Branch and
Bound algorithm associated to GDSF only evicts
the most irrelevant cache results with smallest
cost and which size sum is greater but closest to
the new query which must be put in cache. Thus,
the cache is better managed. For example with a
cache size equal to 20 Mega Bytes, B&BGDSF
Figure 16. Performance gain
performs about 8% better than GDSF.
Figure 13 shows the cache coverage ratio in
percentage. Cache coverage hit shows the effect
of partial matching in semantic caching. This in-
formation is not readable in traditional cache hit
ratio. It is defined as follows (Lee, 1998):
Given a query set consisting of I queries q1,
…qI, let Ai be the number of answers found in
the cache for the query qi, and let Bi be the total
number of answers for the query qi for 1≤ i ≤I.

∑
I
QCRi
CCR = i =1

I

Figure 17. Performance gain where

Ai
QCRi =
Bi

if Bi > 0.
Here again, we can notice that in general
GDSF and B&BGDSF outperform LRU and LFU.
But it is interesting to see that compared to Fig-
ure 12, when the number of hit ratio is equal for
GDSF and B&BGDSF with a cache size of 50 MB
or 60MB, the number of cache coverage ratio is
better using the B&BGDSF algorithm. Figure 16

564
GeoCache
shows the performance gain which is defined as
follows (Brinkhoff, 2002):
disk _ acces _ for _ a _ given _ Policy
gain = −1
disk _ access _ of _ B & BGDSF
where Policy stands for one of the three cache
replacement policies: GDSF, LRU and LFU.
The performance gain shows the reduction
in the number of disk access using the proposed
cache replacement policy, compared to others. In
general, we can notice that the gain is positive us-
ing our proposed algorithm. However, depending
on the cache size, the gain with GDSF is more
or less important, especially when the cache size
becomes small or large. For example in Figure
12 with a cache size equal to 20 or 100 MB,
the performances of GDSF are the same or less
than B&BGDSF. However, if we look at the gain
distribution, the values obtained are about 8 or
18%. Especially with a cache size of 100 MB,
the hit ratio for GDSF and B&BGDSF are roughly
the same, but the gain obtained is different about
18%. These results show that the cache is bet-
ter managed using B&BGDSF. For a cache size
equal to 30 Mega Bytes, we can see that the gain
compared to GDSF is roughly equal to 23%. But
for a cache size between 40 and 90 MB, the gain
compared to GDSF is roughly the same or less
than our proposed algorithm. But the Hit Ratio
and Cache Coverage Ratio show that better cache
management is obtained using the B&BGDSF algo-
rithm. Moreover, for a cache size of 40 or 60 MB
in Figure 12, LRU performs like B&BGDSF; but in
Figure 16 the gain obtained compared to LRU is
the same or better (for example with a cache size
of 40 MB). For a cache size of 50 MB peaks of
gain around 37% with LRU and 25% with LFU
are obtained.
Figures 14, 15 and 17 show other experimen-
tations performed with another query set where
the most recent queries are repeated. In this case,
we notice that LRU outperforms GDSF and LFU,
but B&BGDSF outperforms GDSF, LFU and LRU,
especially when the cache size is smaller (from
10 to 50 MB). This is due to a better selection of
cache queries to evict.
These results show that the cache is better
managed using the B&BGDSF compared to GDSF,
LFU and LRU cache replacement policies. The
cache hit ratio (Figures 12 and 14) show that
B&BGDSF outperforms GDSF, LRU and LFU; but
when the performances of B&BGDSF are roughly
equal to another cache replacement algorithm, the
cache coverage ratio (Figures 13 and 15) show
that a better cache management is obtained us-
ing the proposed algorithm. This is confirmed by
the performance gain in disk access presented in
Figures 16 and 17.
CONCLUSION
In order to optimize the I/O-performance of a da-
tabase system, many techniques have been consid-
ered. One of the most used in spatial database is the
buffering technique associate to the LRU strategy
and its variants. Related works generally focus on
spatial data stored in object-relational databases.
With the popularity of Internet and XML, the OGC
proposed the GML semi-structured language,
which is an extension of XML for the modeling,
transport and storage of geographical information
including both spatial and non-spatial fragments
of geographical data. The GML language is more
and more used in many distributed architectures.
In order to handle efficiently geographical queries
stored in GML documents, we have proposed an
appropriate cache structure and cache replace-
ment policy. The proposed cache structure aims
at considerably reduce the memory used to store
geographical queries. Only non redundant spatial
fragments of a geographical query are stored. Thus,
much more queries can be stored in cache. This
architecture is associated to a semantic cache where
inference rules between spatial predicates are ex-
ploited for a better cache management. Associated
565

with this semantic cache, a new cache replacement
policy is proposed. It takes into account the cost
implied by the different predicates and gives bet-
ter performances than existing cache replacement
policies like GDSF, LRU and LFU. The proposed
cache replacement policy is based on Branch and
Bound and GDSF algorithms. It only evicts the
most irrelevant cache results according to the cost
and size of data stored in cache. The hit ratio, cache
coverage ratio and performance gains show that the
proposed B&BGDSF replacement policy outperforms
GDSF, LRU and LFU algorithms. In perspective,
this work will be extended to spatiotemporal data
stored in semi-structured documents in order to
optimize spatiotemporal queries.
REFERENCES
Aboulnaga, A., & Naughton, J. (2000). Accurate
estimation of the cost of spatial selections. IEEE
International Conference on Data Engineering
(ICDE ’00) (pp. 123-134), San Diego, CA.
Abrams, M., Standbridge, C. R., Adbulla, G.,
Williams, S., & Fox, E. A. (1995). Caching prox-
ies: Limitations and potentials. WWW-4, Boston
Conference.
Arlitt, M., Friedrich, R., Cherkasova, L., Dilley, J.,
& Jin, T. (1999). Evaluating content management
techniques for Web proxy caches. In Proceedings
of the Workshop on Internet Server Performance
(WISP), Atlanta, GA.
Beckmann, N., Kriegel, H. P., Schneider, R., &
Seeger, B. (1990). An efficient and robust access
method for points and rectangles. In Proceeding
ACM SIGMOD International Conference on Man-
agement of Data (pp. 322-331), Atlantic City, NJ.
Boucelma, O., Messid, M., & Lacroix, Z. (2002).
A WFS-based mediation system for GIS interoper-
ability. The 10th ACM International Symposium on
Advances in Geographical Information Systems
(ACM GIS), McLean, VA.
566
GeoCache
Brinkhoff, T. (2002). A robust and self-tuning
page-replacement strategy for spatial database
systems. The 8th International Conference on
Extending Database Technology (EDBT 2002)
(pp. 533-552), Prague, Czech Republic, 2002.
LNCS, Vol. 2287, Springer-Verlag.
Brinkhoff, T., Horn, H., Kriegel, H. P., & Sch-
neider, R. (1993). A storage and access architecture
for efficient query processing in spatial databases.
In Proceedings of the 3rd International Symposium
on Large Spatial Databases (pp. 357-376), Sin-
gapore. LNCS, Vol. 692, Springer.
Cao, P., & Irani, S. (1997). Cost-aware WWW
proxy caching algorithms. Proceedings of USENIX
Symposium on Internet Technologies and Systems
(USITS) (pp. 193-206), Monterey, CA, December.
Chen, L., Wang, S., & Rundensteiner, E. A. (2004).
Replacement strategies for XQuery caching sys-
tems. Data & Knowledge Engineering, 49(2),
145–175. doi:10.1016/j.datak.2003.10.002
Chidlovskii, B., Roncancio, C., & Schneider, M.
L. (1999). Semantic CacheMechanism for hetero-
geneous Web querying. In Proceedings of the 8th
World-Wide WebConference (WWW8).
Chou, H., & DeWitt, D. (1985). An evaluation
of buffer management strategies for relational
database systems.In Proceedings of the 11th VLDB
Conference.
Corocoles, J. E., & Gonzalez, P. (2003). Querying
spatial resources. An approach to the semantic
geospatial Web. CAISE 2003 Workshop, Web,
e-Business, and the Semantic Web (WES: Foun-
dations, Models, Architecture, Engineering and
Applications, LNCS, Springer-Verlag. Dar, S.,
Franklin, M. J., Jonsson, B. T., Srivastava, D., &
Tan, M. (1996). Semantic data caching and replace-
ment. Proceedings of the 22nd VLBD Conference,
Bombay, India.
GeoCache
Gupta, A., Marciano, R., Zaslavsky, I., & Baru,
C. (1999). Integrating GIS and imagery through
XML-based information mediation. In P. Agouris
& A. Stefanidis (Eds.), Integrated spatial data-
bases: Digital images and GIS, LNCS, Vol. 1737
(available from http://www.npaci.edu/DICE/
Pubs/).
Kellerer, H., Pferschy, U., & Pisinger, D. (2004).
Knapsack problems. Springer.
Leclercq, E., Djamal, B., & Yétongnon, K.
(1999). ISIS: A semantic mediation model and
an agent-based architecture for GIS interoper-
ability. Proceedings of the International Database
Engineering and Applications Symposium IDEAS
(pp. 81-92), Montreal, IEEE Computer Society.
Lee, D., & Chu, W. (1998). A semantic caching
schema for wrappers in Web databases. The 1st
ACM International Workshop on Web Information
and Data Management (WIDM), Washington DC.
Lee, D., & Chu, W. (1999). A semantic caching via
query matching for Web sources. In Proceedings
of the 8th ACM International Conference on In-
formation and Knowledge Management (CIKM),
Kansas City, Missouri, USA.
Lorenzetti, P., & Rizzo, L. (1996). Replacement
policies for a proxy cache. Technical report,
Universita di Pisa.
Luo, O., Naughton, J. F., Krishnamurthy, R., Cao,
P., & Li, Y. (2000). Active query caching for da-
tabase {Web} servers. ACM SIGMOD Workshop
on the Web and Databases, WebDB.
Meier, W. (2002). eXist: An open source native
XML database. In A. Chaudri, M. Jeckle, E.
Rahm, R. Unland (Eds.), Web, Web-Services,
and Database Systems. NODe 2002 Web- and
Database-Related Workshops. Springer LNCS
Series, 2593. Erfurt, Germany.
O’Neil, E. J., O’Neil, P. E., & Weikum, G. (1993).
The LRU-K Page replacement algorithm for da-
tabase disk buffering. In SIGMOD (pp. 297-306).
Open GIS Consortium. (1999). Inc., OpenGIS
Simple Features Specification for SQL Revision
1.1. OpenGIS Project Document 99-049.
Open GIS Consortium. (2003). Inc., OpenGIS®
Geography Markup Language (GML) Imple-
mentation Specification, version 3.00, Document
02-023r4.
Pagel, B., Six, H., Toben, H., Widmayer, P. (1993).
Towards an Analysis of Range Query Performance
in Spatial Data Structures. PODS. 214-221.
Savary, L., & Zeitouni, K. (2003). Spatial data
warehouse—A prototype. In The 2nd EGOV In-
ternational Conference (pp. 335-340), Prague.
LNCS 2739.
Savary, L., & Zeitouni, K. (2005). Automated lin-
ear geometric conflation for spatial data warehouse
integration process. The 8th AGILE Conference on
GIScience. Estirl, Portugal.
Silvano, M., & Paolo, T. (1990). Knapsack prob-
lems: Algorithms and computer implementations.
John Wiley & Sons.
Sindoni, G., Tininini, L., Ambrosetti, A., Bedes-
chi, C., De Francisci, S., & Gargano, O. (2001).
SIT-IN: A real-life spatio-temporal information
system. VLDB, 2001, 711–712.
Stoimenov, L., Djordjevic-Kajan, S., & Stoja-
novic, D. (2000). Integration of GIS data sources
over the Internet using mediator and wrapper
technology. MELECON 2000, 10th Mediterranean
Electrotechnical Conference (Vol. 1, 334-336).
Tanenbaum, A. S. (1992). Modern operating
systems. Prentice-Hall.
567
 GeoCache

Theodoridis, Y., & Sellis, T. (1996). A model for Yang, O., Zhang, H., & Zhang, H. (2003). Taylor
the prediction of r-tree performance. In Proceed- series prediction: A cache replacement policy
ings of the 15th ACM Symp. Principles of Database based on second-order trend analysis. The 34th
Systems (pp. 161-171). Annual Hawaii International Conference on
System Sciences, Maui, Hawaii.
Vatsavai, R. R. (2002). GML-QL: A spatial query
language specification for GML.UCGIS Summer, Zhang, M., Javed, S., Shaheen, A., & Gruenwald,
Athens, Georgia. L. (2001). A prototype for wrapping and visual-
izing geo-referenced data in distributed environ-
Voisard, A., & Juergens, M. (1999). Geographical
ments using the XML technology. ACMGIS at
information extraction: Querying or quarrying?
VA (pp. 27-32).
In M. Goodchild, M. Egenhofer, R. Fegeas, C.
Kottman (Eds.), Interoperating geographical in-
formation systems, New York: Kluwer Academic
Publishers.

This work was previously published in International Journal of Data Warehousing and Mining, Vol. 3, Issue 1, edited by D.
Taniar, pp. 67-88, copyright 2007 by IGI Publishing (an imprint of IGI Global).

568
 569

Chapter 35
PDE-Based Image Processing:
Image Restoration

Rajeev Srivastava
Institute of Technology, Banaras Hindu University (ITBHU), India

ABSTRACT
This chapter describes the basic concepts of partial differential equations (PDEs) based image mod-
elling and their applications to image restoration. The general basic concepts of partial differential
equation (PDE)-based image modelling and processing techniques are discussed for image restoration
problems. These techniques can also be used in the design and development of efficient tools for various
image processing and vision related tasks such as restoration, enhancement, segmentation, registration,
inpainting, shape from shading, 3D reconstruction of objects from multiple views, and many more. As
a case study, the topic in consideration is oriented towards image restoration using PDEs formalism
since image restoration is considered to be an important pre-processing task for 3D surface geometry,
reconstruction, and many other applications. An image may be subjected to various types of noises
during its acquisition leading to degraded quality of the image, and hence, the noise must be reduced.
The noise may be additive or multiplicative in nature. Here, the PDE-based models for removal of both
types of noises are discussed. As examples, some PDE-based schemes have been implemented and their
comparative study with other existing techniques has also been presented.

INTRODUCTION resentation of the image. The scientific structure

Image processing is a rapidly growing field which
can be defined as the manipulation of an image
for the purpose of either extracting information
from the image or producing an alternative rep-
DOI: 10.4018/978-1-4666-2038-4.ch035
of any image can be supposed to be based on an
intrinsic principles of mathematics i.e. from im-
age analysis to image processing. Image analysis
includes modeling and analysis of original image
itself, i.e. from image space analysis to different
methods to represent image. The various tools of
image analysis include spectral analysis, wavelets,

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

statistics, level-sets and partial differential equa-
tions (PDEs). On the other hand, image processing
is to modify the original image to improve the
quality or extracting information from the given
image, for example, image restoration, compres-
sion, segmentation, shape and texture analysis.
There are two dual fields that are directly con-
nected to the image processing in contemporary
Computer science. These are Computer vision
which is related to the construction of the 3D world
from the observed 2D images; and another one
is Computer graphics which pursues the opposite
direction in designing suitable 2D scene images to
simulate our 3D world. Image processing can be
considered as the crucial middle way connecting
the vision and graphics fields. Image process-
ing can be considered as an input-output system
shown in Figure 1.
I0 is the input data which represents an observed
or measured single image or image sequences; τ
denotes a typical image processor e.g. restoration,
enhancement, segmentation, compression, inter-
polation, feature extraction, inpainting etc.; and
I denotes output which is also an image or an
image sequence (I1,I2,I3,…..) that contains all the
targeted image features. The problem very often
encountered in this area is to design an efficient
and cost effective and accurate image processor.
Typical design tasks include: de-noising, de-
blurring, edge detection, intensity enhancement,
inpainting, interpolation, compression and de-
compression etc. In addition to these relatively
low-level tasks, there are mid- and high-level
tasks like disocclusion, shape from shading, mo-
tion analysis, image segmentation, and pattern
identification and recognition. In image represen-
tation (Jain, 2006), one is concerned with char-
Figure 1. An image processing system
570
PDE-Based Image Processing
acterization of the quantity that each picture ele-
ment or pixel represents. An image could represent
luminance of objects in a scene such as in digital
photography, the absorption characteristics of the
body tissue as in X-Ray imaging, the radar cross
section of a target in radar imaging, the tempera-
ture profile of a region in infrared imaging or the
gravitational field in an area in geophysical imag-
ing. In general, any 2D function that contains
information can be considered as an image. Image
models give a logical or quantitative description
of the properties of this function. There are three
crucial ingredients of image processing which
include modeling, analysis, and efficient imple-
mentation of processing tools. The design of a
successful image processing technique relies on
having a successful model for images. A funda-
mental issue faced in the design of image analy-
sis techniques is the identification and character-
ization of the image space. The various
approaches to image modeling and analysis in-
clude statistical representations or random field
modeling; spectral and wavelet representations;
and scale-space representations or regularity
spaces.
The statistical approaches treat images as
samples from random fields, which are often mod-
eled by Markov or Gibbs fields or via statistical
learning from an image database. The statistical
properties of the fields are often established
through the filtering technique and learning
theory. Statistical models describe an image as a
member of an ensemble, often characterized by
its mean and covariance functions. This permits
development of algorithms that are useful for an
entire class or an ensemble of images rather than
for a single image. Often the ensemble is assumed
PDE-Based Image Processing
to be stationary so that the mean and covariance
functions can easily be estimated. Random field
modeling is the most appropriate approach for
describing natural images with rich texture patterns
such as trees and mountains. Stationary models
are useful in data compression problems such as
transform coding, restoration problems such as
Wiener filtering (Jain, 2006), and in other applica-
tions where global properties of an ensemble are
sufficient. A more effective use of these models
in image processing is to consider them to be
spatially varying or piecewise spatially invariant.
The spectral and wavelet representations are
also playing an important role in the design and
analysis of an efficient image processor. Wave-
let analysis is about analyzing signal with short
duration finite energy functions. They transform
the signal under investigation into another repre-
sentation which presents the signal in more useful
form. This transformation of the signal is called
Wavelet Transform (Gonzalez, 1987) i.e. Wavelet
Transforms are based on small waves, called wave-
lets, of varying frequency and limited duration.
Unlike the Fourier transform, we have a variety of
wavelets that are used for signal analysis. Choice
of a particular wavelet depends on the type of ap-
plication in hand. Wavelet Transforms provides
time-frequency view i.e. provides both frequency
as well as temporal (localization) information and
exhibits multi-resolution characteristics. Fourier
transform is good for periodic or stationary signals
and Wavelet is good for transients. Localization
property allows wavelets to give efficient repre-
sentation of transients. In Wavelet transforms a
signal can be converted and manipulated while
keeping resolution across the entire signal and still
based in time i.e. Wavelets have special ability to
examine signals simultaneously in both time and
frequency. Wavelets are mathematical functions
that satisfy certain criteria, like a zero mean, and
are used for analyzing and representing signals or
other functions. A set of dilations and Translations
of a chosen mother wavelet is used for the spatial/
frequency analysis of an input signal. The Wavelet
Transform uses overlapping functions of variable
size for analysis. The overlapping nature of the
transform alleviates the blocking artefacts, as each
input sample contributes to several samples of the
output. The variable size of the basis functions,
in addition, leads to superior energy compaction
and good perceptual quality of the decompressed
image. Wavelets Transform is based on the concept
of sub-band coding (Gonzalez, 1987; Soman,
2004; Topiwala, 1998). The current applications
of wavelet include statistical signal processing,
Image processing, climate analysis, financial time
series analysis, heart monitoring, seismic signal
de- noising, de-noising of astronomical images,
audio and video compression, compression of
medical image stacks, finger print analysis, fast
solution of partial differential equations, computer
graphics etc.
The scale-space decomposition is a partial
differential equation (PDE) based approach to the
multi-scale analysis of images. This theory is based
on the linear filtering or heat diffusion process, and
was highlighted by the famous nonlinear model as
proposed by (Perona, & Malik, 1990) for remov-
ing small-scale noise images without losing large
scale structures like edges. There are other novel
theories and trends for image analysis, including
regular-space representations by the bounded
variation (BV) space, free boundary models, and
shape models (Romeny,1994).
The first step of image processing is to con-
struct suitable models for the given task. Currently,
processing models are being developed using tools
like Bayesian decision, variational optimization,
inverse problems and the like. These approaches
arise from fields such as statistical mechanics,
the calculus of variations, nonlinear PDEs, dif-
ferential geometry and topology, harmonic and
functional analysis, and numerical analysis. For
example, numerous PDE based methods (Perona,
& Malik, 1990; Rudin et al., 1992; Witkin, 1983;
Gilboa et al., 2004) have been proposed for image
de-noising problem using PDEs. Once an image
processing model is selected, the next step is its
571

analysis, in order to answer questions of existence
and uniqueness, stability, properties of solutions,
and so on. Many image processing models are
nonlinear or non-convex. Analyzing them requires
new mathematical insights. Relatively few image
processing models have been analyzed in this way.
There is a vast engineering literature of image
processors, which were mostly developed based
on empirical insights and leaps of good intuition.
They require rigorous mathematical analysis that
can be crucial for answering questions of great
practical significance, such as which among the
many techniques proposed for the same task is
superior, what effect the various parameters that
appear in a method have on its behavior, and under
what conditions a given technique can be expected
to perform well. Finally, an efficient algorithm for
implementing the processing model is devised.
The image restoration is a fundamental pre-
processing step of many image processing or
computer vision related applications. The goal
of image restoration problem is to restore the
degraded image for further analysis and use of
image data in a specific application. During the
capture or formation of an image through an
imaging device it may be subjected to various
types of noise that may lead to degradation of
the observed image. The noise may be additive
or multiplicative in nature. The additive noise is
easy to remove in comparison to multiplicative
one. Image restoration techniques are based on
modeling the degradation using a priori knowledge
and applying the inverse process in order to restore
the original image (Gonzalez, & Wintz, 1987).
For some images, the a priori information of the
degradation phenomenon may not be available that
can be further used for modeling the degradation
process. In such cases, blind restoration techniques
are applied for the restoration of the corrupted
image. In blind image restoration technique, the
original image is estimated without the explicit
knowledge of the underlying degradation process.
This process is difficult since information about the
original image or the underlying blurring process
572
PDE-Based Image Processing
is not available in many practical applications such
as in space exploration and astronomy. Therefore,
to retrieve unknown images one has to incorporate
human visual perception mechanism in order to
accurately restore the image i.e. restores the im-
age according to the requirement of the desired
information. The various possible reasons that can
cause the observed image to deviate from the origi-
nal one may include atmospheric irregularities,
instrument aberrations, diffraction limit, detector
noise, thermal noise, electronic noise, and intrinsic
noise such as Poisson noise. These noises may be
additive, multiplicative (speckle noise) or may
be distributed according to a specific probability
distribution (PDF) function in an image. Several
restoration methods such as Weiner Filtering,
Inverse Filtering, Constrained Least Squares, and
Lucy-Richardson iteration have been proposed in
literature (Gonzalez, &Wintz, 1987; Sonka et al,
2007; Lucy, 1974) that remove the noise either
using Fourier Transformation in frequency domain
or by using optimization techniques. The various
methods to estimate the degradation function for
use in restoration are observation, experimenta-
tion and mathematical modeling. In recent past
years, a new idea has been evolved by the research
community for image processing in continuous
domain using partial differential equation (PDEs)
formalism or scale-space decomposition in par-
allel to the statistical modeling. The scale-space
decomposition (Chambolle, 1994; Witkin, 1983)
is a PDE approach to the multi-scale analysis of
images. These PDE based approaches arise from
the fields of mathematical physics such as various
transport phenomenon e.g. heat flow, fluid flow,
statistical mechanics, the calculus of variations,
nonlinear PDEs, and numerical analysis. In PDE
based image processing (Caselles et al., 1998),
the basic idea is to obtain the processed image
as the solution of PDE’s, being in general the
initial condition a function of the original image.
The extensive research on numerical analysis al-
lows us the implementation of these equations,
obtaining accurate image processing algorithms.
PDE-Based Image Processing
When the image is represented as a continuous
signal, PDEs can be seen as the iteration of local
filters with an infinitesimal neighborhood. This
interpretation of PDEs allows to unify and clas-
sify iterated filters as well as to derive new ones
(Caselles, 1998). In literature, authors Price et al.
(1990); Alvarez (1992); Alvarez (1993); Rudin et
al. (1992); Osher, (1990) have proposed various
methods for image restoration and de-noising and
edge detection using PDEs formalism including
the pioneering work of Perona, & Malik (1990).
The proposed chapter is divided in to five
sections. Introduction, gives an introduction of
the topic; PDE-based image processing, presents
general principles, literature survey, examples and
application areas of PDE-based image process-
ing; Performance metrics, gives definitions of
various performance metrics that can be used to
evaluate the performance of a image processing
model; Image restoration, is the main section of
the chapter where various PDE-based techniques
used to restore an image corrupted with additive
and multiplicative noises are discussed. The first
subsection, deals with various PDE-based additive
noise removal techniques. The second subsection,
deals with multiplicative noise reduction from
images. The example of multiplicative noise in-
cludes speckle noise present in various imaging
modalities such as ultrasound imaging, SAR imag-
ing and digital holography. The subsequent next
section, examines the statistics of speckle noise
patterns for various imaging modalities, discusses
the standard speckle reduction techniques avail-
able in literature, PDE-based speckle reduction
techniques, performance analysis and comparison
of PDE-based techniques with other techniques
and finally last section, gives the conclusion of
the chapter.
PDE-BASED IMAGE PROCESSING
General Principles
Partial Differential Equations (PDE’s) and
Geometry-Driven Diffusion in image process-
ing and analysis is a relatively new area which
provides a strong theoretical framework for image
processing and analysis. Attention to these meth-
ods has increased recently due to the important
theoretical and practical results that have been
obtained. The use of PDEs and curve or surface
evolution theory in image modeling and analysis
became a major research topic in recent years. The
PDE-based approach permits to unify and extend
existent algorithms, as well as to develop novel
ones. Extensive research on numerical analysis
allows the implementation of these equations,
obtaining accurate image processing algorithms.
The basic idea is to obtain the processed image as
the solution of PDE’s, being in general the initial
condition a function of the original image.
To explain this (Caselles et al., 1998), suppose
ϕ0: R2→R represent a gray level image, where ϕ0(x,
y) is the gray level value. Introducing an artificial
time t parameter the image deforms according to
∂ϕ
= F [ϕ(x , y, t )]; (1)
∂t
where ϕ(x, y, t)=R2×[0,τ) →R is the evolving
image, F: R→R is an operator characterizing the
algorithm and the image ϕ0(x, y) is the initial
condition. The solution ϕ(x, y, t) of the PDE gives
the processed image at scale t.
In case of vector valued images, a system of
coupled PDEs of the form of (1) is obtained. The
same formalism may be applied to planar curves
(boundaries of planar shapes), where ϕ(x, y) is
a function from R→R2 or surfaces, functions
from R→R3. In this case the operator F must be
restricted to the curve, and all isotropic motions
can be described as a deformation of the curve
573

or surface in its normal direction, with velocity
related to its principal curvature(s) and flow of
the form
∂ϕ →
= F (ki ) N , (2)
∂t
is obtained, where ki’s are the principal curvatures
→ (1978). The notion of scale space i.e. representa-
and N is normal to the curve or surface ϕ(x, y).
A tangential velocity can also be added, which
may help the analysis but does not affect the
geometry of flow.
The PDEs describing the evolution phenom-
enon can be obtained from variational problems
and for an image processing problem it is for-
mulated as
arg{MinϕU (ϕ)}, (3)
where U is a given energy.
Let F(ϕ) denote the Euler-Lagrange derivative
i.e. first variation and under general assumptions,
a necessary condition for ϕ to be minimizer of U
is that F(ϕ)=0, then the local minima may be com-
puted via the steady state solution of the equation
∂ϕ
= F(ϕ), (4)
∂t
where t is an artificial time parameter. One of the
most popular example for energy functional U, for
an image, to be minimized is the Dirichlet integral
2
U (ϕ) = ∫ ∇ϕ (x )dx , (5)
which is associated with the heat equation
∂ϕ
= ∆ϕ(x ). (6)
∂t
574
PDE-Based Image Processing
Extensive research is also being done on the
direct derivation of evolution equations which
are not necessarily obtained from the energy ap-
proaches. This is the case for a number of curvature
equations of the form (2).
The ideas on the use of PDEs in image pro-
cessing (Caselles et al, 1998) were introduced
very back by Gabor (1965); Jain (1977); and Jain
tion of images simultaneously at multiple scales
which is now the basis of most of the research in
PDEs for image processing was introduced by
Koederink (1983) and Witkin (1984) where they
shown that the multiscale image representation is
obtained by Gaussian filtering which is equivalent
to deforming the optimal image via the classical
heat equation, which generates an isotropic dif-
fusion flow. Perona and Malik (1990) work on
anisotropic diffusion proposed to replace Gaussian
smoothing which is equivalent to isotropic diffu-
sion via heat flow, by a directional diffusion that
preserves edges. Their work opened a number of
theoretical and practical questions that continue
to occupy the PDE image processing community.
In the same framework, the very good work of
Osher and Rudin (1990) on shock filters and Rudin
et al. (1992) on total variation (TV) decreasing
methods explicitly stated the importance and the
need for understanding PDEs for image process-
ing applications.
Many other interesting problems in computer
vision can be formulated as minimisation problems
for energy functional (Shen, 2002). In order to
solve these problems, the Euler- Lagrange equa-
tions of the functional are computed, resulting
in a set of necessary conditions. In effect, these
conditions are partial differential equations which
are reformulated as a surface evolution problem.
Among the first successfully utilised for computer
vision problems was Snakes: Active Contours
Models (Kass et al., 1988). While originally
developed for object segmentation in 2D, this
approach can be generalised to 3D. The minimal
PDE-Based Image Processing
surfaces may be employed for 3D reconstruction
of static objects from multiple views as proposed
by Faugeras and Keriven (1998). The basic con-
cepts of curve and surface evolution are as follows
(Osher, &Sethian, 1988; Romeny,1994; Evans, &
Spruck,1991):
Curve Evolution
The curve evolution deals with the task of deform-
ing a curve. Let C0: I→R2 be initial curve and let p
denote its parameterisation. To describe an evolu-
tion in time, a time parameter t≥0 is introduced.
The curve evolution is defined as
→ →
∂ tu (u, v, t ) ,tv (u, v, t ) and n(u, v, t ) denote the
C (p, t ) = α(p, t ) t (p, t ) + β(p, t ) n(p, t ),
∂t
(7)
with the initial curve C(p,0)=C0(p).
The movement of each curve point is written
→ →
in the local coordinate system ( t ; n ) of the curve,
→ →
where t denotes the tangential vector and n
denotes the normal vector. This equation can be
simplified to a curve evolution restricted to the
normal direction, i.e.
→
∂ ing on the Gaussian curvature k. Most of the PDEs
C (p, t ) = β(p, t ) n(p, t ) , (8)
∂t
if the normal velocity 𝛽(p, t) only depends on C(p,
t) and t. If this is the case, the tangential motion
only acts as a re-parameterisation of the curve.
The curve evolution in Equation (8) depends on
the definition of the velocity (p, t). If, for example
(p, t) = 1 for all parameterisations, the evolving
curve at a certain time step t is a dilation of the
initial curve C0. Erosion can thus be modelled by
reverting the sign of (p, t). As another example,
(p, t)= k(p, t) yields the prominent Euclidean
geometric heat flow or curvature motion. Where
k(p, t) denotes the Euclidean curvature.
Surface Evolution
Analogous to curve evolution, one can define a
surface evolution on a surface S0=D→R3. Let u
and v denote the surface parameters and t≥0 the
necessary time parameter. The surface evolution
is then defined by
→
∂
S (u, v, t ) = α1 (u, v, t ) tu (u, v, t )
∂t
→ →
+α2 (u, v, t ) tv (u, v, t ) + β(u, v, t ) n(u, v, t ),
(9)
with S (u, v, 0) = S 0 (u, v ).
→ → →
tangential vectors in u and v direction and the
surface normal, respectively. Similar as in the
case of curves, the tangential components of the
surface evolution act only as re-parameterisations
and are therefore irrelevant for the geometric
shape of the evolving surface. As in the curve
evolution case, the constant velocity flows
∂ →
S = ± n describe dilations and erosions. As
∂t
there are two different concepts of curvature on
surfaces, one can define different flows, one de-
pending on the mean curvature H and one depend-
used in image processing and computer vision
are based on moving curves and surfaces with
curvature based velocities. In this area, the level
set numerical method developed by Osher and
Sethian (1988) played a very important role. The
basic idea is to represent the deforming curve,
surface or image, as the level set of a higher di-
mensional hyper surface. This technique not only
provides more accurate numerical implementa-
tions but also solves topological issues that were
very difficult to treat before. The representation
of objects as level sets (zero-sets) is a fundamen-
tal mathematical technique in mathematical
morphology (Osher, & Fedkiw, 2001; Evans,
575

&Spruck,1991). Another key contribution in the
PDE formalism has been the general segmentation
frame work developed by Mumford and Shah
(1989). Their work has unified a large number of
image segmentation approaches, and opened as
well a large number of theoretical and practical
problems. Kimia et al. (1990) introduced curve
evolution methods in to computer vision for a
computational theory of a planar shape. Refer-
ences to works of other authors in related field
can be found in (Caselles et al., 1998). The frame
work of PDE’s and geometry driven diffusion
have been applied to many problems in image
processing and computer vision, since the seminal
works mentioned above. Examples include con-
tinuous mathematical morphology, invariant shape
analysis, shape from shading, segmentation, object
detection, optical flow stereo, image de-noising,
image sharpening, contrast enhancement, image
interpolation etc.
Advantages of PDEs Evolution
Approaches in Image Analysis
(Caselles et al., 1998)
• They bring out a new concept to the area.
• One can think about image processing in
continuous domain.
• The problem in hand is then approached
as an image deformation task, which help
to arrive to novel solutions of classical
problems.
• When the image is represented as a con-
tinuous signal, PDEs can be seen as the it-
eration of local filters with an infinitesimal
neighborhood. This interpretation of PDEs
allows to unify and classify iterated filters
as well as to derive new ones.
• PDEs approach allows not only to derive
new algorithms, but also to unify previous
ones. PDE formulation is natural in order
to combine algorithms. If two different im-
age processing schemes are given by
576
PDE-Based Image Processing
∂Φ
= F1[Φ(x , y, t )] and
∂t
∂Φ
= F2 [Φ(x , y, t )], then they can be com-
∂t
∂Φ
bined as = αF1 + F2 , where α ∈ R + .
∂t
If F1 and F2 above are the corresponding
Euler-Lagrange operators of two energy
minimization problems with energies U1
and U2, then the flow above minimizes the
energy αU 1 + U 2 .
• Another important advantage of the PDE
approach is the possibility of achieving
high accuracy and stability, with the help
of extensive available research on numeri-
cal analysis.
APPLICATION AREAS
Scale-space theory, anisotropic diffusion, vector-
valued diffusion, systems of coupled PDE’s,
PDE based segmentation, image and video en-
hancement, shape analysis, axiomatic and PDE
based theoretical approaches to image analysis,
variational and PDE’s-from-energy approaches
to image processing, curve and surface evolution,
PDE’s in hybrid systems, comparison of models,
image and noise models and adaptation of PDE’s,
implementations and numerical analysis, Applica-
tions (Mathematical morphology, Invariant shape
analysis, Shape from shading, Segmentation,
Object detection, Optical flow, Image Denois-
ing, Image sharpening, Contrast Enhancement,
Image quantization, and Image Compression).
Examples of some classical PDE-based algo-
rithms and explanation about how the modeling
of curves and surfaces may occur are explained
in following sections.
PDE-Based Image Processing
Image Restoration
A noisy 2D image may be considered for instance
as a surface I: (x, y)→(x, y, I(x, y). With this kind
of model, regularization or smoothing of the im-
age I may be equivalent smooth surface similar
enough to the original noisy one. This can be
done by minimizing energy functional or directly
designing PDE’s with specific regularization
behaviors that evolve the noisy surface (Perona,
& Malik,1990; Rudin,1992; Gilboa et al., 2005;
You, & Kaveh, 2000).
Image Segmentation
A classical way of segmenting images with PDEs
is to model a closed contour with a 2D curve and
then to evolve it from an initial position (random
or user defined) until it fits the exact shape of the
objects present in the picture. Like image resto-
ration purposes, a PDE is describing the curve
Figure 2. Noisy image (left), and restored image
(right)
Figure 4. Image registration treated as the evolution of a displacement field, Direct superposing of two
images (First two images from left), and superposing after image registration (last two images in right)
evolution and may come from energy minimiza-
tion or from pertinent local heuristics (Kass et al.,
1988; Morel, 1988; Caselles, 1997).
Image Registration and Optical Flow
The idea is to find a function allowing to trans-
form an image to another one (Zitova, & Flusser,
2003; Brown, 1992). It is used to detect motions
in video sequences by registering consecutive
frames, or readjust two images in a way that they
fit together such as in application medical image
analysis. Here, a vector field models the pixel
motions between the two images and a PDE is
used to describe its evolution until it converges
to the expected image transformation.
Shape from Shading
The problem is reconstructing a 3D representation
of an object from a single photograph of it. It is
Figure 3. Image segmentation treated as the
evolution of a 2D contour curve, original image
with initialization curve (left), and segmented
image (right)
577
 PDE-Based Image Processing

possible if one looks at the intensity variations of PERFORMANCE MEASUREMENT

the image pixels due to the shadows and the dif-
ferent illumination conditions during the snapshot.
PDEs can describe the flow of an originally flat
3D surface converging to the 3D shape of the real
object. A survey report on shape and shading is
presented in (Zhang, 1999).
METRICS
The various performance measurement metrics
for an image of size m×n used for evaluating the
performance and efficacy of the various proposed
methods are defined as follows:

Image and Surface Inpainting Mean square error

Inpainting refers to technique of modifying an 1 m n 2

image in an undetectable form. It has numerous

MSE =
m ×n
∑ ∑ I (i, j ) − I (i, j )
'

i =1 j =1
applications such as restoration of scratched (10)
photographs, removal of undesirable objects,
reconstruction of 3D surfaces, etc. Inpainting is where I is the observed noisy image and I ' is the
done by propagation of information of surround- filtered image.
ing structure and intensity. The various methods
available in literature includes geometric approach Root mean square error
and iterative algorithm. PDE based methods can
be used for image and surface inpaintings (Chan,
RMSE = MSE (11)
2005).

Peak signal-to-noise ratio

Figure 5. Image inpainting: original scratched
image (left), and inpainted image (right)  255 
PSNR = 20 log10   (12)
 RMSE 
 

Figure 6. Image inpainting: original image (left), and inpainted image (right)

578
PDE-Based Image Processing

Signal-to-noise ratio Average SNR=1/SI. (17)

 m n  Effective Number of Looks (ENL)


 ∑ ∑ I 2 (i, j ) 

SNR = 10 log10  m n i =1 j = 1 
 The number of looks (Dewaele, 1990) in an
 2
 ∑ ∑ (I (i, j ) − I (i, j ))  intensity image is a measure of the statistical
'

 i =1 j =1  fluctuations introduced by speckle resulting from

(13) the interference between randomly positioned
scatterer. Thus ENL gives essentially an idea about
Normalized mean square error the smoothness in the regions on the image that is
supposed to have a homogeneous appearance but
m n
are corrupted by noise. ENL is generally defined as
∑ ∑ (I (i, j ) − I (i, j ))
' 2

i =1 j =1
NMSE = (14)
m n
µt2
∑ ∑ I (i, j ) 2
ENL = (18)
i =1 j =1 σt2

For optimal performance, measured values where t denotes the target area or region of inter-
of MSE, RMSE and NMSE should be small and est, µt and σt are the pixel mean and standard
PSNR and SNR should be large. deviation of a target area of the image. In this
paper, target area is the whole image. A large
Speckle Index value of ENL reflects better quantitative perfor-
mance of the filter.
Since speckle noise is multiplicative in nature,
average contrast of an image may be treated as a Correlation Parameter (CP)
measure of speckle removal. Speckle index (SI)
is defined as (Dewaele, 1990): CP (Salinas et al., 2007) is a qualitative measure
for edge preservation. If one is interested in sup-
var(I ) pressing speckle noise while at the same time
SI = , (15)
E (I ) preserving the edges of the original image then
this parameter can be used. Therefore, to evalu-
ate the performance of the edge preservation or
and its discrete version for an image reads
sharpness, the correlation parameter is defined
as follows
1 m n
σ(i, j )
SI =
mn
∑ ∑ µ(i, j ) , (16)
−
i =1 j =1
m n − ^ ^

∑ ∑ (∆I − ∆ I ) × (∆ I − ∆ I )
where m×n is the size of the image, μ is the mean CP = i =1 j =1
−

m n − m n ^ ^
and σ is the standard deviation.
The speckle index can be regarded as an aver-
∑ ∑ (∆I − ∆ I ) ×∑ ∑ (∆ I − ∆ I )
2 2

i =1 j =1 i =1 j =1

age reciprocal signal-to noise ratio (SNR) with (19)

the signal being the mean value and noise being
the standard deviation.

579

^
where ∆I and ∆ I are high pass filtered versions
^
of original image I and filtered image I obtained
via a 3x3 pixel standard approximation of the
−
− ^
Laplacian operator. The ∆ I and ∆ I are the
^
mean values of I and I , respectively. The cor-
relation parameter should be closer to unity for
an optimal effect of edge preservation.
Structure Similarity
Index Map (SSIM)
SSIM (Wang et al., 2004) is used to compare
luminance, contrast and structure of two different
images. It can be treated as a similarity measure
of two different images. The SSIM of two images
X and Y can be defined as
SSIM (X ,Y ) =
α β
l (X ,Y ) . c(X ,Y ) . s(X ,Y )
γ, (20)
     
where α > 0 , β > 0 and γ > 0 are parameters
and can be used to adjust the relative importance
of three components. SSIM of two images X and
Y can be calculated as
(2µx µy + C 1 ) × (2σxy + C 2 )
SSIM (X ,Y ) =
(µ + µy2 + C 1 ) × (σx2 + σy2 + C 2 )
2
x performance of the blind restoration algorithms
(21)
where µi (i = X or Y) is the mean intensity, σi
(i=X or Y) is the standard deviation, σxy = σx .σy
and Ci (i=1 or 2) is the constant to avoid instabil-
ity when µx2 + µy2 is very close to zero and is
defined as C i = (ki L)2 in which ki << 1 and L
is the dynamic range of pixel values e.g. L=255
for 8-bit gray scale image. In order to have an
overall quality measurement of the entire image,
mean SSIM is defined as
580
PDE-Based Image Processing
m n
1
MSSIM (X ,Y ) =
mn
∑ ∑ SSIM (X ij
,Yij ) .
i =1 j =1
(22)
The MSSIM value should be closer to unity
for optimal measure of similarity.
Histogram Analysis
Though the MSE analysis quantitatively measures
the average difference between the original im-
age and the reconstructed image, it is often as
important and desirable to know how the errors
are distributed in the image. A histogram analysis
can be carried out to quantitatively investigate the
performance of histogram distribution. Based on
the histogram distribution of the zoomed/restored
image, one can evaluate the performance of the
proposed algorithms.
In a blind image restoration problem a priori
information about the degradation process of an
image is not known. Only noisy image is available
as the input and it has to be restored. Since, we
don’t have the original image therefore mean-
square-error (MSE) and peak signal-to-noise
ratio (PSNR) between original (without noise)
and restored image can’t be calculated and these
criterions can’t be used for performance evalu-
ation of the proposed algorithm. Therefore, the
,
can either be evaluated subjectively by a human
observer or by some qualitative procedure that
can be used to compare the various restoration
algorithms from the information available from
noisy image only. In this thesis, the performance
measurement metrics used are blurred signal-to-
noise ratio (BSNR) and average signal-to-noise
ratio (SNR). In addition to BSNR and average
SNR, the histogram analysis technique is also
used to compare the performances of the various
schemes.
PDE-Based Image Processing
Average signal-to-noise ratio
mean(I ) µ noise in 2D B-scan ultrasound images follow
Avg.SNR = = (23)
std .deviation(I )
σn2
A higher value of avg. SNR indicates less
noise i.e. better restoration. Average SNR is also
related to speckle index (SI) which is a measure
of speckle noise. It is related as follows:
1 smoothing. In image regularization or smoothing,
Avg.SNR =
SI
Blurred signal –to-noise ratio (bsnr)
 1

 (M × N ) ∑ ∑ [I orignoisy − I restored ]2 
BSNR = 10 log10 
 σ 2
 n


where M×N is the size of image, Iorignoisy is the
original noisy image; Irestored is the restored filtered
image and σn2 is the variance of noise in original
noisy image (Jain, 2006). A low value of BSNR
indicates less blurring or less noise or better res-
toration.
IMAGE RESTORATION USING
PDES FORMALISM
In this section, 2D image restoration problem
is addressed which can be generalized to 3D
restoration. The noise present in an image may
be additive, multiplicative or may be distributed
according to a specific probability distribution
function (pdf) depending on the application. For
example, additive noise in an image normally
follows the Gaussian distribution. Speckle noise
which is multiplicative in nature may be present
in ultrasound images, synthetic aperture radar
(SAR) images, digital holographic images, optical
coherence tomography (OCT) images etc. Speckle
Rayleigh’s probability distribution; speckle noise
present in SAR imagery follows k-distribution
and noise present in magnetic resonance imaging
(MRI) follow Rician distribution. Depending upon
the type of noise a specific noise removal filter
is required for restoration of a particular image.
Restoration of a digital image from its noisy
version is also referred to as regularization or
(24)
we simplify the image data in a way that only
interesting features are preserved. The regulariza-
∂I
tion term R in PDE formulations like =R
∂t
 introduces additional notion of scale space i.e.
 the data are iteratively regularized or smoothed

 and a continuous sequence of smoother images


 is generated as time t goes by. The regularization
(25) PDEs may be seen as nonlinear filters that sim-
plify the image little by little and minimize the
image variations. The regularization term consti-
tutes the key elements for solving ill-posed com-
puter vision problems such as restoration, seg-
mentation, registration, surface reconstruction
etc. Specifically, the de-noising algorithms are
usually based on a regularization term R coupled
with a data attachment term (I noisy-I) which is also
known as fidelity term and its corresponding PDE
is
∂I
= R + (I noisy − I ) (26)
∂t
Figure 7 shows the examples of various
noise patterns present in an image which may
have specific noise statistics. Figure 8 shows the
histogram plots of various noise patterns in an
image. Histogram plot shows the noise variations
present in an image.
581
 PDE-Based Image Processing

Figure 7. Various noise patterns present in an image

Figure 8. Histogram plots of noise patterns present in an image

582
PDE-Based Image Processing

ADDITIVE NOISE REMOVAL FROM
DIGITAL IMAGES
In PDE based noise removal techniques (Romeny,
1994; Caselles et al., 1998), suppose I is a 2D
scalar noisy image that we want to restore and
the noise can be considered as high frequency
variations σ with low amplitude, added to the
pixels of the regular image
Koenderink (1984) has shown that the solution
of Equation (30) at a particular time t is the con-
volution of the original image Inoisy with a normal-
ized 2D Gaussian kernelGσ with variance
σ = 2t ,
It (x , y ) = ∫∫ I noisy (x − u, y − v ) * Gσ (u, v )dudv,
(31)

I noisy = I regular + σ . (27) 1  x 2 + y 2 

  and σ = 2t
withGσ = exp − 
2πσ 2  2σ 2
 
To regularize Inoisy a common idea is to minimize which indicates that regularization is linear based
its variations estimated by gradient norm of image: on convolution. Here the image is blurred little
by little in an isotropic way during the PDE evo-
lution.
|| ∇I ||= (I x 2 + I y 2 ) . (28)
Some PDE-Based Image Restoration
Then the corresponding variational problem Models for Additive Noise
is the minimization of energy functional
Total Variation (TV) Approach
∫ || ∇I || dΩ . (29)
2
min E (I ) =
I :Ω→R
Ω Rudin et al. (1992) proposed an algorithm for
noise removal based on the minimization of the
The necessary condition for minimizing the total first variation (TV) of image I(x, y, t) given
energy functional E(I) described by Equation (29) by following energy functional:
can be obtained using Euler-Lagrange minimiza-
tion that results in following heat equation E = arg ∫ I x2 + I y2 dxdy = arg ∫ ∇I dxdy.
min image min image

∂I
= c∇2I = c(I xx + I yy ) , (30a)
∂t
I = I noisy with initial condition as the
(t =0)
observed noisy image given as:, (30b)
where ∇2I is Laplacian of image I and c is the
diffusion constant and I (x , y, t ) = I (x , y ) .This
equation describes the isotropic diffusion process.
The basic disadvantage of the isotropic diffusion
is that in addition to noise removal it may also
blur the edges and fine structures present in the
image after certain iterations.
(32)
The minimization is performed under certain
constraints and boundary conditions (zero flow
on the boundary). The constraints they employ
are zero mean and given variance of the noise.
The Euler-Lagrange minimization of above
energy functional leads to following total variation
(TV) based restoration model reads:
 
∂I  ∇I 
= ∇.  , (33)
∂t  ∇I 
 
with initial condition I (x , y; 0) = I noisy .

583

Anisotropic Diffusion Based Model
Perona and Malik (1990) proposed a nonlinear
diffusion method to avoid blurring and localiza-
tion problem of linear diffusion filtering which is
termed as anisotropic diffusion. Anisotropic dif-
fusion is the opposite of isotropic i.e. to designate
a regularization process that does not smooth the
image with the same weight in all the spatial direc-
tions.This achieves both noise removal and edge
enhancement through the use of a non-uniform
diffusion which acts as unstable inverse diffusion
near edges and as linear heat equation like dif-
fusion in homogeneous regions without edges.
In paper (Perona, &Malik, 1990), authors have
used the anisotropic diffusion process which is
the extension of isotropic diffusion process to
remove additive noise from images. In anisotropic
diffusion based filter, the basic idea is that heat
Equation (30) for linear diffusion can be written
in divergence form:
∂I  
= ∇2I = div(gradI ) = ∇.∇I (34)
∂t
The introduction of a conductivity coeffi-
cient c in the above diffusion equation makes it
possible to make the diffusion adaptive to local
image structure:
∂I  
= ∇.c∇I = c∇2I + ∇c.∇I (35)
∂t
The two possible choices for diffusion coef-
ficient c are:
 || ∇I 1 ||
c1 = exp −  and c2 =
;
 k2 
|| ∇I ||2
1+
k2
where k>0 (36)
Both expressions are equal up to first order
approximation and k is a fixed gradient threshold
584
PDE-Based Image Processing
that differentiates homogeneous area and regions
of contours and edges. The value of conductivity
coefficient ranges in between 20-50.
Fourth Order PDE-Based Model
The second order PDEs such as anisotropic dif-
fusion model (Perona, & Malik, 1990) has side
effects of producing the blocky effects during the
evolution process (You, & Kaveh, 2000). The
fourth order PDEs avoids blocky effect while
achieving good trade off between noise removal
and edge preservation. The fourth order PDE
proposed by (You, & Kaveh, 2000) is derived
from a functional which is an increasing function
of the Laplacian of the image intensity function
i.e. ∇2I . The Laplacian of an image at a pixel is
zero if the image is planar in its neighbourhood
and these PDEs attempt to remove noise and
preserve edges by approximating an observed
image with a piecewise planar image. The energy
functional proposed by You and Kaveh (2000)
for additive noise reads
2
arg E (u ) =
min
∫ ∇2I d Ω . (37)
Ω
The Euler-Lagrange (E-L) minimization of
above energy functional leads to following fourth
order PDE for edge enhancement and additive
noise removal from a digital image.
∂I
∂t
(
= ∇2 C ( ∇2I )∇2I , ) (38)
where the diffusion coefficient is defined as:
1
C ( ∇2I ) = . (39)
 ∇2I 
2
 
1 +  
 k 
PDE-Based Image Processing
Complex Diffusion Based Model
In anisotropic diffusion based model (Perona,
&Malik, 1990), if real time factor t is replaced by
complex time factor it and the diffusion coefficient
c( ∇I tn ) by c(Im(I )) then it leads to following
complex diffusion equation (Gilboa et al., 2004)
originally proposed for image enhancement and
additive noise removal from digital images.
∂I
= div (c(Im(I ))∇I ) (40)
∂t
In linear complex diffusion based filter for
image enhancement and de-noising, the authors
(Gilboa et al.,2004) proposed to replace the dif-
fusion coefficient term in equation (40) with a
complex diffusion coefficient c = exp(iθ) , utiliz-
ing the approximation cos θ = 1 + O(θ 2 ) and
sin θ = θ +O(θ 3 ) and after introducing an op-
− by the linear forward diffusion, whereas evolution
erator H = c∆ , the equation (40) can be written
−
as: I t = H I with initial condition I t=0 = I 0 .The
−
solution of this PDE is given as I = exp(t H ) .
The above assumed approximations generates the
following solution
I (x , y, t ) ≈ exp(t ∆)(1 + iθt ∆)I 0 . (41)
Further separating the real and imaginary parts
of the image, I = I R + iI I , and diffusion coef-
ficient, c = cR + icI , the Equation (41) reads
I t = c∇2I = cI xx + cI yy (42)
can be written in x-component form as follows:
I Rt = cR I Rxx − cI I Ixx , I R,t =0 = I 0 , (43a)
I It = cI I Rxx + cR I Ixx , I I ,t =0 = 0, (43b)
where cR = cos θ and cI = sin θ .For small θ,
relation I Rxx >> θI Ixx holds and second term in
RHS of Equation (22a) can be omitted. Therefore,
under small θ approximation above equation re-
duces to
I Rt ≈ I Rxx (44a)
I It ≈ θI Rxx + I Ixx (44b)
Equation (44) describes the evolution of image
using complex diffusion process as proposed by
Gilboa et al. (2004). The initial condition is the
noisy image and after certain number of iterations
the de-noised image is evolved according to Equa-
tion (44). From Equation (44) it can be observed
that evolution of real part of the image is controlled
of imaginary part of the image is controlled by
both the real and imaginary equations. Qualitative
properties of edge detection i.e. second smoothed
derivative is described by the imaginary part of
the image for small value of θ,whereas real values
depicts the properties of ordinary Gaussian scale
-space. For large values of θ, the imaginary part
feeds back in to the real part creating the wave
like ringing effect which is an undesirable prop-
erty. Here, for experimentation purposes value of
π
θ is chosen to be .
30
For nonlinear complex diffusion, the diffu-
sion coefficient is defined as follows (Gilboa et
al., 2004):
e iθ
c(Im(I )) = . (45)
2
 Im(I )
1 +  
 k θ 
585

Here k is the edge threshold parameter. The
value of k ranges from 1 to 1.5 for digital images
having additive noise. The value of k is fine tuned
according to the application in hand.
Other PDE-Based Models
Osher and Rudin (1990) proposed an algorithm
for image enhancement based on shock filters. In
this case the image ϕ(x, y, t) evolves according to
∂ϕ
= − ∇ϕ F (L(ϕ)) , (46)
∂t
where function F(u) satisfies certain technical
conditions and L is a 2nd order nonlinear elliptic
operator. An image evolving according to above
equation develops shocks where L=0. Main goal
of this method is to get as close as possible to the
inverse heat equation.
Alvarez et al. (1992) proposed an algorithm of
image selective smoothing and edge detection. In
this case the image evolves according to
∂ϕ ∇ϕ
= g( G ∗ ∇ϕ ) ∇ϕ div( ), (47)
∂t ∇ϕ
where G is a smoothing kernel (e.g. Gaussian)
and g () is a non increasing function which tends
to zero as t tends to infinity.
In paper (Beck, &Teboulle, 2009), authors
studied gradient-based schemes for image denois-
ing and de-blurring problems based on the dis-
cretized total variation (TV) minimization model
with constraints. They derived a fast algorithm
for the constrained TV-based image de-burring
problem. To achieve this task, they combined an
acceleration of the well known dual approach to
the de-noising problem with a novel monotone
version of a fast iterative shrinkage/ thresholding
algorithm (FISTA). The resulting gradient-based
algorithm shares a remarkable simplicity together
with a proven global rate of convergence which
586
PDE-Based Image Processing
is significantly better than currently known gra-
dient projections-based methods. The results are
applicable to both the anisotropic and isotropic
discretized TV functional. Initial numerical re-
sults demonstrate the viability and efficiency of
the proposed algorithms on image de-blurring
problems with box constraints. Image de-noising
methods based on gradient dependent regularizes
such as Rudin et al.’s total variation (TV) model
often suffer the staircase effect and the loss of fine
details. In order to overcome such drawbacks,
authors of the paper (Chen et al.,2009) presents
an adaptive total variation method based on a new
edge indicator, named difference curvature, which
can effectively distinguish between edges and
ramps. In another recent paper (Chang, & Yang,
2009) authors have proposed a Lattice Boltzmann
method for image de-noising.
Some other PDE-based model includes: Im-
age Restoration from motion blurred image using
PDEs formalism (Srivastava et al., 2009); PDE
based Unsharp Masking, Crispening and High
Boost Filtering of Digital Images (Srivastava et
al., 2009); and Brownian Motion based image
modelling and its applications to image regular-
ization (Srivastava et al., 2009).
Discretization of Models for
Digital Implementations
For digital implementations the PDE based mod-
els discussed in above section can be discretized
using finite differences schemes (Press et al.,1992)
i.e. Finite Time Centered Scheme (FTCS), where
we consider equally spaced points along both the
t-axes and x-axes and denote x j = x 0 + j ∆x ,
j=0,1,2,….,m and tn = t0 + n ∆t , n=0,1,2,….,n
where m×n is the size of image.
The discretized form of anisotropic diffusion
based model, given by Equation (35), reads
 
I n +1 (x , y ) = I n (x , y ) + ∆t.(∇.c∇I ) (48)
PDE-Based Image Processing
The second term in R.H.S. of Equation (48)
can be further discretized using centred difference
scheme as proposed in (Perona, & Malik, 1990).
The discretized form of the fourth order PDE
based model, given by Equation (38), reads
n +1 n has been implemented in MATLAB. The both
I (x , y ) = I (x , y ) + ∆t .
(49)
2 n 2 n
[div(c( ∇ I (x , y ) )∇ I (x , y ))]
The discretized form of the nonlinear complex
diffusion based model given by Equations (40)
and (45) reads
I tn++11 (i, j ) =
( )
I tn (i, j ) + ∆t.(∇. c(Im(I tn (i, j ))∇I nt (i, j )
(50)
For the numerical scheme, given by above
Equations (49-50) to be stable, the von Neumann
analysis (Press et al.,1992), shows that we require
∆t
2
< ¼. If the grid size is set to ∆x = 1 , then
(∆x )
∆t < ¼ i.e. ∆t < 0.25. Therefore, the value of
∆t is set to 0.24 for stability of Equations (48-50).
Figure 9. Results of anisotropic diffusion after 35 iterations for noise variance 0.002 for sample image
lena.jpg (512x512)
RESULTS AND ANALYSIS
Case I: Anisotropic
Diffusion Based Model
The de-noising scheme defined by Equation (35)
choices of diffusion coefficients (c1 and c2) as
defined in Equation (36) have been used. The
value of ∆t was set to 0.24 and value of conduc-
tivity coefficient k was set to 50 and k may attain
any value in between 20-100 (Perona and Malik,
1990). The performance of the anisotropic diffu-
sion method for image de-noising defined by
Equation (35) has been evaluated both qualita-
tively and quantitatively in terms of various
performance measurement metrics such as MSE,
RMSE, NMSE, SNR, PSNR and time for differ-
ent gray images of different resolutions for vary-
ing amount of Gaussian noise variance. The mean
of noise is considered to be zero. From experi-
mentation it has been tested that the scheme defined
by Equation (35) converges to solution at 30-35
iterations for a noise variance of 0.02 for the
sample image in consideration. For less variance
we obtain the processed de-noised image after
20-25 iterations. Figure 9 shows the visual results.
Performance comparisons are shown in Figures
10-11 for both choices of diffusion coefficients.
The execution time is less for option 2 of diffusion
coefficient (Figure 11). NMSE is less and SNR
587
 PDE-Based Image Processing

Figure 10. Performance comparison of anisotropic diffusion: SNR vs noise variance; and NMSE vs noise
variance for both choices of conductivity coeffs (c1 and c2) for different iterations for the image lena.jpg

Figure 11. Performance comparison, Execution time vs noise variance for both choices of conductivity
coeffs (c1 and c2) for different iterations for the image lena.jpg

588
PDE-Based Image Processing

Sample MATLAB code: anisotropic diffusion based model

%K is the conductivity coefficient that varies between 20-100; I is the input

image ; and niter is the number of iterations required for the PDE to converge
to the solution.
function [Ianiso]=PMAnisoDiff (I,k,niter)
It=double(I);
dt=0.25;
t=1;
[x y]=size(I);
epsilon=0.0000000001; % epsilon is added to avoid divide by zero
% k=20-100
for t = 1:niter
[Ix,Iy]=gradient(It);
di=sqrt(Ix.^2+Iy.^2+epsilon);% norm (I)
c=1./(1.+(di./k).^2);
F=(c.*gradient(It));% F=delI/norm(DelI)
Fdiv=gradient(F);% divergence of F
It=It+dt.*Fdiv;
end
Ianiso=It;
End

is high for option 2 of diffusion coefficient (Fig-
ure 10). Hence in terms of NMSE, SNR and ex-
ecution time option 2 of diffusion coefficient is
performing better.
Case II: Complex Diffusion Based
Model
See Figure 12 and Figure 13 for details about
Case II and the complex diffusion based model.
Case III: Comparative Study of PDE-
Based Filters for Additive Noise
Removal
In this section, the comparative studies of four
PDE based filters in consideration are presented
in terms of MSE, PSNR, CP, and MSSIM (see
Table 1). The sample image is lena.jpg of size
200x200. Numbers of iterations for TV-based
model were chosen to 100 for acceptable quality

Figure 12. Original image (Left); Real part of the image (Middle); Imaginary part of the image show-
ing edges (Right)

589
 PDE-Based Image Processing

Figure 13. Results of linear complex diffusion based filter, original noisy microscopic image (left),
restored real image (right)

of results. For anisotropic diffusion based PDE, MULTIPLICATIVE SPECKLE NOISE

fourth order PDE, and nonlinear complex diffu-
sion based PDE, the number of iterations were
set to 50 for the acceptable quality of images.
For more than 50 iterations the performance
decreases. The value of diffusion coefficient k
for anisotropic diffusion based PDE, and fourth
order PDE were set to 20 and that of nonlinear
complex diffusion was set to 1.5. Figures 14 and
15 give performance comparison of the schemes
in consideration.
REDUCTION FROM DIGITAL IMAGES
Speckle Noise: Theoretical
Background
This section describes the general speckle noise
patterns present in various digital images such
as ultrasound images and synthetic aperture ra-
dar (SAR) images. The speckle noise occurs in
coherent imaging of objects whenever surface
roughness of the image being imaged is of the

Table 1. Performance comparison of additive noise removal techniques (Sample image: lena.jpg, (200x200)

Method Noise Variance MSE PSNR [dB] CP MSSIM

Total Variation (TV) 0.0001 290.53 23.499 0.9132 0.5105
0.0025 290.11 23.517 0.91604 0.50941
0.0049 289.63 23.472 0.91668 0.510130.5106
0.0073 294.77 23.436 0.91552 0.51385
0.0097 291.08 23.491 0.91862
Anisotropic Diffusion 0.0001 339.66 22.82 0.93356 0.49915
0.0025 343 22.778 0.93562 0.49918
0.0049 339.82 22.818 0.93311 0.502
0.0073 346.65 22.732 0.93371 0.49723
0.0097 351.63 22.67 0.93374 0.49347
Fourth Order PDE 0.0001 286.74 23.556 0.93839 0.57417
0.0025 287.2 23.549 0.93959 0.57937
0.0049 291.24 23.488 0.94005 0.57761
0.0073 295.06 23.432 0.9399 0.5762
0.0097 300.73 23.349 0.93981 0.57484
Nonlinear Complex Diffusion 0.0001 260.21 23.978 0.935 0.58983
0.0025 259.4 23.991 0.93492 0.58979
0.0049 254.73 24.07 0.93719 0.59331
0.0073 260.33 23.976 0.93478 0.58851
0.0097 263.46 23.924 0.93466 0.58666

590
PDE-Based Image Processing

Figure 14. Performance comparison of additive noise removal filters; MSE(left), and PSNR (right)

Figure 15. Performance comparison of additive noise removal filters; CP(left), and MSSIM (right)

591

order of the wavelength of the incident radiation
(Jain, 2006). The presence of speckle noise in
an imaging system reduces its resolution; espe-
cially for low contrast images and suppression
of speckle noise is an important consideration in
the design of coherent imaging systems. For low
contrast images speckle noise is multiplicative in
nature. Speckle noise may be present in ultrasound
medical images, synthetic aperture radar (SAR)
imagery and digital holographic images and in
other application images. An inherent character-
istic of coherent imaging, including ultrasound
imaging, is the presence of speckle noise which
is a random, deterministic, interference pattern
in an image formed with coherent radiation of a
medium containing many sub-resolution scatter-
ers. The texture of the observed speckle pattern
does not correspond to underlying structure. The
local brightness of the speckle pattern, however,
does reflect the local echogenicity of the underly-
ing scatterers. Speckle has a negative impact on
ultrasound imaging and other types of digital imag-
ing e.g. SAR (Synthetic Aperture Radar) imaging
and digital holography (Srivastava et al., 2009).
Speckle Noise in Ultrasound Images
Speckle noise pattern present in ultrasound im-
ages are multiplicative in nature and distributed
according to Rayleigh’s probability distribution
function. In ultrasound imaging, speckle leads
to reduction in contrast resolution which is re-
sponsible for the poorer effective resolution of
ultrasound compared to x-ray and MRI. Speckle
is present in both RF data and envelope-detected
data. Figure 14 conceptually demonstrates the
impact of speckle noise on information content
(Anderson, 2009). The ultrasound image of a
hypoechoic lesion of 5 mm diameter with -9 dB
contrast is considered. The echogenicity map
corresponding to this object is shown in the top
left panel, opposite the corresponding scattering
function in the top right panel. The scattering func-
tion represents the population of sub-resolution
592
PDE-Based Image Processing
scatterers being imaged, and that are weighted in
amplitude by the echogenicity map. This scattering
function convolved with a point spread function
to produce the RF echo data is shown in the lower
left panel. The RF echo data is zero-mean and thus
does not show what is really of interest, i.e. a map
of local echogenicity, or local echo magnitude.
Envelope detection removes the carrier, producing
the desired image of echo magnitude in the lower
right panel. The differences between this image
and the original echogenicity map arise from
speckle noise. The statistics of fully-developed
speckle (Anderson, 2009; Goodman, 1984) can
be described as follows: For the given stochastic
nature of speckle noise, this noise pattern can be
described statistically. Each of the diffuse scat-
terers in the isochronous volume contributes a
component to the echo signal in a sum is known
as a random walk in the complex plane. If each
step in this walk is considered an independent
random variable, over many such walks we can
apply the Central Limit Theorem to their sum.
Therefore, in fully developed speckle, this com-
plex radio-frequency echo signal from diffuse
scatterers alone has a zero mean, two-dimensional
Gaussian probability density function (PDF) in the
complex plane. The values of the magnitude of
r for many such scatterer populations follow the
Rayleigh PDF. Envelope detection removes the
phase component, creating a signal with Rayleigh
amplitude PDF (Anderson, 2009; Goodman, 1984)
which reads
a a2
pA (a ) = exp(− ), a ≥ 0 (51)
σ2 2σ 2
where a and σ 2 are amplitude and variance of the
back scatterer signal. Speckle brightness is
greater if there are fewer, longer steps in the
random walk than if there are many shorter steps.
This could be accomplished by improving the
spatial resolution of the system. If the scatterer
density is doubled, a 2 increase in brightness
PDE-Based Image Processing

results. When a coherent component is introduced
to the speckle, it adds a constant strong phasor to
the diffuse scatterer’s echoes and shifts the mean
of the complex echo signal away from the origin
in the complex plane. Due to this the Rayleigh
PDF changes into a Rician PDF. The Rician PDF
is defined by the following equation:
a a2 + s2 as
pA (a ) = exp(− )I 0 2 , a ≥ 0
σ 2
2σ 2
σ Gamma distribution and is widely used in optics
(52)
These PDFs are nonzero for a >= 0 only. The
parameter s is the echo strength of the bright scat-
terer, while σ is the standard deviation of the
complex Gaussian described above, i.e. both the
real part and the imaginary part have variances
of σ . I0 is the incomplete Bessel function of zero
order. The Rician PDF is parameterized by the
s
variable k, which is defined as . The Rician
σ
PDF reduces to the Rayleigh PDF for the special
case s = 0. The speckle noise in ultrasound imag-
ing is also approximated as multiplicative noise,
given by following model:
g(x, y)= I(x, y∙ s(x, y), (53)
Here, g(x, y) is the recorded ultrasound image,
I(x, y) is the original ultrasound image and s(x,
y) is the multiplicative speckle noise present in
ultrasound image.
Speckle Noise Pattern
in SAR Imagery
In synthetic aperture radar (SAR) images, the
speckle noise is generally caused by the decline
of echo signals. The image observation model for
SAR imagery is given by
where I0 is the observed SAR image, I is the
original image to be estimated or restored using
some filtering technique and n is the multiplicative
speckle noise. The image I can also be considered
as the radar scattering characteristics of a random
ground object. The speckle noise in a SAR image
follows the k-distribution (Iskander 1999) i.e.
radar clutter models are based on k-distribution.
k-distribution which is also known as Rayleigh-
and radar signal processing for modelling laser
speckles and sea echoes and is found to be use-
ful in describing the statistics of the echo. The
statistics of a k-distributed random variable X are
described by the following distribution function
(Iskander 1999; Iskander et al. 1999):
2  x 
v +1
x 
FX (x ) = 1 −  
 K v +1   ,
a Γ(v + 1)  2a  a 
(55)
where x > 0 and v > -1.
The probability distribution function of the
k-distribution is defined as:
v +1
2  x  x 
fX (x ) =  
 K v   , (56)
a Γ(v + 1)  2a  a 
where x > 0 and v > -1.
Here Γ(.) is the gamma function, K v (.) is
the modified Bessel function of order v and a is
a positive constant. The k-distribution is com-
pletely described by the shape parameter v and
the scale parameter a, which can be estimated
using method of moments or maximum likelihood
methods. If the observed image is I0 and the es-
timated or filtered image is I then the pdf given
by Equation (56) reads

I0=I∙n, (54) v +1
2  I  I 
p(I / I ) = f (I ) = 
  Kv   .
0 I 0 a Γ(v + 1)  2a  a 
(57)

593

Since the acquisition noise on the SAR im-
age is not Gaussian, a least square fit may not
the best choice for such an estimation process.
Furthermore, independent pixel estimation does
not reflect the spatial regularity of the diffusion
function. Hence, it is proposed to tackle these
issues within a variational framework which is
adapted to the noise distribution pattern in the
image and is able to use valuable information
given by the neighbour pixels.
Similarly, speckle noise is also present in
digitally reconstructed holographic images which
are multiplicative in nature and can be reduced
by PDE-based image processing (Srivastava et.
al., 2009).
SPECKLE NOISE REDUCTION
TECHNIQUES
General Approaches
In general the speckle noise (Jain, 2006) has
complex amplitude given as
a(x , y ) = aR (x , y ) + jaI (x , y ) , where aR and aI
are zero mean, independent Gaussian random
variables for each (x, y) with some variance. The
intensity field of speckle noise is given as
2 of the terrain backscatter as well as the nature of
s(x , y ) = a(x , y ) = a 2 + a 2 . (58)
R I
The general image observation model for
speckle noise reads
v(x , y ) = u(x , y )s(x , y ) + η(x , y ) (59)
where v(x, y) is the observed noisy image, u(x, y)
is the original image to be restored, s(x, y)is the
intensity of speckle noise and 𝜂(x, y)is detector
noise which is additive in nature. Assuming de-
tector noise to be zero, the general observation
model for speckle noise reads
594
PDE-Based Image Processing
v(x , y ) = u(x , y )s(x , y ) . (60)
The speckle noise model given by Equa-
tion (60) is a multiplicative noise model. This
multiplicative noise model can be converted to
additive noise by using homomorphic filtering
approach which is easier to implement. Other,
effective filters may also be used for the reduction
of speckle noise.
The speckle noise can be reduced by multi-
look processing, spatial filtering or homomorphic
filtering (Jain, 2006). The multi-look processing
is usually done during data acquisition stage and
speckle reduction by spatial and homomorphic
filtering is performed on the image after its acqui-
sition. Irrespective of the methods used to reduce
the speckle noise from images, the ideal speckle
reduction method must preserve radiometric
information and the textural information i.e. the
edges between different areas and spatial signal
variability. The spatial filters are of two types i.e.
adaptive and non-adaptive. Non-adaptive filters
take the parameters of the whole image signal into
consideration and leave out the local properties of
the terrain backscatter or the nature of the sensor.
These kinds of filters are not appropriate for non-
stationary scene signal. Fast Fourier Transform
(FFT) is an example of such filters. The adaptive
filters accommodate changes in local properties
the sensor. In adaptive filters, the speckle noise
is considered as being stationary but the changes
in the mean backscatters due to changes in the
type of target are taken into consideration. Adap-
tive filters reduce speckles while preserving the
edges and these filters modify the image based
on statistics extracted from the local environment
of each pixel. Adaptive filter varies the contrast
stretch for each pixel depending upon the Digital
Number (DN) values in the surrounding moving
kernel. A filter that adapts the stretch to the re-
gion of interest produces a better enhancement.
Examples of adaptive filters are: Mean, median,
Wiener filter (Jain, 2006), Lee filter (Lee, 1983),
PDE-Based Image Processing

Frost filter (Frost et al., 1982), Kuan filter (Kuan where mean = average of pixels in a moving
et al., 1987), Hybrid filter (Rajan et al., 2008) and window,
Speckle Reducing Anisotropic Diffusion (SRAD)
based filter (Yu, &Acton, 2002). In a recent paper var(x )
k= , (62)
(Juan et al., 2009), the authors investigate and (mean )2 σ 2 + var(x )
compile some of the techniques mostly used in
the smoothing or suppression of speckle noise in
ultrasound images. The basic concepts of these and
adaptive filters are as follows:
 σ + µ2 
var(x ) =  w 2 w 
 − µw .
2
(63)
Mean Filter  σ + 1 

The Mean Filter is a simple one and does not

remove the speckles but averages it into the data
and it is the least satisfactory method of speckle
noise reduction as it results in loss of detail and
resolution. However, it can be used for applica-
tions where resolution is not the first concern.
Median Filter
The Median filter is also a simple one and removes
pulse or spike noises. Pulse functions of less than
one-half of the moving kernel width are suppressed
or eliminated but step functions or ramp functions
are retained.
Lee-Sigma and Lee Filters
The Lee-Sigma and Lee filters (Lee, 1981; Lee,
1983) utilize the statistical distribution of the DN
values within the moving kernel to estimate the
value of the pixel of interest. These two filters
assume a Gaussian distribution for the noise in
the image data. The Lee filter is based on the as-
sumption that the mean and variance of the pixel of
interest is equal to the local mean and variance of
all pixels within the user-selected moving kernel.
The scheme for computing digital number output
(DNout) is as follows:
(DNout)=[mean]+k[(DNin−mean] (61)
µw and σw are the mean and variances of pixels
within chosen window. The Sigma filter is based
on the probability of a Gaussian distribution. It
is assumed that 95.5% of random samples are
within a two standard deviation range. This noise
suppression filter replaces the pixel of interest
with the average of all DN values within the mov-
ing kernel that fall within the designated range.
Frost Filter
The Frost filter (Frost et al., 1982) replaces the
pixel of interest with a weighted sum of the values
within the n × n moving kernel. The weighting
factors decrease with distance from the pixel of
interest. The weighting factors increase for the
central pixels as variance within the kernel in-
creases. This filter assumes multiplicative noise
and stationary noise statistics and follows the
following formula:
−α t
DN = ∑ k αe , (64)
n×n
4 σ2
where α = ( )() . (65)
nσ2 I 2
Here k = normalization constant, I = local
mean, σ =local variance, σ = image coefficient
of variation value, t = X − X 0 + Y −Y0 , and
n = moving kernel size.

595

Kuan Filter
Kuan filter (Kuan et al., 1987) first transforms the
multiplicative noise model into a signal-dependent
additive noise model. Then the minimum mean
square error criterion is applied to the model. The
resulting filter has the same form as the Lee filter
but with a different weighting function. Because
Kuan filter made no approximation to the original
model, it can be considered to be superior to the
Lee filter.
The resulting grey-level value R for the
smoothed pixel is:
R = I c *W + I m * (1 −W ) , (66)
where:
W = (1 − C u2 C i2 ) / (1 + C u2 )
1
Cu =
NumberofLooks
S provides the derivation of speckle reducing
Ci = .
Im
Ic= center pixel in filter window, Im = mean
value of intensity within window, and S = standard
deviation of intensity within window.
The Kuan filter is used primarily to filter
speckled radar data. It is designed to smooth out
noise while retaining edges or shape features in
the image. Different filter sizes will greatly affect
the quality of processed images. If the filter is too
small, the noise filtering algorithm is not effective.
If the filter is too large, subtle details of the image
will be lost in the filtering process. A 7x7 filter
usually gives the best results. The NumberofLooks
parameter is used to estimate noise variance and
it effectively controls the amount of smoothing
applied to the image by the filter. Theoretically,
596
PDE-Based Image Processing
the correct value for NumberofLooks should be
the effective number of looks of the radar image.
It should be close to the actual number of looks,
but may be different if the image has undergone
re-sampling. The user may experimentally adjust
the NumberofLooks value so as to control the ef-
fect of the filter. A smaller NumberofLooks value
leads to more smoothing; a larger NumberofLooks
value preserves more image features.
PDE Based Filters
In recent years, several PDE based methods have
been developed for removal of additive noise from
images (Perona, &Malik, 1990; Gilboa et al.,
2004; You, & Kaveh, 2004) which can be used by
homomorphic filters to reduce speckle noise. The
basic idea behind PDE based noise removal based
on energy minimization technique as discussed in
section image restoration using PDEs formalism.
Speckle Reducing Anisotropic
Diffusion (SRAD) Filter
In this paper (Yu, &Acton, 2002), the authors
anisotropic diffusion (SRAD), a diffusion method
tailored to ultrasonic and radar imaging applica-
tions. SRAD is the edge-sensitive diffusion for
speckled images, in the same way that conven-
tional anisotropic diffusion is the edge-sensitive
diffusion for images corrupted with additive noise.
At first authors had shown that the Lee and Frost
filters can be cast as partial differential equations,
and then SRAD filter is derived by allowing edge-
sensitive anisotropic diffusion within this context.
SRAD exploits the instantaneous coefficient of
variation, same as the Lee and Frost filters utilize
the coefficient of variation in adaptive filtering.
The instantaneous coefficient of variation is a
function of the local gradient magnitude and
Laplacian operators.
PDE-Based Image Processing

PDE-Based Homomorphic Filtering 3. Finally, the restored holographic image,

Approach for Speckle Reduction Irestored, can be obtained by taking the expo-
nentiation of output obtained in step ii.
Based on the discussions in section Speckle
noise: Theoretical background, it can be said that I restored = exp(I (x , y )). (69)
the speckle noise is multiplicative in nature for
ultrasound images and SAR images. Assuming
detector noise to be zero, the general observation
EXAMPLES
model for speckle noise reads
Non-Linear Complex Diffusion Based
v(x, y)=u(x, y)s(x, y). (67)
Homomorphic Filter for Speckle
Since, the direct reduction of multiplicative
Reduction from Digital Images
noise is a difficult task therefore the multiplica-
In this section (Srivastava et al., 2009), the nonlin-
tive noise is first converted in to additive noise
ear complex diffusion based homomorphic filter
by taking the logarithm of equation (67) i.e. by
is proposed to be used for speckle reduction from
applying homomorphic transform and reduction
ultrasound and synthetic aperture radar (SAR) im-
of this additive noise by some available filters and
ages. The performance of the nonlinear complex
finally taking the exponentiation of filter output
diffusion based homomorphic filter is evaluated
to produce the speckle reduced image.
in terms of MSE and PSNR and a comparative
The homomorphic filtering approach for
study of this scheme with other standard speckle
speckle reduction can be described as (Srivastava
reduction techniques such as Lee filter, Frost fil-
et al., 2009):
ter, Kuan filter and SRAD filter is presented. The
obtained results show that the nonlinear complex
1. Apply the logarithmic transform on equa-
diffusion based homomorphic filter outperforms
tion(67) to convert the multiplicative noise
all schemes in consideration.
in to additive one which reads

log v(x , y ) = log u(x , y ) + log s(x , y ) Methods and Models

(68)
⇒ w(x , y ) = I (x , y ) + ηs (x , y )
As discussed earlier, the nonlinear complex dif-
fusion based PDE reads
Where w(x , y ) = log v(x , y ) is the observed
hologram image in log domain, ∂I → →
I (x , y ) = log u(x , y ) is the noiseless image = ∇ .(D(Im(I ) ∇ I )) . (70)
∂t
in log domain that is to be recovered and
ηs (x , y ) = log s(x , y ) the amount the of the
speckle noise which is now an additive noise Expressing the above PDE in its tangent and
and is to be minimized. normal forms, the presented model reads
2. In this step, an additive noise removal filter
(e.g. Wiener filter, median filter, PDE based ∂I
= ∇2I (ξ, ξ ) + D(Im(I ))∇2I (η, η) (71a)
diffusion filters and other filters) is applied ∂t
to remove or minimize the additive noise
ηs (x , y ) .

597

I (x , y, 0) = I 0 (Initial Condition) (71b)
The first term in equation (71a) diffuses the
image, I, only in the orthogonal direction of the
gradient and the second term diffuses the image,
I, in the direction of gradient. The terms ξ and
η are unit vectors respectively orthogonal and in
the direction of the gradient vector ∇I of the
image I (t, ) .These unit vectors are defined
(Perona, 1998) as follows
∇I ⊥ I y , −I x
ξ= = and
∇I I x2 + I y2
∇I I x , Iy
η= = (72)
∇I I x2 + I y2
The term D(Im(I ) is the complex diffusion
coefficient and Im(I ) is the imaginary part of the
image. The diffusion coefficient D(Im(I ) is de-
fined as follows (Gilboa et al.,2004)
iθ
D(Im(I )) = e (73)
 Im(I )

1 +  
 λθ 
Here λ is a threshold parameter and θ is the
small phase angle(θ << 1) . The imaginary value
of image I i.e. Im(I ) divided by θ is used for
controlling the diffusion process. For small t, this
term vanishes and allows stronger diffusion to
reduce the noise; with time its influence increas-
es and preserves the ramp features of the image.
The function D(Im(I ) , known as conductivity
coefficient, is a function of local image differen-
tial structure that depends on local partial deriva-
tives (Gilboa et al., 2004). The conductivity coef-
ficient controls the diffusion process within the
image that can be varied according to image
structure i.e lower the diffusion where edges oc-
curs to preserve the edges and enhance the diffu-
598
PDE-Based Image Processing
sion in smooth region of image. The function D(.)
is chosen with values between 0 an 1 according
to Equation (73). When D(.) = 1, the two diffusion
terms in Equation (71a) have equal weights and
the Equation (71a) converts to a 2D isotropic heat
equation i.e. Laplacian. When D(.) = 0, only the
diffusion in the orthogonal direction of the gradi-
ent operates. In homogeneous zones of the image,
g (.) will have values near 1.
For simplicity of representation, suppose
I ξξ = ∇2I (ξ, ξ )represents the diffusion along the
normal of the edge and I ηη = ∇2I (η, η) represents
the diffusion along the tangent of the edge. The
diffusion along the normal diffuses across the
edges and diffusion along the tangent continues
along the edges. The blurring of edges of image
or area of high frequency content is mainly due
to diffusion along the normal. To control the dif-
fusion the diffusion across the edges, I ηη , is
lowered in comparison to the diffusion along the
tangent and this is achieved by varying the values
of the diffusion coefficient D(Im(I )) according
to Equation (73). Further, I ξξ and I ηη are defined
(Perona, 1998) as follows-
I xx I x2 − 2 ∗ I xy I x I y + I yy I y2
I = (74)
ξξ
I x2 + I y2
I xx I x2 + 2 ∗ I xy I x I y + I yy I y2
I ηη = .
I x2 + I y2
(75)
Discretization of the Model
For digital implementations, the Equation (71) can
be discretized using forward time centered scheme
(FTCS) and centered difference scheme. Here, we
consider equally spaced points along both the t-
axes and x-axes and denote x j = x 0 + j ∆x ,
j=0,1,2,….,m and tn = t0 + n ∆t , n=0,1,2,….,n
PDE-Based Image Processing
where m×n is the size of image. Using the FTCS,
the discrete form of Equation (70) reads
I in, j+1 = I in, j + ∆t.[∇2I (ξ, ξ ) + D(Im(I ))∇2I (η, η)]
i.e.
I in, j+1 = I in, j + ∆t.[I + D(Im(I ))I ηη ] .
ξξ
(76a)
I (x , y, 0) = I 0 (Initial Condition) (76b)
For discretizing I ξξ and I ηη as defined in Equa-
tion (74-75), centred difference scheme have been
used. For the numerical scheme given by Equation
(75) to be stable, the von Neumann analysis (Press
∆t
et al.,1992) shows that we require < ¼.
(∆x )2
If the grid size is set to ∆x = 1 , then ∆t < ¼ i.e.
∆t < 0.25. Hence, ∆t = 0 − 0.25 for stability
of Equation (76) during implementation. The
basis for selection of the stopping time is that
running the diffusion equation on an image I at
scale t is equivalent to convolution with Gaussian
kernel of standard deviation 2t .
The final proposed algorithm reads as follows.
Algorithm- Complex Diffusion
Based Homomorphic Filtering
1. In first step, logarithm transformation is ap-
plied on input given by equation (67) which
reads
log v(x , y ) = log u(x , y ) + log s(x , y )
⇒ w(x , y ) = I (x , y ) + η(x , y )
2. In second step, input image in log domain
is processed using complex diffusion based
filter to remove the additive noise. For niter
= 1 to n
win, j+1 = win, j + ∆t.[w + D(Im(w ))w ηη ] ,
ξξ
with w(x , y, 0) = w
End
3. Finally the speckle reduced image, u, is
obtained by exponentiation of the output w
of step (2) i.e. u = ew .
RESULTS AND PERFORMANCE
COMPARISON
Test Case I: Speckle Reduction
from Ultrasound Images
The first category of images considered for speckle
noise reduction is ultrasound images. The proposed
algorithm was applied on several ultrasound
images and visual results for one sample image
Gyn_normal_intrauterine_IUD.jpg (465x589)
for varying amount of speckle noise variance
e.g. 0.001, 0.002, 0.004, 0.01, 0.02 and 0.04 are
shown in Figure15 for speckle noise variance
0.04. The performance of the proposed scheme is
also compared with other state of the arts methods
available in literature for speckle reduction as
shown in Table 2 for test image Gyn_normal_in-
trauterine_IUD.jpg. Figure16 shows performance
comparison of various speckle reduction methods
with the proposed scheme for the sample image
Gyn_normal_intrauterine_IUD.jpg (465x589) in
terms of speckle variance Vs RMSE and speckle
variance Vs PSNR and respectively. From Fig-
ure16 it can be observed that the proposed homo-
morphic complex diffusion process for speckle
reduction is associated with minimum RMSE
and maximum PSNR in comparison to other
techniques reported in literature for the sample
ultrasound image taken in to consideration and
the performance trend remained same for other
images too. Therefore, the results show that the
proposed complex diffusion based homomorphic
599
 PDE-Based Image Processing

Table 2. Performance comparison of speckle reduction methods (Sample ultrasound image: Gyn_nor-
mal_intrauterine_IUD.jpg, (465x589)

Method Speckle Variance RMSE PSNR [dB]

Lee Filter 0.001 4.0559 35.9691
0.002 4.3344 35.3922
0.004 4.8315 34.4491
0.01 5.9573 32.6298
0.02 7.0466 31.1712
0.04 7.9918 30.0779
Frost Filter 0.001 6.1928 32.2930
0.002 6.8972 31.3573
0.004 8.1355 29.9231
0.01 11.0139 27.2920
0.02 14.6570 24.8099
0.04 19.8619 22.1704
Kuan Filter 0.001 4.0580 35.9646
0.002 4.3325 35.3961
0.004 4.8404 34.4331
0.01 5.9526 32.6367
0.02 7.0544 31.1616
0.04 7.9911 30.0787
Hybrid Filter 0.001 7.0480 31.1695
(After 02 iterations) 0.002 7.5056 30.6231
0.004 8.3987 29.6465
0.01 10.5484 27.6671
0.02 17.4995 23.2703
0.04 17.5058 23.2672
SRAD (After 05 iterations) 0.001 9.0476 29.0001
(Anisotropic diffusion based 0.002 9.0810 28.9681
filter) 0.004 11.6356 26.8151
0.01 10.9302 27.3583
0.02 12.9925 25.8570
0.04 16.0786 24.0059
Proposed Method (After 10 0.001 3.3114 37.7305
iterations) 0.002 3.3804 37.5516
0.004 3.5280 37.1801
0.01 3.8814 36.3509
0.02 4.2837 35.4945
0.04 4.9452 34.2471

filtering scheme performs better than other exist-
ing scheme for ultrasound images.
Test Case II: Speckle Reduction from
SAR Images
The second category of images considered for
speckle noise reduction is synthetic aperture radar
(SAR) images. The proposed algorithm was ap-
plied on several SAR images and results for one
sample image, SAR.gif (516x416) for varying
amount of speckle noise variance e.g. 0.001, 0.002,
0.004, 0.01, 0.02 and 0.04 are presented here.
Figure17 shows the visual results for the sample
image for speckle noise variance 0.04. Figure18
show performance comparison of various speckle
reduction methods with the proposed scheme in
terms of speckle variance Vs RMSE and speckle
variance Vs PSNR, respectively. From Figure18 it
can be observed that the proposed homomorphic
complex diffusion process for speckle reduction
is associated with minimum RMSE and maxi-
mum PSNR in comparison to other techniques in
consideration for the sample SAR image. Also,

600
PDE-Based Image Processing

Figure 16. Impact of speckle noise on information content for ultrasound images

this performance trend remained same for other
test images. Therefore, the results show that the
proposed complex diffusion based homomorphic
filtering scheme performs better than other exist-
ing scheme for SAR images too.
Therefore, from above results and discussions
it can be concluded that the nonlinear complex
diffusion based filter performs better in com-
parison to other speckle reduction methods for
both category of images. This method also per-
forms better for digital holographic images (Sriv-
astava et al., 2009).
FUTURE RESEARCH DIRECTIONS
The various possible areas where PDE-based
image modelling techniques can be applied

Figure 17. Original ultrasound image Gyn_normal_intrauterine_IUD.jpg (465x589) (top-left); Speckled

image with variance 0.04 (top-right); De-speckled image using proposed scheme (bottom)

601
 PDE-Based Image Processing

Figure 18. Performance comparison of speckle reduction methods: Speckle Variance Vs RMSE (left),
Speckle Variance Vs PSNR(right) for sample ultrasound image: Gyn_normal_intrauterine_IUD.jpg,
465x589

include: Scale-space theory, image restoration,
image sharpening, contrast enhancement, image
analysis and feature extraction, segmentation,
video processing, shape analysis, axiomatic
and theoretical approaches to image analysis,
variational and PDE’s-from-energy approaches
to image processing, curve and surface evolution,
PDE’s in hybrid systems, comparison of models,
implementations and numerical analysis, math-
ematical morphology, object detection, optical
flow, image compression, shape from shading, 3D
reconstruction of views from 2D images, pattern
recognition and vision related applications.
PDE-based image modeling techniques can
play an important role in analyzing the images
obtained from various application areas such as:

Figure 19. Original SAR image (Courtesy ESA-ESRIN), SAR.gif (top-left); Speckled image with variance
0.04 (top-right); De-speckled image using proposed complex diffusion based homomorphic filter (bottom)

602
PDE-Based Image Processing

Figure 20. Performance comparison of speckle reduction methods: Speckle Variance Vs RMSE (left),
Speckle Variance Vs PSNR (right) for sample image: SAR.gif, 516x416

Medical Imaging • Biometrics: Some open problems in this

domain include development of efficient
The efficient and accurate PDE-based image PDE-based models for segmentation, fin-
processing tools can be developed which can aid ger prints enhancement & restoration, fea-
in to accurate diagnosis of the disease. The some ture extraction, face detection, identifica-
problems of medical imaging that needs further tion etc.
attention include: • Optical Information Processing:
Restoration of optical images, speckle re-
• Medical image enhancement and duction and restoration of 3D holograms
de-noising and image processing of microscopic
• Rician noise removal from MRI images images.
• Speckle reduction from optical coherence • Other application areas: where the PDE-
tomography (OCT) images based image modeling can be applied are:
• Segmentation of an object of interest e.g. ◦◦ Biosignal processing (ECG etc)
Segmentation for Alzheimer’s disease de- ◦◦ Confocal Microscopy
tection, segmentation of Iris, segmentation ◦◦ Seismic analysis
of tissue. ◦◦ Geology and weather monitoring and
• Medical image registration etc. control
• Synthetic Aperture Radar (SAR) ◦◦ Robotic vision and many others.
Imaging and Remote Sensing: Some
open problems in this domain include de-
velopment of efficient PDE-based models
for speckle reduction & segmentation, en-
hancement, restoration etc.

603

CONCLUSION
In this chapter, partial differential equation
(PDE)-based approaches for image modelling and
processing was discussed for image restoration
task. The PDE-based approaches can be used in
various image processing and computer vision
applications such as restoration, segmentation,
shape from shading, inpainting, registration, 3D
reconstruction of surfaces from multiple views and
many more. This chapter is oriented towards the
image restoration problem which is an important
pre-processing task for many applications such
as 3-D surface geometry and reconstruction and
many more. The introduction part, discussed the
basic concepts of image processing and various
image modelling techniques. General principles
of PDE based image processing and concepts
of curve and surface evolution were discussed.
Advantages and application areas of PDE-based
image processing were presented. The various
performance measurement metrics used for
comparison purposes were also given. Further,
various methods and techniques for removal and
reduction of additive and multiplicative noises
were discussed. For additive noise removal vari-
ous popular PDE-based filters such as TV-based
approach, anisotropic diffusion based approach,
Fourth order PDEs, and complex diffusion based
approach were discussed. The performances of
some filters were also evaluated. The complex
diffusion based filter is considered to be more
efficient filter in comparison to other filters in
consideration. For multiplicative or speckle noise,
the speckle noise pattern in various imaging
modalities such as ultrasound imaging and SAR
imaging are discussed. Further, the various meth-
ods for speckle reduction were also discussed and
finally a comparative analysis of various methods
were presented for two cases which concludes
that complex diffusion based homomorphic filter
performs better in comparison to other methods
in consideration. The schemes discussed in this
chapter justify the applicability of the PDE-based
604
PDE-Based Image Processing
models in image processing and computer vision
which may form the basis for further research in
this area for other image processing tasks such as
image registration, shape from shading and 3D
surface reconstruction.
REFERENCES
Alvarez, L., Guichard, F., Lions, P. L., & Morel,
J. M. (1992). Image selective smoothing and
edge detection by nonlinear diffusion. SIAM
Journal on Numerical Analysis, 29, 845–866.
doi:10.1137/0729052
Alvarez, L., Guichard, F., Lions, P. L., & Morel,
J. M. (1993). Axioms and fundamental equations
of image processing. Archives of Rational Me-
chanics, 23, 199–257. doi:10.1007/BF00375127
Anderson, M. E., & Trahey, G. E. (2009). A semi-
nar on k-space applied to medical ultrasound.
Retrieved June 10, 2009, from http://dukemil.
egr.duke.edu/Ultrasound/k-space
Barbara, Z., & Jan, F. (2003). Image registra-
tion methods: A survey. Image and Vision
Computing, 21, 977–1000. doi:10.1016/S0262-
8856(03)00137-9
Beck, A., & Teboulle, M. (2009). Fast gradient-
based algorithms for constrained total variation
image de-noising and de-blurring problems.
IEEE Transactions on Image Processing, 18(11),
2419–2434. doi:10.1109/TIP.2009.2028250
Brown, L. G. (1992). A survey of image registra-
tion techniques. [CSUR]. ACM Computing Sur-
veys, 24(4), 325–376. doi:10.1145/146370.146374
Caselles, V., Catte, F., Coll, T., & Dibos, F. (1997).
A geometric model for active contours. Interna-
tional Journal of Computer Vision, 22, 61–69.
doi:10.1023/A:1007979827043
PDE-Based Image Processing
Caselles, V., Morel, J., & Sapiro, G. (1998). Intro-
duction to the special issue on partial differential
equations and geometry driven diffusions in image
processing. IEEE Transactions on Image Process-
ing, 7(3), 269–273. doi:10.1109/TIP.1998.661176
Chambolle, A. (1994). Partial differential equa-
tions and image processing. IEEE International
Conference on Image Processing, Austin, TX.
Chan, T. F., & Shen, J. (2005). Variational image
inpainting. Communications on Pure and Ap-
plied Mathematics, 58(5), 579–619. doi:10.1002/
cpa.20075
Chen, Q., Montesinos, P., Sun, Q. S., Heng, P.
A., & Xia, D. S. (2009). Adaptive total variation
denoising based on difference curvature. Image
and Vision Computing, 28(3). doi: 10.1016/j.
imavis. 2009.04.012
Dewaele, P., Wambacq, P., Oosterlinck, A., &
Marchand, J. L. (1990). Comparison of some
speckle reduction techniques for SAR images.
Geoscience and Remote Sensing Symposium,
IGARSS ‘90, (pp. 2417-2422).
Evans, L. C., & Spruck, J. (1991). Motion of level
sets by mean curvature. International Journal of
Differential Geometry, 33, 635–681.
Frost, V. S., & Stiles, J. A. (1982). A model for radar
Images and its application to adaptive digital filter-
ing of multiplicative noise. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 4(2),
157–166. doi:10.1109/TPAMI.1982.4767223
Gabor, D. (1965). Information theory in elec-
tron microscopy. Laboratory Investigation, 14,
801–807.
Gilboa, G., Sochen, N., & Zeevi, Y. Y. (2004).
Image enhancement and denoising by complex
diffusion processes. IEEE Transactions on Pat-
tern Analysis and Machine Intelligence, 25(8),
1020–1036. doi:10.1109/TPAMI.2004.47
Gonzalez, R. C., & Wintz, P. (1987). Digital image
processing (2nd ed.). New York, NY: Academic
Press.
Goodman, J. W. (1984). Statistical properties of
laser speckle patterns. In Dainty, J. C. (Ed.), Laser
speckle and related phenomena (pp. 9–75). Berlin,
Germany: Springer-Verlag.
Iskander, D. R. (1999). Estimation of the param-
eters of the k-distribution using higher order and
fractional moments. IEEE Transactions on Aero-
space and Electronic Systems, 35(4), 1453–1457.
doi:10.1109/7.805463
Iskander, D. R., Zoubir, A. M., & Boashash, B.
(1999). A method for estimating the parameters
of k-distribution. IEEE Transactions on Signal
Processing, 47, 880–884. doi:10.1109/78.752614
Jain, A. K. (1977). Partial differential equations
and finite difference methods in image process-
ing, part I: Image representation. Journal of
Optimization Theory and Applications, 23, 65–91.
doi:10.1007/BF00932298
Jain, A. K. (1978). Partial differential equations
and finite difference methods in image processing,
Part II: Image restoration. IEEE Transactions on
Automatic Control, 23(5), 817–834. doi:10.1109/
TAC.1978.1101881
Jain, A. K. (2006). Fundamentals of digital image
processing. India: PHI.
Kass, M., Witkin, A., & Terzopoulos, A. (1988).
Snakes: active contour models. International
Journal of Computer Vision, 1, 321–331.
doi:10.1007/BF00133570
Kimia, B. B., Tanenbaum, A., & Zucker, S. W.
(1990). Lecture Notes in Computer Science: Vol.
427. Toward a computational theory of shape: An
overview. New York: Springer-Verlag.
Koenderink, J. J. (1984). The structure of images.
Biological Cybernetics, 50, 363–370. doi:10.1007/
BF00336961
605

Kuan, D. T., & Sawchuk, A. A. (1987). Adaptive
restoration of images with speckle. IEEE Transac-
tions on Acoustics, Speech, and Signal Processing,
35, 373–383. doi:10.1109/TASSP.1987.1165131
Lee, J. S. (1981). Speckle analysis and smooth-
ing of synthetic aperture radar images. Com-
puter Graphics and Image Processing, 17, 24–32.
doi:10.1016/S0146-664X(81)80005-6
Lee, J. S. (1983). Digital image smoothing and the
sigma filter. Computer Vision Graphics and Im-
age Processing, 24, 255–269. doi:10.1016/0734-
189X(83)90047-6
Lucy, L. B. (1974). Image restoration of high
photometric quality. In R.J Hanisch & R. L. White
(Eds.), Proceedings of the Restoration of HST
Images and Spectra STSci, (pp. 79-85).
Mateo, J. L., & Fernández, C. A. (2009). Finding
out general tendencies in speckle noise reduc-
tion in ultrasound images. Expert Systems with
Applications, 36, 7786–7797. doi:10.1016/j.
eswa.2008.11.029
Milan, S., Vaclav, H., & Boyle, R. (2007). Image
processing, analysis and machine vision (2nd ed.).
Washington, DC: PWS Publishing.
Morel, J. M., & Solimini, S. (1988). Segmentation
of images by variational methods: A constructive
approach. Revista Matematica de Universidad
Complutense Madrid, 1, 169–182.
Mumford, D., & Shah, J. (1989). Optimal ap-
proximations by piecewise smooth functions
and variational problems. Communications on
Pure and Applied Mathematics, 42, 577–685.
doi:10.1002/cpa.3160420503
Osher, S., & Fedkiw, R. P. (2001). Level set meth-
ods: An overview and some recent results (pp.
1–65). California, USA: IPAM GBM Tutorials.
606
PDE-Based Image Processing
Osher, S., & Rudin, L. I. (1990). Feature oriented
image enhancement using shock filters. SIAM
Journal on Numerical Analysis, 27, 919–940.
doi:10.1137/0727053
Osher, S., & Sethian, J. A. (1988). Fronts propa-
gating with curvature dependent speed: Algo-
rithms based on Hamilton-Jacobi formulations.
Journal of Computational Physics, 79, 12–49.
doi:10.1016/0021-9991(88)90002-2
Perona, P. (1998). Orientation diffusions. IEEE
Transactions on Image Processing, 7, 457–467.
doi:10.1109/83.661195
Perona, P., & Malik, J. (1990). Scale space and
edge detection using anisotropic diffusion. IEEE
Transactions on Pattern Analysis and Machine
Intelligence, 12, 629–639. doi:10.1109/34.56205
Press, H., Teukolsky, S., Vetterling, T., & Flan-
nery, B. (1992). Numerical recipes in C: The art
of scientific computing (2nd ed.). Cambridge
University Press.
Price, C. B. (1990). Image enhancement and
analysis with reaction diffusion paradigm. Pro-
ceedings of the Institution of Electrical Engineers,
137, 136–145.
Rajan, J., Kannan, K., & Kaimal, M. R. (2008).
An improved hybrid model for molecular image
de-noising. Journal of Mathematical Imaging
and Vision, 31, 73–79. doi:10.1007/s10851-008-
0067-4
Romeny, B. ter H. (1994). Geometry driven dif-
fusion in computer vision. Boston, MA: Kluwer.
Rudin, L. I., Osher, S., & Fatemi, E. (1992). Non
linear total variation based noise removal algo-
rithms. Physica D. Nonlinear Phenomena, 60,
259–268. doi:10.1016/0167-2789(92)90242-F
PDE-Based Image Processing
Salinas, H., & Fernandez, D. C. (2007). Compari-
son of PDE-based nonlinear diffusion approaches
for image enhancement and de-noising in optical
coherence tomography. IEEE Transactions on
Medical Imaging, 26(6), 761–771. doi:10.1109/
TMI.2006.887375
Shen, T. F., & Vese, L. (2002). Variational PDE
models in image processing. Joint Math Meeting,
San Diego.
Soman, K. P., & Ramachandran, K. I. (2004).
Insight in to wavelets from theory to practice.
India: PHI.
Srivastava, R. (2011). A complex diffusion based
nonlinear filter for speckle reduction from opti-
cal coherence tomography images. International
Conference on Communication, Computing &
Security (ICCCS-2011), Rourkela, India, (pp.
259-264). ACM Press. ISBN-978-1-4503-0464-1
Srivastava, R., & Gupta, J. R. P. (2010). A PDE-
based nonlinear filter adapted to Rayleigh’s
speckle noise for de-speckling 2D ultrasound
images. International Conference on Contempo-
rary Computing (IC3-2010), Communications
in Computer and Information Science (CCIS),
CCIS-94, India, (pp. 1-12). Berlin, Germany:
Springer-Verlag
Srivastava, R., Gupta, J. R. P., & Parthasarathy, H.
(2009). Comparison of PDE based and other tech-
niques for speckle reduction from digitally recon-
structed holographic images. Optics and Lasers
in Engineering, 48(5), 626–635. doi:10.1016/j.
optlaseng.2009.09.012
Srivastava, R., Gupta, J. R. P., Parthasarathy, H., &
Srivastava, S. (2011). An adaptive nonlinear PDE
based speckle reduction technique for ultrasound
images. International Journal of Biomedical
Engineering and Technology, 6(3). doi:10.1504/
IJBET.2011.041468
Srivastava, R., Gupta, J. R. P., & Parthasarthy,
H. (2009). Complex diffusion based speckle re-
duction from digital images. IEEE International
Conference on Methods and Models in Computer
Science (ICM2CS-09), Delhi, India, (pp. 43-49).
Topiwala, P. N. (Ed.). (1998). Wavelet image and
video compression. Kluwer Academic Publishers.
Wang, Z., Bovik, A. C., Sheikh, H. R., & Celli,
E. P. S. (2004). Image quality assessment: From
error visibility to structural similarity. IEEE
Transactions on Image Processing, 13(4), 1–14.
doi:10.1109/TIP.2003.819861
Witkin, A. P. (1983). Scale space filtering. Interna-
tional Joint Conference on Artificial Intelligence,
Germany, (pp. 1019-1023).
You, Y. L., & Kaveh, M. (2000). Fourth order
partial differential equations for noise removal.
IEEE Transactions on Image Processing, 9,
1723–1730. doi:10.1109/83.869184
Yu, Y., & Acton, S. T. (2002). Speckle reducing
anisotropic diffusion. IEEE Transactions on Im-
age Processing, 11(11), 1260–1270. doi:10.1109/
TIP.2002.804276
Zhang, R., Tsai, P.-S., Cryer, J. E., & Shah, M.
(1999). Shape from shading: A survey. IEEE Trans-
actions on Pattern Analysis and Machine Intel-
ligence, 21(8), 690–706. doi:10.1109/34.784284
KEY TERMS AND DEFINITIONS
Additive Noise: It follows Gaussian distribu-
tion.
Homomorphic Filter: This filter is used for
speckle reduction from digital images using filters
defined for additive noise removal in logarithmic
domain. In logarithmic domain, the multiplicative
noise converts in to additive noise.
PDE-Based Filters: Partial differential equa-
tion based filters derived using variational calculus
607
 PDE-Based Image Processing

by minimizing the energy functional of the image
defined in terms of gradient norm of the image.
Performance Metrics: The various perfor-
mance measurement metrics are MSE, PSNR,
CP, MSSIM, speckle index and effective number
of looks.
Restoration: This is one of the digital image
processing tasks that removes or reduces the noise
from images to improve its visual quality.
Speckle Noise: A multiplicative noise that
appears in digital holographic images, ultrasound
imaging, SAR imaging and many other applica-
tions. It degrades the visual quality of the image.
This noise is introduced during the formation of
image. Speckle noise in ultrasound images follows
Rayleigh’s distribution.
Speckle Reduction: It deals with the methods
to reduce speckle noise.

This work was previously published in 3-D Surface Geometry and Reconstruction: Developing Concepts and Applications,
edited by Umesh Chandra Pati, pp. 49-89, copyright 2012 by Information Science Reference (an imprint of IGI Global).

608
 609

Chapter 36
Multicast over Location-
Based Services
Péter Hegedüs
Budapest University of Technology and Economics, Hungary

Mihály Orosz
Budapest University of Technology and Economics, Hungary

Gábor Hosszú
Budapest University of Technology and Economics, Hungary

Ferenc Kovács
Budapest University of Technology and Economics, Hungary

ABSTRACT
This chapter details the potential found in combining to different technologies. The two basically dif-
ferent technologies, LBSs in mobile communication and the well-elaborated multicast technology are
merged in the multicast via LBS solutions. As this chapter demonstrates, this emerging new area has a
lot of possibilities, which have not been completely utilized.

INTRODUCTION or a given group of people in order to process

Currently, an important area of mobile communi-
cation is ad-hoc computer networks, where mobile
devices need base stations however they form an
overlay without any Internet-related infrastruc-
ture, which is a virtual computer network among
them. In this case, the selective, location-related
communication has not been solved completely.
Traditional Location-Based Services (LBSs)
determine the current location of a given person
DOI: 10.4018/978-1-4666-2038-4.ch036
location-dependent information. This use does
not cover the full range that is conceivable for
these services. This article introduces so-called
Zone Services as a new sub-category of LBSs. In
contrast to traditional LBSs, Zone Services collect
information about persons currently located in a
given geographic area. For these services, new
considerations regarding data collection, privacy,
and efficiency have to be made. Hence, it has to
be determined what techniques or mechanisms
common in traditional LBSs or in other areas like

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

databases or mobile communication systems can
be reused and what concepts have to be developed.
One of the various communication models
among software entities is the one-to-many data
dissemination, called multicast. The multicast
communication over mobile ad-hoc networks
has increasing importance (Hosszú, 2005). The
article described the fundamental concepts and
solutions on the area of LBSs and the possible
multicasting over the LBS systems. This kind
of communication is in fact a special case of the
multicast communication model, called geocast,
where the sender disseminates data to a subset of
the multicast group members that are in a specific
geographical area. This chapter shows that this
special kind of multicast utilizes the advantages of
LBSs, since multicast is based on location-aware
information that is available in location-based
solutions.
The two basically different technologies, LBSs
in mobile communication and the well-elaborated
multicast technology are merged in the multicast
via LBS solutions. As the chapter demonstrates,
this emerging new area has a lot of possibilities,
which has not been completely utilized.
BACKGROUND
The positioning technologies in the LBS solutions
are based on the various distances of the commu-
nication mobile from the different base stations.
With advances in automatic position sensing and
wireless connectivity, the application range of
mobile LBSs is rapidly developing, particularly
in the area of geographic, tourist and local travel
information systems (Ibach et al., 2005). Such
systems can offer maps and other area-related
information. The LBS solutions give the capability
to deliver location-aware content to subscribers on
the basis of the positioning capability of the wire-
less infrastructure. The LBS solutions can push
location-dependent data to mobile users according
to their interest or the user can pull the required
610
Multicast over Location-Based Services
information by sending a request to a server that
provides location-dependent information.
LBSs process information with respect to the
location of one or several persons, also referred to
as targets before presenting it to the user. In recent
years, LBSs have become increasingly important
and have helped accelerate the development
towards ubiquitous computing environments.
Traditional LBSs map targets to locations (e.g.,
Where is person X located?), i.e., they find the
position of a specific person or group of people.
This type of LBS is denoted as Tracking Services.
There are a lot of location positioning meth-
ods and technologies, such as the satellite-based
Global Positioning System (GPS) that is widely
applied (Hofmann-Wellenhof et al., 1997). The
location determination methods that do not use
the GPS can be classified into three categories:
Proximity, Triangulation (lateration), and Scene
analysis or pattern recognition (Hightower &
Borriello, 2001). Signal strength is frequently
applied to determine proximity. As a proximity
measurement, if a signal is received at several
known locations, it is possible to intersect the
coverage areas of that signal to calculate a location
area. If one knows the angle of bearing (relative
to a sphere) and distance from a known point to
the target device, then the target location can be
accurately calculated. Similarly, if somebody
knows the range from three known positions
to a target, then the location of the target object
can be determined. A GPS receiver uses range
measurements to multiple satellites to calculate
its position. The location determination methods
can be server-based or client-based according
to the place of computation (Hightower & Bor-
riello, 2001).
LBSs utilize their ability of location-awareness
to simplify user interactions. With advances in
wireless connectivity, the application range of
mobile LBSs is rapidly developing, particularly in
the field of tourist information systems - telematic,
geographic, and logistic information systems.
However, current LBS solutions are incompat-
Multicast over Location-Based Services
ible with each other since manufacturer-specific
protocols and interfaces are applied to aggregate
various components for positioning, networking,
or payment services. In many cases, these com-
ponents form a rigid system. If such a system has
to be adapted to another technology, e.g., moving
from GPS based positioning to in-house IEEE
802.11a-based Wireless Local-Area Network
(WLAN) or Bluetooth based positioning, it has
to be completely redesigned (Haartsen, 1998).
As such, the ability of interoperation of differ-
ent resources under changeable interconnection
conditions becomes crucial for the end-to-end
availability of the services in mobile environments
(Ibach, & Horbank, 2004).
Chen et al. (2004) introduces an enabling
infrastructure, which is a middleware, in order to
support LBSs. This solution is based on a Location
Operating Reference Model (LORE) that solves
many problems of constructing LBSs, including
location modeling, positioning, tracking, location-
dependent query processing and smart location-
based message notification. Another interesting
solution is the mobile yellow page service.
An interesting development is the Compose
project, which aims to overcome the drawbacks
of the current solutions by pursuing a service inte-
grated approach that encompasses pre-trip and on-
trip services where on-trip services could be split
into in-car and last-mile services (Bocci, 2005).
The pre-trip service means the 3D navigation of
the users in a city environment, and the on-trip
service means the in-car and the last-mile services
together. The in-car service is a composition of
an LBS and a satellite broadcasting/multicasting
method. In this case, the user has wireless-link
access by Personal Digital Assistant (PDA) to
broadcast or multicast. The last-mile service helps
the mobile user with PDAs to receive guidance
during the final part of the journey.
The article focuses on the multicast solutions
over the current LBS solutions. This kind of com-
munication is in fact a special case of the multicast
communication model, called geocast, where the
sender disseminates the data to a subset of the
multicast group members that are in a specific
geographical area.
MULTICASTING
The models of multicast communication differ in
the realization of the multiplication function in
the intermediate nodes. In the case of Datalink-
Level the intermediate nodes are switches, on
the Network-Level they are routers and on the
Application-Level the fork points are applica-
tions on hosts.
The Datalink-Level based multicast is not flex-
ible enough for new applications, which is why it
has no practical importance. The Network-Level
Multicast (NLM), known as IP-multicast, is well
elaborated and sophisticated routing protocols
are developed for it. However, it has not yet been
widely deployed since routing among the Autono-
mous Systems (AS) has not been solved perfectly.
The application level solution gives less efficiency
compared to the IP-multicast, however, its deploy-
ment depends on the application itself and it has
no influence on the operation of the routers. That
is why the Application-Layer Multicast (ALM) is
currently increasing in importance.
There are a lot of various protocols and imple-
mentations of the ALM, some of which are suit-
able for communication over wireless networks,
which enhance the importance of the ALM. The
reason for this is that in the case of mobile devices
the importance of ad-hoc networks is increasing.
Ad-hoc is a network that does not need any infra-
structure. Such networks are Bluetooth (Haartsen,
1998) and Mobile Ad Hoc NETwork (MANET),
which comprise a set of wireless devices that can
move around freely and communicate in relaying
packets on behalf of one another (Mohapatra et
al., 2004).
In computer networking, there is a weaker
definition of this ad-hoc network. Ad-hoc is a
computer network that does not need a routing
611

infrastructure. It means that the mobile devices
that use base stations can create ad-hoc computer
networks. In such situations, the usage of Applica-
tion-Level Networking (ALN) technology is more
practical than IP-Multicast. In order to support this
group communication, various multicast routing
protocols are developed for the mobile environ-
ment. The multicast routing protocols for ad-hoc
networks differ in terms of state maintenance,
route topology and other attributes.
The simplest ad-hoc multicast routing methods
are flooding and tree-based routing. Flooding
is very simple, which offers the lowest control
overhead at the expense of generating high data
traffic. This situation is similar to the traditional IP-
Multicast routing. However, in a wireless ad-hoc
environment, the tree-based routing fundamen-
tally differs from a wired IP-Multicast situation,
where tree-based multicast routing algorithms
are obviously the most efficient ones, such as in
the Multicast Open Shortest Path First (MOSPF)
routing protocol (Moy, 1994). Though tree-based
routing generates optimally small data traffic on
the overlay in the wireless ad-hoc network, the
tree maintenance and updates need a lot of control
traffic. That is why the simplest methods are not
scalable for large groups.
A more sophisticated ad-hoc multicast rout-
ing protocol is the Core-Assisted Mesh Protocol
(CAMP), which belongs to the mesh-based mul-
ticast routing protocols (Garcia-Luna-Aceves &
Madruga, 1999). It uses a shared mesh to support
multicast routing in a dynamic ad-hoc environ-
ment. This method uses cores to limit the control
traffic needed to create multicast meshes. Unlike
the core-based multicast routing protocol as the
traditional Protocol Independent Multicast-Sparse
Mode (PIM-SM) multicast routing protocol (Deer-
ing et al., 1996), CAMP does not require that all
traffic flow through the core nodes. CAMP uses
a receiver-initiated method for routers to join
a multicast group. If a node wishes to join the
group, it uses a standard procedure to announce
its membership. When none of its neighbors are
612
Multicast over Location-Based Services
mesh members, the node either sends a join request
toward a core or attempt to reach a group member
using an expanding-ring search process. Any mesh
member can respond to the join request with a join
Acknowledgement (ACK) that propagates back to
the request originator.
Compared to the mesh-based routing protocols,
which exploit variable topology, the so-called
gossip-based multicast routing protocols exploit
randomness in communication and mobility.
Such multicast routing protocols apply gossip
as a form of randomly controlled flooding to
solve the problems of network news dissemina-
tion. This method involves member nodes to
talk periodically to a random subset of other
members. After each round of talk, the gossipers
can recover their missed multicast packets from
each other (Mohapatra et al., 2004). Compared
to the deterministic approaches, this probabilistic
method will better work in a highly dynamic ad
hoc network because it operates independently of
network topology and its random nature fits the
typical characteristics of the network.
THE LOCATION-AWARE MULTICAST
The geocasting can be combined with flooding.
Such methods are called forwarding zone meth-
ods, which constrain the flooding region. The
forwarding zone is a geographic area that extends
from the source node to cover the geocast zone.
The source node defines a forwarding zone in the
header of the geocast data packet. Upon receiving
a geocast packet, other machines will forward
it only if their location is inside the forwarding
zone. The Location-Based Multicast (LBM) is
an example for such geocasting-limited flooding
(Ko & Vaidya, 2002).
An interesting type of ad-hoc multicasting is
the geocasting. The host that wishes to deliver
packets to every node in a certain geographical area
can use such a method. In this case, the position
of each node with regard to the specified geocast
Multicast over Location-Based Services
region implicitly defines group membership. Ev-
ery node is required to know its own geographical
location. For this purpose they can use the GPS.
The geocasting routing method does not require
any explicit join or leave actions. The members
of the group tend to be clustered both geographi-
cally and topologically. The geocasting method
of routing exploits the knowledge of location.
LBM GEOCASTING AND
IP MULTICASTING
Using LBM in a network where routers are in
fixed locations and their directly connected hosts
are within a short distance, the location of these
hosts can be approximated with the location of
their router. These requirements are met by most of
the GSM (Global System for Mobile Communica-
tions), UMTS (Universal Mobile Telecommunica-
tion System), WIFI (Wi-Fi Certified), WIMAX
(Worldwide Interoperability for Microwave Ac-
cess) and Ethernet networks, therefore a novel
IP layer routing mechanism can be introduced.
This new method is a simple kind of the
geocasting-limited flooding, extending the normal
Multicast Routing Information Base (MRIB) with
the geological location of the neighbor routers.
Every router should know its own location, and a
routing protocol should be used to spread location
information between routers. The new IP protocol
is similar to the User Datagram Protocol (UDP)
protocol, but it extends it with a source location and
a radius parameter. The source location parameter
is automatically assigned by the first router. When
a router receives a packet with empty source loca-
tion, it assigns its own location to it. The radius
parameter is assigned by the application itself, or
it can be an administratively defined parameter
This method requires changes in routing
operational systems, but offers an easy way to
start geocasting services on an existing IP infra-
structure without using additional positioning
devices (e.g., GPS receiver) on every sender and
receiver. The real advantages of the method are
that geocasting services can be offered for all
existing mobile phones without any additional
device or infrastructure.
FUTURE TRENDS
The multicast communication over mobile ad-
hoc networks has increasing importance. The
article has described the fundamental concepts
and solutions. It especially focused on the area
of LBSs and the possible multicasting over them.
It was shown that a special kind of the multicast,
called geocast communication model utilizes
the advantages of LBSs, since it is based on the
location-aware information made available in the
location-based solutions.
There are two known issues of this IP level
geocasting. The first problem is the scalability, the
flooding type of message transfer is less robust as
compared to multicast tree based protocols, but this
method is more efficient in a smaller environment
than using tree allocation overhead of multicast
protocols. The second issue is that the source must
be connected directly to the router that is physically
in the center position in order to become source of
a session. The proposed geocasting-limited flood-
ing protocol should be extended to handle those
situations where the source of a session and the
target geological location are in different places.
CONCLUSION
The two basically different technologies, the
Location-Based Services in the mobile commu-
nication world and the well-elaborated multicast
communication technology of the computer
networking are jointed in the multicast over LBS
solutions. As it was described, this emerging new
area has a lot of possibilities, which have not been
completely utilized.
613

As a conclusion it can be stated that despite
the earlier predicted slower development rate of
the LBS solutions, nowadays the technical possi-
bilities and the consumers’ demands have already
met. The geospatial property of LBSs provides
technical conditions to apply a specialized type
of the multicast technology, called geocasting,
which gives an efficient and user group targeted
solution for one-to-many communication.
REFERENCES
Bocci, L. (2005). Compose Project Web Site,
Retrieved from http://www.newapplication.it/
compose
Chen, Y., Chen, Y. Y., Rao, F. Y., Yu, X. L., Li,
Y., & Liu, D. (2004). LORE: an infrastructure to
support location-aware services. IBM Journal of
Research and Development, 48(5/6), 601–615.
Deering, S. E., Estrin, D., Farinacci, D., Jacobson,
V., Liu, C-G., & Wei, L. (1996). The PIM architec-
ture for wide-area multicast routing. IEEE/ACM
Trans. on Networking, 4(2), 153-162.
Garcia-Luna-Aceves, J. J., & Madruga, E. L.
(1999, August). The core-assisted mesh protocol.
IEEE Journal on Selected Areas in Communica-
tions, 1380–1394. doi:10.1109/49.779921
Haartsen, J. (1998). The universal radio interface
for ad hoc, wireless connectivity. Ericsson Re-
view,3. Retrieved 2004 from http://www.ericsson.
com/review
Hightower, J., & Borriello, G. (2001 Aug.). Loca-
tion Systems for Ubiquitous Computing. IEEE
Computer, 57-65.
Hofmann-Wellenhof, B., Lichtenegger, H., &
Collins, J. (1997). Global Positioning System:
Theory and Practice. Fourth Edition, Springer-
Verlag Wien, New York, NY.
614
Multicast over Location-Based Services
Hosszú, G. (2005). Reliability Issues of the
Multicast-based Mediacommunication. In Pagani,
M. (Ed.), Encyclopedia of Multimedia Technology
and Networking (pp. 875-881). Hershey, PA: Idea
Group Reference.
Ibach, P., & Horbank, M. (2004). Highly-Available
Location-based Services in Mobile Environments.
Paper presented at the International Service Avail-
ability Symposium 2004, Munich, Germany, May
13-14.
Ibach, P., Tamm, G., & Horbank, M. (2005).
Dynamic Value Webs in Mobile Environments
Using Adaptive Location-Based Services. In
Proceedings of the 38th Hawaii International
Conference on System Sciences (9 pages). IEEE.
Ko, Y.-B., & Vaidya, N. H. (2002). Flooding-
Based Geocasting Protocols for Mobile Ad Hoc
Networks. Proceeding of the Mobile Networks and
Applications. Kluwer Academic, 7(6), 471-480.
LocatioNet. (2006). LocatioNet and Ericsson En-
ter Into Global Distribution Agreement. Retrieved
from http://www.locationet.com
Mohapatra, P., Gui, C., & Li, J. (2004). Group com-
munications in mobile ad hoc networks. Comput-
er, 37(2), 52–59. doi:10.1109/MC.2004.1266296
Moy, J. (March 1994). Multicast extensions to
OSPF. Network Working Group RFC 1584.
Uppuluri, P., Jabisetti, N., Joshi, U., & Lee, Y.
(2005). P2P Grid: Service Oriented Framework
for Distributed Resource Management. In Pro-
ceedings of the IEEE International Conference
on Web Services, July 11-15, Orlando, FL, USA.
Weiss, D. (2006). Zone Services — A New Ap-
proach of Location-based Services, Retrieved
from http://www.pervasive2006.org/
Multicast over Location-Based Services
KEY TERMS AND DEFINITIONS
Ad-Hoc Computer Network: Mobile devices
that require base stations can create the ad-hoc
computer network if they do not need routing
infrastructure.
Application-Layer Multicast (ALM): A
novel multicast technology, which does not require
any additional protocol in the network routers,
since it uses the traditional unicast (one-to-one)
IP transmission. Its other name: Application-Level
Multicast (ALM).
Application-Level Network (ALN): The
applications, which are running in the hosts, can
create a virtual network from their logical con-
nections. This is also called overlay network. The
operations of such software entities are not able to
understand without knowing their logical relations.
In most cases these ALN software entities use the
P2P model (see below), not the client/server (see
below) for the communication.
Autonomous System (AS): A network with
common administration; it is a basic building
element of the Internet. Each AS is independent
from the others.
Client/Server Model: It is a communicating
model, where one hardware or software entity
(server) has more functionalities than the other
entity (the client), whereas the client is responsible
to initiate and close the communication session
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 239-245,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
towards the server. Usually the server provides
services that the client can request from the server.
Its alternative is the P2P model (see below).
Geocast: One-to-many communications
among communicating entities, where an entity
in the root of the multicast distribution tree sends
data to that certain subset of the entities in the
multicast dissemination tree, which are in a spe-
cific geographical area.
IP-Multicast: Network-level multicast tech-
nology, which uses the special class-D IP-address
range. It requires multicast routing protocols in the
network routers. Its other name: Network-Level
Multicast (NLM).
Multicast Routing Protocol: In order to
forward the multicast packets, the routers have
to create multicast routing tables using multicast
routing protocols.
Multicast Tree: A virtual graph, which gives
the paths of sending multicast data from the source
(the root of the tree) to the nodes of the tree. Its
other name: Dissemination tree or Distribution
Tree.
Peer-to-Peer (P2P): It is a communication
model where each node has the same author-
ity and communication capability. They create
a virtual network, overlaid on the Internet. Its
members organize themselves into a topology
for data transmission.
615
616

Chapter 37
A New Image Distortion
Measure Based on Natural
Scene Statistics Modeling
Abdelkaher Ait Abdelouahad Hocine Cherifi
University of Mohammed V-Agdal, Morocco University of Burgundy, France

Mohammed El Hassouni Driss Aboutajdine

University of Mohammed V-Agdal, Morocco University of Mohammed V-Agdal, Morocco

ABSTRACT
In the field of Image Quality Assessment (IQA), this paper examines a Reduced Reference (RRIQA)
measure based on the bi-dimensional empirical mode decomposition. The proposed measure belongs
to Natural Scene Statistics (NSS) modeling approaches. First, the reference image is decomposed into
Intrinsic Mode Functions (IMF); the authors then use the Generalized Gaussian Density (GGD) to
model IMF coefficients distribution. At the receiver side, the same number of IMF is computed on the
distorted image, and then the quality assessment is done by fitting error between the IMF coefficients
histogram of the distorted image and the GGD estimate of IMF coefficients of the reference image,
using the Kullback Leibler Divergence (KLD). In addition, the authors propose a new Support Vector
Machine-based classification approach to evaluate the performances of the proposed measure instead
of the logistic function-based regression. Experiments were conducted on the LIVE dataset.

INTRODUCTION sentially on human observers’ opinion. A popular

Performances of image processing algorithms can
be evaluated essentially based on the perceived
image quality. As a result, an image quality as-
sessment method is required. Indeed, subjective
measures can be involved. They are based es-
DOI: 10.4018/978-1-4666-2038-4.ch037
method for assessing image quality involves ask-
ing people to quantify their subjective impressions
by selecting one of the five classes: Excellent,
Good, Fair, Poor, Bad, from the quality scale
(ITU, 2000), then these opinions are converted
into scores. Finally, the average of the scores is
computed to get the Mean Opinion Score (MOS).
However, subjective measures cannot be applied

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A New Image Distortion Measure Based on Natural Scene Statistics Modeling
in real-time applications due to the time required
for subjective experiments.
Beside subjective measures, objective mea-
sures assess the quality automatically with no need
to human observers. But up to now, an objective
measure which can replace the human observer
does not exist.
General-purpose objective measures belong
to Full Reference (FR) approaches. Thanks to the
availability of the reference image, they can use
a pixel-to-pixel comparison like the well-known
Peak-Signal-to-Noise (PSNR) or introduced sta-
tistical features-based similarity measure (Wang,
Bovik, Sheikh, & Simoncelli, 2004). Nevertheless,
in real-world applications (like transmission) the
reference image is not always available. In that
case, No-Reference (NR) approaches are more
suitable. They attempt to assess the quality of a
processed image without any cue from its unpro-
cessed version. For this goal, they need a prior
information on the distortion. As consequence,
most of NR approaches are conceived for specific
distortion type and cannot be generalized to other
distortions (Wang, Sheikh, & Bovik, 2002). Re-
duced Reference (RR) approaches provide a good
compromise between FR and NR, as only partial
information is involved. They rely on perceptually
relevant features that have to be extracted from
the reference image. These features are used at the
receiver side for detecting visual quality degrada-
tion. Consequently, the measure proposed in this
paper falls in the RR approaches.
Recently, a number of authors have success-
fully introduced RR methods based on: image
distortion modeling (Gunawan & Ghanbari, 2003;
Kusuma & Zepernick, 2003), human visual system
(HVS) modeling (Carnec, Le Callet, & Barba,
2003, 2005), or finally natural image statistics
modeling (Wang & Simoncelli, 2005; Ait Abde-
louahad, El Hassouni, Cherifi, & Aboutajdine,
2011). Wang and Simoncelli (2005) introduced
a RRIQA measure called WNISM and based
on Steerable pyramids (a redundant transform
of wavelets family). Although, this method has
known some success when tested on five types of
distortion, it suffers from some weaknesses. First,
steerable pyramids are a non-adaptive transform,
and depend on a basis function. This later cannot
fit all signals, and when this happens, a wrong
time-frequency representation of the signal is
obtained. Consequently it is not sure that steer-
able pyramids will achieve the same success for
other type of distortions. Furthermore, the wavelet
transform provides a linear representation which
cannot reflect the nonlinear masking phenom-
enon in human visual perception (Foley, 1994).
A novel decomposition method was introduced
by Huang et al. (1998), named Empirical Mode
decomposition (EMD). It aims to decompose non
stationary and non linear signals to finite number
of components: Intrinsic Mode Functions (IMF),
and a residue. It was first used in signal analysis,
and then it attracted more researchers’ attention.
A few years later Nunes, Bouaoune, Delechelle,
Niang, and Bunel (2003) proposed an extension
of this decomposition in the 2D case Bi-dimen-
sional Empirical Mode Decomposition (BEMD).
A number of authors have benefited from the
BEMD in several image processing algorithms:
image watermarking (Taghia, Doostari, & Taghia,
2008), texture image retrieval (Andaloussi et al.,
2009), and feature extraction (Wan, Ren, & Zhao,
2008). In contrast to wavelet, EMD is nonlinear
and adaptive method, it depends only on data
since no basis function is needed.
Motivated by the advantages of the BEMD,
and to remedy the wavelet drawbacks discussed
above, here we propose the use of BEMD as a
representation domain. As distortions affects IMF
coefficients and also their distribution. The inves-
tigation of IMF coefficients marginal distribution
seems to be a reasonable choice. In the literature,
most RR methods use a logistic function-based
regression method to predict mean opinion scores
from the values given by an objective measure.
These scores are then compared in term of cor-
relation with the existing subjective scores. The
higher is the correlation, the more accurate is the
617
 A New Image Distortion Measure Based on Natural Scene Statistics Modeling
objective measure. In addition to the objective
measure introduced in this paper, an alternative
approach to logistic function-based regression is
investigated. It is an SVM-based classification,
where the classification was conducted on each
distortion set independently, according to the
visual degradation level. The better is the clas-
sification accuracy the higher is the correlation
of the objective measure with the HVS judgment.
To verify the relevance of the adaptivity, we
used the LIVE dataset which contains several
distortion types with different distortion levels.
We believe that with such dataset we can study
the genericity of the proposed method.
This paper is organized as follows. First, we,
present the proposed IQA scheme. The BEMD
and its algorithm are presented afterwards. Then,
we describe the distortion measure, followed by a
section that explains how we conduct the experi-
ments and presents some results of a comparison
with existing methods. Finally, we give some
concluding remarks.
IQA PROPOSED SCHEME
In this paper, we propose a new IQA scheme
based on the BEMD decomposition. This scheme
provides a distance between a reference image
and its distorted version as an output. This dis-
tance represents the error between both images
and should have a good consistency with human
judgment.
The scheme consists in two stages as men-
tioned in Figure 1. First, a BEMD decomposition
is employed to decompose the reference image
at the sender side and the distorted image at the
receiver side. Second, the features are extracted
from the resulting IMFs based on modeling natural
image statistics. The idea is that distortions make
a degraded image appearing unnatural and affect
image statistics. Measuring this unnaturalness can
lead us to quantify the visual quality degradation.
One way to do so is to consider the evolution of
618
marginal distribution of IMF coefficients. This
implies the availability of IMF coefficient histo-
gram of the reference image at the receiver side.
Using the histogram as a reduced reference raises
the question of the amount of side information
to be transmitted. If the bin size is coarse, we
obtain a bad approximation accuracy but a small
data rate while when the bin size is fine, we get
a good accuracy but a heavier RR data rate. To
avoid this problem it is more convenient to assume
a theoretical distribution for the IMF marginal
distribution and to estimate the parameters of the
distribution. In this case the only side informa-
tion to be transmitted consists of the estimated
parameters and eventually an error between the
empirical distribution and the estimated one. The
GGD model provides a good approximation of
IMF coefficients histogram and this only with
the use of two parameters. Moreover, we consider
the fitting error between empirical and estimated
IMF distribution. Finally, at the receiver side we
use the extracted features to compute the global
distance over all IMFs.
THE BI-DIMENSIONAL EMPIRICAL
MODE DECOMPOSITION
The Empirical Mode Decomposition (EMD)
has been introduced (Huang et al., 1998) as a
driven-data algorithm, since it is based purely
on the properties observed in the data without
pre-determined basis functions. The main goal
of EMD is to extract the oscillatory modes that
represent the highest local frequency in a signal,
while the remainders are considered as a residual.
These modes are called Intrinsic Mode Functions
(IMF). An IMF is a function that satisfies two
conditions:
1. The function should be symmetric in time,
and the number of extrema and zero cross-
ings must be equal, or at most differ by one.
A New Image Distortion Measure Based on Natural Scene Statistics Modeling
Figure 1. The deployment scheme of the proposed RRIQA measure
2. At any point, the mean value of the upper
envelope and the lower envelope must be
zero.
The so called “sifting process” works itera-
tively on the signal to extract each IMF. Let x (t )
be the input signal, the algorithm of the EMD is
summarized as follows:
Empirical mode decomposition algorithm
1. Identify all extrema of x (t ).
2. Interpolate between minima (resp.
maxima), ending up with some enve-
lope emin (t ) (resp. emax (t ) ).
3.Compute the mean
m (t ) =
(e (t ) + e
min max )
(t )
2
4. Extract the detail d (t ) = x (t ) − m(t ).
5. Iterate on the residual m(t ) .
The sifting process consists in iterating from
step 1 to 4 upon the detail signal d (t ) until this
later can be considered as zero mean. The resultant
signal is designated as an IMF and then the re-
sidual will be considered as the input signal for
the next IMF. The algorithm terminates when a
stopping criterion or a desired number of IMFs
is reached.
After IMFs are extracted through the sifting
process, the original signal x (t ) can be repre-
sented by:
n
x (t ) =∑Imf j (t ) +m (t ) (1)
j =1
Where Imf j is the j th extracted IMF and n
is the total number of IMFs.
In two dimensions (BEMD), the algorithm
remains the same as for a single dimension with
a few changes: the curve fitting for extrema in-
terpolation will be replaced with a surface fitting;
this increases the computational complexity for
identifying extrema and specially for extrema
interpolation.
619
 A New Image Distortion Measure Based on Natural Scene Statistics Modeling
To extract IMFs, several two dimensions EMD
versions have been developed (Linderhed, 2005;
Damerval, Meignen, & Perrier, 2005; Bhuiyan,
Adhami, & Khan, 2008). They differ mainly by
the interpolation method they use. The most recent
one proposed an interpolation based on statistical
order filters. From a computational cost standpoint,
it is a fast implementation. Indeed only no iteration
is required for computing each IMF. This a good
reason enough for using it in our experiments.
Figure 2 illustrates an application of the BEMD.
The resulting IMFs from a BEMD retain the
highest frequencies at each decomposition level.
The frequency content decrease as the order of
the IMF increases. For example, the first IMF
contains higher frequencies than the second one.
DISTORTION MEASURE
The resulting IMFs from a BEMD show the high-
est frequencies at each decomposition level, this
frequencies decrease as the order of the IMF in-
Figure 2. An example of image decomposition using the BEMD
620
creases. For example, the first IMF contains higher
frequencies than the second one. Furthermore, in a
particular IMF the coefficients histogram exhibits
a non Gaussian behavior, with a sharp peak at zero
and heavier tails than the Gaussian distribution as
can be seen in Figure 3 (a). Such a distribution can
be well fitted with a two parameters Generalized
Gaussian Density (GGD) model given by:
  β 
β   x  
p (x ) = exp −    (2)
1   α  
2αΓ( )    
β
∞
where Γ (z ) = ∫e dt , z > 0 represents the
−t z −1
t
0
Gamma function, α is the scale parameter that
describes the standard deviation of the density,
and β is the shape parameter.
In the conception of an RR method, we should
consider a transmission context, where an image
in the sender side with a perfect quality has to be
A New Image Distortion Measure Based on Natural Scene Statistics Modeling

Figure 3. Histograms of IMF coefficients under various distortion types: (a) original image, (b) addi-
tive Gaussian noise contaminated image, (c) Gaussian blurred image, (d) transmission errors distorted
image. Solid curves: histogram of IMF coefficients. Dashed curves: GGD model fitted to the histogram
of IMF coefficients in the original image. The horizontal axis represents the IMF coefficients, while the
vertical axis represents the frequency of these coefficients.

transmitted to a receiver side. The RR method
consists in extracting relevant features from the
reference image and uses them as a reduced de-
scription. However, the selection of features is a
critical step. On one hand, extracted features
should be sensitive to a large type of distortions
to guarantee the genericity, and also be sensitive
to different distortion levels. On the other hand,
extracted features should have a minimal size as
possible. Here, we propose a marginal distribution-
based RR method since the marginal distribution
of IMF coefficients changes from a distortion type
to another as illustrated in Figures 3 (b), (c), (d),
and (e). Let us consider IMFO as an IMF from
the original image and IMFD its corresponding
from the distorted image. To quantify the quality
degradation, we use the Kullback Leibler Diver-

621
 A New Image Distortion Measure Based on Natural Scene Statistics Modeling
gence (KLD) which is recognized as a convenient
way to compute divergence between two Prob-
ability Density Functions (PDFs). Assuming that
p (x ) and q (x ) are the PDFs of IMFO and IMFD
respectively, the KLD between them is defined
as:
p (x )dx
d (p || q ) = ∫ p (x ) log (3)
q(x )
For this aim, the histograms of the original
image must be available at the receiver side. Even
if we can send the histogram to the receiver side
it will increase the size of the feature signifi-
cantly and causes some inconvenients. The GGD
model provides an efficient way to get back coef-
ficients histogram, so that only two parameters
are needed to be transmitted to the receiver side.
In the following, we note pm (x ) the approxima-
tion of p (x ) using a 2- parameters GGD model.
Furthermore, our feature will contains a third
characteristic which is the prediction error defined
as the KLD between p (x ) and pm (x ):
pm (x )dx
d (pm || p) = ∫ p (x ) log
m
p(x )
 (4)
In practice, this quantity can be computed as
it follows:
Pm (i )
d (pm || p) = ∑ i =1 Pm (i ) log
L
dx (5)
p (i )
Where P (i ) and Pm (i ) are the normalized
heights of the i th histogram bins, and L is the
number of bins in the histograms.
Unlike the sender side, at the receiver side we
first compute the KLD between q (x ) and pm (x )
(Equation (6)). We do not fit q (x ) with a GGD
622
model cause we are not sure that the distorted
image is still a natural one and consequently if
the GGD model is still adequate. Indeed the dis-
tortion introduced by the processing can greatly
modify the marginal distribution of the IMF coef-
ficients. Therefore it is more accurate to use the
empirical distribution of the processed image.
pm (x )dx
d (pm || q ) = ∫ pm (x ) log q(x )
(6)
Then the KLD between p (x ) and q (x ) are
estimated as:
d(p || q ) = d (pm || q ) − d (pm || p 
) (7)
Finally the overall distortion between an origi-
nal and distorted image is as it follows:
 
1 
D = log2 1 +
 Do
∑
K
k =1 (
d k p k || q k )  (8)

where K is the number of IMFs, p k and q k are
the probability density functions of the k th IMF
in the reference and distorted images, respec-

tively. d k is the estimation of the KLD between
p k and q k , and Do is a constant used to control
the scale of the distortion measure.
The proposed method is a real RR one thanks
to the reduced number of features used. The image
is decomposed into four IMFs and from each IMF
we extract only three parameters {α, β, d (pm || p )}
so that 12 parameters in the total. Increasing the
number of IMF will increase the computational
complexity of the algorithm and thus the size of
the feature set. To estimate the parameters ( α, β )
we used the moment matching method (Van de
Wouwer, Scheunders, & Van Dyck, 1999), and for
extracting IMFs we used a fast and adaptive BEMD
A New Image Distortion Measure Based on Natural Scene Statistics Modeling
(Bhuiyan, Adhami, & Khan, 2008) based on sta-
tistical order filters, to replace the sifting process
which is time consuming.
To evaluate the performances of the proposed
measure, we use the logistic function-based re-
gression which takes the distances and provides
the objective scores. Another alternative to the
logistic function-based regression is proposed
and it is based on SVM classifier. More details
about the performance evaluation are given in
the next section.
EXPERIMENTAL RESULTS
Our experimental test was carried out using the
LIVE database (Sheikh, Wang, Cormack, & Bo-
vik, 2005). It is constructed from 29 high resolu-
tion images and contains seven sets of distorted
and scored images, obtained by the use of five
types of distortion at different levels. Set1 and
2 are JPEG2000 compressed images, set 3 and
4 are JPEG compressed images, set 5, 6 and 7
are respectively: Gaussian blur, white noise and
transmission errors distorted images. The 29 refer-
Figure 4. The 29 reference images of the LIVE database
ence images shown in Figure 4 have very different
textural characteristics, various percentages of
homogeneous regions, edges and details.
To score the images one can use either the
MOS or the Differential Mean Option Score
(DMOS) which is the difference between “refer-
ence” and “processed” Mean Opinion Score. For
LIVE database, the MOS of the reference images
is equal to zero, then the difference mean opinion
score (DMOS) and the MOS are the same.
To illustrate the visual impact of the different
distortions, Figure 5 presents the reference image
and the distorted images. In order to examine how
well the proposed metric correlates with the hu-
man judgement, the given images have the same
subjective visual quality according to the DMOS.
As we can see, the distance between the distorted
images and their reference image is of the same
order of magnitude for all distortions.
In Figure 6, we show an application of the
measure in Equation (8) to five white noise con-
taminated images, as we can see the distance
increases as the distortion level increases; this
demonstrates a good consistency with human
judgment.
623
 A New Image Distortion Measure Based on Natural Scene Statistics Modeling

Figure 5. An application of the proposed measure to different distorted images: (a) White noise, D =
9.36, DMOS =56.68), (b) Gaussian blur, D= 9.19, DMOS =56.17), (c) Transmission errors, D= 8.07,
DMOS =56.51)

Figure 6. An application of the proposed measure to different levels of Gaussian white noise contami-
nated images

624
A New Image Distortion Measure Based on Natural Scene Statistics Modeling
The tests consist in choosing a reference image
and one of its distorted versions. Both images are
considered as entries of the scheme given in Fig-
ure 1. After feature extraction step in the BEMD
domain a global distance is computed between
the reference and distorted image as mentioned
in Equation (8). This distance represents an objec-
tive measure for image quality assessment. It
produces a number and that number needs to be
correlated with the subjective MOS. This can be
done using two different protocols:
Logistic Function Based-Fegression
The subjective scores must be compared in term
of correlation with the objective scores. These
objective scores are computed from the values
generated by the objective measure (the global
distance in our case), using a nonlinear function
according to the Video Quality Expert Group
(VQEG) Phase I FR-TV (VQEG, 2000). Here, we
use a four parameter logistic function given by:
Figure 7. Scatter plots of (DMOS) versus the model prediction for the JPEG2000, Transmission errors,
White noise and Gaussian blurred distorted images
γ1 − γ2
logistic (γ, D ) = + γ2 ,
 D −γ 3 
− 
 γ 
1 +e 4
where γ = (γ1, γ2 , γ 3 , γ 4 ) . Then,
MOS p = logistic (γ, D ) .
Figure 7 shows the scatter plot of DMOS
versus the model prediction for the JPEG2000,
Transmission errors, White noise, and Gaussian
blurred distorted images. We can easily remark
how well is the fitting especially for the Transmis-
sion errors and the white noise distortions.
Once the nonlinear mapping is achieved, we
obtain the predicted objective quality scores
(DMOSp). To compare the subjective and objec-
tive quality scores, several metrics were introduced
by the VQEG. In our study, we compute the cor-
relation coefficient to evaluate the accuracy
prediction and the Rank order coefficient to
evaluate the monotonicity prediction. These met-
rics are defined as follows:
625
 A New Image Distortion Measure Based on Natural Scene Statistics Modeling

CC = As we can see, our method ensures better

( )(
MOS (i ) − MOS MOS p (i ) − MOS p
N
∑ i =1
(MOS (i ) − MOS ) ∑ (MOS
2 2
N N
∑ i =1 i =1
(MOS (i ) − MOS (i))
N 2
6∑
ROCC = 1 − i =1
N (N − 1)
where the index i denotes the image sample and
N denotes the number of samples.
We planned two experiments to evaluate the
performances of the proposed RRIQA measure.
The first experiment aims to evaluate performanc-
es in term of accuracy by the use of the Pearson
linear correlation coefficient, while the second
experiment involves the Spearman rank order
correlation coefficient to evaluate performances
in term of monotonicity. In Table 1, we report
the results obtained for two RR metrics (BEMD,
WNISM) and two FR metrics (PSNR, MSSIM).
As the FR metrics use more information we can
expect than they should be more performing than
RR metrics. This is true for MSSIM but not for
the PSNR that perform poorly as compared to the
RR metrics for all the types of degradation except
for the noise perturbation.
)
prediction accuracy (higher correlation coeffi-
cients), better prediction monotonicity (higher
)
(i ) − MOS
p p Spearman rank-order correlation coefficients)
(9) than the WNISM method, and this for the white
noise. While for the remaining distortion sets the
proposed measure supplies results comparable to
p
the WNISM ones. Also comparing to PSNR which
2
is a FR method, we can observe a significant
(10) improvements for the blur and transmission errors
distortions.
We notice that we carried out other experi-
ments for using the KLD between probability
density functions (PDFs) by estimating the GGD
parameters at the sender and the receiver side, but
the results were not satisfying comparing to the
proposed measure. This can be explained by the
strength of the distortion that makes reference
image lose its naturalness and then an estimation
of the GGD parameters at the receiver side is not
suitable. To go further, we thought to examine
how an IMF behaves with a distortion type. For
this aim, we conducted the same experiments as
above but on each IMF separately. Table 2 shows
the results for three distortions from the LIVE
database.
As observed, the sensitivity of an IMF to the
quality degradation changes depending on the
distortion type and the order of the IMF. For in-

Table 1. Performance evaluation for the quality measure using LIVE database

Dataset JP2(1) JP2(2) JPG1 JPG2 Noise Blur Error

Correlation Coefficient (prediction accuracy)

BEMD 0.8265 0.8695 0.9137 0.7866 0.9332 0.8405 0.9176
WNISM 0.9329 0.9502 0.9295 0.8613 0.8902 0.8874 0.9221
PSNR 0.9337 0.8948 0.9015 0.9136 0.7742 0.9866 0.8811
MSSIM 0.9675 0.9668 0.9646 0.9856 0.9706 0.9361 0.9439
Rank-Order Correlation Coefficient (prediction monotonicity)
BEMD 0.8285 0.8702 0.8704 0.7846 0.9068 0.8349 0.9065
WNISM 0.9356 0.9453 0.8623 0.8475 0.8699 0.9147 0.9210
PSNR 0.9231 0.8816 0.8907 0.8077 0.9855 0.7729 0.8785
MSSIM 0.9566 0.9677 0.9572 0.9441 0.9718 0.9421 0.9497

626
A New Image Distortion Measure Based on Natural Scene Statistics Modeling

Table 2. Performance evaluation using IMFs separatel

White Noise Gaussian Blur Transmission Errors

IMF1 CC= 0.91 ROCC= 0.90 CC= 0.74 ROCC= 0.75 CC= 0.87 ROCC= 0.87

IMF2 CC= 0.75 ROCC= 0.73 CC= 0.82 ROCC= 0.81 CC= 0.86 ROCC= 0.85

IMF3 CC= 0.85 ROCC= 0.87 CC= 0.77 ROCC= 0.73 CC= 0.75 ROCC= 0.75

IMF4 CC= 0.86 ROCC= 0.89 CC= 0.41 ROCC= 0.66 CC= 0.75 ROCC= 0.74

stance, the performance decreases for the “Trans-
mission errors” distortion as the order of the IMF
increases. Also, some IMFs are more sensitive for
one set, while for the other sets it is not. A weight-
ing factor according to the sensitivity of the IMF
seems to be a good way to improve the accuracy
of the proposed method. The weights are chosen
in a way to give more importance for the IMFs
which give better correlation values. To do so, the
weights have been tuned experimentally, since no
emerging combination can be applied in our case.
Let us take the “Transmission errors” set for ex-
ample, if w1,w2 ,w 3 ,w 4 are the weights for the
IMF1, IMF2 , IMF3 , IMF4 respectively, then we
should have w1 >w2 > w 3 >w 4 . We change the
value of wi ,i = 1, …, 4 until reaching better results.
Some improvements have been obtained, but only
for the Gaussian blur set as CC=0.88 and
ROCC=0.87. This improvement around 5% is
promising as the weighing procedure is very rough.
One can expect further improvement by using a
more refined combination of the IMF. Detailed
experiments on the weighting factors remain for
future work.
SVM-Based Classification
Traditionally, RRIQA methods use the logistic
function-based regression to obtain objective
scores. In this approach one extracts features
from images and trains a learning algorithm to
classify the images based on the feature extracted.
The effectiveness of this approach is linked to the
choice of discriminative features and the choice
of the multiclass classification strategy. One can
ask, why Support vector machine classifier?
Nowadays, several classifiers exist. Among
these classifiers, the Support Vector Machine
(SVM) have been shown to provide better results
than traditional techniques such as neural network.
The SVM rely on two main keys:
1. The SVM uses a hyperplane or a set of
hyperplanes for the separation purpose.
Thus, a good separation is achieved with
the hyperplane that has the largest distance
to the nearest training data of any class.
2. In order to discriminate sets which are not
linearly separable in their original space, a
mapping into another space with higher or
infinite dimension is used. The separation
problem becomes then easier in that new
space.
However, SVM is a binary classifier. In other
words, it is designed to separate data into two
classes; while real-world applications classifica-
tion problems are multiclass. Then, an extension
of the SVM to multiclass case is required. Two
strategies were introduced. The first consists in
considering all data in a single optimization; it is

627
 A New Image Distortion Measure Based on Natural Scene Statistics Modeling

named All In One (AIO). The second, consists give a better classification accuracy. The entries
in combining several binary SVM classifiers, it of the SVM are formed by the distances com-
is termed devide and combine. For the second puted in Equation (7). For the i th distorted image,
strategy we have three main methods: One Against X i = d1, d2 , d3 ,d4  represents the vector of fea-
One (OAO), Directed Acyclic Graph (DAG), One
tures (only four IMFs are used). Table 3 shows
Against All (OAA). The difference between these
the classification accuracy per set of distortion
methods comes from the number of SVM binary
As illustrated in Table 3, the classification
classifiers involved and the strategy used to com-
accuracy varies between 83.42 for the JPG2 dis-
bine them. Nevertheless, OAO, DAG, OAA pres-
tortion set and 96.55 for the white noise distortion
ent some limitations, since they do not consider
set. In the case of logistic function-based regres-
the full problem directly. Beside the devide and
sion, the top value of the correlation coefficient
combine strategy, the AIO approach seems to be
that we can obtain is equal to 1 as a full correla-
more natural one. Demirkesen and Cherifi (2008)
tion between objective and subjective scores while
have made a complete comparison of multiclass
for the classification case, the classification ac-
SVM classification methods. Results show that
curacy can be interpreted as the probability by
the AIO strategy outperforms the three methods
which we are sure that the objective measure
of devide and combine strategy. Convinced by
correlates well with the human judgment, thus a
the results of this study we have chosen the AIO
classification accuracy that equal to 100% is
strategy in our classification problem.
equivalent to a CC that equal to 1. This leads to
The 7 sets of distorted images from the LIVE
a new alternative of the logistic function-based
dataset are used. Set1 and set 2: JPEG2000 com-
regression with no need to predicted MOS. Thus,
pression, set 3 and set 4: JPEG compression, set
one can ask which one is more preferable? The
5: white noise, set 6: Gaussian blur, set 7: fast
logistic function-based regression or the SVM-
fading. The determination of the training and the
based classification. From the first view, the
testing sets has been realized thanks to the cross
SVM-based classification seems to be more pow-
validation (leave one out). Let us consider a spe-
erful. Nevertheless this gain on performances is
cific set (e.g., white noise). Since the DMOS
obtained at the price of an increasing complexity.
values are in the interval [0,100], this later was
On the one hand a complex training is required
divided into five equal intervals ]0,20], ]20,40],
before one can use this strategy. On the other hand
]40,60], ]60,80], ]80,100] corresponding to the
when this training step has been done the clas-
quality classes: Bad, Poor, Fair, Good Excellent,
sification is straightforward.
respectively. Thus the set of distorted images is
divided into five subsets according to the DMOS
associated to each image in the set. Then at each Table 3. Classification accuracy (%) for LIVE
iteration we trained a multiclass SVM (five database
classes) using the leave one out cross validation.
Type of distortion Classification accuracy
In other words each iteration involves using a
JP(1) 84.63
single observation from the original sample as
JP(2) 88.59
the validation data, and the remaining observations
JPG1 94.87
as the training data. This is repeated such that
JPG2 83.42
each observation in the sample is used once as
Noise 96.55
the validation data. The Radial Basis Function
Blur 89.55
(RBF) kernel was utilized and a feature selection
step was carried out to select its parameters that Error 93.10

628
A New Image Distortion Measure Based on Natural Scene Statistics Modeling
CONCLUSION
A reduced reference method for image quality
assessment is introduced, it’s a new one since it
is based on the BEMD, and also the classifica-
tion framework is proposed as an alternative of
the logistic function-based regression. This later
produces objective scores in order to verify the
correlation with subjective scores, while the
classification approach provides an accuracy
rates which explain how the proposed measure is
consistent with the human judgment. Promising
results are given demonstrating the effectiveness
of the method especially for the white noise dis-
tortion. As a future work, we expect to increase
the sensitiveness of the proposed method to other
types of degradations to the level obtained for
the white noise contamination. We plan to use an
alternative model for the marginal distribution of
BEMD coefficients. The Gaussian Scale Mixture
seems to be a convenient solution for this purpose.
We also plan to extend this work to other types of
distortion using a new image database.
REFERENCES
Ait Abdelouahad, A., El Hassouni, M., Cherifi,
H., & Aboutajdine, D. (2011). Image quality as-
sessment based on IMF coefficients modeling. In
Proceedings of the International Conference on
Digital Information and Communication Technol-
ogy and its Applications, Dijon, France (Vol. 166,
pp. 131-145).
Andaloussi, J., Lamard, M., Cazuguel, G., Tairi,
H., Meknassi, M., Cochener, B., & Roux, C.
(2009). Content based medical image retrieval:
Use of generalized Gaussian density to model
BEMD IMF. In Proceedings of the World Congress
on Medical Physics and Biomedical Engineering,
Munich, Germany (Vol. 25, pp. 1249-1252).
Bhuiyan, S., Adhami, R., & Khan, J. (2008). A
novel approach of fast and adaptive bidimensional
empirical mode decomposition. In Proceedings of
the IEEE International Conference on Acoustics,
Speech and Signal Processing, Huntsville, AL
(pp. 1313-1316).
Carnec, M., Le Callet, P., & Barba, D. (2003). An
image quality assessment method based on per-
ception of structural information. In Proceedings
of the IEEE International Conference on Image
Processing, France (Vol. 3, pp. 185-188).
Carnec, M., Le Callet, P., & Barba, D. (2005).
Visual features for image quality assessment with
reduced reference. In Proceedings of the IEEE
International Conference on Image Processing,
Paris, France (Vol. 1, pp. 421-424).
Damerval, C., Meignen, S., & Perrier, V. (2005).
A fast algorithm for bidimensional EMD. IEEE
Signal Processing Letters, 12(10), 701–704.
doi:10.1109/LSP.2005.855548
De Forges, J. R. O. (1999). Locally adaptative
method for progressive still image coding. In Pro-
ceedings of the IEEE International Symposium on
Signal Processing and its Applications, Brisbane,
Australia (Vol. 2, pp. 825-829).
Demirkesen, C., & Cherifi, H. (2008). A compari-
son of multiclass SVM methods for real world
natural scenes. In J. Blanc-Talon, S. Bourennane,
W. Philips, D. Popescu, & P. Scheunders (Eds.),
Proceedings of the 10th International Conference
on Advanced Concepts for Intelligent Vision
Systems, Juan-les-Pins, France (LNCS 5259, pp.
752-763).
Foley, J. (1994). Human luminance pattern mecha-
nisms: Masking experiments require a new model.
Journal of the Optical Society of America. A, Op-
tics, Image Science, and Vision, 11(6), 1710–1719.
doi:10.1364/JOSAA.11.001710
629
 A New Image Distortion Measure Based on Natural Scene Statistics Modeling
Gunawan, I. P., & Ghanbari, M. (2003). Reduced
reference picture quality estimation by using local
harmonic amplitude information. In Proceedings
of the London Communications Symposium, Lon-
don, UK (pp. 137-140).
Huang, N. E., Shen, Z., Long, S. R., Wu, M. C.,
Shih, H. H., & Zheng, Q. (1971). …Liu, H. H.
(1998). The empirical mode decomposition and the
Hilbert spectrum for non-linear and non-stationary
time series analysis. Proceedings - Royal Society.
Mathematical, Physical and Engineering Sci-
ences, 454, 903–995. doi:10.1098/rspa.1998.0193
International Telecommunications Union (ITU).
(2000). Recommendation BT. 500-10: Methodol-
ogy for the subjective assessment of the quality of
television pictures. Geneva, Switzerland: Author.
Kusuma, T. M., & Zepernick, H.-J. (2003). A
reduced-reference perceptual quality metric for
in-service image quality assessment. In Proceed-
ings of the Joint First Workshop on Mobile Future
and Symposium on Trends in Communications,
Nedlands, WA, Australia (pp. 71-74).
Linderhed, A. (2005). Variable sampling of the
empirical mode decomposition of two dimen-
sional signals. International Journal of Wavelets,
Multresolution, and Information Processing, 3(3),
435–452. doi:10.1142/S0219691305000932
Nunes, J., Bouaoune, Y., Delechelle, E., Niang,
O., & Bunel, P. (2003). Image analysis by bidi-
mensional empirical mode decomposition. Im-
age and Vision Computing, 21(12), 1019–1026.
doi:10.1016/S0262-8856(03)00094-5
Sheikh, H., Wang, Z., Cormack, L., & Bovik,
A. (2005). LIVE image quality assessment data-
base. Retrieved from http://live.ece.utexas.edu/
research/quality
This work was previously published in the International Journal of Computer Vision and Image Processing (IJCVIP), Volume
2, Issue 1, edited by Muhammad Sarfraz, pp. 1-15, copyright 2012 by IGI Publishing (an imprint of IGI Global).
630
Taghia, J., Doostari, M., & Taghia, J. (2008). An
image watermarking method based on bidimen-
sional empirical mode decomposition. In Proceed-
ings of the IEEE Congress on Image and Signal
Processing (Vol. 5, pp. 674-678).
Van de Wouwer, G., Scheunders, P., & Van Dyck,
D. (1999). Statistical texture characterization
from discrete wavelet representations. IEEE
Transactions on Image Processing, 8(4), 592–598.
doi:10.1109/83.753747
VQEG. (2000). Final report from the video qual-
ity experts group on the validation of objective
models of video quality assessment. Retrieved
from http://www.vqeg.org/
Wan, J., Ren, L., & Zhao, C. (2008). Image
feature extraction based on the two-dimensional
empirical mode decomposition. In Proceedings
of the Congress on Image and Signal Processing,
Harbin, China (Vol. 1, pp. 627-631).
Wang, Z., Bovik, A., Sheikh, C. H. R., & Simon-
celli, E. P. (2004). Image quality assessment:
From error visibility to structural similarity.
IEEE Transactions on Image Processing, 13(4),
1624–1639. doi:10.1109/TIP.2003.819861
Wang, Z., Sheikh, H. R., & Bovik, A. C. (2002).
No-reference perceptual quality assessment of
JPEG compressed images. In Proceedings of the
IEEE International Conference on Image Process-
ing, Rochester, NY (Vol. 1, pp. 477- 480).
Wang, Z., & Simoncelli, E. (2005). Reduced-
reference image quality assessment using a
wavelet-domain natural image statistic model. In
Proceedings of the SPIE Conference on Human
Vision and Electronic Imaging, San Diego, CA
(Vol. 5666, pp. 149-159).
 631

Chapter 38
Modern Navigation Systems
and Related Spatial Query
Wei-Shinn Ku
Auburn University, USA

Haojun Wang
University of Southern California, USA

Roger Zimmermann
National University of Singapore, Singapore

ABSTRACT
With the availability and accuracy of satellite-based positioning systems and the growing computational
power of mobile devices, recent research and commercial products of navigation systems are focusing
on incorporating real-time information for supporting various applications. In addition, for routing
purposes, navigation systems implement many algorithms related to path finding (e.g., shortest path
search algorithms). This chapter presents the foundation and state-of-the-art development of navigation
systems and reviews several spatial query related algorithms.

INTRODUCTION ability and accuracy of satellite-based positioning

Navigation systems have been of growing interest
in both industry and academia in recent years. The
foundation of navigation systems is based on the
concept of utilizing radio time signals sent from
some wide-range transmitters to enable mobile
receivers to determine their exact geographic
location. Based on this precise location, mobile
receivers are able to perform location-based
services (Shekhar, et al 2004). With the avail-
systems and the growing computational power of
mobile devices, recent research, and commercial
products of navigation systems are focusing on
incorporating real-time information for support-
ing various applications. In addition, for routing
purposes navigation systems implement many
algorithms related to path finding (e.g., shortest
path search algorithms). An increasing number
of useful applications are implemented based on
these fundamental algorithms.

DOI: 10.4018/978-1-4666-2038-4.ch038

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Modern Navigation Systems and Related Spatial Query
MODERN NAVIGATION SYSTEMS
A navigation system is an integration of position
and orientation devices, computation devices,
communication hardware and software for guiding
the movement of objects (e.g., people, vehicles,
etc.) from one location to another. In general, the
infrastructure of navigation systems can be classi-
fied into two subsystems: positioning signal trans-
mission systems and positioning signal receivers.
The positioning signal transmission system allows
the signal receiver to determine its location (lon-
gitude, latitude, and altitude) using timing signals.
Positioning signal receivers range from hand-held
devices, cellular phones, to car-based devices.
These devices typically include some storage
of map data and the computing capabilities of
spatial operations, such as calculating directions.
Additionally, in some novel geoinformatics ap-
plications, the receiver also relies on some server
components for various services, such as real-time
traffic information. In such a scenario, a server
infrastructure is introduced which includes a Web
server, a spatial database server, and an applica-
tion server to provide these services. The signal
receiver communicates with the server via wired
or wireless networking infrastructures.
Positioning Signal
Transmission Systems
Positioning signal transmitters, such as satellites
and base stations, broadcast precise timing signals
by radio to receivers, allowing them to determine
exact geographic locations and then dynamically
display and update their current position on digital
maps. As of 2006, the Global Positioning System
(GPS) is the only fully functional satellite-based
positioning signal transmission system.
Global Positioning System
The invention of GPS has had a huge influence
on modern navigation systems. GPS was devel-
632
oped by the U.S. Department of Defense in the
mid-1980s. Since it became fully functional in
1994, GPS has acted as the backbone of modern
navigation systems around the world.
The GPS consists of a constellation of 24
satellites in circular orbits at an altitude of 20,200
kilometers (Leick, 1995). Each satellite circles
the Earth twice a day. Furthermore, there are six
orbital planes with four satellites in each plane.
The orbits were designed so that at least four
satellites are always within line-of-sight from
most places on the earth (Langley, 1991). The
trajectory of the satellites is measured by five
monitoring stations around the world (Ascension
Island, Colorado Springs, Diego Garcia, Hawaii,
and Kwajalein). The master control station, at
Schriever Air Force Base, processes the monitor-
ing information and updates the onboard atomic
clocks and the ephemeris of satellites through
monitoring stations (El-Rabbany, 2002).
Each GPS satellite repeatedly broadcasts radio
signals traveling by line-of-sight, meaning that
they will pass through air but will not penetrate
most solid objects. GPS signals contain three
pieces of information (Hofmann-Wellenhof et
al, 1994): a pseudo random sequence, ephemeris
data, and almanac data. The pseudo random se-
quence identifies which satellite is transmitting the
signal. Ephemeris data allows the GPS receiver
to determine the location of GPS satellites at any
time throughout the day. Almanac data consists of
information about the satellite status and current
time from the onboard atomic clock of the satellite.
The GPS receiver calculates its location based
on GPS signals using the principle of trilateration
(Kennedy, 2002). First, the GPS receiver calcu-
lates its distance to a GPS satellite based on the
timing signal transmission delay from the satellite
to the receiver multiplied by the speed of radio
signals. After measuring its distance to at least
four satellites, the GPS receiver calculates its
current position at the intersection of four abstract
spheres, one around each satellite, with a radius of
the distance from the satellite to the GPS receiver.
Modern Navigation Systems and Related Spatial Query
GPS Accuracy
As a positioning signal transmission system,
the accuracy of GPS is a very important issue.
However, GPS was initially introduced with a
feature called Selective Availability (or SA) that
intentionally degraded the accuracy by introducing
an error of up to 100 meters into the civil timing
signals. Improved accuracy was available to the
United States military and a few other users who
were given access to the undegraded timing sig-
nal. On May 1, 2000, SA was finally turned off,
resulting in a substantial improvement of the GPS
accuracy (Conley, 2000).
Additionally, the accuracy of GPS can be af-
fected by the atmospheric conditions (e.g., Iono-
sphere, Troposphere) as well as reflections of the
radio signal off the ground and the surrounding
structures close to a GPS receiver. The normal
GPS accuracy is about 30 meters horizontally and
52 meters vertically at the 95% probability level
when the SA option is turned off (Kennedy, 2002).
There are several approaches that have been used
to improve the accuracy of GPS.
Differential GPS (DGPS) (Kennedy, 2002)
uses a network of stationary GPS receivers on the
ground acting as static reference points to calculate
and transmit correction messages via FM signals
to surrounding GPS receivers in a local area. The
improved accuracy provided by DGPS is equal to
0.5 m to 1 m near the reference point at the 95%
probability level (Monteiro et al. 2005). Before
the SA option was turned off by the Department
of Defense, DGPS was used by many civilian
GPS devices to improve the accuracy.
The Wide Area Augmentation System (WAAS)
(Loh, Wullschleger et al. 1995) has been widely
embedded in GPS devices recently. WAAS uses
25 ground reference stations across the United
States to receive GPS signals and calculate cor-
rection messages. The correction messages are
uploaded to a geosynchronous satellite and then
broadcast from the satellite on the same frequency
as GPS to the receivers. Currently WAAS only
works for North America as of 2006. However,
the European Geostationary Navigation Overlay
Service (EGNOS) and the Multi-Functional Sat-
ellite Augmentation System (MSAS) are being
developed in Europe and Japan, respectively. They
can be regarded as variants of WAAS.
The Local Area Augmentation System (LAAS)
(United States Department of Transportation,
FAA, 2002) uses a similar approach where cor-
rection messages are calculated, transmitted,
and broadcast via VHF data link within a local
area where accurate positioning is needed. The
transmission range of these correction messages
is typically about a 30 to 50 kilometer radius
around the transmitter.
GLONASS and Galileo
Positioning System
The GLObal NAvigation Satellite System
(Global’naya Navigatsionnaya Sputnikovaya
Sistema, GLONASS) is a satellite-based position-
ing signal transmission system developed by the
Russian government as a counterpart to GPS in
the 1980’s. The complete GLONASS consists
of 24 satellites in circular orbits at an altitude of
19,100 kilometers. Each satellite circles the Earth
in approximately 11 hours, 15 minutes. The orbits
were designed such that at least five satellites are
always within line-of-sight at any given time.
Based on measurements from the timing signal of
four satellites simultaneously, the system is able
to offer location information with an accuracy
of 70 meters.
There were 17 satellites in operation by
December 2005 offering limited usage. With
the participation of the Indian government, it is
expected that the system will be fully operational
with all 24 satellites by 2010.
GALILEO (Issle et al. 2003) is being developed
by the European Union as an alternative to GPS
and GLONASS. GALILEO is intended to provide
positioning signals with a precision higher than
GPS to both civil and military users. Moreover, it
633
 Modern Navigation Systems and Related Spatial Query
improves the coverage of satellite signals at high
latitude areas. The constellation of GALILEO
consists of 30 satellites in circular orbits at an
altitude of 23,222 kilometers. The GALILEO
system is expected to be fully operational by 2010.
Positioning Signal Receivers
Most positioning signal receiving devices are
designed for the use with the GPS system. These
devices have been manufactured in a wide variety
for different purposes, from devices integrated
into cars, personal digital assistants, and phones,
to dedicated devices such as hand-held GPS
receivers. The most popular variants are used in
car-based navigation systems that visualize the
position information calculated from GPS signals
to locate an automobile on a road retrieved from
a map database.
In these car-based systems, the map database
usually consists of vector information of some
area of interest. Streets and points of interest are
encoded and stored as geographic coordinates.
The client is able to find some desired places
through searching by address, name, or geographic
coordinates. The map database is usually stored
on some removable media, such as a CD or flash
memory. A common approach is to have a base
map permanently stored in the ROM of GPS de-
vices. Additional detailed information of areas of
interest can be downloaded from a CD or online
by the user in the future.
Integrating the positioning data from a GPS
receiver with the Geographic Information Sys-
tem (GIS) involves data retrieval, data format
transformation, multi-layer data display, and
data processing. With GPS, it is possible to col-
lect the positioning data in either the real-time or
post-processed mode. The digital format of GPS
data is then converted into a compatible format
used in the GIS applications (Steede-Terry 2000;
Kennedy, 2002). Together with other spatially
referenced data (e.g., the digital road map data),
the GIS application consists of a collection of
634
layers that can be analyzed for a wide variety of
purposes, such as calculating the route from the
current position to a destination.
NAVIGATION RELATED SPATIAL
QUERY ALGORITHMS
As mentioned earlier, many location-based end-
user applications can be provided after a position-
ing signal receiver calculates the position of a user.
There are several spatial query algorithms which
are commonly utilized by modern positioning
signal receivers (e.g., GPS devices) for supporting
location-based services and shortest path routing.
We broadly categorize them into location-based
query algorithms and shortest path search algo-
rithms in this section. Table 1 summarizes the
symbolic notations used throughout this section.
Location-Based Query Algorithms
Point Query: The term point query (PQ) can
be defined as: given a query point q, find all the
spatial objects O which contain q.
PQ(q ) = {O | ∀oi ∈ O, q ∈ oi }
The query processing efficiency of a point
query can be improved by utilizing spatial indices,
e.g., the R-tree (Guttman, 1984) or the Quadtree
(Samet, 1984). With a spatial index, all the spatial
objects are represented by geometric approxi-
mations such as an MBR (Minimum Bounding
Rectangle). Consequently, determining whether
Table 1. Symbolic notation
Symbol Meaning
q The location of a query point
O A set of spatial objects
dist(a,b) The Euclidean distance between two objects a and
b
D A set of nodes
E A set of edges
Modern Navigation Systems and Related Spatial Query
the query point is in an MBR is less expensive
than checking if the query point is in an irregular
polygon. After retrieving all the MBRs which
overlap with the query point as candidates, the
exact geometry of each element in the candidate
set is examined. Point queries can be applied to
determine the overlapping regions (e.g., admin-
istrative divisions) of navigation system users.
Nearest Neighbor Query: The term nearest
neighbor query (NNQ) can be defined as: given
a query point q and a set of spatial objects O, find
the spatial object oi ∈ O which has the shortest
distance to q.
NNQ(q ) = {oi | ∀o j ∈ O, dist (q, oi ) ≤ dist (q, o j )} there is a path from the vertex of interest to each
R-trees and their derivatives (Sellis et al. 1987;
Beckmann et al. 1990) have been a prevalent
method to index spatial data and increase query
performance. To find nearest neighbors, branch-
and-bound algorithms have been designed that
search an R-tree in either a depth-first (Rousso-
poulos et al. 1995) or best-first manner (Hjaltason
& Samet, 1999) to detect and filter out unqualified
branches. Both types of algorithms were designed
for stationary objects and query points. They may
be used when moving objects infrequently pose
nearest neighbor queries.
Range Query: The term range query (RQ) can
be defined as: given a query polygon q and a set
of spatial objects O, find all the spatial objects in
O which intersect with q.
RQ(q ) = {O | ∀oi ∈ O, oi ∩ q ≠ ϕ} are executed based on static distance informa-
Range queries can be solved in a top-down re-
cursive procedure utilizing spatial index structures
(e.g., the R-tree). The query region is examined
first against each branch (MBR) from the root.
If the query polygon overlaps with any branch,
the search algorithm is employed recursively
on sub-entries. This process terminates after it
reaches the leaf nodes of the index structure. The
selected entries in the leaves are used to retrieve
the records associated with the selected spatial
keys (Shekhar et al. 2004).
Shortest Path Search Algorithms
Dijkstra’s Algorithm: One important function
of navigation systems is to find the shortest route
to a user specified destination. The well-known
Dijkstra’s algorithm (Dijkstra, 1959) provides an
ideal solution for finding single-source shortest
paths in a graph of vertices connected through
edges. We present the algorithm, assuming that
of the other vertices. It is a simple modification to
handle the case where this is not so. We initialize
a set of vertices D to contain only the node whose
shortest paths are to be determined and assume the
vertex of interest is v1. We also initialize a set E
of edges to being empty. First we choose a vertex
vi that is closest to v1 and add it to D. In addition,
we also add the edge < v1, vi > to E. That edge is
clearly a shortest path from v1 to vi. Then we check
the paths from v1 to the remaining vertices that
allow only vertices in D as intermediate vertices.
A shortest of these paths is a shortest path. The
vertex at the end of such a path is added to D and
the edge that touches that vertex is added to E.
This procedure is continued until D covers all the
vertices. At this point, E contains the edges for the
shortest paths (Neapolitan & Naimipour, 1998).
Adaptive Shortest Path Search Algorithm:
Most existing shortest path searching algorithms
tion: pre-defined road segments with fixed road
conditions are used in the computation. However
any real-time events (e.g., detours, traffic conges-
tions, etc.) affecting the spatial network cannot
be reflected in the query result. For example, a
traffic jam occurring on the route to the computed
destination most likely elongates the total driving
time. More drastically, the closure of a restaurant
which was found as the destination according to
635
 Modern Navigation Systems and Related Spatial Query
its network distance might even invalidate a query
result. In other words, finding the shortest path
in terms of travel time is more important than
the actual distance. Therefore, we need adaptive
shortest path search algorithms which can integrate
real-time events into the search/routing procedure.
Ku et al. (Ku et al. 2005) proposed a novel travel
time network that integrates both road network
and real-time traffic event information. Based on
this foundation of the travel time network, they
developed an adaptive shortest path search algo-
rithm that utilizes real-time traffic information to
provide adaptive shortest path search results. This
novel technique could be implemented in future
navigation systems.
CONCLUSION
We have presented the foundation and state of
the art development of navigation systems and
reviewed several spatial query related algorithms.
GPS has been increasingly used in both military
and civilian applications. It can be forecast that
GPS will be extensively used and its applicability
expanded into new areas of applications in the
future. Meanwhile, additional civilian frequencies
will be developed and allocated to ease the conges-
tion of civil usage. GPS developers are anticipating
the advent of the European GALILEO system that
will introduce the birth of the Global Navigation
Satellite System (GNSS) infrastructure, which
combines the functionality of GPS and GALILEO
together (Gibbons 2004). The interoperation of
GPS and GALILEO will benefit the users with
more signal availability, more signal power, and
improved signal redundancy around the world.
In addition, several websites (e.g., MapQuest,
Yahoo! Maps, etc.) have integrated shortest path
search algorithms into on-line services. Users
can conveniently search the shortest path to their
destinations by utilizing these services.
636
ACKNOWLEDGMENT
This article was made possible by the NSF grants
ERC Cooperative Agreement No. EEC-9529152,
CMS-0219463 (ITR), and IIS-0534761. Any
opinions, findings and conclusions or recom-
mendations expressed in this material are those
of the authors and do not necessarily reflect those
of the National Science Foundation.
REFERENCES
Beckmann, N., Kriegel, H.-P., Schneider, R., &
Seeger, B. (1990). The R*-Tree: An Efficient and
Robust Access Method for Points and Rectangles.
Proc. ACM SIGMOD Int’l Conf. Management of
Data (pp. 322–331). ACM Press.
Conley, R. (2000). Life After Selective Avail-
ability. U.S. Inistitute of Navigation Newsletter,
10(1), Berlin, Germany: Springer-Verlag.
Dijkstra, E. W. (1959). A note on two problems in
connexion with graphs. Numerische Mathematik,
1, 269–271. doi:10.1007/BF01386390
El-Rabbany, A. (2002). Introduction to GPS: The
Global Positioning System. Boston, MA: Artech
House.
Gibbons G. (2004). Compatible with the Future.
GPS World, 15(4).
Guttman, A. (1984). R-Trees: A Dynamic Index
Structure for Spatial Searching. Proc. ACM
SIGMOD Int’l Conf. Management of Data (pp.
47–57). ACM Press.
Hjaltason, G. R., & Samet, H. (1999). Distance
browsing in spatial databases. ACM Transac-
tions on Database Systems, 24(2), 265–318.
doi:10.1145/320248.320255
Hofmann-Wellenhof, B., Lichtenegger, H., & Col-
lins, J. (1994). Global Positioning System: Theory
and Practice. Berlin, Germany, Springer-Verlag.
Modern Navigation Systems and Related Spatial Query
Issler, J. L., Hein, G., Godet, J., Martin, J. C., Er-
hard, P., Lucas-Rodriguez, R., & Pratt, T. (2003).
Galileo frequency and signal design. GPS World,
14(6), 30–37.
Kennedy, M. (2002), The Global Positioning
System and GIS: An Introduction, 2nd Edition.
New York: Taylor and Francis.
Ku, W.-S., Zimmermann, R., Wang, H., & Wan,
C.-N. (2005). Adaptive Nearest Neighbor Queries
in Travel Time Networks. Proc. 13th ACM Int’l
Symp. Geographic Information Systems (ACM-
GIS 2005) (pp. 210–219). ACM Press.
Langley, R. B. (1991). The orbits of GPS satel-
lites. GPS World, 2(3), 50–53.
Leick, A. (1995). GPS Satellite Surveying, 2nd
Edition. New York: John Wiley & Sons.
Loh, R., Wullschleger, V., Elrod, B., Lage, M., &
Haas, F. (1995). The U. S. wide-area augmentation
system (WAAS). Navigation, 42(3).
Monteiro, L. S., Moore, T., & Hill, C. (2005). What
is the accuracy of DGPS? Journal of Navigation,
58, 207–225. doi:10.1017/S037346330500322X
Neapolitan, R., & Naimipour, K. (1998). Foun-
dations of Algorithms. Jones and Bartlett. (pp.
234-241).
Roussopoulos, N., Kelley, S., & Vincent, F.
(1995). Nearest Neighbor Queries. Proc. ACM
SIGMOD Int’l Conf. Management of Data (pp.
71–79). ACM Press
Samet, H. (1984). The quadtree and related hi-
erarchical data structures. ACM Comput. Surv.
ACM, 16(2), 187–260.
This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 285-292,
copyright 2009 by Information Science Reference (an imprint of IGI Global).
Sellis, T. K., Roussopoulos, N., & Faloutsos, C.
(1987). The R+-Tree: A Dynamic Index for Multi-
Dimensional Objects. Proc. of Very Large Date
Base Conf., VLDB, (pp. 507–518).
Shekhar, S., Vatsavai, R. R., Ma, X., & Yoo, J. S.
(2004). Navigation Systems: A Spatial Database
Perspective. Location-Based Services. Morgan
Kaufmann Publishers, (pp. 41–82).
Steede-Terry, K. (2000). Integrating GIS and
the Global Positioning System. Redlands, CA:
ESRI Press.
United States Department of Transportation,
Federal Aviation Administration, (2002). Local
Area Augmentation System Ground Facility,
Specification FAA-E-2937A.
KEY TERMS AND DEFINITIONS
Ephemeris: Refers to the relative positions
of the planets, or satellites in the sky at a given
moment.
Geosynchronous Satellite: A satellite whose
orbital track lies over the equator.
GIS: A system for creating, integrating, ana-
lyzing and storing managing geographical data
and associated features. In general, GIS provides
users with an interface to query, retrieve, and edit
the spatial data in an efficient way.
Shortest Path Search: Finding the shortest
or least cost path through an underlay network.
Trilateration: Computing the relative posi-
tions of an object using the geometry of sphere
intersections. To accurately determine the relative
position of an object in 2D, trilateration uses at
least 3 reference points, and the measured distance
between the object and each reference point.
637
638

Chapter 39
Managing Schema Evolution
in a Federated Spatial
Database System
Xiaoying Wu Geoff West
Curtin University, Australia Curtin University, Australia

Jianhong (Cecilia) Xia Lesley Arnold

Curtin University, Australia Landgate, Australia
Bert Veenendaal
Curtin University, Australia

ABSTRACT
A Federated Spatial Database System (FSDBS) is the integration of multiple spatial data sources and
the realisation of effective spatial data sharing. These are becoming increasingly popular as more and
more spatial and non-spatial datasets are integrated, especially those across a number of independent
organisations. However, in a FSDBS environment, database schemas are subject to change and the
management of these changes is complex and inefficient. This is because schema changes in one local
database will invalidate applications built against this local schema, but also applications built against
the global schema.

The research is motivated by developments in the Shared Land Information Platform, built by Landgate
in Western Australia as a Spatial Data Infrastructure enabler that has been running since 2007. The
more than 350 datasets in SLIP are from and controlled by 20 organisations with queries built over the
different datasets. Changes in the various databases require schema updating to be streamlined.

In this chapter, an Automatic Schema Evolution Framework is explored and developed to more effec-
tively manage schema evolution in a FSDBS. This framework provides a Schema Element Dependency
Meta-Model, a set of Schema Change Templates, and incorporates view generation, view rewriting, and
query rewriting as solutions. These developed methods ensure applications can accommodate schema
changes and hence remain valid.

DOI: 10.4018/978-1-4666-2038-4.ch039

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Managing Schema Evolution in a Federated Spatial Database System
INTRODUCTION
The traditional database can be described by
schemas that define relations, attributes, etc. A
spatial database schema is an extension of the
traditional schema to include spatial descriptions
and behaviours required for a spatial database
(Yeung & Hall, 2007). Spatial database schemas,
like traditional schemas, are subject to change or
evolution due to changes in perception of reality
and application requirements. Changes can also
occur as a consequence of integration with other
systems, compliance to new regulations and the
implementation of new security requirements.
The task of managing schema evolution is to
ensure schemas, data and application consistency
and integrity after change. Schema changes often
result in the applications built against the schemas
being invalid.
In a Federated Spatial Database System (FS-
DBS), in which spatial means geographically
spatial, the spatial data are shared by multiple
organizations and applications across a number
of databases. A local database schema change
will affect not only the applications locally but
also applications built on the federated schemas.
Effective management is problematic and there are
several challenges to managing schema evolution
in a FSDBS including:
• How to evaluate the impact of schema
changes on local and global databases;
• How to propagate the schema changes
across databases in a federated environ-
ment; and,
• How to automatically rewrite run time que-
ries, and rewrite and generate views (spatial
views and traditional views) so that the ap-
plications are immune to schema changes.
The objective is to explore and develop an
Automatic Schema Evolution (ASE) framework
for managing schema evolution in a FSDBS. The
main contribution of this chapter is the proposal
of semi-automatic methods that, in future work,
will lead to the development of fully automatic
methods.
Of particular interest is the Shared Land In-
formation Platform (SLIP) built by Landgate (au-
thoritative agency for land information in Western
Australia) as a Spatial Data Infrastructure (SDI)
enabler (2011). SLIP went into production in 2007
and consists of more than 350 datasets from 20
different agencies. As such it is a federated database
and queries are built over the different datasets.
Schema updating is needed because the datasets
are agency owned and controlled allowing local
changes to occur.
For schema evolution, the mismatch between
evolved schemas and queries need to be overcome.
Queries here include queries generated by applica-
tions and stored queries (views and spatial views)
as seen by users. The ASE framework defines the
components and methods for automatic run time
query and view rewriting, as well as new view
generation when schemas change.
BACKGROUND
Before explaining the proposed ASE framework,
it is necessary to define various aspects required
for the ASE Framework, explore the various is-
sues and concepts involved in the framework, and
critique the current state of the art in research into
federated and spatial databases.
Spatial Database
Management Systems
In order to model the spatial aspects of real world
objects, a number of different approaches have been
proposed and developed. For example, the MADS
(Modeling Application Data with Spatio-Temporal
features) model defines constructs and language
to describe an application schema from different
aspects including thematic and spatial-temporal
data structures, and multi-representation aspects
639
 Managing Schema Evolution in a Federated Spatial Database System
(Parent, Spaccapietra, & Zimányi, 2006). Spatial
object types supported in MADS include both
feature and coverage types. ISO 19109—Rules
for application Schema (2005)—defines a general
feature model to describe the feature object types
and their properties including attributes, operations
and associations. Application schema developed
by INSPIRE (2009) adopted and augmented ISO
19101(2002) and ISO 19109 (2005) standards
by defining a modelling language and introduc-
ing the coverage spatial object type. Application
schemas are conceptual schemas that are platform
independent. The value domain of a spatial attri-
bute in the conceptual spatial schemas belong to
defined spatial data types including points, lines
and polygons.
With the development of database technolo-
gies, spatial object types and their properties de-
scribed by the conceptual spatial schema can be
implemented in a Spatial Database Management
System (SDBMS). A SDBMS manages and ma-
nipulates spatial data including vector and raster
data as well as non-spatial data (Rigaux, Scholl,
& Voisard, 2002).
Spatial database systems are distinguished from
non-spatial database systems in two ways (Yeung
& Hall, 2007):
• Storage of spatial data objects: points, lines,
and polygons; and
• Functionality to process spatial data such as
spatial indexes, extended spatial query lan-
guages and effective algorithms for spatial
operations.
The implementation of feature objects is sup-
ported by contemporary DBMSs. The Simple
Features Access (SFA) specification (also called
ISO 19125) defines the SQL implementation of the
geometry object model that is consistent with ISO
19109—Rules for application Schema (2005)—
and ISO 19107—Spatial Schema (2003). SFA spec-
ifies the SQL schema for storage, query, retrieval
and update of spatial data via a SQL interface.
640
SQL implementations include the implementation
of primitive data types in the including the binary
data type; and support for extended geometry data
types(OGC, 2006a, 2006b). However, coverage
objects are not included in the SFA specification
so storage, query and management of coverage
data are not supported by the SQL interface.
Federated Spatial Database Systems
Increasing spatial data and service sharing is
motivated by (1) cost savings on data collection,
maintenance, and software; (2) improved spatial
data usability and consistency; and (3) enhanced
inter-organisational relationships and the best use
of scarce GIS resources.
A Spatial Data Infrastructure (SDI) facilitates
spatial data sharing and can be built with advanced
relevant technologies and a set of standards,
specifications and policies (Nebert, 2004). The
key point of a SDI is interoperability and integra-
tion that enables users from different areas such
as government, the private sector and the general
public to discover and access spatial data and
services provided by different organisations. In
the last decade, SDI has become the global trend
in the Geographic Information Science (GIS)
industry. Examples include INSPIRE (2011) in
Europe, the National Spatial Data Infrastructure
(NSDI) in the United State (FDGC, 2011) and the
Australia Spatial Data Infrastructure (ASDI) in
Australia and New Zealand (ANZLIC, 2011). Of
particular interest is the SDI enabler (SLIP) build
by Landgate in WA that has been in operation since
2007. More than350 datasets from 20 different
organisations are available via SLIP. Importantly
each agency acts as custodian for their data and
will make changes to the database structures as
needed by their business requirements. Hence,
queries built on one or more of the datasets needs
to be aware of such changes and the queries modi-
fied appropriately.
A Federated Spatial Database System (FSDBS)
integrates various geographically distributed spa-
Managing Schema Evolution in a Federated Spatial Database System
tial data sources and provides a unified data access
mechanism in order to facilitate spatial data sharing
(Yeung & Hall, 2007). This important aspect of
a SDI has been achieved through advancements
in networks and communications, distributed
computing technologies, and conformance with
standards and policies. Federated database systems
have been the subject of active research for deal-
ing with database heterogeneity (Litwin, Mark,
& Roussoupoulos, 1990; Sheth & Larson, 1990).
A basic FSDBS shown in Figure 1 consists of:
1. AFederated Spatial Database Management
System: An ordinary DBMS that includes a
server, a federated database and the system
catalog (Yeung & Hall, 2007). This virtual
system receives requests from applications,
Figure 1. Architecture of a FSDBS
analyses queries, discomposes them into
sub-queries, sends sub-queries to relevant
data sources, composes results from data
sources and sends them back to the end users
and applications. The system catalog contains
schema information for the federated data-
base as well as matching and mapping in-
formation between schemas;
2. Spatial data sources: Which are autonomous
and heterogeneous; and
3. Applications: Including desktop and web
applications. For example, the federated
server can work with a web feature server
to provide web-mapping services.
Building a FSDBS is more challenging than
building a non-spatial DBMS because of the need
641
 Managing Schema Evolution in a Federated Spatial Database System
to accommodate vector and raster spatial data as
well as non-spatial data. To integrate spatial data
from various data sources, firstly a federated
schema which integrates various structurally and
semantically different source schemas needs to
be built along with schema level correspondences
between schemas; secondly, the instance level
correspondences need to be detected by match-
ing the semantics of the features and matching
their geometries (Goesseln & Sester, 2003). For
example, the same real object may be represented
differently in different datasets and needs to be
matched in the FSDBS.
In a FSDBS, the federated schema is dependent
on local schemas so when a local schema changes,
applications built on local schemas and federated
schemas and applications built on the federated
schemas will be affected. Managing schema evolu-
tion overcomes the mismatch between applications
Figure 2. A simple example of a FSDBS from Landgate
642
and the evolved schemas, and keeps consistency
of the correspondences between schemas after
schema changes.
Figure 2 is an example of a FSDBS taken from
a Landgate database. There are two local schemas
S1, S2, and a federated schema G. S1 consists of
two feature object classes, ExistingMainRoads and
ProposedMainRoads, and one non-spatial object
class, RoadType. A map application A1 built on S1
displays all highways. S2 consists of one feature
object class, LocalRoads and one non-spatial object
class, FeatureNames. The federated schema G
includes one feature object class, RoadSegments.
A web map application A2 displays all road seg-
ments including both main roads and local roads.
Schema changes in S1 will affect application A1
built on S1, as well as Schema G and application
A2 built on G. For example, if ExistingMainRoads
and ProposedMainRoads in S1 are merged into
Managing Schema Evolution in a Federated Spatial Database System
one feature object class MainRoads, applications
A1 and A2 will both be affected.
Query and Spatial Query
In a DBMS, applications generate queries that
are then sent to the server hosting the DBMS to
retrieve data. Simply speaking, a query is a ques-
tion imposed on data (Ramakrishnan & Gehrke,
2003). In a relational DBMS, a query is a function
that results a relation consisting of a set of tuples.
Structured Query Language (SQL) is used in
most commercial DBMSs. SQL is a declarative
language in which a user only specifies what they
want and do not have to worry about how to get
it. The basic form of a SQL query is Select-From-
Where.
A spatial query is more complex than a non-
spatial query because it deals with data of two or
three dimensions. A spatial query includes support
of geometry data types, and spatial operators and
spatial joins on these data types.
Spatial operators classified by the OGC (2006a,
2006b) include basic, query and analysis operators.
Basic operators deal with the general properties of a
geometric object. Examples are boundary, envelope
and spatial reference. Query operators test spatial
relationships including overlap, touch and cross
between geometric objects. Analysis operators are
used to perform spatial analysis such as distance,
difference and buffering. A spatial join is similar
to a join in a regular relational query that compares
two or more feature classes in terms of their loca-
tion using spatial operators, such as overlay, cross
and difference (Yeung & Hall, 2007).
A spatial database uses Spatial SQL that ex-
tends traditional SQL. While there is increased
functionality through the ability to deal with spatial
objects and the incorporation of spatial operators
and spatial joins, spatial SQL still retains the
simplicity of traditional SQL (Egenhofer, 1994).
Views and Spatial Views
In a relational DBMS, a view is defined as a stored
query, which can be accessed as a derived table that
contains the resulting tuples for the query. Views
can be used to provide a flexible representation
of a database by providing users with only data
of interest and concealing the rest (Date, 2003).
A spatial view extends a view by adding a spatial
column. Spatial views are created the same way
as traditional views as stored spatial queries that
contain a spatial column and spatial operations.
Spatial selection and spatial joins on the underlying
spatial data retrieves the results for applications.
Query Rewriting
Query rewriting is the process of converting query
statements into different expressions while still
keeping the logical structure of the query (Deutsch,
Popa, & Tannen, 2006). Equivalent query rewrit-
ing occurs when the result of a rewritten query
is equivalent to the results of the original query.
Query rewriting has been widely studied for differ-
ent applications such as query optimisation, data
integration, data exchange and schema evolution
(Halevy, 2001). For example, rewriting queries
with materialised views has attracted attention for
performance improvement of a database system.
Different query rewriting algorithms have been
developed to rewrite queries with views (Duschka
& Genesereth, 1997; Levy, Rajaraman, & Ordille,
1996; Popa, 2001). Rewriting queries against
one schema into equivalent ones against another
schema has been adopted as a solution to schema
integration, schema transformation and schema
evolution (Curino, Moon, Ham, & Zaniolo, 2009;
Tatarinov & Halevy, 2004; Ullman, 2000).
Schema Mapping
Schema mapping specifies the correspondence
between schemas (Bernstein, Green, Melnik, &
Nash, 2008). More specifically, a schema mapping
643
 Managing Schema Evolution in a Federated Spatial Database System
is a set of mapping rules between two schemas,
which maps one schema to another. Schema
mapping specifies the relationships between two
schemas including both structural and semantic
relationships.
Schema mapping is fundamental for query
rewriting. In an integrated system, the global
schema provides a unified interface to end users.
A query imposed on the global schema is typi-
cally decomposed into a number of sub-queries
imposed on local schemas. The decomposition
here depends on mapping information typically
stored as metadata.
In data integration, views are often used to
represent the mapping between schemas. There
are two main approaches: Global-As-View (GAV)
and Local-As-View (LAV) (Cali, Calvanese, Gia-
como, & Lenzerini, 2002). GAV represents the
global schema in terms of local schemas while
LAV represents local schemas in terms of the
global schema.
The form of a GAV mapping is: ∀x(φ(x)
→P(x)), where φ(x) is a conjunction of atoms
over source schemas and P(x) is the atom over
the target schema.
The LAV mapping is of the form: ∀ x(Q(x)
→∃y ψ (x,y)),where Q(x) is the atom over the
source schemas and ψ (x,y) is the conjunction of
atoms over the target schema.
The advantage of GAV is simplicity as query
rewriting can be achieved simply by unfolding and
substituting. That is, each schema element of one
schema can be replaced by associated elements of
another schema. However, GAV doesn’t have same
extensibility as LAV (Cali, Calvanese, Giacomo,
& Lenzerini, 2002).
GAV and LAV can be generalised into a mixed
approach, Global-Local-As-View (GLAV) map-
ping, also known as tuple-generating-dependency
(Lenzerini, 2002). The form of GLAV is ∀x(φ(x)
→∃y ψ (x,y)) which has more expressive power
than GAV and LAV. In fact, GAV and LAV are
special cases of GLAV.
644
In a FSDBS, schema mapping consists of (i)
mapping between local database schemas and the
federated schema, and (ii) mapping between dif-
ferent schema versions of one system.
Related Work
Schema evolution has been widely studied (Bounif
& Pottinger, 2006; Zhou, Rundensteiner, & Shin,
1997). Research on the impact of schema changes
has generated different methodologies and tools
to identify the impact of schema evolution on ap-
plications (Karahasanovic & Sjoberg, 2001; Maule,
Emmerich, & Rosenblum, 2008; Sjoberg, 1993).
Rather than rewriting applications, different
approaches have been developed to provide trans-
parent evolution to users. For example, schema
versioning uses a version-control mechanism in
which different versions of schemas and their cor-
responding data exist concurrently so applications
against different and legacy schemas are left un-
touched and remain valid (Monk & Sommerville,
1993; Roddick, 1995).
A view mechanism, based on the concept of ver-
sioning, is another approach that provides schema
evolution transparency to end users (Bellashsene,
1996; Young-Gook & Rundensteiner, 1997). In
this approach, different views are created for dif-
ferent applications. When a change is needed for
an application, only the view corresponding to that
application will be updated to reflect the changes
required. Other views remain unchanged and thus
other applications are also unchanged.
Curino (2009) automated schema evolution for
the Wikipedia system by adopting a query rewriting
approach. A predictive approach has been proposed
by Bounif (2006), which made a database schema
ready for evolution by predicting and anticipating
further possible change requirements before they
could occur.
In a federated database system, schema evolu-
tion in one database will affect other databases as
the schema mapping between the two schemas will
Managing Schema Evolution in a Federated Spatial Database System
become invalid. To overcome this inconsistency,
a number of approaches have been developed.
Version Management has been proposed in
a federated system to produce a new version
of mapping between local and federated sche-
mas (Schonhoff, Strassler, & Dittrich, 2001). A
meta-model of Model Correspondence Assertions
(MoCAs) has been developed which relates one
schema to others in a federated system (Busse &
Pons, 2001). Schema evolution can be propagated
using the MoCAs model and evolution action
specifications. An incremental schema integration
approach propagates change of the local schema to
the federated schema by identifying sub-schema
affected that are then re-integrated (Motz, 2005).
Most research on schema evolution has been
theoretical. There has not been much development
on semi-automatic and automatic procedures;
hence, schema evolution still requires a high degree
of manual work. In addition, very little research has
been carried out on schema evolution for spatial
database systems, and particularly for FSDBSs. As
Figure 3. The ASE framework for managing schema evolution in a FSDBS
a consequence, this chapter focuses on improving
on manual methods by managing schema evolution
in a FSDBS in a semi-automatic manner.
CONCEPTUAL AUTOMATIC SCHEMA
EVOLUTION FRAMEWORK
This section proposes and explains the conceptual
Automated Schema Evolution (ASE) Framework
and its component parts. The main task of the ASE
framework is to provide application independence
to schema changes in a FSDBS. The main compo-
nents shown in Figure 3 grouped into three layers:
the database layer, the middle layer and the user
interface, delineated by horizontal dashed lines.
The database layer includes the global database
and all local spatial databases participating in the
federation. A metadata repository is also included
to store all metadata needed for this framework.
The middle layer provides application immu-
nity to changes that occur in the database layer.
645
 Managing Schema Evolution in a Federated Spatial Database System
Two techniques are included: (1) view generation
or query rewriting for applications built on base
tables to keep applications unaffected by database
schema changes; and (2) view redefinition against
the new schemas for applications built against a
view to keep consistency between the view and
the evolved database schemas.
The user interface layer mainly deals with
processes involving interaction with the Data-
base Administrator (DBA) who manages schema
changes by specifying the schema changes and
performing the schema change analysis. Once a
schema change is initiated, scripts for the schema
change and data migration can be generated auto-
matically. Schema mapping is generated automati-
cally according to the mapping rules defined by
SCTs, and rewriting view definitions and update
of SEDs is semi-automatic with interaction from
the DBA.
The steps to perform a schema change include:
1. Schema change analysis: By specifying
the schema change and consulting SED
information, the DBA evaluates the impact
of the schema change.
2. Schema change design: If the DBA decides
to move forward with the change planned,
they can design the implementation of the
changes using a sequence of SCTs. SCTs
are developed according to schema change
taxonomy and represent different changes in
the spatial database.
3. Schema change implementation: Scripts to
implement schema changes can be generated
by specifying SCTs and participating schema
elements.
4. Schema mapping generation and com-
position: Each SCT has schema mapping
rules. These rules and participating schema
elements produce the schema mapping
between the old and new versions of the
schema when it changes. The schema map-
ping can be derived by combining the old
and new schema mappings. For example,
646
in the Landgate example of Figure 2 dis-
cussed previously, schema S1 changes to
S2 produced mapping M12. Schema S2 then
changes to S3 and produces M23. M12 and M23
are combined into mapping M13 between S1
and S3. Mapping combination also provides
the mapping changes between the federated
schema and the evolved local schema.
5. Mapping validation: Mappings between
schema versions can be reviewed by users
for correctness and overridden if needed.
6. View rewriting: According to SED and
schema mapping information, views that
have dependencies on the changed schema
elements can be detected and rewritten into
the equivalent expressions under the new
schema.
7. SED update: After a view is redefined, the
dependency information is updated replacing
the old one stored in the metadata repository.
8. View generation: Generating a new view
might occur for certain kinds of schema
change such as renaming a table.
9. Query rewriting: For applications built on
base tables, schema changes may break the
link between queries and the old version of
the schemas. For example, for a map applica-
tion, this may result in the removal of features
from the map. To solve this, run time query
rewriting is required. When a query against
an old version of the schema is processed,
the query will be translated into an equivalent
query expressed against the new version ac-
cording to the schema mapping.
Spatial Data Repositories
Spatial Data Repositories include local spatial
databases and the federated spatial database. Local
spatial databases contain base tables that store spa-
tial and non-spatial data and views. The federated
spatial database also consists of tables and views
as a virtual database built on the local databases.
Managing Schema Evolution in a Federated Spatial Database System
Schema Change Templates
A key objective is to identify the schema change
scenarios that can occur in a spatial database en-
vironment and generate the set of SCTs to satisfy
different schema changes. These SCTs can be
specified at the conceptual spatial schema level
and then mapped into schema changes on the
underlying SQL schema (Relational or Object-
Relational DBMS) according to mapping rules
between these two data models. Using the Simple
Features Access (SFA) specification (ISO 19125)
(OGC, 2006a, 2006b), Table 1 shows the mapping
between the conceptual spatial schema and the SQL
schema. See the section on the SED meta-model
for more details.
Table 2 lists the schema change taxonomy in
a conceptual schema and corresponding examples
for the Landgate example shown in Figure 2.
According to the mapping rules listed in Table
1, a set of SCTs can be generalised in the SQL
schema, as listed in Table 3 in which A, B, C, and
D are sets of attributes, a, b, c, and d are single
attributes, and R, T, and S are tables. Each SCT
consists of one or more ordered schema changes.
Schema primitive changes include addition or
deletion of columns, addition or deletion of tables
Table 1. Mapping between the conceptual spatial
schema and the SQL schema
Conceptual Spatial SQL Schema
Schema
Feature Object Class Table (one or more tables depend-
ing on the data types supported by
the DBMS)
Non-Spatial Object Class Table
Spatial Attributes Geometry Column or Geometry
table with reference to the feature
table
Non-Spatial Attribute Column
Relationship Foreign Key and Primary Key
(1:1 or 1:M)
Table with two Foreign Keys
(M:N)
etc. For example, moving a column involves add-
ing a column in the destination table, copying the
data from the original table and deleting the column
in the original table. For Merge Tables, In order
to be able rewrite queries, a column is added as
the condition when two tables are merged. For
Join Tables, to avoid data loss, an outer join is
needed.
To simplify schema changes, each SCT can
generate scripts automatically. SCTs capture the
semantics of change and can specify whether or
not a view should be generated. For example,
after renaming a table, a view with the same name
of the original table can be generated. In this re-
search, SCTs are the basis for schema mapping
because schema change is treated as the mapping
as explained later in the section on Schema Map-
ping.
Metadata Repository
Metadata is treated as the first class for schema
evolution. Currently, the most common strategy
for metadata management is to create a repository
to store metadata that can store different types
of metadata such as structural, descriptive, and
administrative metadata.
A metadata repository provides a consistent
and united access mechanism to data to improve
the effectiveness of information management. It
also performs other functions such as metadata
definition, transformation, mapping, synchroniza-
tion and versioning, as well as impact analysis of
metadata change (Sen, 2004).
There are four parts in the metadata reposi-
tory: (1) SEDs; (2) schema mapping; (3) schema
change history; and (4) metadata from which the
schema element dependency and schema mapping
are derived.
Schema Catalog
The system catalog in a RDBMS stores metadata
related to the relational database. For each of the
647
 Managing Schema Evolution in a Federated Spatial Database System

Table 2. Examples of conceptual spatial schema change taxonomy based on Figure 2

Conceptual Spatial Schema Examples

Feature Object Addition of a feature object class Add a Line feature object class Rails in S1
Class
Deletion of a feature object class Delete ProposedMainRoads in S1
Modification of Rename a Feature Object Class Rename LocalRoads to Roads in S2
a feature Object
Add an Attribute Add RoadCustodian in LocalRoads in S2
class 
Delete an attribute Delete RoadUsage in ProposedMainRoads in S1
Rename an attribute Change LaneCount to LaneNumber in S2
Non-spatial Addition of a non-spatial Object class  Add object class RoadName in S1
Object class
Deletion  of a non-spatial Object class  Delete FeatureName in S2
Modification Rename a non-spatial Object class  Rename FeatureNames to RoadNames in S2
of a non-spatial
Add an attribute Add attribute Creator in FeatureName in S2
Object class
Delete an attribute Delete FeatureType in FeatureNames in S2
Rename an attribute Rename Name to RoadName in FeatureNames in S2
Join feature object class with non-spatial object class Join LocalRoads with FeatureNames in S2
Split a feature object into a feature object class and non-spatial object class Slipt ExistingMainRoads into ExistingMainRoads and
RoadNames (new) in S1
Compose feature object classes into one  Compose ProposedMainRoads and Existing MainRoads
into MainRoads in S1
Decompose an object class into two or more object classes Decomposed FeatureNames into RoadNames and Other-
Names in S2
Decompose a feature object class into two or more feature object classes Decompose LocalRoads into ExistingLocalRoads and
ProposedLocalRoads in S2
Move an attribute Move Suburb from FeatureNames to LocalRoads in S2
Relationship Class Add a relationship Add relationship between ExistingMainRoads and
RoadType in S1
Delete a relationship Delete relationship between FeatureNames and Local-
Roads in S2
Modify the cardinality Modify the cardinality between FeatureNames and Loc-
alRoads in S2 from 1:1 to 1:n

Table 3. Schema change templates

local databases, the metadata contains informa-
Schema Operation Input Schema Output Schema
tion on the database schema including database
Add a column R(A) R(A,b)
tables, views, attributes, relationships, indexes,
Add a table T(A)
and security.
The federated spatial database system catalog
Rename a table R(A) T(A)
also stores metadata on the mapping between the
Merge tables R(A), S(A) T(A,b)
federated database schema and local database
Join tables R(a,B), S(a,C) T(a,B,C)
schema.
Split a table R(a,B,C) S(a,B), T(a,C)
Both types of catalog are needed for schema
Decompose a table R(A) S(A), T(A)
evolution in a FSDBS.
Move a column R(a,B,c), S(a,D) R(a,B), S(a,c,D)
Split a column R(a,B) R(c,d,B)
Merge columns R(a,b,C) R(d,C)

648
Managing Schema Evolution in a Federated Spatial Database System
The Schema Element
Dependency Meta-Model
The SED meta-model (illustrated in Figure 4
using UML syntax) is used to represent column
level schema element dependencies in a FSDBS.
This conceptual meta-model consists of three
parts: the schema elements of a conceptual spatial
schema, the schema elements of a SQL schema,
and the SEDs.
The conceptual spatial schema is a conceptual
data model adapted from the generic feature
model from ISO 19109 (2005). Schema elements
defined in the model include object classes consist-
ing of spatial feature and non-spatial object
classes, attributes including spatial attributes and
Figure 4. SED meta-model
non-spatial attributes, and relationships. A feature
object class represents a collection of vector objects
of the same type. A non-spatial object class rep-
resents a category of objects with no spatial char-
acteristics. An object class has a name and attributes
including spatial (for a spatial object class) and
non-spatial attributes. Each attribute has its own
properties including name, data type etc. Spatial
attribute data types are points, lines, or polygons.
Object classes can have relationships between
them.
The SQL schema is an implementation of
a conceptual spatial schema according to SFA
specifications (OGC, 2006a, 2006b). Schema ele-
ments include tables and views that are supported
by contemporary relational or object-relational
649
 Managing Schema Evolution in a Federated Spatial Database System
DBMSs. A table contains the name of the table,
names of all columns and data types for these
columns. A view includes the name of the view
and names of columns, required because a view
is a virtual table that is derived from base tables
or views by means of stored queries. A view may
also include other clauses in its definition such as
where and having.
There are three types of tables in a DBMS:
feature table, geometry table and non-spatial table.
A feature table stores a collection of features of
the same type. Attributes of a feature object class
including the spatial attributes are columns of a
feature table. A geometry table stores the geometric
information of the geometric objects. Tables with-
out any spatial element are called non-spatial tables.
The relationship between the conceptual spatial
schema and the SQL schema can be seen in Figure
4. Object classes are stored as tables in the relational
database. More specifically, the implementation
of a non-spatial object class is a non-spatial table
while the implementation of a feature object class
varies depending on the data types supported by the
underlying DBMS. For example, a feature object
class uses a feature table and a geometry table, with
a key reference between them if the predefined data
types in the DBMS are used. However, only the
feature table is needed if geometry data types are
supported by the DBMS. Attributes of an object
class are stored as columns of the corresponding
table. A relationship between two object classes is
implemented as either a table or the primary key
and foreign key column of the two tables.
Schema element dependencies consist of sche-
ma elements and the relationships between them.
Schema elements are divided into Compound and
Part. The Part component denotes atomic elements
such as a column. The Compound component
represents the elements that contain parts such as
tables and views. Relationships include Referential
Integrity (RI), DerivedFrom and ReferencedBy. A
relationship connects two schema elements. The
dependency information, termed a column level
dependency, can be derived from the relationships
650
between the source and the target elements that
are connected by the relationship. Depending on
the databases containing the source and target
elements, the dependency will be identified as an
internal or external dependency.
In a DBMS, some schema elements may
refer to other elements when they are defined.
For example, a view is defined by a query that
refers to one or more tables or one or more views.
Normally, the element that is being referred to
is called a referenced element. The element that
references the other element is called a dependent
element, because it depends on the other element
(or referencing element).
If one schema element is changed, the depen-
dent elements may become invalid. For example,
when a table is renamed, the views that are defined
on this table will become invalid if the definitions
of the views are not updated to the new name of
the corresponding table.
There are two kinds of dependencies between
database schema elements, direct dependency
and indirect dependency. Direct dependency is
where the referenced element is referred by the
dependent element directly. Indirect dependency
is where the referenced element is referred to by
an element that is referred to by another element.
The dependencies derived from the metadata
according to the SED meta-model are direct depen-
dencies. Indirect dependencies can be determined
iteratively based on the direct dependencies.
Schema Mapping
Schema mapping in the metadata repository con-
sists of horizontal and vertical schema mapping.
Horizontal mapping refers to mapping between
different systems such as the global schemas and
local schemas while vertical mapping refers to
mapping between an old version and a new ver-
sion of particular schema in one database system.
In this research describing schema change as
schema mapping is used. When a database schema
changes, the schema change itself is treated as the
Managing Schema Evolution in a Federated Spatial Database System

schema mapping between the old and new ver- Roads, and, the second one is to split Main-
sion. Each change operation has a corresponding Roads into MainRoads and RoadNames.
mapping rule between schema versions before and Table 5 lists the schema changes and their
after the change. corresponding schema mappings.
The corresponding schema mapping of each
SCT is shown in Table 4 in which F(), F1(), and Before the schema change, the schema mapping
F2() are functions for splitting or merging columns. MSG between the federated schema G and the local
In order to avoid loss of information, deletion of schema S1 and S2 is:
schema elements is not allowed. Under such a
condition, the expressive power of GAV (Global- G.RoadSegments (n, g, lc, rt, sl)
as-View that represents the global schema in terms
of local schema as described in the section on ← (S2.LocalRoads (g, lc, rt, st, nID) ⋈
Schema Mapping) is enough to describe schema S2.FeatureNames (n, nID))
mapping. The language used to describe mapping
here is relational algebra but other representations ∪ ((S11. ExistingMainRoads (n, g, lc, sl,rtID) ⋈
could be used such as relational calculus. Schema S11.RoadType (tID, tn))
change can be achieved by applying one or more
SCTs. All the mapping rules from SCTs form the ∪ (S11.ProposedMainRoads (n, g, lc, sl, rtID) ⋈
S11.RoadType (tID, tn)))
schema mapping between the old and new versions
of the schema.
where n, g, lc, ru, sl, rs, rtID, nID and rn denote
Name, Geometry, LaneCount, RoadUsage, Speed-
Example 1. Suppose there are two schema
Limit, RoadStatus, RoadTypeID, NameID and
changes in schema S1, as shown in Figure
RoadName respectively.
5. The first change is to merge ExisingMain-
Schema evolution can result in the schema
Roads and ProposedMainRoads into Main-
mapping being invalid. For example, in Figure 5,
the schema mapping MSG becomes invalid when
Table 4. Schema change and corresponding S11 changes. In order to ensure the consistency
schema mapping of schema mapping, schema mapping combines
two schema mappings into one (Nash, Bernstein,
Schema change Schema Mapping & Melnik, 2007). This approach can be applied
Add a Column π R(A) ← π R(A,b) to vertical and horizontal mapping. When a data
Add a Table source schema changes, a new schema mapping
Rename a Table R(A) ← T(A) between the global schemas and local schemas
Merge Tables
R(A) ← σ (T(A,b)), S(A) ← can be derived from the original mapping and the
σ(T(A,b))
change schema mapping. Likewise, for the local
R(a,B)← π T(a,B,C)
Join Tables
S(a,C)← π T(a,B,C)
schemas, schema mapping between older versions
and the current one can be derived from the schema
Split a Table R(a,B,C)←S(a,B) ⋈ T(a,C)
changes and the old schema mapping.
Decompose a Table R(A) ← S(A) ∪ T(A)
In Example 1, mapping M13 between S11 and
R(a,B,c) ← π σ (R(a,B) ⋈ S(a,c,D))
Move a Column
S(a,D) ← π S(a,c,D)
S13 can be derived from M12 and M23. Mapping
Split a Column R(a) ←F(R(c,d))
MSG between the global schema G and local
R(a) ←F1(R(d)),
schema S1 and S2 has to be adapted to keep con-
Merge Columns sistency of the mapping. Similarly, MSG′ and MSG″
R(b) ←F2(R(d))

651
 Managing Schema Evolution in a Federated Spatial Database System

Table 5. Schema changes and corresponding schema mapping in Example 1

Mapping Name Schema Change Schema Mapping

M12 (mapping be- merge ExisingMainRoads S11.ExistingMainRoads (n, g, lc, ru, sl, rtID) ← σ rs =
tween S11 and S12) and ProposedMainRoads into ’Existing’ (S12.MainRoads (n, g, lc,ru, sl, rs, rtID))
MainRoads
S11.ProposedMainRoads (n, g, lc,ru, sl,rtID) ← σ rs =
’Proposed’ (S12.MainRoads (n, g, lc,ru, sl, rs, rtID))
M23(mapping be- split MainRoads into Main- S12.MainRoads (n, g, lc,ru, sl, rs, rtID) ← S13.MainRoads (nID, g, lc,ru, sl, rs,
tween S12 and S13) Roads and RoadNames rtID) ⋈ S13.RoadNames (nID, rn)
n, g, lc, ru, sl, rs, rtID, nID and rn denote Name, Geometry, LaneCount, RoadUsage, SpeedLimit, RoadStatus, RoadTypeID, NameID, and
RoadName, respectively.

can be derived from schema mapping (M12 and
M23) and the original mapping (MSG and MSG′
later). Table 6 lists mappings that are generated
by mapping combination for Example 1.
Schema Change History
In the metadata repository, the schema change
history keeps a record of each schema change
providing clear information of schema versions
and how the schema evolves. The semantics of the
schema change can be determined and examining
the history records if any errors occur can reverse
schema changes.
Query/View Rewriting and
View Generation
For a spatial database system, applications such as
map viewing access data stored in the underlying
database. The users can build queries against the
base tables or views (spatial views if there is a
spatial column).

Figure5. Example of schema changes

652
Managing Schema Evolution in a Federated Spatial Database System
Table 6. Schema mapping derived from combination in Example 1
Mapping Name Mapping Combination
M13 (mapping between S11 M13 = M12 • M13
and S13)
MSG′ (mapping between MSG′ = MSG • M12
G and S1 and S2 when S1
changes from S11 to S12)
MSG″ (mapping between MSG″ = MSG′ • M23
G and S1and S2 when S1
changes from S12 to S13)
n, g, lc, ru, sl, rs, rtID, nID and rn denote Name, Geometry, LaneCount, RoadUsage, SpeedLimit, RoadStatus, RoadTypeID, NameID, and
RoadName, respectively.
In order to manage schema evolution in a
spatial database system, two kinds of scenarios
have to be considered: query using base tables and
query using views. For a query using views, the
views need to be redefined with the change of the
schemas. When a query is built against the base
tables, two methods can be used in order to support
old queries expressed against old schema. They
are (1) query rewriting and (2) the generation of
a new view. Query rewriting is used to ensure the
queries generated from the applications against the
old schema can be translated into queries against
the new schema. By doing this, applications will
be still valid. View generation does not apply to all
schema changes. Only certain changes can gener-
ate a view to keep the application impervious to
schema change such as the renaming of a table.
If an application is built on views, changes
of schemas might invalidate the view definition.
Therefore, view rewriting is used to redefine and
recompile the view after schema changes. Since a
view is a stored query, view rewriting is very similar
to query rewriting. However, there are differences
between them. Firstly, view rewriting is a one-off
operation. It only occurs when schema changes
occur, whereas query rewriting happens every
time the query is processed. This is because any
affected view can be detected by SEDs while the
queries from the application can only be detected
Derived Mapping
S11.ExistingMainRoads (n, g, lc,ru, sl,rtID) ← σ rs =’ Existing’(S13.Main-
Roads (nID, g, lc,ru, sl, rs, rtID) ⋈ S13.RoadNames (nID, rn))
S11.ProposedMainRoads (n, g, lc,ru, sl,rtID) ← σ rs =’ Proposed’(S13.
MainRoads (nID, g, lc,ru, sl, rs, rtID) ⋈ S13.RoadNames (nID, rn))
G.RoadsSegments (n, g, lc, rt, sl) ← (S2.LocalRoads (g, lc, rt, st, nID) ⋈
S2.FeatureNames (n, nID)) ∪(σ rs =’ Existing’∪ rs =’ Proposed’ ((S12.
MainRoads(n, g, lc, sl,rtID) ⋈ S12.RoadType (tID, tn)))
G.RoadsSegments (n, g, lc, rt, sl) ←(S2.LocalRoads (g, lc, rt, st, nID) ⋈
S2.FeatureNames (n, nID)) ∪ (σ rs =’ Existing’∪ rs =’ Proposed’(S13.
MainRoads (nID, g, lc,ru, sl, rs, rtID) ⋈ S13.RoadNames (nID, rn) ⋈ S13.
RoadType (tID, tn)))
when they are processed. Secondly, view rewrit-
ing must ensure the view schema is unchanged,
as the change of the view schema will invalidate
the applications built on that view. For example,
the view name and its column names should not
be changed. Nonetheless, query rewriting only
needs to provide equivalent results.
Different algorithms have been developed
for query rewriting such as inverse rules, bucket
algorithm and the Chase and BackChase (C&B)
algorithm (Bai, Hong, McTear, & Wang, 2006;
Deutsch, et al., 2006; Duschka & Genesereth,
1997b). The C&B algorithm has attracted most
attention and was first developed to rewrite que-
ries under a set of constraints in order to improve
query optimisation (Deutsch, Popa, & Tannen,
1999; Deutsch, et al., 2006).
In this research, query rewriting is simply
achieved by unfolding and substituting. As de-
scribed in the section on Schema Mapping, GAV
represents the global schemas in terms of local
schemas. The advantage of GAV is the simplicity
for query rewriting and the combination of two
GAV mappings is still a GAV mapping. Therefore,
schema mapping in this research is restricted to this
form of GAV mapping. Under such circumstance,
a query can be rewritten by unfolding (Cali, 2003).
With substitution, queries expressed against the
global schema can be rewritten as equivalent ones
653
 Managing Schema Evolution in a Federated Spatial Database System
against the source schemas. Similarly, queries
expressed against old schemas can be rewritten
against the new schemas.
Example 2. Suppose application A1 in Figure
2 is the map application that displays all
highways from S1 and Q1 is the query gen-
erated by A1. When S1 changes from S11 to
S12, as shown in Figure 5, by substituting
ExistingMainRoads(n, g, lc, sl,rtID) and
ProposedMainRoads(n, g, lc, sl, rtID) accord-
ing to schema mapping M12, Q1 against S11
can be rewritten against S12. Table 7 shows
the query Q1 (expressed in relational algebra)
before and after rewriting.
CONCLUSION
Schema evolution is inevitable in a FSDBS
because of the likelihood of its dependency on
databases owned and maintained by other or-
ganisations. Managing schema evolution in such
an environment is a significant challenge. This
research aims to develop the methodologies to
semi-automatically manage schema evolution in
a FSDBS environment so that schema changes
will be transparent to applications and users. To
achieve this, view generation/rewriting and query
rewriting are proposed as the solution.
In order to implement the methodology, a SED
meta-model and a set of SCTs have been developed.
The SED meta-model defines the constructs to
build column level dependencies in a FSDBS. The
schema element dependencies can then be derived
from the SED meta-model and the metadata of the
spatial database.
Table 7. Query Q1 before and after rewriting
Before Rewriting (against S11)
σ tn =’highway’((S11.ExistingMainRoads(n, g, lc, sl,rtID) ∪ S11.
ProposedMainRoads (n, g, lc, sl, rtID)) ⋈ S11.RoadType (tID, tn))
654
Schema element dependencies provide the ba-
sis for impact analysis and determine which view
definition needs to be redefined when the schema
changes. SCTs are used to define a rich bounded
set of schema change scenarios in a spatial data-
base environment and provide the mapping rules
for each change.
By incorporating the mapping composition,
schema mappings between the older versions and
the new version of a local schema can be derived.
Schema mapping adaptation between the global
and the local schema can be achieved as well. SCTs
also provide the solution as to whether or not a
view should be generated. Based on the schema
mapping information, run time queries and views
that are expressed against an old schema version
can be rewritten into expressions in terms of the
current schema by unfolding and substituting. The
end result is that applications will be immune to
schema changes.
Effective management of schema evolution in
a FSDBS is an integral part of an SDI. It ensures
discovery and access to spatial data and services
is continuous even though schema changes have
occurred. It also provides schema evolution trans-
parency to end-users and minimises modification of
applications thus saving time and money. Effective
management of schema evolution also increases
the reusability of spatial data; ensures the longevity
of spatial data and services well into the future;
improves spatial data and service management;
and engenders collaboration and interoperability
between data custodians and application develop-
ers and users.
Future developments include methods to deal
with federated fully functional spatial databases
as well as developing the semi-automatic methods
After Rewriting (against S12)
σ tn =’highway’ ((σ rs =’Existing’ ∪ σ rs =
’Proposed’ (S12.MainRoads (n, g, lc,ru, sl, rs, rtID) ⋈ S12.Road-
Type (tID, tn))
Managing Schema Evolution in a Federated Spatial Database System
described in this chapter to be fully automatic.
Semi-automatic and fully automatic methods
will significantly advance federated databases
and move away from the more community-based
methods.
REFERENCES
ANZLIC. (2011). Australian spatial data infra-
structure. Retrieved from http://www.anzlic.org.
au /infrastructure_ASDI.html.
Bai, Q., Hong, J., McTear, M. F., & Wang, H.
(2006). A bucket-based approach to query re-
writing using views in the presence of inclusion
dependencies. Journal of Research and Practice
in Information Technology, 38(3), 251–266.
Bellashsene, Z. (1996). Views for schema evolu-
tion in object-oriented database systems. IEEE
Transactions on Knowledge and Data Engineer-
ing, 9(4), 600–624.
Bernstein, P. A., Green, T. J., Melnik, S., & Nash,
A. (2008). Implementing mapping composition.
The VLDB Journal, 17(2), 333–353. doi:10.1007/
s00778-007-0059-9
Bounif, H., & Pottinger, R. (2006). Schema re-
pository for database schema evolution. Paper
presented at the the 17th International Conference
on Database and Expert Systems Applications.
New York, NY.
Busse, S., & Pons, C. (2001). Schema evolution
in federated information system. Datenbanksys-
teme in Büro [BTW]. Technik und Wissenschaft,
9, 26–43.
Cali, A. (2003). Reasoning in data integration
systems: Why LAV and GAV are siblings. Berlin,
Germany: Springer.
Cali, A., Calvanese, D., Giacomo, G. D., & Len-
zerini, M. (2002). On the expressive power of
data integration systems. Paper presented at the
Proceedings of the 21st International Conference
on Conceptual Modeling. New York, NY.
Curino, C., Moon, H., Ham, M., & Zaniolo, C.
(2009). The PRISM workwench: Database schema
evolution without tears. Paper presented at the
the 2009 IEEE International Conference on Data
Engineering. New York, NY.
Date, C. J. (2003). An introduction to database
system (8th ed.). Boston, MA: Addison Wesley.
Deutsch, A., Popa, L., & Tannen, V. (1999). Physi-
cal data independence, constraints, and optimiza-
tion with universal plans. Paper presented at the
25th International Conference on Very Large Data
Bases. New York, NY.
Deutsch, A., Popa, L., & Tannen, V. (2006). Query
reformulation with constraints. Paper presented at
the SIGMOD Rec. New York, NY.
Duschka, O. M., & Genesereth, M. R. (1997).
Answering recursive queries using views. In Pro-
ceedings of the Sixteenth ACM SIGACT-SIGMOD-
SIGART Symposium on Principles of Database
Systems. Tucson, AZ: ACM.
Egenhofer, M. J. (1994). Spatial SQL: A query
and presentation language. IEEE Transactions on
Knowledge and Data Engineering, 6(1), 86–95.
doi:10.1109/69.273029
FDGC. (2011). National spatial data infrastruc-
ture. Retrieved from http://www.fgdc.gov /nsdi/
nsdi.html.
Goesseln, G., & Sester, M. (2003). Semantic and
geometric Integration of geoscientific data sets
with ATKIS - Applied to geo-objects from geology
and soil science. Paper presented at the ISPRS
Commission IV Joint Workshop “Challenges in
Geospatial Analysis, Integration and Visualization
II”. Stuttgart, Germany.
655
 Managing Schema Evolution in a Federated Spatial Database System
Halevy, A. Y. (2001). Answering queries using
views: A survey. The VLDB Journal, 10(4),
270–294. doi:10.1007/s007780100054
INSPIRE. (2009). INSPIRE generic conceptual
model. Geneva, Switzerland: INSPIRE Infrastruc-
ture for Spatial Information in Europe.
INSPIRE. (2011). About INSPIRE. Retrieved
from http://inspire.jrc.ec.europa.eu /index.cfm/
pageid/48.
ISO. (2002). ISO 19101: Geographic information
- Reference model. Geneva, Switzerland: ISO.
ISO. (2003). ISO 19107: Geographic information
-Spatial schema. Geneva, Switzerland: ISO.
ISO. (2005). ISO 19109: Geographic informa-
tion -- Rules for application schema. Geneva,
Switzerland: ISO.
Karahasanovic, A., & Sjoberg, D. (2001). Visu-
alizing impacts of database schmea changes - A
controlled experiment. Paper presented at the the
IEEE 2001 Symposia on Human Centric Comput-
ing Languages and Environments (HCC 2001).
New York, NY.
Lenzerini, M. (2002). Data integration: A theoreti-
cal perspective. In Proceedings of the Twenty-First
ACM SIGMOD-SIGACT-SIGART Symposium on
Principles of Database Systems. Madison, WI:
ACM.
Levy, A. Y., Rajaraman, A., & Ordille, J. J. (1996).
Querying heterogeneous information sources using
source descriptions. Paper presented at the 22nd
International Conference on Very Large Data
Bases. New York, NY.
Litwin, W., Mark, L., & Roussoupoulos, N. (1990).
Interoperability of multiple autonomous data-
bases. ACM Computing Surveys, 22(3), 267–293.
doi:10.1145/96602.96608
656
Maule, A., Emmerich, W., & Rosenblum, D.
S. (2008). Impact analysis of database schema
changes. ICSE, 10-17. Retrieved from http://
research.microsoft.com /pubs/118211/Andy%20
Maule%20-%20Thesis.pdf.
Monk, S., & Sommerville, I. (1993). Schema evolu-
tion in OODBS using class versioning. SIGMOD
Record, 22(3), 16–22. doi:10.1145/163090.163094
Motz, R. (2005). Dynamic maintenance of an
integrated schema. In From Integrated Publica-
tion and Information Systems to Information and
Knowledge Environments (Vol. 3379, pp. 21–30).
Berlin, Germany: Springer. doi:10.1007/978-3-
540-31842-2_3
Nash, A., Bernstein, P. A., & Melnik, S. (2007).
Composition of mappings given by embedded
dependencies. ACM Transactions on Database
Systems, 32(1), 4. doi:10.1145/1206049.1206053
Nebert, D. D. (2004). Developing spatial data
infrastructure: The SDI cookbook. The Global
Spatial Data Infrastructure (GSDI). Retrieved
from http://www.gsdi.org.
OGC. (2006a). OpenGIS® implementation specifi-
cation for geographic information - Simple feature
access - Part 1: Common architecture. Retrieved
from http://www.opengeospatial.org.
OGC. (2006b). OpenGIS® implementation specifi-
cation for geographic information - Simple feature
access - Part 2: SQL option. Retrieved from http://
www.opengeospatial.org.
Parent, C., Spaccapietra, S., & Zimányi, E. (2006).
Conceptual modeling for traditional and spatio-
temporal applications, the MADS approach.
Berlin, Germany: Springer.
Popa, L. (2001). Object/relational query optimiza-
tion with chase and backchase. PhD Disseration.
Philadelphia, PA: University of Pennsylvania.
Managing Schema Evolution in a Federated Spatial Database System
Ramakrishnan, R., & Gehrke, J. (2003). Database
management systems (3rd ed.). New York, NY:
McGraw-Hill.
Rigaux, P., Scholl, M., & Voisard, A. (2002).
Spatial database - With application to GIS. New
York, NY: Morgan Kaufmann.
Roddick, J. F. (1995). A survey of schema ver-
sioning issues for database systems. Informa-
tion and Software Technology, 37(7), 383–393.
doi:10.1016/0950-5849(95)91494-K
Schonhoff, M., Strassler, M., & Dittrich, K. R.
(2001). Version propagation in federated data-
base systems. Paper presented at the International
Databases and Applications Symposium (IDEAS
2001). Grenoble, France.
Sen, A. (2004). Metadata management: Past, pres-
ent and future. Decision Support Systems, 37(1),
151–173. Retrieved from http://www.sciencedirect.
com /science/article/B6V8S-47X1YN5-1/2/ d75b-
b1ab996404d75a0203f7e97c6041 doi:10.1016/
S0167-9236(02)00208-7
Sheth, A. P., & Larson, J. A. (1990). Feder-
ated database systems for managing distrib-
uted, heterogeneous, and autonomous databases.
ACM Computing Surveys, 22(3), 183–236.
doi:10.1145/96602.96604
Sjoberg, D. (1993). Quantifying schema evolu-
tion. Information and Software Technology, 35(1),
35–44. doi:10.1016/0950-5849(93)90027-Z
SLIP. (2011). Welcome to SLIP enabler. Retrieved
from https://www2.landgate.wa.gov.au /web/
guest.
Tatarinov, I., & Halevy, A. (2004). Efficient query
reformulation in peer data management systems.
Paper presented at the 2004 ACM SIGMOD In-
ternational Conference on Management of Data.
Paris, France.
This work was previously published in Discovery of Geospatial Resources: Methodologies, Technologies, and Emergent
Applications, edited by Laura Díaz, Carlos Granell and Joaquín Huerta, pp. 56-77, copyright 2012 by Information Science
Reference (an imprint of IGI Global).
Ullman, J. (2000). Information integration
using logical views. Theoretical Computer
Science, 239(2), 189–210. doi:10.1016/S0304-
3975(99)00219-4
Yeung, A. K. W., & Hall, G. B. (2007). Spatial
database systems - Design, implementation and
project management. Berlin, Germany: Springer.
Young-Gook, R., & Rundensteiner, E. (1997). A
transparent schema evolution system based on
object-oriented view technology. IEEE Transac-
tions on Knowledge and Data Engineering, 9(4),
600–624. doi:10.1109/69.617053
Zhou, L., Rundensteiner, A. E., & Shin, G. K.
(1997). Schema evolution of an object-oriented re-
al-time database system for manufacturing automa-
tion. IEEE Transactions on Knowledge and Data
Engineering, 9(6), 956. doi:10.1109/69.649319
KEY TERMS AND DEFINITIONS
Federated Spatial Database System: An
integrated spatial database system. It integrates
multiple autonomous spatial data sources into a
single federated database and provides a unified
data access mechanism to applications.
Spatial Query: A query deals with data of 2 or
3 dimensions by supporting spatial data types, spa-
tial operators, spatial operations and spatial joins.
Spatial View: A stored query that contains a
spatial column.
Schema Mapping: Specification of the cor-
respondences between schemas.
Schema Mapping Composition: To compose
two schema mappings into one mapping.
657
658

Chapter 40
Map Matching Algorithms for
Intelligent Transport Systems
Mohammed A. Quddus
Loughborough University, UK

ABSTRACT
Map matching algorithms integrate positioning data with spatial road network data to support the
navigation modules of intelligent transport systems requiring location and navigation data. Research
on the development of map matching algorithms has significantly advanced over the last few years. This
article looks at different methods that have been adopted in map matching algorithms and highlights
future trends in map matching and navigation research.

INTRODUCTION Services. Two essential components required for

A map matching algorithm integrates positioning
data with spatial road network data for the iden-
tification of the correct link on which a vehicle is
travelling and to determine the physical location
of the vehicle on the link. This algorithm could
be used as a key component in the navigation
modules of intelligent transport systems (ITS)
that require locational data of a relatively high
quality. Such ITS applications include navigation
and route guidance, bus priority at junctions, fleet
management, road user charging, accident and
emergency management, and Location-Based
DOI: 10.4018/978-1-4666-2038-4.ch040
the navigation module of an ITS are: (1) a device
to determine the geometric position of the vehicle
and (2) a spatial road network database for spatial
(physical) reference of the vehicle location.
The most common geometric positioning
devices used to obtain inputs to a map matching
algorithm are Dead Reckoning (DR) motion sen-
sors, global navigation satellite systems (GNSS)
such as the Global Positioning System (GPS),
and integrated navigation systems such as the
integration of GPS and DR. The review of the
literature suggests that the integrated GPS and
DR system increases the performance of a map
matching algorithm if all other factors are equal,
especially in the case for dense and built-up urban

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Map Matching Algorithms for Intelligent Transport Systems
areas. However, even with good sensor calibration
and sensor fusion technologies, inaccuracies in
the GPS/DR system are inevitable. This is mainly
due to the errors in both GPS and DR sensors.
The spatial road network data are the other
important component of a map matching algo-
rithm. Since land vehicles primarily travel on
known road networks, the spatial road network
data are used as a spatial reference to locate them.
This, for example, assists drivers to relate their
observed positions obtained from the navigation
system with a physical location in the real-world
and hence can guide them along a pre-calculated
route. However, the spatial road network data also
contain errors arising mainly from the processes
of creation and digitization of maps. The errors
can be estimated using either the scale of the map
or field experiments.
The performance of a map matching algorithm
can be improved in two ways: (1) by controlling the
errors in both positioning sensors and spatial road
network data and (2) by improving the technique
used in the map matching process. A number of
map matching algorithms have been developed by
researchers around the world using different tech-
niques such as topological analysis of spatial road
network data, probabilistic theory, Kalman filter,
fuzzy logic, and belief theory. The performance of
these algorithms has improved over the years due
to the application of advanced techniques in the
map matching processes and to the improvements
of the quality of both positioning data and spatial
road network data over time. The purpose of this
article is to present an in-depth literature review
on map matching algorithms and to suggest the
future trends in map matching research.
The rest of the article is structured as follows.
The next section describes methodologies used in
map matching algorithms. This is followed by a
description of future trends in the development
of map matching algorithms. Finally, the conclu-
sion is given.
METHODOLOGIES USED IN MAP
MATCHING ALGORITHMS
The general purpose of a map matching algorithm
is to identify the correct road segment on which the
vehicle is travelling and to determine the vehicle
location on that segment. The parameters used to
select a precise road segment are mainly based
on the proximity between the position fix and the
road, the degree of correlation between the vehicle
trajectory derived from the position fixes and
the road centreline, and the topology of the road
network. Orthogonal projection of the position fix
onto the selected road segment is normally used
to calculate the vehicle location on the segment.
Figure 1 shows a general map matching process
(see Quddus, 2006 for details) which takes inputs
from an integrated GPS/DR such as easting (E),
northing (N), speed (v), and heading (θ) and the
error variances associated with them. The map
matching process also takes inputs from a spatial
digital road network database. The outputs of the
algorithm are the correct link on which the vehicle
is travelling and the location of the vehicle () and
the error variances associated with them.
Approaches for map matching processes found
in the literature can be categorised into three
groups: geometric, topological and advanced.
These are briefly discussed below.
Geometric Map Matching Algorithms
A geometric map matching algorithm makes use
of the geometric information of the digital road
network by considering only the shape of the links
(Greenfeld, 2002). It does not consider the way
links are connected to each other. The most com-
monly used geometric map matching algorithm is
a simple search algorithm. In this approach, each
of the positioning fixes matches to the closest
‘node’ or ‘shape point’ of a road segment. This is
known as point-to-point matching (White et al.,
2000). This approach is both easy to implement
and very fast. However, it is very sensitive to
659
 Map Matching Algorithms for Intelligent Transport Systems
Figure 1. A map matching algorithm
the way in which the network was digitised and
hence has many problems in practice. That is,
other things being equal, arcs with more shape
points are more likely to be properly matched. In
a straight arc with two end nodes, all positioning
points that fall above the arc only match with the
arc’s end nodes.
Another geometric map matching approach
is point-to-curve matching (White et al., 2000).
In this approach, the position fix obtained from
the navigation system is matched onto the closest
curve in the network. Each of the curves is com-
prised of a set of piecewise linear line segments.
Distance is calculated from the position fix to each
of the line segments. The line segment which gives
the smallest distance is selected as the one on which
the vehicle is apparently travelling. Although this
approach gives better results than point-to-point
matching, it does have several shortcomings that
make it inappropriate in practice. For example, it
gives very unstable results in urban networks due
to the high road density. Moreover, the closest link
may not always be the correct link.
The other geometric approach is to compare
the vehicle’s trajectory against known roads. This
is also known as curve-to-curve matching (White
660
et al., 2000; Phuyal, 2002). This approach first
identifies the candidate nodes using point-to-
point matching. Then, given a candidate node, it
constructs piecewise linear curves from the set
of paths that originate from that node. Secondly,
it constructs piecewise linear curves using the
vehicle’s trajectory i.e., the positioning points,
and determines the distance between this curve
and the curve corresponding to the road network.
The road arc which is closest to the curve formed
from positioning points is taken as the one on
which the vehicle is apparently travelling. This
approach is quite sensitive to outliers and depends
on point-to-point matching, sometimes giving
unexpected results.
Topological Map Matching
Algorithms
In GIS, topology refers to the relationship be-
tween entities (points, lines, and polygons). The
relationship can be defined as adjacency (in the
case of polygons), connectivity (in the case of
lines), or containment (in the case of points in
polygons). Therefore, a map matching algorithm
which makes use of the geometry, connectivity, and
Map Matching Algorithms for Intelligent Transport Systems
contiguity of the links is known as a topological
map matching algorithm (Greenfeld, 2002, and
Meng et al., 2003).
Greenfeld (2002) reviews several approaches
for solving the map matching problem and pro-
poses a weighted topological algorithm. This is
based on a topological analysis of a road network
and uses only coordinate information on observed
positions of the user. It does not consider any
heading or speed information determined from
GPS. This method is very sensitive to outliers as
these can lead to inaccurate vehicle heading cal-
culations. The other deficiency of this algorithm
is not to take into account the vehicle speed as
GPS position fixes are scattered randomly when
the speed is less than 3.0 m/sec (Taylor et al., 2001
and Ochieng et al., 2004).
An enhancement of the topological map match-
ing algorithm is proposed by Quddus et al. (2003).
The basic characteristic of this algorithm is the use
of navigation data from either GPS or an integrated
GPS/DR system that includes position, speed, and
heading. Data on the historical trajectory of the
vehicle (i.e., data on earlier map-matched posi-
tions) is used to avoid sudden switching of mapped
locations between unconnected road links. The
topological aspects of the road network such as
link connectivity and shape points (representing
link curvature) are used to improve the perfor-
mance of the algorithm. Heading and speed data
from the positioning sensors are also employed
to further improve the algorithm’s performance,
especially at junctions.
Advanced Map Matching Algorithms
Advanced map matching algorithms are referred to
as those algorithms that use more refined concepts
such as a probabilistic theory (Zhao, 1997, Ochieng
et al., 2004), a Kalman Filter (e.g., Tanaka et al.,
1990, Kim et al., 2000), the Dempster-Shafer’s
theory (also known as Belief Theory) (e.g., Najjar
and Bonnifait, 2003; Yang et al., 2003), a fuzzy
logic model (e.g., Zhao, 1997; Kim et al., 1998,
Kim and Kim 2001, Syed and Cannon, 2004,
Quddus et al., 2006), or the application of Bayes-
ian interference (Pyo et al., 2001). The details of
these algorithms can be found in Quddus (2006).
However, the probabilistic and fuzzy logic-based
algorithms are briefly discussed below.
The probabilistic approach to a map match-
ing algorithm was first introduced by Honey et
al. (1989) for the case of a navigation solution
obtained from a DR sensor. Zhao (1997) extended
this concept for the case of a navigation solution
obtained from GPS. Other studies which used a
probabilistic approach or a part of this approach
are, for example, Kim et al. (2000) and Pyo et
al. (2001. The probabilistic algorithm requires
the definition of an elliptical or rectangular
confidence region around a position fix obtained
from a navigation sensor. The error region is then
superimposed on the road network to identify a
road segment on which the vehicle is travelling.
If an error region contains a number of segments,
then the evaluation of candidate segments are
carried out using heading, connectivity, and
closeness criteria. Indeed, there are many other
parameters e.g., speed of the vehicle, distance to
the downstream junction, etc. that can be used to
improve the probabilistic map matching process.
Ochieng et al. (2004) develops an enhanced
probabilistic map matching algorithm. In this
algorithm, the elliptical error region is only
constructed when the vehicle travels through a
junction, in contrast to constructing it for each po-
sition fix as suggested by Zhao (1997). Therefore,
there is no need to create the error region when
the vehicle travels along a link. This makes the
algorithm faster as there are a number of processes
involved in the creation of the error region and
hence the identification of the correct link. The
method is more reliable as the construction of an
error region in each epoch may lead to incorrect
link identification.
Fuzzy logic based map matching algorithms
have been developed by a number of researchers
(e.g., Zhao, 1997; Kim et al., 1998, Fu et al., 2004,
661
 Map Matching Algorithms for Intelligent Transport Systems
Syed and Cannon, 2004) over the last few years.
The limitations of these algorithms include: (a)
ignoring most of the available inputs to the fuzzy
logic model, (b) overlooking the connectivity
among road links and the historical trajectory of
the vehicle, which can enhance the performance
of map matching algorithms, (c) ignoring the error
sources associated with the spatial road network
and the navigation sensor, and (d) failing to vali-
date the map matching algorithms to assess their
performance.
Quddus et al. (2006) developed an improved
fuzzy logic-based map matching algorithm that
took account of these limitations. Specifically, a
number of new input variables (at no extra cost)
were included in the fuzzy logic map matching
algorithm. These were: the speed of the vehicle,
the connectivity among road links, the quality
of position solution, e.g., Horizontal Dilution
Of Precision (HDOP), and the position of a fix
relative to a candidate link. These inputs were
incorporated into the rules in order to improve
the performance of the algorithm. The output of
the fuzzy inference system (FIS) was the link on
which the vehicle is travelling. The membership
functions were then optimised using an accurate
input/output dataset. This dataset was obtained
from high accuracy GPS carrier phase observa-
tions. Moreover, three sets of knowledge-based
fuzzy rules were formulated when the navigation
data came from GPS/DR. The first set (six rules)
was for an initial map-matching process (IMP).
The second set (thirteen rules) was for subsequent
map-matching on a link (SMP-Link). The third set
(four additional rules) was for subsequent map-
matching at a junction (SMP-Junction). The results
showed that the algorithm is able to identify 99.2%
of the links correctly with horizontal accuracy of
5.5m (95%) in suburban area of London.
In summary, it can be said that different algo-
rithms have different strengths and weaknesses.
Some algorithms may perform very well within
suburban areas but may not be appropriate for
urban areas and vice versa. Existing map match-
662
ing algorithms are not capable of satisfying the
requirements of all ITS applications and services.
For instance, bus priority at junctions require a
positioning accuracy of 5 m (95%) and none of
the existing algorithms can meet this positioning
requirement, especially, within dense urban areas.
FUTURE TRENDS
Further improvements in map matching algorithms
are essential as existing algorithms are not always
capable of satisfying user requirements. The per-
formance and speed of a map matching algorithm
largely depends on the technique used in the algo-
rithm. For instance, it has been found that the fuzzy
logic based map matching algorithms provide better
performance compared with other methods for the
same inputs. Therefore, future research on map
matching algorithms will focus on the technique
used in the algorithm. Other potential techniques
would be to employ a pattern recognition approach,
a hieratical fuzzy inference system optimised by a
genetic algorithm (GA), or a hybrid method. Other
potential areas of research would be to improve the
inputs to a map matching algorithm. For instance,
the deployment of European Galileo and the mod-
ernization of GPS over the next few years will have
a profound impact on future GNSS receiver design
which will enhance positioning data. The accuracy
of spatial road network data is also continuing to
improve over time. It is therefore expected that
future map matching algorithms together with the
use of these improved inputs will be able to sup-
port the navigation function of ITS services with
a horizontal accuracy of 5 m (95%) even in dense
and built-up urban areas.
CONCLUSION
Map matching algorithms play a key role to
support the navigation module of an ITS. The
performance of a map matching algorithm largely
Map Matching Algorithms for Intelligent Transport Systems
depends on the quality of the spatial road network
data, the quality of the positioning data, and
the methodologies used in the algorithm. This
article has mainly discussed techniques used by
map matching algorithms. It has been revealed
that the performance of a fuzzy logic based map
matching algorithm is better when compared to
other algorithms. However, existing map match-
ing algorithms cannot meet the requirements for
some applications, especially in built-up urban
areas and hence further improvement is necessary.
This article then discusses the future trends in the
development of map matching algorithms. It is
envisaged that a future map matching algorithm
will also emphasize the reliability of the position-
ing solution.
REFERENCES
Fu, M., Li, J., & Wang, M. (2004). A hybrid map
matching algorithm based on fuzzy comprehen-
sive Judgment, IEEE Proceedings on Intelligent
Transportation Systems, 613-617.
Greenfeld, J. S. (2002). Matching GPS observa-
tions to locations on a digital map. In proceedings
of the 81st Annual Meeting of the Transportation
Research Board, January, Washington D.C.
Honey, S. K., Zavoli, W. B., Milnes, K. A., Phil-
lips, A. C., White, M. S., & Loughmiller, G. E.
(1989). Vehicle navigational system and method,
United States Patent No., 4796191.
Kim, S., & Kim, J. (2001). Adaptive fuzzy-net-
work based C-measure map matching algorithm
for car navigation system. IEEE Transactions
on Industrial Electronics, 48(2), 432–440.
doi:10.1109/41.915423
Kim, S., Kim, J.-H., & Hyun, I.-H. (1998). De-
velopment of a map matching algorithm for car
navigation system using fuzzy Q-factor algorithm.
In proceedings of the World Congress Intelligent
Transport System, October, Seoul, Korea.
Kim, W., Jee, G., & Lee, J. (2000). Efficient use
of digital road map in various positioning for
ITS. IEEE Symposium on Position Location and
Navigation, San Deigo, CA.
Meng, Y. (2006). Improved Positioning of Land
Vehicle in ITS Using Digital Map and Other Ac-
cessory Information. PhD Thesis, Department of
Land Surveying and Geoinformatics, Hong Kong
Polytechnic University.
Najjar, M. E., & Bonnifait, P. (2003). A roadmap
matching method for precise vehicle Localization
using belief theory and Kalman filtering. The 11th
International Conference in Advanced Robotics,
Coimbra, Portugal, June 30 - July 3.
Ochieng, W. Y., Quddus, M. A., & Noland, R.
B. (2004). Integrated positioning algorithms for
transport telematics applications. In proceedings
of the Institute of Navigation (ION) annual confer-
ence, 20-24 September, California, USA.
Phuyal, B. (2002). Method and use of aggre-
gated dead reckoning sensor and GPS data for
map matching. In proceedings of the Institute
of Navigation (ION) annual conference, 20-27
September, Portland, OR.
Pyo, J., Shin, D., & Sung, T. (2001). Development
of a map matching method using the multiple
hypothesis technique. IEEE Proceedings on
Intelligent Transportation Systems (pp. 23-27).
Quddus, M. A. (2006). High integrity map match-
ing algorithms for advanced transport telematics
applications. PhD Thesis, Centre for Transport
Studies, Department of Civil and Environmental
Engineering, Imperial College London.
Quddus, M. A., Ochieng, W. Y., & Noland,
R. B. (2006). A high accuracy map match-
ing algorithm for road transport. Journal of
Intelligent Transportation Systems: Technol-
ogy, Planning, and Operations, 10(3), 103–115.
doi:10.1080/15472450600793560
663
 Map Matching Algorithms for Intelligent Transport Systems
Quddus, M. A., Ochieng, W. Y., Zhao, L., &
Noland, R. B. (2003). A general map matching
algorithm for transport telematics applications.
GPS Solutions, 7(3), 157–167. doi:10.1007/
s10291-003-0069-z
Syed, S., & Cannon, M. E. (2004). Fuzzy logic-
based map matching algorithm for vehicle naviga-
tion system in urban canyons. In proceedings of the
Institute of Navigation (ION) national technical
meeting, 26-28 January, California, USA.
Tanaka, J., Hirano, K., Itoh, T., Nobuta, H., &
Tsunoda, S. (1990). Navigation system with
map-matching method. Proceeding of the SAE
International Congress and Exposition (pp 40-50).
Taylor, G., Blewitt, G., Steup, D., Corbett, S.,
& Car, A. (2001). Road Reduction Filtering for
GPS-GIS Navigation. Transactions in GIS, ISSN
1361-1682, 5(3), 193-207.
White, C. E., Bernstein, D., & Kornhauser, A.
L. (2000). Some map matching algorithms for
personal navigation assistants. Transportation
Research Part C, Emerging Technologies, 8,
91–108. doi:10.1016/S0968-090X(00)00026-7
Yang, D., Cai, B., & Yuan, Y. (2003). An improved
map-matching algorithm used in vehicle naviga-
tion system. IEEE Proceedings on Intelligent
Transportation Systems, 2, 1246–1250.
Zadeh, L. A. (1965). Fuzzy sets. Information
and Control, 8, 338–353. doi:10.1016/S0019-
9958(65)90241-X
Zhao, L., Ochieng, W. Y., Quddus, M. A., & No-
land, R. B. (2003). An Extended Kalman Filter
algorithm for Integrating GPS and low-cost Dead
reckoning system data for vehicle performance
and emissions monitoring. Journal of Navigation,
56, 257–275. doi:10.1017/S0373463303002212
Zhao, Y. (1997). Vehicle Location and Navigation
System. Artech House, Inc., MA.
664
KEY TERMS AND DEFINITIONS
Accuracy: Accuracy is defined as the nearness
of a measurement to the standard or true value i.e.,
a highly accurate navigation system will provide
measurements very close to the standard, true or
known values. A rigorous statement of accuracy
includes statistical measures of uncertainty and
variation. Accuracy is generally represented by
standard deviation of errors (difference between
measurements on the map and the true value).
Dead-Reckoning (DR): DR is the process of
estimating the position of a vehicle (or a moving
object) with respect to a known position using
information on heading, speed, time, and distance
travelled by the vehicle.
Fuzzy Logic: Fuzzy logic is a superset of con-
ventional (Boolean) logic that has been extended
to handle the concept of partial truth i.e., truth
values between “completely true” and “com-
pletely false”. It was introduced in the 1960’s by
Zadeh (1965). It is suitable to deal with problems
involving knowledge expressed in vague, and
linguistic terms.
Global Positioning System (GPS): GPS is
a satellite-based radio-navigation, positioning,
and time-transfer system. It is designed, financed,
and deployed by the US Department of Defense
(US DoD) and operated jointly by the US DoD
and the Department of Transportation (US DoT).
Intelligent Transportation Systems (ITS):
ITS can be defined as the integrated application of
advanced sensors, computers, electronics, naviga-
tion, and communication technologies to vehicles
and roadways that increase safety, reduce conges-
tion, enhance mobility, minimize environmental
impact, increase energy efficiency, and promote
economic productivity for a healthier economy.
Map Matching (MM): MM is a process to
match the positioning data (i.e., longitude and
latitude or Easting and Northing) onto the digital
road map data.
Map Matching Algorithms for Intelligent Transport Systems

Topology: Topology is the relationship be- case of polygons), connectivity (in the case of
tween entities (points, lines, and polygons). The lines), or containment (in the case of points in
relationship can be defined as adjacency (in the polygons).

This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 302-309,
copyright 2009 by Information Science Reference (an imprint of IGI Global).

665
666

Chapter 41
Describing and Selecting
Collections of Georeferenced
Media Items in Peer-to-Peer
Information Retrieval Systems
Daniel Blank
University of Bamberg, Germany

Andreas Henrich
University of Bamberg, Germany

ABSTRACT
The ever-increasing amount of media items on the World Wide Web and on private devices leads to a
strong need for adequate indexing and search techniques. Trends such as personal media archives, social
networks, mobile devices with huge storage space, and networks with high bandwidth capacities make
distributed solutions and in particular Peer-to-Peer (P2P) Information Retrieval (IR) systems attractive.
On the other hand, when designing effective media search applications, various search criteria have to
be addressed. Hereby, geospatial information is frequently used as well as other criteria, such as text,
audio or visual media content, and date and time information.

In this chapter, the authors outline how collections of georeferenced media items can be indexed and
searched in P2P IR systems. They discuss different types of P2P IR systems and focus in detail on an
approach based on collection description and selection techniques. This approach tries to adequately
describe and select collections of georeferenced media items. Finally, the authors discuss its broad ap-
plicability in various application fields.

DOI: 10.4018/978-1-4666-2038-4.ch041

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Describing and Selecting Collections of Georeferenced Media Items
INTRODUCTION
In recent years, the availability—and with it the
usefulness—of geospatial metadata has increased
dramatically. Digital cameras and mobile phones
are nowadays often equipped with GPS sensors
at affordable cost. Hence, such devices are able
to capture georeferenced information in the per-
sonal lives of millions of people from all over the
world. In addition, geo-tagging tools with rich user
interfaces have emerged in different domains and
large geo-tagging initiatives try to georeference
textual resources such as in case of Wikipedia.
As a consequence, an increased importance of
geospatial information in the context of search
can be recognized.
Obviously, geospatial information is not the
only search criterion. When searching for media
items other criteria such as textual content, time-
stamps, and (low-level) audio or visual content
information can be used as well—often in an
integrative way. A combination of these criteria
can allow for the effective retrieval of text, image,
audio, and video documents.
As we all know, the amount of media items
on the World Wide Web and on private devices
steadily increases. Service providers such as Flickr,
YouTube, or Facebook (Beaver, et al., 2010) have
to maintain huge hardware infrastructures in order
to keep up with the tremendous increase in data
volumes. So far, it is unclear if existing server-
centered solutions will also suit our needs in the
future. Hence, a need for alternative indexing and
search techniques might arise.
Peer-to-Peer (P2P) Information Retrieval (IR)
systems consist of computers from all over the
world. These computers can act as both clients
and servers. By applying a scalable P2P IR pro-
tocol, a “service of equals” for the administration
of media items can be established in contrast to
existing client/server-based solutions. No expen-
sive infrastructure has to be maintained and idle
computing power in times of inactivity can be
used to maintain, analyze, and enrich media items.
P2P IR systems offer the benefit that media items
can remain on individual devices since there is no
need for storing them on remote servers hosted
by third party service providers. Crawling which
consumes large amounts of web traffic (Bockting
& Hiemstra, 2009) can thus be avoided. In addi-
tion, dependency from service providers acting
as informational gatekeepers can be reduced,
because they are no longer able to decide which
information can be retrieved or accessed and which
cannot. In times of a strong market concentration
in internet search and social network applications
as well as public debates addressing the privacy
of data, P2P IR could offer some benefits.
As our primary use case, georeferenced images
are administered in a P2P IR system. The images
of a certain user are stored locally on the user’s
personal device(s) and a scalable P2P IR protocol
is applied in order to facilitate retrieval. An image
can hereby be described by various criteria: textual
metadata, (low-level) visual content features, a
timestamp, and a geographic coordinate. Personal
media collections containing multiple images can
thus be represented by corresponding collection
descriptions allowing for efficient and effective
collection selection when processing a given query.
We assume in the following that at least some of
the images of a peer are geo-tagged. A resource
description capturing the geographic footprint of
an image collection can thus be generated from
the set of georeferenced images. In this chapter,
we focus on geospatial query processing. In par-
ticular, we address geospatial k nearest neighbor
(k-NN) queries—finding the k closest media items
according to a given query location.
In literature, many approaches for P2P IR can
be found. The following section entitled Peer-to-
Peer Information Retrieval Systems for Geospatial
Search will give an overview on different types of
P2P IR systems and outline how georeferenced me-
dia items can be indexed in a P2P setting in general.
We additionally describe how a comprehensive
indexing of the abovementioned search criteria
(geospatial, textual, date and time, and audio or
667
 Describing and Selecting Collections of Georeferenced Media Items
visual content information) can be achieved. In
addition, we discuss associated consequences for
query processing in a distributed scenario.
A P2P IR system which might use the above-
mentioned criteria and which is based on resource
(i.e. collection) selection is presented in detail in
section Resource Selection based on Descriptions
of Geospatial Footprints. Resource selection
techniques are applied to determine a ranking
of promising resources based on descriptions of
their geospatial footprint. Peers are contacted in
ranked order to retrieve the media items with the
closest locations according to a user-given query
location. It is important to note that the core of
our P2P IR system—its resource description and
selection techniques addressing georeferenced
media items—is not limited to P2P IR systems.
Different application fields and usage scenarios of
these resource description and selection techniques
are therefore discussed at the end of this chapter.
In a Future Work section, we will focus on
different open aspects w.r.t. geospatial resource
description and selection before we end with a
Conclusion section.
PEER-TO-PEER INFORMATION
RETRIEVAL SYSTEMS FOR
GEOSPATIAL SEARCH
P2P IR systems are mostly classified as being
structured or unstructured overlay networks. In
addition, we introduce the distinction of data-
independent and data-dependent overlays as a
secondary classification criterion to reflect if, for
example, a peer’s content or query profiles affect
overlay generation. This distinction provides
helpful insights in order to pinpoint different
characteristics in a more organized way. In the
following subsections, we will briefly discuss
unstructured and structured as well as hybrid and
super-peer approaches.
668
Unstructured Topologies
Data-independent: Main protocols in this group
are PlanetP (Cuenca-Acuna, et al., 2003) and its
extension Rumorama (Müller, et al., 2005). In
Rumorama, a peer sees the network as a single,
small PlanetP network (called subnet) with con-
nections to other peers, which see other PlanetP
subnets. Each peer can choose the size of its
subnet according to local processing power and
bandwidth capacity. Within a subnet, a peer knows
data summaries of all other peers in the subnet.
Gossiping techniques are used to disseminate the
data summaries. In a subnet, summary-based re-
source selection allows for semantic query routing.
Additionally, a peer maintains a small set of links
pointing to neighboring peers in other subnets in
order to be able to forward queries outside the
boundaries of its own subnet.
In its original form, peers are assigned to
subnets arbitrarily, i.e. independent of the peers’
content. However, Rumorama can be easily
extended by a grouping of peers—similar to the
content-dependent overlays described in the fol-
lowing—for example by content or by geospatial
proximity of the host, which operates the peer.
A further benefit of the resource descriptions is
their possibility of being visualized. Thus, they
might be applied for interactive retrieval, e.g. by
providing—with low bandwidth requirements—a
visual overview of peer data for a large number
of peers.
Routing indexes in various forms represent
aggregated information in an unstructured P2P IR
system maintained at a peer for all its neighboring
peers in order to decide in which direction queries
should be forwarded. Initially designed for one-
dimensional values in order to avoid network
flooding in the early days of P2P computing,
they have for example been extended to allow for
multi-dimensional queries. Here, bounding boxes
are used to summarize the content of neighboring
peers (for references cf. Doulkeridis, et al., 2009).
Describing and Selecting Collections of Georeferenced Media Items
Data-dependent: Many Semantic Overlay
Networks (SONs; for references and a detailed
description cf. Doulkeridis, et al., 2010) are data-
dependent, unstructured P2P networks. Here, the
content of a peer’s data or information about past
queries defines a peer’s place in the network. Thus,
summaries of a peer’s content or query profiles are
needed. Two types of links are usually maintained
in such systems: short links grouping peers with
similar content or query profiles into so called
“Clusters of Interest” (COIs) and long links which
are established between different COIs. During
query processing, the query has to be forwarded
to the most promising COI(s). Clustering, clas-
sification, and gossiping techniques can be applied
in order to form COIs.
Indexing of Multiple Criteria in Unstruc-
tured Topologies: Many of the unstructured,
data-independent P2P IR systems are based on
resource descriptions which summarize a single
peer’s content, summarize the content of multiple
peers reachable when following a certain direc-
tion, or for example summarize information about
past queries. In general, it is possible to apply one
summary type and a corresponding resource selec-
tion technique per feature type. Feature-specific
peer rankings can be combined by applying an
algorithm for the merging of ranked lists (Belkin,
et al., 1995; Ilyas, et al., 2008). As an alternative
for creating independent resource descriptions per
search criterion, summaries and resource selection
algorithms integrating multiple feature types are
also possible (cf. Hariharan, et al., 2008).
In section Resource Selection based on De-
scriptions of Geospatial Footprints we will focus
on resource description and selection techniques
for geospatial information. The design of resource
description and selection techniques for textual
data is for example addressed in Cuenca-Acuna
(2003). Summaries for high-dimensional feature
vectors in order to summarize content-based media
features are outlined in Blank et al. (2007).
In data-dependent unstructured networks mul-
tiple search criteria can be addressed when forming
the COIs. Since many approaches are based on
a similarity measure between resources’ content,
different criteria can be integrated when determin-
ing the similarity of peers in order to group them
together. Alternatively, multiple overlays might be
maintained, i.e. one overlay per search criterion.
Structured Topologies
Data-independent: Structured P2P IR systems
are based on distributed indexing structures.
Distributed Hash Tables (DHTs) represent the
most prominent class member. Every peer in
the network is usually responsible for a certain
range of the feature space. Thus, when entering
the network or updating local content, index data
has to be transferred to remote peers according
to the peers’ responsibilities. In case of data-
independent, structured P2P IR systems, terms (cf.
Bender, et al., 2005) or high-dimensional feature
vectors for content-based image retrieval (cf.
Novak, et al., 2008; Lupu, et al., 2007; Vu, et al.,
2009) are usually mapped to one-dimensional or
multi-dimensional keys. They can be indexed in a
classical DHT such as Chord (Stoica, et al., 2001)
or CAN (Ratnasamy, et al., 2001) respectively. It
has to be noted that there is a large variety of such
P2P protocols. Very detailed information with
references pointing to relevant research articles
can be found in Shen et al. (2010).
Data-dependent: SONs—as described
above—can also be implemented on top of a DHT
in order to enhance query routing (Doulkeridis,
et al., 2010). Various forms of clustering, clas-
sification together with gossiping techniques can
be applied in order to establish links to peers with
similar content.
Indexing of Multiple Criteria in Structured
Topologies: In structured, data-independent sys-
tems, correlations between different criteria are
difficult to exploit when indexing multiple feature
types (e.g. geospatial and image content informa-
tion). If we for example assume an image from
the Sahara Desert with shades of beige sand and
669
 Describing and Selecting Collections of Georeferenced Media Items
blue sky, different peers might be responsible for
indexing the geospatial and the image content in-
formation. Thus, when distributing the index data
of the Sahara image, querying for it, or removing
it from the network, (at least) two different peers
have to be contacted. Within SONs, the simulta-
neous indexing of multiple criteria would again
require the definition of a similarity between peers’
content by combining for example geospatial and
image content information. Alternatively, multiple
overlays might be maintained.
Hybrid and Super-Peer Approaches
In unstructured P2P IR systems, a peer only ad-
ministers index data of media items, which belong
to its user. Thus, when entering the network or
updating media items, full index data does not
have to be transferred to remote peers. Peer au-
tonomy is better respected compared to structured
networks (Doulkeridis, et al., 2010). On the other
hand, structured systems offer query processing
with logarithmic cost. In order to reduce the load
imposed on the network when inserting new
media items in structured systems, super-peer
architectures (Papapetrou, et al., 2007) as well as
DHT-based indexing of compact data summaries
instead of full index data have been proposed (cf.
Lupu, et al., 2007).
In general, there is a convergence of structured
and unstructured P2P IR systems with many hybrid
approaches. We have for example evaluated an
approach where index data is transferred amongst
peers in certain rounds in order to make peers more
focused and—as a consequence—summaries
more selective. More selective summaries with
peers having specialized on a certain range of
the feature space lead to more efficient resource
selection (Eisenhardt, et al., 2008).
There is plenty of work addressing super-peer
architectures (for references cf. Doulkeridis, et al.,
2009). They are designed in order to overcome
some limitations of “true” P2P IR systems and
make use of increased capabilities such as storage
670
capacity, processing power, or available network
bandwidth. Often, concepts known from “true”
P2P IR systems are extended and transferred to
super-peer architectures. Also within super-peer
networks the convergence of different approaches
can be seen. Doulkeridis et al. (2009) for example
apply multi-dimensional routing indexes on a
super-peer level and additionally group similar
super-peers close together in order to allow for
improved query routing.
RESOURCE SELECTION
BASED ON DESCRIPTIONS OF
GEOSPATIAL FOOTPRINTS
Geospatial Resource Description
and Selection in Rumorama-
Like P2P IR Systems
It is important to note that the resource selection
techniques presented in the following are not re-
stricted to data-independent, unstructured P2P IR
systems such as Rumorama. The summaries can
also be applied in data-dependent, unstructured
P2P IR systems when forming COIs and in struc-
tured networks in order to be indexed for example
in a DHT. In addition, summaries can be used by
super-peers for selecting “normal” peers or other
super-peers. Here, it might be possible to relax the
strict compliance of some design parameters. A
system based on super-peers might for example
allow for less space efficient resource descriptions
in order to achieve better retrieval performance.
In this section, we will present four different
resource description and corresponding selection
techniques for geospatial information. Further ap-
plication fields of the analyzed geospatial resource
description and selection techniques outside the
P2P IR context are also possible. We will focus
on this aspect in the next section which is entitled
Resource Descriptions and Selection Techniques
in different Application Fields.
Describing and Selecting Collections of Georeferenced Media Items
In this chapter, we outline and extend earlier
work from Blank and Henrich (2009, 2010) mainly
according to two major directions. First, we ana-
lyze peer ranking selectivity in more detail by not
only providing the average fraction of contacted
peers as the main performance measure. We ana-
lyze the distribution of the fraction of contacted
peers over the set of queries, which gives additional
insights. Second, we show that also on a global
scale (i.e. based on a data collection of geo-tagged
Flickr images from different areas of the world),
it is sufficient in our scenario to approximate
distances by the use of Euclidean distance when
searching for the 20 nearest neighbors according
to a given query location.
We use the terms image and document inter-
changeably in order to refer to concrete media
items. This is due to the fact that our experiments
in Blank and Henrich (2009, 2010) and also in
this very chapter are based on image documents.
Nevertheless, we believe that our approach can
be extended to other media types. Of course, by
restricting our analysis on images, we implicitly
allow only a single lat/lon-coordinate per media
item. This might be different in case of textual
documents for example where several locations
could be referenced within a single document. A
detailed analysis of these and related challenges
could be part of future work.
Bounding Boxes (BB)
When using bounding boxes as resource descrip-
tions, every peer computes a minimum bounding
Figure 1. Visualizing summary creation for BB (left), GRID4 (middle), and HFS8/UFS8 (right: ■ cor-
responds to reference points). Four documents indicated as x are geo-tagged.
box over the geospatial coordinates of its image
collection (refer to. Figure 1). We encode a lati-
tude/longitude-pair (for short: lat/lon-pair) with 8
bytes, 4 for latitude and 4 for longitude. Therefore,
we require 16 bytes of raw data for the bounding
box (i.e. two lat/lon-pairs, e.g. the lower left and
upper right corner).
Peer ranking is performed as follows. If a peer
pa contains the query location in its bounding box
whereas peer pb does not, peer pa is ranked
higher than peer pb and vice versa. In case the
query location lies in the bounding box of both
peers pa and pb, the size of a peer (i.e. the number
of documents a peer administers) is used as an
additional criterion. Peers with more documents
are ranked higher. In earlier studies, several al-
ternatives have been evaluated (cf. Blank &
Henrich, 2010) and this strategy turned out to be
beneficial. If neither peer pa nor peer pb contain
the query within its bounding box, the peer with
the smaller minimum distance from the query
location to its bounding box is preferred. We as-
sume a spherical model of the earth with a radius
of 6371 km. If not stated otherwise, we use Ha-
versine distance (Sinnott, 1984) to compute the
distance between two points on the sphere.
Of course, it is also possible to represent
the geospatial footprint of a peer by multiple
bounding boxes. Becker et al. (1991) present
an algorithm for summarizing a set of bounding
boxes by two bounding boxes which minimize
the area that is covered. Chen et al. (2006) pro-
pose several threshold-based algorithms to split a
single bounding box into several smaller ones in
671
 Describing and Selecting Collections of Georeferenced Media Items
order to reduce the space within a bounding box
which is not covered by any index data. These
approaches demonstrate that there is optimization
potential w.r.t. the representation of geo-regions
in geospatial indexing. However, the mentioned
approaches stick with bounding boxes and the ap-
proaches presented later on in this chapter might be
interesting alternatives in this application field. A
detailed comparison might be part of future work.
Grid-Based Summaries (GRIDr)
In a second approach, the lat/lon-coordinates are
mapped to a grid (see. Figure 1). A parameter r
represents the number of rows of the grid. The
number of columns is twice the number of rows
since longitude range is twice as big as latitude
range. The range of a grid cell (in degrees) is de-
termined by 180° / r = 360° / (2r) in the latitude
and longitude domain respectively. This simplified
view is for example also applied in Dolin et al.
(1997) and results in non-uniform grid cell sizes
on the sphere. We gain selectivity and retrieval
performance by increasing the number of grid
cells at the price of additional storage overhead
partially compensated through compression tech-
niques (cf. section Experiments). Every grid cell is
represented by a single bit. If one or more image
locations fall into a certain cell, the corresponding
bit is set to 1. Otherwise, it remains 0. Bit posi-
tions in the summary are determined horizontally
from left to right and from bottom to top. Effects
of alternative strategies on compression could be
evaluated in future work.
During peer ranking the grid cell containing
the query location is determined. If peer pa has an
image in this cell whereas peer pb has not, peer pa is
ranked higher than peer pb and vice versa. We also
consider neighboring grid cells. If either both or
none of peer pa and peer pb have an image located
within the cell containing the query location, the
GRID approach considers the neighboring cells
recursively until a ranking decision can be made.
So, in the first round the ranking decision is always
672
based on a single cell; in the second round it is in
most cases based on 1+8=9 cells and in the third
round on 1+8+16=25 cells and so on. This is not
always the case since there might be no neighbor-
ing cells in a certain direction, e.g. as soon as a
cell in the north or south is reached. Of course,
at the 180° meridian we assume that there is no
boundary and neighborhood relations are valid
in both directions. The ranking criterion in every
round is the number of grid cells containing one
or more image location(s)—the more the better.
Of course, there are alternative approaches in
literature for mapping the spherical coordinates
to a grid for example in order to achieve uniform
grid cell sizes on the sphere (Putman, et al., 2007).
In various domains, cubed grids (for references cf.
Putman, et al., 2007), or triangular meshes (Szalay,
et al., 2005) have been proposed. These techniques
might also be considered in future work, although
we do not expect large improvements compared
to our naïve approach—neither in terms of rank-
ing selectivity nor in terms of reduced summary
sizes. We gain ranking selectivity by increasing
the number of grid cells. On the other hand, this
increase does only lead to a sublinear increase
of resource description sizes since compression
techniques are applied.
Highly Fine-Grained Summaries (HFSn)
This approach is based on resource descriptions
originally designed for summarizing visual con-
tent information of images, for example the color
distribution or texture of an image (Blank, et al.,
2007). A set of n predetermined image locations
are used as reference points. This set of reference
points is known to all peers and built from external
sources (a Gazetteer is used in our case). How to
adequately obtain the reference points is outlined
in the remainder of this chapter. So far, it is suffi-
cient to note that every image location of a peer’s
local image collection is assigned to the closest
reference point according to Haversine distance
(cf. Figure 1, right). Hereby, a cluster histogram is
Describing and Selecting Collections of Georeferenced Media Items
computed counting how many image locations of
a peer’s collection are closest to a certain reference
point, i.e. cluster center cj (1 ≤ j ≤ n).
Peer ranking is performed as follows. Refer-
ence points cj are sorted in ascending order ac-
cording to Haversine distance to the query. The
first element of the sorted list L corresponds to the
cluster center being closest to the query. Peers with
more documents in this so called query cluster are
ranked higher than peers with fewer documents in
the query cluster. If two peers administer the same
amount of documents in the analyzed cluster, the
next element out of L is chosen and both peers
are ranked according to the number of documents
within the very cluster. This procedure continues
until either a ranking decision can be made, which
favors one peer over the other, or the end of L
is reached. In the latter case, a random decision
would be made.
Ultra Fine-Grained Summaries (UFSn)
In contrast to HFS, UFS are based on a bit vector
with the bit at position j indicating if center j is
the closest center to one or more of a peer’s image
locations. Therefore, we obtain a bit vector of size
n. Of course, there is some loss of information
when switching from HFS to UFS with n staying
constant. However, UFS have the potential of
resulting in more space efficient resource descrip-
tions. Potentially, this allows for more reference
points being used, which might result in similar
or even improved retrieval performance compared
to HFS. Among other aspects, this is evaluated
in Blank and Henrich (2010) and will be briefly
summarized in the following experimental sec-
tion. Before doing so, we will describe the data
collections we use.
The abovementioned resource description
techniques as well as variations and combinations
of them (e.g. combining grid and bounding box
based representations) are also used in the context
of multidimensional and metric index structures
(Samet, 2006). The decision of choosing the best
subtree is similar to the resource selection prob-
lem. Summaries in the P2P context correspond for
example to aggregations maintained in the nodes
of a tree, for example bounding boxes in the case
of an R-tree (Guttman, 1984). We therefore point
the interested reader to Samet (2006).
Experiments
Data Collections
Two data collections of geo-tagged images are
used in our experiments:
1. Geoflickr: During the year 2007 a large
amount of publicly available images was
crawled which had been uploaded to Flickr
(http://www.flickr.com). In our scenario, ev-
ery Flickr user operates a peer of its own. We
therefore assign images to peers by means of
the Flickr user ID. All of the crawled images
are geo-tagged. After some data cleansing
the Geoflickr collection consists of 406450
geo-tagged images from 5951 different users/
peers.
2. Geograph: Geograph (http://www.geo-
graph.org.uk/) “aims to collect geographi-
cally representative photographs and infor-
mation for every square kilometre of Great
Britain and Ireland.” We downloaded the
geo-tagged images and distributed them to
peers again in a user-centric approach. In
our scenario every Geograph participant
operates a single peer: 2609 peers administer
246937 images and thus image locations in
total.
The distribution of the number of images per
peer is displayed for both collections in Figure
2. For both collections the distribution of the
number of images per peer is very skew which is
typical for many P2P settings (Cuenca-Acuna, et
al., 2003). Few peers administer large amounts of
the collection. On the other hand, there are many
673
 Describing and Selecting Collections of Georeferenced Media Items
Figure 2. Number of images per peer for the Geoflickr and Geograph collection
peers which store only few images. A more detailed
analysis of the distribution of the peer sizes, i.e.
the number of images per peer, can be found in
Blank and Henrich (2010).
A visualization of the geographic distribution
of the image locations can be found in Blank and
Henrich (2010), too. The Geoflickr collection
consists of photos taken in various parts of the
world with hotspots in North America, Europe,
and Japan. In contrast, images of the Geograph
collection are limited to the UK and Ireland with
images more densely located around urban areas
such as London.
Experimental Settings
In our experiments, we use 200 image locations
as queries. These are randomly selected from the
Figure 3. Image locations used as queries for Geoflickr
674
underlying data collection. The query locations are
visualized in Figure 3. For HFS and UFS where
the outcome of the experiments is affected by the
selection of reference points, we run at least ten
experiments with the 200 queries each. Since we
do not remove the image with the query location,
it is—on average—more likely that a big peer
contributes to the retrieval result than a small peer
because—on average—it is more likely to choose
the query from a big peer than from a small peer.
An additional strategy for selecting the queries is
analyzed in Blank and Henrich (2010).
Space efficiency of different resource descrip-
tions is measured by analyzing average summary
sizes. For compressing the summaries we apply
Java’s gzip implementation with default param-
eter values. Our measurements include serializa-
tion overhead necessary in order to distribute the
Describing and Selecting Collections of Georeferenced Media Items
resource descriptions in the network. The product
of average summary size times the number of
peers indicates the network load, which is imposed
by a single round of gossiping (i.e. every peer
sends its summary once to all other peers in the
PlanetP-like subnet).
In earlier studies (Blank & Henrich, 2010),
in order to measure peer ranking selectivity we
determine the fraction of peers which needs to be
contacted on average to retrieve a certain fraction
of the top-k image locations (k = 20). In this chap-
ter, we take a closer look at ranking selectivity
by analyzing the distribution of the fraction of
contacted peers over all queries.
The top-k geo-locations are computed using
Vincenty distance (Vincenty, 1975). Since we are
interested in the selectivity of the resource descrip-
tion and selection techniques, we analyze all of a
peer’s image locations as soon as it is contacted,
because the top-k image locations of a peer de-
termined using Haversine distance might differ
from the top-k image locations computed using
Vincenty distance. In a real-world application,
only the top-k image locations will be transferred
(together with some additional information such
as peer ID, etc.)
Figure 4. Ranking selectivity for different approaches
Empirical Analysis of Approaches
Peer ranking selectivity for both collections is
displayed in Figure 4 with the help of boxplots
including minimum and maximum values. More
experimental results with a focus on the average
number of peers, which are contacted during query
processing, can be found in Blank and Henrich
(2010). For reasons of brevity, we focus on a set-
ting with n = 8192 for UFS and HFS here. For the
GRID approach we partition the data space in 32
rows and 64 columns. Thus, in an uncompressed
way, GRID will also result in a bit vector of size
8192.
From Figure 4 can be observed that HFS per-
forms slightly better than UFS. This seems reason-
able since HFS encodes frequency information
which can be beneficial for peer ranking. If we
assume that the query is closest to a certain refer-
ence point c*, then it is obvious to contact peer
pa before peer pb if peer pa has assigned more
image locations to c* than peer pb. In case of UFS,
this information gets lost. Here, peer pb with
fewer image locations assigned to c* might be
selected before peer pa. Although there might be
cases where such a strategy will lead to better
ranking selectivity, these cases will be excep-
tional.
675
 Describing and Selecting Collections of Georeferenced Media Items
For the Geoflickr collection (Figure 4, left),
both HFS and UFS perform better than BB and
GRID. Interquartile ranges of UFS and HFS are
much smaller than for GRID and especially BB.
GRID offers better ranking selectivity than BB.
In addition to the average number of contacted
peers (cf. Blank & Henrich, 2010), also median
and 75th percentile of GRID are clearly below
corresponding values for BB. Also interquartile
range is smaller. Nevertheless, at least for some
queries, it is very difficult for GRID to offer
adequate ranking selectivity as indicated by the
maximum value of almost 18%. This is the case
for queries, which lie in a very populated grid cell,
i.e. a grid cell where many peers assign documents
to. We therefore looked at queries, which offer
poor ranking selectivity for GRID in more detail.
It can be observed that many of these queries lie
in cells where many of the documents reside, i.e.
cells, which contain metropolises such as London
for example.
In case of Geograph (Figure 4, right), the grid
is not adapted to the boundaries of the United
Kingdom (of course this can easily be done). We
did not adapt it in order to show the effects of a
skew distribution of geospatial image locations
on a global scale. HFS and UFS are better suited
for such scenarios than GRID, because they bet-
ter adapt to the data which is used. Also for HFS
Figure 5. Average summary sizes for different approaches
676
and UFS reference points are chosen on a global
scale and are not restricted to the boundaries of
the UK. They are selected from a Gazetteer and
correspond to various locations from all over the
world.
Summaries based on a single bounding box
per peer lack peer ranking selectivity and seem
to be too coarse in order to be able to compete
with HFS or UFS, but, bounding boxes can be
represented in a very space efficient way. They
require only 45 byte per resource description as
can be observed from Figure 5.
For both collections, HFS affords—on aver-
age—approximately twice as big resource descrip-
tions as UFS. If we take a closer look at the
Geoflickr collection, average summary sizes of
UFS are slightly bigger than in case of GRID. On
the other hand, further experiments show that
even GRID with more cells (for example 96 rows
and 192 columns) cannot outperform UFS in terms
of ranking selectivity. GRID96 contacts on average
more peers than UFS8192 in order to retrieve the
Top-20 with a bigger average summary size.
A more detailed analysis of ranking selectivity
for HFS and UFS is displayed in Figure 6. It visu-
alizes the number of contacted peers for different
values of n. Medians in case of HFS are always
smaller than in case of UFS. It can be confirmed
that HFS outperforms UFS. But, performance
Describing and Selecting Collections of Georeferenced Media Items
gaps decrease with increasing values of n. Dif-
ferences in peer ranking selectivity diminish with
increasing n, since the corresponding histograms
become more and more similar with many zeros
and some summary bin values set to 1. Of course,
for HFS, the values of some summary bins might
still be bigger than 1, but with increasing n this
becomes rarer and rarer. Also interquartile ranges
for HFS and UFS become more and more similar
when increasing n. This indicates that ranking
selectivity of HFS and UFS equal more and more.
In the following, we will limit ourselves to the
analysis of UFS since this approach seems to of-
fers a good compromise between ranking selectiv-
ity and summary size. In order to further trade-off
these two factors under the influence of other
aspects, we refer to the general cost model pre-
sented in Blank and Henrich (2010) for a more
detailed analysis.
Obtaining the Reference Points
For obtaining the reference points, we employ
Geonames gazetteer (http://www.geonames.org/)
as well as United Nations’ per country statistics
obtained through Worldmapper (http://www.
worldmapper.org/). Within this chapter, we focus
on statistics about men’s income, Gross Domestic
Product (GDP), population and WWW usage.
Various other statistics are evaluated in Blank
and Henrich (2010). Based on the statistics we
Figure 6. Ranking selectivity of HFS and UFS for different values of n
proportionally select the number of reference
points from a certain country. Reference points are
selected amongst all populated places of a certain
country at random. So, for example, if x% of the
world’s GDP comes from a certain country, x% of
the reference points are randomly chosen amongst
all populated places of the specific country.
Figure 7 shows ranking selectivity in terms of
the average fraction of peers which are contacted
in order to retrieve the top-20 image locations.
We plot UFS results for different values of n. An
analysis of HFS offers similar characteristics and
is omitted here for reasons of brevity. We can see
that the strategy based on GDP performs best.
This confirms the finding in Blank and Henrich
(2010) where only small values of n were ana-
lyzed. It should be noted here that we used the
GDP-based approach in the preceding experiments
of this chapter.
Analysis of Distance Measures
In our experiments, the 20 closest image locations
according to a given query location are always
computed using Vincenty distance (Vincenty,
1975) in order to achieve good accuracy, since
we are interested in the “true” nearest neighbors.
On the other hand, in our algorithms we have
used Haversine distance (Sinnott, 1984) which
is computationally more efficient than Vincenty
distance at the price of computing less accurate
677
 Describing and Selecting Collections of Georeferenced Media Items
Figure 7. Choosing reference points for UFS from external sources
distance values. Whereas Vincenty distance is
based on an ellipsoidal shape of the earth, Ha-
versine distance assumes a spherical model. In
addition to Haversine and Vincenty distance,
we analyzed two versions of Euclidean distance
(the use of lat/lon-pairs as 2-dimensional vectors
as well as 3-dimensional vectors resulting from
a projection of the lat/lon-coordinates onto a
sphere). These distances might be used in order to
further approximate distance values and speed-up
query processing. As for Haversine distance, we
use a rather coarse earth radius of approximately
6731 km. We observed in experiments that all
approaches offer similar retrieval performance.
Hence, it is sufficient to apply Euclidean distance.
In the following section, we will describe
how geospatial resource description and selection
techniques, which were presented and evaluated in
the preceding sections, could be applied in differ-
ent application fields. By doing so we will show
that their use is not limited to the P2P IR domain.
Resource Descriptions and
Selection Techniques in
Different Application Fields
In addition to P2P IR, geospatial resource summa-
rization and selection techniques can also be used
in traditional distributed IR applications. Personal
meta-search is a novel application of distributed
678
IR, where all the online resources of a person are
queried. Web users frequently administer various
e-mail accounts, bookmarks of web pages, image
collections, databases, etc. These resources are
typically heterogeneous in size, media type and
update frequency (Thomas & Hawking, 2009) pos-
sibly requiring space efficient and at the same time
selective (geospatial) resource descriptions. Thus,
applications might be built providing a unified
search service (similar to a meta-search engine)
over all these resources. This could prevent web
users from the time consuming task of querying
all resources “manually.”
Spatial Data Infrastructures (SDIs; for a general
description see for example Nogueras-Iso et al.
[2005]) might also be an interesting application
field for geospatial resource description and selec-
tion techniques. Chen et al. (2010) identify three
critical problems with current solutions which are
for example based on the Catalogue Service for
the Web (CSW; http://www.opengeospatial.org/
standards/cat). First, server-based solutions offer
the problem of a single point of failure. Second, it
is the task of the users who search for services to
identify the proper portal. Third, users who want
to provide data have to find a suitable portal to
register to, which might also be a time consum-
ing task. In order to overcome these issues, it is
broadly recognized that P2P technology might
be beneficial for the discovery of geospatial web
Describing and Selecting Collections of Georeferenced Media Items
services and the publishing of geographic data
(Xiujun, et al., 2006; Chen, et al., 2010). What
remains is the question for the best P2P technology
in the context of SDIs. Due to the high maintenance
costs of structured solutions with nodes joining or
leaving the network, Chen et al. (2010) argue for
a hybrid solution. Initially nodes join an unstruc-
tured network and only if the node stays in the
network for a longer time period, it can move to
the structured network. While the administration of
lat/lon-coordinates in a structured network might
for example be based on a quadtree-like (Chen,
et al., 2010) or R+-tree-like (Xiujin, et al., 2006)
structure, or a space-filling curve based mapping
to one dimensional values (Memon, et al., 2009),
the unstructured part of the network might use
a resource description and selection based ap-
proach for web service discovery. Here, rather
than minimum bounding boxes such as in case
of (Xiujin, et al., 2006), grid-based or Voronoi-
based space partitioning schemes as presented in
this chapter might also be beneficial to represent a
geospatial web service profile. Such profiles might
be generated by different resources/peers by the
invocation of the getCapabilities() method which
is defined by many OGC standards such as Web
Map Service (WMS; http://www.opengeospatial.
org/standards/wms), Web Feature Service (WFS;
http://www.opengeospatial.org/standards/wfs),
or Web Coverage Service (WCS; http://www.
opengeospatial.org/standards/wcs). Currently, for
example in case of WCS and WFS, Voronoi-based
coverage summaries are not intended.
Resource description and selection techniques
might also be applied within sensor networks
(Elahi, et al., 2009). In sensor networks, limited
processing power, bandwidth, and energy capaci-
ties necessitate aggregation techniques which are
based on local information with a clear focus on
space efficiency. One might think of in-situ or
remote sensing applications, which aggregate
and summarize gathered data before transmitting
them in the network.
Lupu et al. (2007) present an approach for
information sharing in mobile ad hoc networks.
When people meet at certain events or places only
for a limited amount of time, there might not be
enough time to transfer full index data when a
person for example searches for media items in
the proximity of a given query location.
Compact geospatial resource descriptions
might also be valuable for focused crawling
(cf. Ahlers & Boll, 2009). If a service provides
summaries of the geospatial extend of a certain
website or media archive, a crawler could esti-
mate the potential usefulness of this resource for
its focused crawling task before actually visiting
the source. This way, crawl efficiency can be im-
proved by preventing the crawler from analyzing
too many irrelevant pages. Web traffic imposed
by downloading large irrelevant data volumes can
thus be avoided.
Distributed IR techniques can also be used
for vertical selection within aggregated search
(Arguello, et al., 2009). Vertical selection is the
task of identifying relevant verticals, i.e. focused
search services such as image, news, video or
shopping search. A user issuing the textual query
“Beatles Hamburg” might also be interested in
music videos captured in Hamburg and thus the
results of video search or small previews should
be integrated in result presentation of classical
web search if video files match the geospatial
restriction. In this context, a vertical can be in-
terpreted as a resource and the task of selecting
relevant verticals is similar to resource selection
in distributed IR requiring adequate features, i.e.
resource descriptions, and corresponding selec-
tion mechanisms.
Space efficient geospatial resource descriptions
might also be beneficial in the context of recom-
mender systems and social search for example in
order to compute the similarity between different
users of social network sites. Similar users can
be determined not only based on having the same
friends, using the same tags, bookmarking the
679
 Describing and Selecting Collections of Georeferenced Media Items
same media items, etc. (Guy, et al., 2010), but
also depending on the similarity of geospatial
footprints which are obtained from the media
items a user administers.
For many scenarios, it might be necessary to
enhance the basic description and selection tech-
niques presented in this chapter in order to sum-
marize for example trajectories or more complex
objects such as polygons. Also in this context, the
literature on multidimensional and metric index
structures provides a good starting point for further
studies (cf. Samet, 2006). Many multidimensional
access methods are capable of administering lines
and polygons, which are aggregated for example
in the inner nodes of tree-based access methods
in an adequate way.
FUTURE RESEARCH DIRECTIONS
So far, our heuristic approach lacks the avail-
ability of an adequate stopping criterion. In our
experiments we analyze how many peers need
to be contacted in order to retrieve the 20 closest
locations according to a given query location. Of
course, we could use these empirical findings in
order to derive a mechanism which tries to guar-
antee that for example in 90% of the queries all
20 nearest neighbors can be found. Nevertheless,
we doubt if such an approach adapts to differ-
ent media types and collection sizes. Thus, it is
necessary to design algorithms for k-NN query
processing which can successfully prune peers
from query processing if they do not contribute
relevant documents.
Another interesting aspect might be to further
optimize the GRID and BB approaches. For BB,
techniques, which use more than one bounding
box per resource, might be promising (cf. Becker,
et al., 1991; Chen, et al., 2006). For the GRID
approach, triangular partitions as well as a parti-
tioning which results in cells of equal size might
be considered (cf. Szalay, et al., 2005; Putman,
et al., 2007).
680
It might also be interesting to further analyze
ranking selectivity when very remote places on
earth are used as query locations. If they repre-
sent places in certain countries with low GDP,
this might lead to a loss in ranking selectivity for
UFS and HFS. In these cases, other strategies for
obtaining the centroids might be used. In addition,
BB or GRID might be better suited in such cases.
As mentioned before, we also plan to use UFS
and HFS to enhance centralized index structures.
In general, there is a need for resource descrip-
tion and selection techniques based on local image
features. Here, image content is represented by
several feature vectors per image. Local features
are also used in video retrieval. A further research
direction might be to apply HFS and UFS in this
context.
Since resource selection techniques for text
and media content information are available, one
might take a closer look at combined queries.
If a user queries for “Beatles Hamburg” it is
not sufficient to rank peers high, which contain
documents addressing the Beatles or the city of
Hamburg. A user might expect documents which
match both criteria.
CONCLUSION
In this chapter, we have outlined different P2P ap-
proaches for the large-scale administration of media
content. We identified geospatial information as an
important search criterion and discussed how it can
be applied in the various P2P approaches. Later in
the chapter, we presented and evaluated different
resource description and selection techniques for
geospatial k-NN queries and identified UFS as a
promising approach. Here, binary histograms are
used which capture if a certain reference location
is closest to one or more of a resources media loca-
tions. We also outlined how UFS can be applied in
a P2P scenario. In addition, we qualitatively argued
that geospatial resource description and selection
techniques such as UFS might also be promising
in various other application fields.
Describing and Selecting Collections of Georeferenced Media Items
REFERENCES
Ahlers, D., & Boll, S. (2009). Adaptive geospa-
tially focused crawling. In D. Cheung, I. Song,
W. Chu, X. Hu, J. Lin, J. Li, & Z. Peng (Eds.),
Proceedings of the 18th ACM International Con-
ference on Information and Knowledge Manage-
ment, (pp. 445-454). Hong Kong, China: ACM.
Arguello, J., Diaz, F., Callan, J., & Crespo, J.-F.
(2009). Sources of evidence for vertical selection.
In M. Sanderson, C. Zhai, J. Zobel, J. Allan, &
J.-A. Aslan (Eds.), Proceedings of the 32nd In-
ternational ACM SIGIR Conference on Research
and Development in Information Retrieval, (pp.
315-322). Boston, MA: ACM.
Beaver, D., Kumar, S., Li, H. C., Sobel, J., &
Vajgel, P. (2010). Finding a needle in haystack:
Facebook’s photo storage. In Proceedings of the
9th USENIX Conference on Operating Systems
Design and Implementation (OSDI 2010). Berke-
ley, CA: USENIX Association.
Becker, B., Franciosa, P. G., Gschwind, S., Ohler,
T., Thiemt, G., & Widmayer, P. (1991). An optimal
algorithm for approximating a set of rectangles
by two minimum area rectangles. In H. Bieri &
H. Noltemeier (Eds.), Proceedings of the Inter-
national Workshop on Computational Geometry,
(pp. 13-25). Berlin, Germany: Springer.
Belkin, N. J., Kantor, P., Fox, E. A., & Shaw, J.
A. (1995). Combining the evidence of multiple
query representations for information retrieval.
Information Processing & Management, 31(3),
431–448. doi:10.1016/0306-4573(94)00057-A
Bender, M., Michel, S., Weikum, G., & Zimmer,
C. (2005). The minerva project: Database selec-
tion in the context of P2P search. In G. Vossen,
F. Leymann, P. C. Lockemann, W. Stucky (Eds.),
Proceedings of Datenbanksysteme in Business,
Technologie und Web, 11: Fachtagung des GI-
Fachbereichs “Datenbanken und Informations-
systeme”, (pp. 125-144). Karlsruhe, Germany:
DIS.
Blank, D., El Allali, S., Müller, W., & Henrich, A.
(2007). Sample-based creation of peer summaries
for efficient similarity search in scalable peer-to-
peer networks. In R. Lienhart, A. R. Prasad, A.
Hanjalic, S. Choi, B. P. Bailey, & N. Sebe (Eds.),
Proceedings of the 15th International Confer-
ence on Multimedia, (pp. 143-152). Augsburg,
Germany: ACM.
Blank, D., & Henrich, A. (2009). Summariz-
ing geo-referenced photo collections for image
retrieval in P2P networks. In Proceedings of the
International Workshop on Geographic Infor-
mation on the Internet, (pp. 55-60). Retrieved
from http://georama-project.labs.exalead.com /
workshop/GIIW-proceedings.pdf.
Blank, D., & Henrich, A. (2010). Description
and selection of media archives for geographic
nearest neighbor queries in P2P networks. In
A. R. Doherty, C. Gurrin, G. J. F. Jones, & A. F.
Smeaton (Eds.), Proceedings of Information Ac-
cess for Personal Media Archives Workshop, (pp.
22-29). Received from http://doras.dcu.ie/15373/.
Bockting, S., & Hiemstra, D. (2009). Collection
selection with highly discriminative keys. In
Proceedings of the 7th International Workshop on
Large-Scale Distributed Systems for Information
Retrieval. Retrieved from http://lsdsir09.isti.cnr.
it/ lsdsir09-1.pdf.
Chen, S., Liang, S., & Wang, M. (2010). A local-
ity-aware peer-to-peer approach for geospatial
web services discovery. In Canadian Geomatics
Conference. Retrieved from http://www.isprs.
org /proceedings/XXXVIII/part1/13/ 13_01_Pa-
per_65.pdf.
Chen, Y.-Y., Suel, T., & Markowetz, A. (2006).
Efficient query processing in geographic web
search engines. In S. Chaudhuri, V. Hristidis, &
N. Polyzotis (Eds.), Proceedings of the 25th In-
ternational Conference on Management of Data,
(pp. 277-288). Chicago, IL: ACM.
681
 Describing and Selecting Collections of Georeferenced Media Items

Cuenca-Acuna, F., Peery, C., Martin, R. P., & Guttman, A. (1984). R-trees: A dynamic index
Nguyen, T. D. (2003). PlanetP: Using gossiping structure for spatial searching. In B. Yormark
to build content addressable peer-to-peer informa- (Ed.), Proceedings of ACM SIGMOD Conference,
tion sharing communities. In Proceedings of IEEE (pp. 47-57). Boston, MA: ACM.
International Symposium on High Performance
Guy, I., Jacovi, M., Perer, A., Ronen, I., & Uziel,
Distributed Computing, (pp. 236-246). Seattle,
E. (2010). Same places, same things, same people?
WA: IEEE Press.
Mining user similarity on social media. In Pro-
Dolin, R., Agrawal, D., Abbadi, A. E., & Dil- ceedings of the 22nd International Conference
lon, L. K. (1997). Pharos: A scalable distributed on Computer Supported Cooperative Work, (pp.
architecture for locating heterogeneous informa- 41-50). Savannah, GA: ACM.
tion sources. In F. Golshani & K. Makki (Eds.),
Hariharan, R., Hore, B., & Mehrotra, S. (2008).
Proceedings of the 6th International Conference
Discovering GIS sources on the web using sum-
on Information and Knowledge Management, (pp.
maries. In R. Larsen, A. Paepcke, J. L. Borbinha,
348-355). Las Vegas, Nevada: ACM.
& M. Naaman (Eds.), Proceedings of the 8th ACM/
Doulkeridis, C., Vlachou, A., Nørvåg, K., Kotidis, IEEE Joint Conference on Digital Libraries, (pp.
Y., & Vazirgiannis, M. (2009). Multidimensional 94-103). Pittsburgh, PA: ACM Press.
routing indices for efficient distributed query
Ilyas, I. F., Beskales, G., & Soliman, M. A. (2008).
processing. In D. Cheung, I. Song, W. Chu, X.
A survey of top-k query processing techniques in
Hu, J. Lin, J. Li, & Z. Peng (Eds.), Proceedings
relational database systems. ACM Computing Sur-
of the 18th ACM International Conference on
veys, 40(4), 1–58. doi:10.1145/1391729.1391730
Information and Knowledge Management, (pp.
1489-1492). Hong Kong, China: ACM Press. Lupu, M., Li, J., Ooi, B. C., & Shi, S. (2007).
Clustering wavelets to speed-up data dissemina-
Doulkeridis, C., Vlachou, A., Nørvåg, K., & Vazir-
tion in structured P2P manets. In Proceedings of
giannis, M. (2010). Distributed semantic overlay
the 23rd International IEEE Conference on Data
networks. In Shen, X., Yu, H., Buford, J., & Akon,
Engineering, (pp. 386-395). Istanbul, Turkey:
M. (Eds.), Handbook of Peer-to-Peer Networking.
IEEE Press.
Berlin, Germany: Springer. doi:10.1007/978-0-
387-09751-0_17 Memon, F., Tiebler, D., Dürr, F., Rothermel, K.,
Tomsu, M., & Domschitz, P. (2009). Scalable
Eisenhardt, M., Müller, W., Blank, D., El Allali, S.,
spatial information discovery over distributed hash
& Henrich, A. (2008). Clustering-based, load bal-
tables. In Proceedings of the Fourth International
anced source selection for CBIR in P2P networks.
ICST Conference on COMmunication System
International Journal of Semantic Computing,
softWAre and middlewaRE (COMSWARE 2009).
2(2), 235–252. doi:10.1142/S1793351X08000439
New York, NY: ACM.
Elahi, B. M., Römer, K., Ostermaier, B., Fahr-
Müller, W., Eisenhardt, M., & Henrich, A. (2005).
mair, M., & Kellerer, W. (2009). Sensor ranking:
Scalable summary based retrieval in P2P networks.
A primitive for efficient content-based sensor
In A. Chowdhury, N. Fuhr, M. Ronthaler, H.-J.
search. In Proceedings of the 8th International
Schek, & W. Teiken (Eds.), Proceedings of the
Conference on Information Processing in Sensor
14th ACM International Conference on Informa-
Networks, (pp. 217-228). Washington, DC: IEEE.
tion and Knowledge Management, (pp. 586-593).
Bremen, Germany: ACM Press.

682
Describing and Selecting Collections of Georeferenced Media Items
Nogueras-Iso, J., Zarazaga-Soria, F. J., & Muro-
Medrano, P. R. (2005). Geographic informa-
tion metadata for spatial data infrastructures:
Resources, interoperability and information
retrieval. Secaucus, NJ: Springer-Verlag.
Novak, D., Batko, M., & Zezula, P. (2008). Web-
scale system for image similarity search: When
the dreams are coming true. In Proceedings of
6th International Workshop on Content-Based
Multimedia Indexing, (pp. 446-453). London,
UK: IEEE.
Papapetrou, O., Siberski, W., Balke, W.-T., &
Nejdl, W. (2007). DHTs over peer clusters for
distributed information retrieval. In Proceedings
of the 21st International Conference on Advanced
Information Networking and Applications, (pp.
84-93). Niagara Falls, Canada: IEEE.
Putman, W. M., & Lin, S.-J. (2007). Finite-volume
transport on various cubed-sphere grids. Jour-
nal of Computational Physics, 227(1), 55–78.
doi:10.1016/j.jcp.2007.07.022
Ratnasamy, S., Francis, P., Handley, M., Karp,
R., & Schenker, S. (2001). A scalable content-
addressable network. In Proceedings of the An-
nual Conference on Applications, Technologies,
Architectures, and Protocols for Computer Com-
munication, (pp. 161-172). San Diego, CA: ACM.
Samet, H. (2006). Foundations of multidimen-
sional and metric data structures. San Francisco,
CA: Morgan Kaufmann Publishers Inc.
Shen, X., Yu, H., Buford, J., & Akon, M. (2010).
Handbook of peer-to-peer networking. Berlin,
Germany: Springer. doi:10.1007/978-0-387-
09751-0
Sinnott, R. (1984). Virtues of the haversine. Sky
and Telescope, 68(2), 158.
Stoica, I., Morris, R., Karger, D., Kaashoek, F.,
& Balakrishnan, H. (2001). Chord: A scalable
peer-to-peer lookup service for Internet applica-
tions. In Proceedings of the Annual Conference
on Applications, Technologies, Architectures, and
Protocols for Computer Communication, (pp.
149-160). San Diego, CA: ACM.
Szalay, A., Gray, J., Fekete, G., Kunszt, P., Ku-
kol, P., & Thakar, A. (2005). Indexing the sphere
with the hierarchical triangular mesh. Technical
Report: MSR-TR-2005-123. Retrieved from
http://research.microsoft.com /pubs/64531/tr-
2005-123.pdf.
Thomas, P., & Hawking, D. (2009). Server selec-
tion methods in personal metasearch: A compara-
tive empirical study. Information Retrieval, 12(5),
581–604. doi:10.1007/s10791-009-9094-z
Vincenty, T. (1975). Direct and inverse solutions
of geodesics on the ellipsoid with application of
nested equations. Survey Review, 22(176), 88–93.
Vu, Q. H., Lupu, M., & Wu, S. (2009). Simpson:
Efficient similarity search in metric spaces over
P2P structured overlay networks. In H. Sips,
D. Epema, & H.-X. Lin (Eds.), Proceedings of
the 15th International Euro-Par Conference on
Parallel Processing, (pp. 498-510). Delft, The
Netherlands: Springer.
Xiujun, M., Gang, L., Kunqing, X., & Meng, S.
(2006). A peer-to-peer approach to geospatial
web services discovery. In Proceedings of the 1st
International Conference on Scalable Information
Systems (InfoScale 2006). New York, NY: ACM.
KEY TERMS AND DEFINITIONS
Peer-to-Peer (P2P) System: A P2P system is
made up of distributed resources (i.e. computing
devices) which cooperate in order to provide and
consume certain services in a decentralized fash-
683
 Describing and Selecting Collections of Georeferenced Media Items

ion. In contrast to traditional client/server systems,
a peer can act as both, a client and a server.
Peer-to-Peer Information Retrieval (P2P
IR) System: A P2P system with a focus on the ad-
ministration and retrieval of media items. Hereby,
content-based information retrieval techniques are
applied in order to search for media items.
Resource Description: An important task in
distributed IR which provides the basis for query
routing. The resources have to describe their con-
tent in an adequate way—optimizing the trade-off
between selectivity and space efficiency of the
data summaries (also called resource descriptions).
Resource Selection: The process of determin-
ing the order in which distributed resources should
be contacted to fulfill a certain information need.
Besides knowing the order in which resources
should be contacted it is also important to deter-
mine when it is no longer beneficial to contact
further resources.
Structured P2P IR Systems: P2P IR systems
which are based on distributed index structures
with distributed hash tables (DHTs) being the
most prominent class member. In structured P2P
IR systems, every peer is responsible for a certain
region of the feature space.
Summary: The term summary also refers
to resource descriptions. It emphasizes the need
for space efficiency of the resource descriptions
in certain application contexts, e.g. unstructured
P2P IR systems.
Unstructured P2P IR Systems: In contrast
to structured P2P IR systems, connections in un-
structured P2P IR overlays do not emerge from
a distributed index structure and are thus formed
more arbitrarily. Here, every peer usually admin-
isters the index data of its own media objects.

This work was previously published in Discovery of Geospatial Resources: Methodologies, Technologies, and Emergent
Applications, edited by Laura Díaz, Carlos Granell and Joaquín Huerta, pp. 1-20, copyright 2012 by Information Science
Reference (an imprint of IGI Global).

684
 685

Chapter 42
Cognitive Mapping and
GIS for Community-Based
Resource Identification
Lyn Kathlene
Colorado State University, USA

ABSTRACT
This chapter describes and analyzes the effectiveness of two methodological techniques, cognitive
mapping and geographical information systems (GIS), for identifying social service resources. It also
examines the processes used to integrate hand-drawn map information into geocoded data points and
provides recommendations for improving efficiency and precision. As a first step to integrate Jefferson
County social service delivery into community-based child welfare “systems of care” (SOC), both formal
and informal services had to be identified. Cognitive mapping, a process by which participants draw
visual representations of geographical areas, was conducted with 247 participants in Jefferson County,
Colorado. Over 3,500 resources were identified and entered into a GIS to analyze the availability, capac-
ity, and distribution of social services in the county and within communities. Identification of community
resources via cognitive mapping and GIS analysis provide: (1) a comprehensive database of existing
services; (2) a basis to build communication networks and cooperation among government and com-
munity providers; (3) the ability to create an efficient system that avoids duplication of efforts; (4) an
understanding of the geographical distribution of resources; (5) the identification of resources lacking in
the county and specific communities; and (6) knowledge differences among diverse participant groups.

INTRODUCTION Division of Human Services, to conduct a re-

In December, 2003, the Colorado Institute of
Public Policy (CIPP) at Colorado State University
was contracted by Jefferson County, Colorado,
DOI: 10.4018/978-1-4666-2038-4.ch042
source identification analysis. The project was
one component in the first year of a five-year
Health and Human Services — Children’s Bureau
grant to create a “systems of care” (SOC) in child
welfare social service delivery.1 Jefferson County,
Colorado, was one of eight pilot sites awarded an

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Cognitive Mapping and GIS for Community-Based Resource Identification
SOC grant. The CIPP component was to identify
services available at the community-level and
discover services that were lacking.
SOC is a major paradigm shift in social service
delivery. It removes the locus of authority away
from one individual, the social service worker,
and replaces it with a group of service providers,
family, and community members to develop col-
lectively a comprehensive plan to move the child
and family out of crisis. The provision of services
are to be coordinated, community-based, cultur-
ally competent and individualized (Stroul, 1986).
To integrate Jefferson County social service
delivery into community-based comprehensive
child welfare SOC, both formal and informal
services had to be identified. Informal services
are of particular interest since these are likely the
least well-known (there was no official directory)
and serve populations at a community SOC level
(rather than county-wide). For definition purposes,
informal services were identified for participants
as private or not-for-profit programs, including
services such as church soup kitchens, non-profit
agencies providing transportation services for the
elderly, and in-home daycare providers not regis-
tered with the county. Formal services are public
programs at the state, county, and local level, such
as Jefferson County Mental Health Services, Title
XX daycare providers, public schools, public
transportation, and park and recreation programs.
To identify existing resources at the com-
munity level, cognitive mapping, a process by
which participants draw visual representations
of geographical areas, was conducted with 247
participants in Jefferson County, Colorado. Par-
ticipant groups in the mapping included social
service, non-profit, and faith-based providers,
social service clients (Temporary Assistance for
Needy Families (TANF) recipients, youths, foster
care providers, and adoptive parents), residents
and ethnic/racial enclaves (Latino, Eastern Euro-
pean, Native American, and African American).
In addition, all resources listed in the Jefferson
County resource guides were included in the
686
resource database. Over 3,800 unique resources
were identified and entered into a GIS — Arc-
Map, a component of ArcView — to analyze the
availability, capacity, and distribution of social
services in the county and within communities.
Census data was overlaid to identify high-need
areas and ethnic enclaves.
Here, a novel application of GIS for designing
improved social service delivery systems is de-
scribed. The chapter also discusses complications
involved in working with human service agencies,
and reconsiders the processes developed to merge
cognitive mapping information into ArcMap.
Cognitive Mapping: Origin and Uses
Cognitive mapping has not been comprehensively
integrated with GIS to the degree discussed in this
project, although elementary integration has been
done in some previous projects (Fulton, Horan, &
Serrano, 1997; Kathlene, 1997; Kathlene & Horan,
1998; Horan, Serrano, & McMurran, 2001).2 The
potential usefulness to the human services sectors
through projects that use the combined method-
ologies is substantial. To better understand this
“fit,” a brief review of cognitive mapping follows.
Cognitive mapping did not originate from
research on humans. Rather, the term “cognitive
map” was originally used to describe the mental
representations that rats develop as they navigate
the same maze multiple times (Tolman, 1948).
Quickly, researchers became interested in using
the information from cognitive maps in the human
context, and the resulting collection of methods
became known as cognitive mapping. Later, the
term expanded to include mental depictions of
more abstract entities, like ideas or chains of
events. Since its inception, cognitive mapping
has been used as an approach to a number of
real-world issues.
Cognitive mapping’s first practical application
was in the field of urban planning when Kevin
Lynch (1960) found that certain places or elements
in a city generated a positive emotional reaction
Cognitive Mapping and GIS for Community-Based Resource Identification
among its residents. This type of research, he
felt, could be used to better design cities so as to
make them more memorable. Jack Nasar (1988)
extended Lynch’s study by including a model
of how city attributes influenced the affective
responses of residents and visitors.
Cognitive mapping is also used to evaluate
mental representations of smaller areas. Map-
ping of specific neighborhoods (Quaiser-Pohl,
Lehmann, & Eid, 2004; Uzzell, Pol, & Badenas,
2002), college campuses (Hardwick, Wooldridge,
& Rinalducci, 1983; Holahan & Dobrowolny,
1978; Sholl, 1987), and buildings (Moeser, 1988;
O’Laughlin & Brubaker, 1998; O’Neill, 1991)
adds to knowledge about how spatial abilities
develop and are utilized, the skills related to map
creation, and the mapping abilities of various
groups. Numerous overlying maps of these smaller
areas have been used to program autonomous
mobile robots (Yoshino, 1991).
Almost anyone can successfully complete a
mapping exercise. Studies have found the quality
of female and male’s maps are similar (Evans,
1980; Magana & Norman, 1980; O’Laughlin &
Brubaker, 1998); nearly all ages can successfully
participate (Quaiser-Pohl, Lehmann, & Eid, 2004);
and map quality is not influenced by drawing skill
(Evans, 1980; Hardwick, Wooldridge, & Rinalduc-
ci, 1983). The only criteria that affects map quality
is familiarity with the target area, where quality
increases with familiarity (Evans, 1980; Fridgen,
1987; Unger & Wandersman, 1985). Familiarity,
however, can distort maps. For example, locations
of importance to the participant are often drawn
larger or more centrally than other map elements
(Holahan & Dobrowolny, 1978; Kathlene, 1997).
Finally, at the coding stage, the use of multiple
coders without knowledge of the study hypothesis
is commonly used (Daniels & Johnson, 2002);
yet, studies have found nearly perfect agreement
among multiple coders (Quaiser-Pohl, Lehmann,
& Eid, 2004).
For this project, the above issues did not present
problems. Each of the participants was directed
to draw the area they self-identified as their most
recognizable; therefore, they had high levels of
familiarity. The maps required particularly little
drawing skill as participants were only asked to
draw boundaries (for which they are provided a
ruler), mark their home with an “X,” and identify
landmarks with squares. Since the objective in the
study was to identify resources of importance to
the participant; having the maps drawn to scale
was irrelevant because the exact geographic co-
ordinates for each location are determined during
the coding process. Coding the maps involved
recording objective location information, thereby
removing the problem of coder subjectivity. In
short, information gleaned from the maps was
not dependent upon the drawing skills of the
participants.
ADAPTATION OF MAPPING
METHODS FOR RESOURCE
IDENTIFICATION
Each use of mapping described above required a
unique adaptation of the method. Participants in
these studies were asked to sketch maps, arrange
pictures of pieces of paper according to their loca-
tion or importance, locate points on a map, circle
areas on a map, recognize features on an aerial
photograph, or label features on a blank map.
The literature effectively suggests that as many
methods of mapping exist as purposes for using it.
In this project, participants attending a map-
ping workshop were first provided background
on the overall goals of the SOC project as well
as their role as a mapper in achieving those goals.
The hope was that understanding the importance
of the project would encourage conscientious
participation. Then, participants were provided
with a large blank sheet of paper and asked to
indicate the street or physical boundaries within
which they most commonly travel. Street maps
of the area were available to help them in this and
later tasks. Workshops consisted of three to sixteen
687
 Cognitive Mapping and GIS for Community-Based Resource Identification
people, and at each session they were encouraged
to work cooperatively with their fellow workshop
participants. Following a visualization exercise
to assist them in accessing non-verbal memories,
they were asked to fill in details such as streets,
natural features, and resources they commonly
use. It was explained that these maps were not
meant to be artistic works; simple boxes and lines
drawn with their rulers would suffice to represent
objects and locations. Once the maps were com-
plete, participants were asked to label various
types of resources with colors corresponding to
commercial businesses, community resources,
government agencies, community strengths, and
community weaknesses. They also completed a
survey asking for detailed description of the im-
portant resources and features on their maps and
to provide any additional information regarding
resources lacking in their area. Each map was
then entered into a GIS database. Figure 1 is an
example of a cognitive map drawn for the Jef-
ferson County Project.
Figure 1. Example of hand drawn cognitive map
688
GIS ANALYSIS TECHNIQUE
The GIS for this cognitive mapping project pro-
vides aggregated spatial analysis of the maps.
GIS combines the thousands of data points col-
lected from the participant-drawn maps, which
greatly enhances the value of cognitive mapping
information. In addition, each distinct physical
address has dozens of descriptors that identify
not only who provided the information but also
secondary data such as the participants’ percep-
tions of a particular resource or service (Longley,
Goodchild, Maguire, & Rhind, 2001). The data
is accessible in Excel-type tables that can be eas-
ily queried and displayed on a digital street map
(Longley, Goodchild, Maguire, & Rhind, 2001).
This analysis allows for preliminary interpretation
of the data set, such as the dispersion or clustering
of resources in geographic areas (Heagerty & Lele,
1998; Reich & Davis, 2003). GIS can incorporate
external data sources, such as census data, which
can be layered with the cognitive mapping data.
Cognitive Mapping and GIS for Community-Based Resource Identification
Integrating the cognitive mapping data with GIS
software creates nearly limitless ways to analyze
the data and allows for easily-interpreted visual
results.
Technical Capacity for the Project
For this project the particular GIS capabilities
were:
• Visual representation of addresses and
areas via data points, data lines, and data
shapes.
• The ability to layer data points for sorting,
categorizing, and selection purposes.
• Accompanying tables of descriptive vari-
ables for each data layer with the ability
to search and sort individual data points
based on particular variable(s).
• The ability to export the data to conduct
advanced spatial analysis.
The cognitive mapping process was greatly en-
hanced, especially for analysis purposes, by using
GIS to create unique, specialized parameters (for
a theoretical model, see Jordan, Raubal, Gartrell,
& Egenhofer, 1998). The potential outcomes of
such a union include:
• The capability to export a large quantity
of data in an Excel-type format that can
be searched and sorted based on any given
number of criteria.
• The ability to organize and sort data in a
spatial manner (i.e. in the form of maps).
• The opportunity to create an easily-updat-
ed database for organizing additional data
gathered by the cognitive mapping process
and through other resource identification
processes.
Structure of the GIS for the Project
The structure of the GIS system for this project
was based on three principles:
1. The structure needed to be flexible and al-
low for the maximum amount of data to be
stored and easily accessed.
2. The structure needed to be relatively simple
such that an inexperienced GIS user can un-
derstand the sorting, searching, and selecting
potential of the system.
3. The GIS system needed to be able to organize
data from a variety of sources and be easily
updated over time.
The first step to organizing the data was to
create basic social service or resource categories
(in GIS terms these categories are referred to as
layers). Each layer needed to be broad enough to
encompass a variety of unique data points (iden-
tified resources) but narrow enough to create a
good classification system for a beginning search
and find procedure. These basic layers were the
essential framework of the GIS system and the
search capabilities of the system that providers
could use; therefore, they had to match either the
basic social service field breakdown of resource
types and/or be intuitively organized.
Ultimately, 24 layers were created for the proj-
ect including Housing, Mental Health Services,
Transportation, Health Care, Emergency Services,
Domestic Violence, and Victim Services. Within
each layer were key fields with similar categories
across all the layers. For example, the data field
“Resource Type” exists in each of the 24 layers and
contains only one of five values: Governmental,
Community Non-Profit, Faith-Based, Private, or
Unknown. This allowed queries and linkages to
be made across layers. As an illustration, a user
could search for all the community non-profit
sponsored resources or services in two or more of
the layers and combine these results on one map.
689
 Cognitive Mapping and GIS for Community-Based Resource Identification
Search or selection functions work on a layer
by layer level in ArcMap. Within the Housing
layer there were attributes such as Resource
Name, Resource Address, and Resource Hours
of Availability. These attributes are easily sorted.
For example, there should only be one Lakewood
Housing Authority (LHA), but if there are two
locations for LHA, they can be distinguished from
one another in the data table due to the different
values entered in the Resource Address cell. How-
ever, not all searches are done based on a known
name or location; therefore, a more extensive list
of attributes must exist to properly describe each
resource. Some examples of other attribute identi-
fiers were Resource Type (government-sponsored,
private, faith-based, or community non-profit)
or Funding Source. Another type of attribute is a
binary descriptor such as Used by Respondent Yes/
No. This would apply to data points identified by
clients and useful in various types of qualitative
analysis. Below is a simplified example of a layer
table structure and associated attributes.
It is easy to see in Table 1 the large number of
possible searches based on only seven attributes.
For example, a provider might want to find all
government-sponsored housing resources. The
search would use “Resource Type=Government.”
The first two rows would be selected from the
search. If housing resources provided by com-
munity non-profits is of interest, the search would
be written as “Respondent Category=Client and
Resource Type=Community Non-Profit and Used
Table 1. Housing layer attribute table
Survey ID Respondent Resource Resource type
Category Name
0105 Client LHA Government
0431 Provider LHA Government
0302 Community Lakeside Apart- Private
Member ments
0746 Client Allison Care Community Non-
Center Profit
690
by Respondent=Yes.” In the table above, this
would select only the last row.
Each row represents one unique data point on
the GIS map and therefore represents one unique
resource. Resources can be identified and placed
on the map multiple times, as in the above ex-
ample, with the first and second row both being
the Resource Type “LHA.” What distinguishes
the two data points, however, is not the location
on the map but the multiple attributes in the table
listed above. Here, LHA was identified once by
a provider and once by a client, information that
indicates common knowledge. Alternatively, if
only clients identify the Allison Care Center, a
knowledge gap in Jefferson County Human Ser-
vices providers and resource manuals is revealed,
which can be easily remedied. Creating a com-
prehensive database with separate rows of mul-
tiple criteria allows for flexibility in the analysis
of resources ranging from a simple resource search
engine to a complex spatial analysis tool to nu-
merically describe the layout of resources and
services in Jefferson County.
GIS ANALYSIS OF
COMMUNITY RESOURCES
To do the GIS analysis, the resources identified
by the mapping and focus group participants were
verified using the 2004 Denver Metro phone book
(www.dexonline.com) and/or MapQuest (www.
mapquest.com). The addresses were entered into
Hours of Funding Type Used by
Availability Respondent
9am-5pm Government No
9am-5pm Government N/A
N/A N/A Yes
N/A Private Yes
Cognitive Mapping and GIS for Community-Based Resource Identification
an Excel spreadsheet, the type of resource and
service was coded into one or more of the 24
types of resources, and the participant group and
demographics were entered.3 The spreadsheet was
imported into ArcMap and geo-coded to place the
identified resources with known addresses on a
common geographical coordinate system.
The 24 layers of resource and service types
were created by: (1) working with participants
in four pilot sessions to understand how they
conceptualized services; (2) presenting the pilot
workshop information to Jefferson County Human
Service employees to add categories that aligned
with their current resource guides and experiences;
and (3) aligning with the Aires taxonomy used by
Colorado 2-1-1 system.4 The resulting twenty-four
layers allowed for some resources/services to be
coded into more than one layer, thereby increas-
ing the ability of information seekers to find the
desired resources. For example, a service dedicated
to providing transportation to elderly individuals
would be included in Aging and Adult/Senior
Services as well as Transportation Services. The
dual coding in this project and used by the 2-1-1
system produces a comprehensive searchable da-
tabase that can reach a specific resource through
multiple avenues.
At the onset of the project, the research team
decided that a method for determining when the
mapping process had reached a saturation level
of redundant information was needed. A novel
method, a “repeat rate,” was created. The repeat
rate was set at 80% based on the time and cost
involved in identifying additional resources. The
repeat rate estimates the projected number of new
points to be expected from each additional map.
So, for example, on the first map every identified
resource is unique and a new point in the database,
the repeat rate is zero. By the 100th map, if 50%
of the points collected were previously identified
on any of the previous 99 maps, the non-repeated
points for that map would be 50%. At some point,
theoretically, no new information will be gleaned
from additional maps, resulting in a 100% repeat
rate. In practical terms, full saturation cannot be
reached; but more importantly, a trade-off must be
made between additional information and cost to
acquire the information. In this project, there was
an average of 20 resource points per map. Each
map took approximately four hours to process
(from address coding the points to data entry
into ArcMap). An 80% threshold repeat rate was
chosen as a break-even level. At this rate, only one
additional new resource would be identified per
one hour of coding time. Less than one additional
new resource per hour of work was deemed an
inefficient use of resources.
SELECTED RESULTS
While the cognitive mapping process identified
formal and informal resources, without the abil-
ity to do aggregated and stratified analyses the
collected data would only have provided a list
of resources known by the mappers. To create a
comprehensive SOC child welfare system requires
detailed knowledge about the geographical distri-
bution of the density of community resources and
the types available in the community. Additionally,
to understand and correct the community-level
information gaps among providers, community
members and clients, a geographical analysis
stratified by resource categories and information
sources (e.g., county resource guides, social ser-
vice workers, community providers, faith-based
providers, and clients) was needed. GIS is able
to provide such geographical analysis. Impor-
tantly, data presented in visual form facilitated
information dissemination among a wide variety
of stakeholders.
Seven Jefferson County cities were the focus
of the mapping project. Clients and community
resident groups were drawn from Lakewood,
Arvada, Wheat Ridge, Golden, Littleton, and the
mountain communities of Conifer and Evergreen.
A total of 42 workshops were held at locations
in each of these communities. Special sessions
691
 Cognitive Mapping and GIS for Community-Based Resource Identification
Figure 2. Latino population growth in Jefferson
County, 1990-2000
were held with members of the dominant ethnic
groups, identified using census data imported into
ArcMap. The groups included Native American,
Russian, and Latino community residents,5 with
Latinos being the largest ethnic population in Jef-
ferson County. To determine the areas in Jefferson
County with the highest growth rate of Latinos,
1990 and 2000 census data was imported into Ar-
cMap and the percent change was calculated. The
highest concentrations of the Latino population
were found in the central eastern area of Jefferson
County (Lakewood and Arvada) on the border of
Denver County. Figure 2 shows the Latino growth
patterns from 1990 to 2000. Numbers on the map
are the percentage of the area’s Latino population
in the 2000 Census.
Current and past users of Jefferson County
social services were identified by Jefferson
County. Since the database was confidential, Jef-
ferson County would not allow CIPP to pull a
stratified sample. In addition, over half the client
records were missing complete address informa-
692
tion. Because of these difficulties, all complete
client files with zip codes in the study area were
mailed invitations by Jefferson County to par-
ticipate in the study. Fortunately, the clients,
foster, and adoptive parents who participated were
distributed throughout the study area. The follow-
ing map shows the home location cross-streets of
the participants and the extrapolated hypothetical
boundaries of their identified communities, rang-
ing from one-third mile to one mile around their
residence. The buffer area was based on actual
community boundaries drawn on the maps, which
ranged from larger than one mile to a handful that
were smaller than one-third mile. Figure 3 indi-
cates that the mapping participants who were
current or previous users of Jefferson County
social services had collective knowledge of
nearly all of the study area (see Appendix A for
demographic information of the participants).
Figure 4 shows the 3,845 resources identified
and mapped. The number of unique points is 1,819
after removing the repeated information. Identified
resources are distributed across all the Jefferson
County study areas with a scattering throughout
Denver County and a dense clustering along the
main arterial east-west street, Colfax.
Next are selected categories of resources to
illustrate how GIS was used to inform SOC plan-
ning efforts.
• Childcare: One hundred and eleven dif-
ferent childcare providers were identified.
As Figure 5 shows, there is very little over-
lap in knowledge between the Jefferson
County resource guides and the partici-
pants in the mapping workshops. This
could indicate: (1) clients and providers
are not familiar with the childcare resourc-
es which Jefferson County has identified;
and (2) Jefferson County is not effectively
disseminating this information. Many of
the childcare resources identified in the
Jefferson County resource guides are be-
fore-school and after-school care programs.
Cognitive Mapping and GIS for Community-Based Resource Identification

Figure 3. Client home locations and range boundaries

In contrast, the cognitive mapping partici-
pants identified pre-school childcare facili-
ties rather than before and after-school pro-
grams. Based on this analysis, the current
guides are lacking in pre-school childcare
resources. It is likely this information is
housed with a specific organization rather
than listed within Jeffco’s resource guides;
however, a consolidated database will fa-
cilitate resource integration, information
referral, and client access.

Figure 4. All Jefferson County, Colorado, resources identified by participants and resource guides

693
 Cognitive Mapping and GIS for Community-Based Resource Identification
• Education: There were 366 education
resources identified. This is the category
in the Jefferson County guides with the
most resources. As would be expected, the
Jefferson County Public Schools Resource
Guide has all the primary and secondary
public schools, as well as a few private
schools in the county. Figure 6 shows there
is very little additional information provid-
ed by the mapping sessions with regard to
education resources in Jefferson County;
however, the mapping identified a signifi-
cant number of resources in Denver that
are known and used by Jefferson County
residents. The Jefferson County education
resources not identified in the resource
guides were typically services such as
breast-feeding classes located at hospitals,
ESL classes available at community cen-
ters and/or libraries, and other educational
services available at libraries, these impor-
tant educational services are missing in
Jefferson County’s guides.
• Health care resources: Of the 190 health
care resources, 140 were identified by the
cognitive mapping process with only one-
third of the services identified by both par-
Figure 5. Childcare resources identified by participants vs. Jefferson County resource guides
694
ticipants and the Jefferson County resource
guides. As seen in the Figure 7, however,
the Jefferson County resource guides also
have information about several health care
services in Denver. Most of the Denver and
Jefferson County services were also iden-
tified by the cognitive mapping process.
Information from the cognitive mapping
process increases the resource base by
two-thirds, adding significantly to health
care knowledge of Jefferson County.
Three of the 24 resource categories have been
provided above for sake of brevity. However,
there were several trends that appeared in almost
all of the resource categories. First, the Jefferson
County resource guides failed to provide adequate
information. Except in education and childcare,
the resource guides fell far short of the number
of resources identified by the mapping participants.
While there are many childcare resources in the
Jefferson County information guides, there is
little overlap between the childcare listed in the
Jefferson County guides and the childcare identi-
fied by the cognitive mapping. The GIS maps
effectively demonstrate such knowledge gaps.
Cognitive Mapping and GIS for Community-Based Resource Identification

Figure 6. Education resources: Cognitive mapping vs. resource guides

Second, there are a significant number of
resources in Denver County (east of Jefferson
County) that providers and clients identify. Rea-
sonable accessibility to Denver County, as well as
lack of availability of the resources in Jefferson
County, likely accounts for this trend. Building
a community-based SOC will require Jefferson
County to find ways to offer some of these services
locally, a challenge that will require developing
community partnerships to overcome the financial
constraints which the County faces.
Third, opposite of the previous trend, Jefferson
County resource guides provide mainly Denver
locations for some types of resources, even though
the same resources exist in numerous places in
Jefferson County. Available resources closer to
Jefferson County residents are a fundamental
component of SOC and, in this trend, require only
disseminating the information effectively, which
is a low-cost method to improve community-based
service delivery.

Figure 7. Health care resources: Cognitive mapping vs. resource guides

695
 Cognitive Mapping and GIS for Community-Based Resource Identification
Finally, there is a large disparity in knowledge
between clients and providers. With the exception
of 3 of the 24 categories, Education, Recreation,
and Commercial Resources, the providers and
clients did not overlap significantly in knowledge
about resources. Providers know more about
traditional resources such as other agencies or
governmentally-supported social services, while
clients know about resources of a less traditional
nature, such as churches, motels, and parks where
teenagers gathered to socialize and engage in rec-
reational sports activities. Although these informal
resources are not referral services that providers
typically pass along to clients, they are important
community-based resources to share with clients.
In creating a community-based SOC, providers
need to be aware of the alternative methods cli-
ents use to meet their needs. In some instances,
this new information will lead to the creation of
government/community partnerships to more
effectively and efficiently deliver services. In
other circumstances, the additional knowledge of
resources will provide clients with options and/
or fill gaps in needs that traditional government
and community providers cannot meet.
LESSONS LEARNED
Several problems directly and indirectly related
to the GIS component of the project became ap-
parent and required adjustments to the procedures
or accommodations to the expected output. These
include research procedures that are incompatible
with social service agencies’ capacity, issues of
client confidentiality, repeat rates, incomplete and/
or inaccurate databases for coding resource loca-
tions, coding protocols, and mapping accuracy.
First, as has been found before, many county
and local agencies lack leadership that under-
stands the value of GIS in policy decision-making
(Greene, 2000; Nedovic-Budic, 1996; Ventura,
1995; Worrall & Bond, 1997). Hence, many agen-
cies lack the technical ability to employ GIS and,
696
consequently, also lack the understanding to work
effectively and efficiently with the researchers.
Furthermore, because social service agencies
typically do not have a GIS analyst on staff, data
and map files have limited usefulness beyond the
initial analysis as presented in the final report. Fi-
nally, human service agencies have organizational
procedures that create significant barriers in imple-
menting research projects, barriers that need to be
addressed in the project planning stages (Ventura,
1995). Jefferson County Human Services suffered
from all three impediments and was exacerbated
by the high turnover of the staff. In the first year,
two-thirds of the project staff left. By the middle
of the second year, only one person out of nine key
project staff remained. Those who left included
the project manager and the principal investigator,
both of who had been replaced twice. Within 18
months, none of the people who conceptualized
and wrote the HHS grant were involved in the
project. Institutional memory was wiped clean and
new staff was unfamiliar and wary of many com-
ponents laid out in the grant proposal, including
the untraditional resource identification method.
Higher administrative support for the innovative
project waned, and “business as usual” reasserted
itself as the dominant paradigm. It became clear
that the resource database developed through the
mapping process would not be updated on a regular
basis and, perhaps, not disseminated throughout
the organization if left to Jefferson County. The
CIPP sought out a more stable organization to
house the resource data, Colorado 2-1-1, with the
permission of the first project manager.
Second, human service agencies as well as
educational institutions cannot share client/stu-
dent data. This presents a significant research
barrier when the project requires participation of
these populations. Ideally, individuals within the
organizations would have both the access to the
data and sophistication to manipulate the data in
accordance with standard research protocols. This
is unlikely to be the case in institutions which
are financially strapped and lack the vision or
Cognitive Mapping and GIS for Community-Based Resource Identification
political will to invest in trained personnel and
needed research tools. To ameliorate these condi-
tions, project planning must include agreed-upon
protocols for effectively and efficiently handling
confidential data.
Third, unique to this project was the creation
of a “repeat rate” to set a standard for data density.
The 80% repeat rate was selected for efficiency of
resources, based on an extrapolation of the aver-
age number of points per map and time needed to
code and enter the data for each map. Unknown
was how many participants/maps were needed to
reach the 80% repeat rate in each of the 24 cat-
egories. Initially, the CIPP recommended target
was 450 participants. This number was revised
downward by Jefferson County Human Services
to a maximum of 250 participants. From the 247
actual participants, the 80% repeat rate was reached
in only two of the 24 resource categories. The
average repeat rate was 55% across all categories,
indicating that more than 250 participants were
needed to reach 80%. Whether 450 participants
were ultimately required is unknown. More im-
portantly, did the lower repeat rate significantly
affect the quality of the project? Certainly, fewer
resources were identified at the 55% rate; but 1,480
resources not in Jefferson County resource guides
were identified; not an insignificant contribution
to building a more comprehensive social services.
Fourth, in the process of coding the maps
and sorting the data to find repeated addresses
or groupings by type of provider, and so forth, it
was discovered that precise alphanumeric coding
was critical. With the large number of data fields
(attributes) assigned to each participant, there were
inconsistencies in some of the categories. The data
cleaning was more extensive than anticipated.
Future projects should utilize numeric coding in
attributes to the fullest extent possible and develop
strict alphanumeric standards for addresses, orga-
nizational names, and other alpha fields.
Finally, to find resource addresses, MapQuest
and the Denver metro area phone book were used.
MapQuest was the most efficient method but had
the most errors, as discovered when the address
was imported into ArcMap. A cross-check with
the phone books corrected most of these errors.
Nine percent of the mapping points were
unidentifiable due to a combination of missing
information in MapQuest and the phone book, and
poor location information on the hand drawn maps.
The latter accounted for a greater proportion of
the unidentified points, especially resources such
as neighborhood parks and unnamed resources
such as “soup kitchen.” Rather than rely solely
on participants naming the nearest cross streets
to such resources, the closest known commercial
entity should be identified. This redundancy will
reduce missing data due to participant error in
naming streets.
FUTURE TRENDS
While this project was limited to identifying
resources, spatial patterns of resource locations,
and knowledge gaps, the collected data can be
mined further. More specific uses can be created,
such as a searchable Web-based provider resource
database and the identification of physical and/or
service areas with inadequate resources in relation
to socio-economic deprivation areas. The latter
allows providers to demonstrate specific needs,
important for several reasons, including the pursuit
of future programmatic funding. These specific
uses are described in greater detail as follows:
• Provider resource database: In the future,
the Web-based database can be converted
into a tool for social service providers to
identify available resources and the most
accessible locations for clients (Worrall &
Bond, 1997). The end user (a case-worker)
would be able to search for particular re-
sources based on any number of criteria
or a combination of criteria. For example,
one might enter necessary criteria such as
Rental Assistance Housing Resource lo-
697
 Cognitive Mapping and GIS for Community-Based Resource Identification
cated within three miles of a given loca-
tion that also caters to Spanish-speaking
clientele. After these attributes or criteria
are entered into the appropriate locations
on the Webpage, a list of all the resources
or providers that fit the criteria could be
retrieved, similar to the business name
search feature available through a site such
as MapQuest. Finally, digital maps could
be generated with driving directions for the
case-worker to print out for the client. It is
also possible to map the public transporta-
tion routes to services.
• Needs assessments: The database can be
used to conduct comprehensive, quantifi-
able, and defensible needs assessments.
A social service provider administrator or
grant writer could search the data described
above in conjunction with Census data and
the County’s client locations to reveal ar-
eas of need or areas of excess (Bond &
Devine, 1991; Worrall & Bond, 1997).6 A
strategic plan could be developed to deter-
mine where a new office or access point for
a particular resource should be located to
serve the greatest number of clients. This
type of spatial analysis based on quantifi-
able numbers and distances can be used to
justify a particular course of action either
for internal/external accountability or to
acquire funding for various projects aimed
at community resource and social service
distribution.
CONCLUSION
Cognitive mapping combined with GIS analysis
is a powerful method for identifying community
resources by providing: (1) a comprehensive
database of existing services; (2) a basis to build
communication networks and cooperation among
government and community providers; (3) the
ability to create an efficient system that avoids
698
duplication of efforts; (4) an understanding of
the geographical distribution of resources; (5) the
identification of resources lacking in the county
and specific communities; and (6) knowledge
differences among diverse participant groups.
The addition of 1,480 resource locations within
the seven study areas (only a portion of Jefferson
County) nearly tripled the number of resources
and services listed in the Jefferson County guides.
Ultimately, service delivery in SOC is about
building partnerships across the multiple services
and bringing in new, even sometimes untraditional,
community partners. Family involvement is the
key in this collaborative arrangement. Similar to
untraditional community partners and resources,
families as partners do not fit easily within cur-
rent social service delivery structures, values,
and beliefs. Recognizing, valuing, and partnering
with resource providers identified by clients and
community members is one important step toward
shifting practices. Cognitive mapping with GIS
provides a tool for taking the first critical steps.
ACKNOWLEDGMENT
The author would like to thank April Smith,
Department of Psychology, Colorado State Uni-
versity, and Mary Tye, Department of Psychol-
ogy, Colorado State University, for running the
workshops and coding the data; David Wallick,
Colorado Institute of Public Policy, Colorado
State University, for conducting the GIS analysis;
and Juliana Hissrich for providing administrative
support to the project.
REFERENCES
Bond, D., & Devine, P. (1991). The role of
geographic information systems in survey
analysis. The Statistician, 40, 209–215.
doi:10.2307/2348493
Cognitive Mapping and GIS for Community-Based Resource Identification
Daniels, K., & Johnson, G. (2002). On trees and
triviality traps: Locating the debate on the con-
tribution of cognitive mapping to organizational
research. Organization Studies, 23(1), 73–81.
Evans, G. W. (1980). Environmental cogni-
tion. Psychological Bulletin, 88(2), 259–287.
doi:10.1037/0033-2909.88.2.259
Fridgen, J. D. (1987). Use of cognitive maps to
determine perceived tourism regions. Leisure
Sciences, 9(2), 101–117.
Fulton, W., Horan, T., & Serrano, K. (1997). Put-
ting it all together: Using the ISTEA framework to
synthesize transportation and broader community
goals. Claremont Graduate University, University
Research Institute, Claremont, CA.
Greene, R. W. (2000). GIS in public policy: Us-
ing geographical information for more effective
government. Redlands, CA: ESRI Press.
Hardwick, D. A., Wooldridge, S. C., & Rinalducci,
E. J. (1983). Selection of landmarks as a correlate
of cognitive map organization. Psychological
Reports, 53(3), 807–813.
Heagerty, P. J., & Lele, S. R. (1998). A composite
likelihood approach to binary spatial data. Journal
of the American Statistical Association, 93(443),
1099–1111. doi:10.2307/2669853
Hjortso, C. N., Christensen, S. M., & Tarp, P.
(2005). Rapid stakeholder and conflict assessment
for natural resource management using cognitive
mapping: The case of Damdoi Forest Enterprise,
Vietnam. Agriculture and Human Values, 22,
149–167. doi:10.1007/s10460-004-8275-z
Hobbs, B. F., Ludsin, S. A., Knight, R. L., Ryan, P.
A., Biberhofer, J., & Ciborowski, J. J. H. (2002).
Fuzzy cognitive mapping as a tool to define
management objectives for complex ecosys-
tems. Ecological Applications, 12, 1548–1565.
doi:10.1890/1051-0761(2002)012[1548:FCMA
AT]2.0.CO;2
Holahan, C. J., & Dobrowolny, M. B. (1978).
Cognitive and behavioral correlates of the
spatial environment: An interactional analysis.
Environment and Behavior, 10(3), 317–333.
doi:10.1177/0013916578103003
Horan, T., Serrano, K., & McMurran, G. (2001).
GIS for livable communities: Examiniation of
community perceptions of assets, liabilities and
transportation improvements. San Jose, CA: San
Jose University, Mineta Transportation Institute,
College of Business.
Jordan, T., Raubal, M., Gartrell, B., & Egenhofer,
M. J. (1998, July). An affordance-based model of
place in GIS. In Eighth International Symposium
on Spatial Data Handling ’98 Conference Pro-
ceedings, Vancouver, BC, Canada (pp. 98-109).
Kathlene, L. (1997). 29th street greenway corridor
citizen survey panel: Results of mapping exercise,
phase 3. Minneapolis, MN: University of Min-
neapolis, Humphrey Institute of Public Affairs.
Kathlene, L., & Horan, T. (1998). GIS survey of
29th street corridor, Minneapolis, MN. Minneapo-
lis, MN: University of Minneapolis, Humphrey
Institute of Public Affairs.
Longley, P. A., Goodchild, M. F., Maguire, D. J.,
& Rhind, D. W. (2001). Geographic information
systems and science. New York: John Wiley and
Sons, LTD.
Lynch, K. (1960). The image of the city. Cam-
bridge, MA: MIT Press.
Magana, J. R., & Norman, D. K. (1980). Meth-
odological inquiry into elicitation procedures:
Cognitive mapping and free listing. Perceptual
and Motor Skills, 51(3), 931–934.
Moeser, S. D. (1988). Cognitive mapping in a
complex building. Environment and Behavior,
20(1), 21–49. doi:10.1177/0013916588201002
Nasar, J. L. (1988). The evaluative image of the
city. Thousand Oaks, CA: Sage Publications.
699
 Cognitive Mapping and GIS for Community-Based Resource Identification
Nedovic-Budic, Z., & Godschalk, D. R. (1996).
Human factors in adoption of geographical infor-
mation systems: A local government case study.
Public Administration Review, 56, 554–567.
doi:10.2307/977254
O’Laughlin, E. M., & Brubaker, B. S. (1998).
Use of landmarks in cognitive mapping: Gender
differences in self report versus performance.
Personality and Individual Differences, 24(5),
595–601. doi:10.1016/S0191-8869(97)00237-7
O’Neill, M. J. (1991). Evaluation of a con-
ceptual model of architectural legibility. En-
vironment and Behavior, 23(3), 259–284.
doi:10.1177/0013916591233001
Quaiser-Pohl, C., Lehmann, W., & Eid, M. (2004).
The relationship between spatial abilities and rep-
resentations of large-scale space in children — a
structural equation modeling analysis. Person-
ality and Individual Differences, 36(1), 95–107.
doi:10.1016/S0191-8869(03)00071-0
Reich, R. M., & Davis, R. (2003). Spatial sta-
tistical analysis of natural resources (Tech. Rep.
No. NR512). Fort Collins, CO: Colorado State
University.
Sholl, M. J. (1987). Cognitive maps as orienting
schemata. Journal of Experimental Psychol-
ogy. Learning, Memory, and Cognition, 13(4),
615–628. doi:10.1037/0278-7393.13.4.615
Stroul, B. (1996). Profiles of local systems of care.
In B. A. Stroul and R. M. Friedman (Eds.), Systems
of care for children’s mental health (pp. 149-176).
Baltimore: Paul H. Brookes Publishing Co.
Tolman, E. C. (1948). Cognitive maps in rats and
men. Psychological Review, 55(4), 189–208.
doi:10.1037/h0061626
Unger, D. G., & Wandersman, A. (1985). The
importance of neighbors: The social, cognitive,
and affective components of neighboring. Ameri-
can Journal of Community Psychology, 13(2),
139–169. doi:10.1007/BF00905726
700
Uzzell, D., Pol, E., & Badenas, D. (2002). Place
identification, social cohesion, and environmental
sustainability. Environment and Behavior, 34(1),
26–53. doi:10.1177/0013916502034001003
Ventura, S. J. (1995). The use of geographi-
cal information systems in local government.
Public Administration Review, 55, 461–467.
doi:10.2307/976770
Worrall, L., & Bond, D. (1997). Geographical
information systems, spatial analysis, and public
policy: The British experience. International
Statistical Review, 65, 365–379.
Yoshino, R. (1991). A note on cognitive maps: An
optimal spatial knowledge representation. Jour-
nal of Mathematical Psychology, 35, 371–393.
doi:10.1016/0022-2496(91)90053-V
ENDNOTES
1
The project was supported by grant
#90CA1715/01, CFDA #93.570 from the
Federal Department of Health and Human
Services through Jefferson County, Colo-
rado.
2
The term cognitive mapping is used for a vari-
ety of techniques, including “fuzzy cognitive
mapping,” a technique that builds mental
maps of perceptions from focus-group and
interviews (Hjortso, Christensen, & Tarp,
2005; Hobbs et al., 2002). In this project,
cognitive mapping means hand-drawn
maps of tangible community resources and
locations, a geographical data collection
technique new to GIS.
3
Nine percent of the mapping points could
not be accurately located and were dropped
from the analysis. Of the remaining 89%, two
possible location errors could occur in trans-
ferring the cognitive map information into a
database for ArcMap. First, multiple coders
could use different alphanumeric codes,
thereby making the same resource appear
Cognitive Mapping and GIS for Community-Based Resource Identification
as a different resource. To correct this error,
the data was cleaned by conducting sorts on
multiple columns in the excel spreadsheet to
reveal unknown duplicates. For example, a
search on “Research Name” might find the
same resource with inconsistent address
codes. If the address did not match exactly
(e.g., one was coded with “St.” and another
coded with “Street,” the coding was corrected
to be consistent. Similar searches were done
on other categories such as street address,
street name, and zip code. The data was
cleaned accordingly. The second error was
from incorrect addresses in the MapQuest
and/or Dex directory. The Dex directory is
the official metropolitan phone and address
directory and should have a high level of
reliability; however, the actual reliability
rate is unknown. To correct for possible
errors, all identified social services not in
the Jefferson County resource guides (e.g.,
soup kitchens, English as a Second Language
courses, support groups, etc.) were called
to verify the address. It was assumed that
the Jefferson County resource guides had
accurate information.
4
All identified resources were provided
to Colorado’s 2-1-1 system, which is the
national abbreviated dialing code for free
access to health and human services infor-
mation and referral (I&R). 2-1-1 is an easy-
to-remember and universally-recognizable
number that makes a critical connection
between individuals and families in need and
the appropriate community-based organiza-
tions and government agencies. Housing the
data with 2-1-1 allows statewide access to
resources and bi-annual updating to keep the
information current. Colorado 2-1-1 system
is the depository for the resources collected
in this project. Web searchable database of
resources can be found at http://211colorado.
org/
5
CIPP provided Jefferson County with the
ethnic enclave areas based on the 2000 Cen-
sus. The Asian communities fell outside the
project boundaries set by Jefferson County
(see Figure 1) and, unlike Russians, Latinos,
and Native Americans, Jefferson County did
not request mapping with the Asian com-
munity.
6
For example, it might be found that 65% of
all users of a certain type of resource (this
data would be collected by cognitive map-
ping alone) live “x” number of miles away
(analysis performed by the GIS system) from
a particular needed or frequently-accessed
resource (gathered through cognitive map-
ping and other sources).
701
 Cognitive Mapping and GIS for Community-Based Resource Identification

APPENDIX

Only forty percent of the participants provided demographic information, which limits the ability to
determine the gender, age, and ethnicity/race of the participants. However, there is no way to determine
the representativeness of the sample on these traditional demographics since the population character-
istics are unknown. Even among the clients, the demographics are not available because most of the
client records were incomplete. Unlike many social research projects, demographic representation is
less of a concern. For the identification of resources, a cross-section of the types of people who use or
provide services and the geographical distribution of their knowledge was most important, of which
both criteria were met.

Table 2. Demographics of participants (n=100)

Demographic All participants Providers Clients Community

characteristic (n=100) (n=19) (n=72) Residents (n=9)
Number and percent female 85% 90% 82% 100%
Average age 34.39 39.75 31.86 44.83
Number and percent Caucasian 62% 68% 64% 33%
Number and percent Latino 19% 5% 24% 11%
Number and percent African American 6% 0% 4% 33%
Number and percent Native American 9% 21% 4% 22%
Number and percent Other 4% 5% 3% 0%

This work was previously published in Emerging Spatial Information Systems and Applications, edited by B. Hilton, pp. 326-
350, copyright 2007 by IGI Publishing (an imprint of IGI Global).

702
 703

Chapter 43
Complementary Part
Detection and Reassembly
of 3D Fragments
Vandana Dixit Kaushik
Harcourt Butler Technological Institute, India

Phalguni Gupta
Indian Institute of Technology Kanpur, India

ABSTRACT
This chapter presents an algorithm for identifying complementary site of objects broken into two parts.
For a given 3D scanned image of broken objects, the algorithm identifies the rough sites of the broken
object, transforms the object to a suitable alignment, registers it with its complementary part which
belongs to the same object, and finds the local correspondence among the fragmented parts. The pre-
sented algorithm uses multiple granularity descriptors and edge extraction to detect the exact location
of multiple cleavage sites in the object. It greatly reduces the amount of information to be matched and
also helps in identification of the parts; as a result it reduces the computational time in the processing.
It is also applicable to all triangulated surface data even in the presence of noise.

INTRODUCTION 2D image processing, it is essential to have the

Recent advances in the field of computer vision,
image processing and pattern recognition have
led to substantial research in object identifica-
tion and feature extraction from 2D images for
different applications. In order to design efficient
methods for solving problems in the field of
DOI: 10.4018/978-1-4666-2038-4.ch043
good quality of image which is dependent on the
image acquisition system and the environment.
Generally, 2D methods are found to be efficient
under controlled environment where the input
images are acquired in almost fixed orientation,
distance, camera calibration, illumination, fixed
atmospheric condition etc. However, in reality it
is expected that the acquired images which are to
be processed do not possess good quality and fixed

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Complementary Part Detection and Reassembly of 3D Fragments
environment. Further, information available in 2D
images may not be sufficient to generate needful
result for a given application. This has motivated
people to explore the possibility of analyzing 3D
images. The most obvious difference between a
3D image and a 2D image of a particular object
is its depth information. The 2D image provides
only monocular depth information while a 3D
image consists of both monocular and binocular
depth information. This binocular depth informa-
tion should also be considered for the study on
images. Moreover, a 2D image is very sensitive to
changes due to illumination or lighting conditions
and change in orientation of image being captured.
Since any 3D image considers the overall geom-
etry of the image being captured, therefore, it is
insensitive to the above mentioned changes. How-
ever, there are many challenges associated with
3D images; out of which handling large volume
of data, high computational cost to process these
data, inconvenient 3D acquisition systems and the
digital archiving of 3D image data (Terry, 1995)
are most important. There exists a large number
of 3D scanners (Boehler, & Marbs, 2002) that
can be used to scan any type of objects - from a
small-sized object to a life-sized artifact (Levoy,
2000). Features are extracted from this scanned
data for further processing.
This chapter deals with the problem of detect-
ing the complementary part and to reassemble the
broken 3D fragments. It has many applications in
the fields like archeology where multiple broken
fragments are obtained from the sites and parts
belonging to the same object are to be identified.
Along with this, automatic reassembly is required
in art restoration, forensics, computer-aided de-
sign, chemistry, medicine etc. (Barequet, & Sharir,
1997). Generally, the reconstruction of arbitrary
objects from their fragments may have many
constraints. Some of these are mentioned below.
• Arbitrary shapes of parts (fragments)
• Unknown shape and number of the final
objects
704
• Arbitrary number of fractured faces per
fragment
• Missing fragments
• Probably flawed or weathered surfaces
• Non-existence of strict assemblage rules
Solutions to any assembly problem should
be able to handle above mentioned challenges.
The problem considered in this chapter can be
formally defined as follows: Consider a 3D im-
age consisting of multiple objects scanned using
a 3D laser scanner and each object is broken
into two pieces. The problem is to detect the
complement of a broken part of a given object.
Detection of complementary parts in the multiple
broken objects is an important phase during the
reassembly. Here, the rough site/ cleavage site in
any fragment is identified and then an attempt is
made to detect its complementary rough site in
the other fragment. The detected complementary
parts are aligned for reassembly. For example
Figure 1 shows the multiple broken fragments.
Fragment A and A’ belong to one object. This
chapter presents an algorithm for the detection
of A and A’ which are the complementary parts
of same object. The result of detection is shown
in Figure 2 (a) and (b) respectively.
BACKGROUND
There have been lot of studies in the area of reas-
sembly and reconstruction of broken 3D objects
Figure 1. Multiple broken fragments
Complementary Part Detection and Reassembly of 3D Fragments
Figure 2. (a) Fragment A (b) Fragment A’ [Complementary of A]
due to advances in the field of 3D scanners and
3D imaging devices. In the last decade there have
been significant technological advances in the
design of tools for digitizing and modeling 3D
shape of objects. Modeling of 3D shapes also
includes regeneration and reassembly of broken
objects. Given multiple pieces of some 3D scanned
objects, the problem considered in this chapter is
to detect all complementary parts of each broken
object and to reassemble these detected parts.
In order to identify and reassemble the comple-
mentary parts, it considers the property of broken
fragments which leave the cleavage site rougher
as compared to remaining surface and based on
this property, the rough site of each broken frag-
ment is determined.
The work described here is mainly based on an
efficient surface inspection algorithm for cleavage/
rough site detection in broken 3D objects. The
algorithm captures differences in the geometric
structure in different parts of the 3D object rep-
resented by polygon meshes. Most of the work
in the field of roughness analysis of 3D meshes
has been done in context of visualization, finding
surface defects to assess the visual quality of water-
marked 3D objects. Rogowitz & Rushmeier have
proposed the perceptual metric which is based on
the estimation of roughness of the surface of the
watermarked 3D objects (Rogowitz, & Rushmeier,
2001). Perceptual considerations have been used
in Computer Graphics for various operations such
as mesh simplification (Bolin, & Meyer, 1998,
Lindstrom, & Turk, 2000, Nicolas, et. al. 2002,
Scopigno, et. al., 1998, Williams, et. al., 2003) and
perceptually guided rendering. Mesh simplifica-
tion aims at reducing the number of vertices of a
polygonal mesh model while preserving as much
as possible its visual appearance. In general, the
simplification process is driven by a similarity
metric that measures the impact of changes of the
model after each simplification step. Hausdorff
distance is often used as a metric to evaluate the
distortions due to mesh simplification. Nicolas et.
al. have proposed two tools for geometric mesh
comparison (Nicolas, et. al. 2002). This chapter
takes triangular meshes representing the surfaces
in terms of triangles [known as facets] as inputs
and determines all rough facets [obtained through
some threshold value] which form the cleavage
site of the fragment. This cleavage site is used to
detect the complementary part of a given broken
object. The meshes representing surfaces at dif-
ferent levels are shown in Figure 3(a-d)
Wu, et. al. have obtained the roughness of a
single facet using a dihedral angle based approach
(Wu, et. al., 2001). In this work, some conclusion
can be drawn on cleavage sites on the surfaces
by assigning a roughness level to each facet. This
metric does not take into account the scale of
roughness. In other words, The proposed metric
measures ‘bumpiness’ of a surface with granular-
ity close to the size of single facet and this is one
705
 Complementary Part Detection and Reassembly of 3D Fragments

Figure 3. Meshes at different levels

the area of neighboring facets. It does not take
into account the direction of neighbouring facets
which is very important as this may lead to wrong
detection of smooth surface as rough surface. In
order to explain it more clearly, let us consider
Figure 5 representing a smooth surface and a
rough surface. Corsini et. al. have ignored the
direction of neighbouring facets to determine the
roughness value and it provides same roughness
value for both the cases. Since both the metrics
discussed above focus on perceptual quality of
3D meshes, they may not be suitable for cleavage
site detection. However, Kaushik, et. al have
proposed a metric which makes uses of roughness
detection at single facet level along with different
granularity levels (Kaushik, et. al., 2007). Further,
it also takes into account the direction of neigh-
bouring facets thereby assigning a different rough-
ness values to a smooth surface and a rough
major drawback of this metric. Corsini et. al. have
surface. Therefore, it is found to be suitable for
proposed a new metric which takes into account
cleavage site detection.
multi-scale roughness along with the roughness
The detected cleavage sites in 3D fragments
at single facet level (Corsini, Gelasca, Ebrahimi,
help to reassemble 3D broken objects. This chap-
2005). Multi-scale roughness refers to bump size
ter first determines cleavage site or sites in broken
in a rough surface. For example, bumps or grains
3D scanned fragments, and then identifies comple-
in a brick are larger than those in a ceramic mate-
mentary parts in two broken objects at the cleaved
rial. Figure 4 illustrates the rough surfaces at
sites which reduce the area to be matched con-
various levels of granularity.
siderably and finally proposes a new process to
Generally, roughness of a facet can be obtained
register these complementary parts for reassembly.
as a function of the dihedral angles made with
It can be noted here that registration plays an
neighboring facets, variation of these dihedral
important role in 3D model acquisition, object
angles and the area of facets under consideration.
recognition, and geometry processing. For the
In order to obtain the roughness of a facet, the
two sets of points representing two scanned frag-
above said metric only considers the area of
ments, the registration process tries to find suitable
facet whose roughness is being calculated and not

Figure 4. Roughness at various levels of granu- Figure 5. Different surfaces

larity

706
Complementary Part Detection and Reassembly of 3D Fragments

Figure 6. Registration process

transformation composed of a rotation matrix and
a translation vector that aligns a scanned fragment
to get best match with another one. Figure 6(a)
shows two sets P and Q of points. After applying
the registration process the set P of points is aligned
to Q as shown in Figure 6(b).
There exist several well known algorithms for
registration. Most of them are based on Iterative
Closest Points [ICP] (Besl, & McKay, 1992).
Masuda, & Yokoya have used Least Median of
Squares (Rogowitz, & Rushmeier 2001) in the
ICP based algorithm to overcome the original
method’s intolerance to outlier data (Masuda, &
Yokoya, 1995). Even though the above algorithms
along with ICP can work for registration of
similar surfaces, but they may not work for
complementary surfaces. Another drawback with
the ICP is that it is very slow to determine the
correspondence between two sets of points and
may not give real time results. Also, the ICP based
algorithm involves processing of all the points in
the set which makes the processing time ex-
tremely high.
graphics applications. Since the scanner can scan
the objects having maximum height of 16 inch and
diameter of 10 inch, we have selected fragments
accordingly. It has a door to insert objects to be
scanned to place on a rotating plate. It scans the
object and creates a 3D point cloud data. Figure
7 shows the image of the scanner.
ROUGHNESS DETECTION
DESCRIPTOR
An efficient algorithm for detecting the rough
site of a broken fragment has been presented. It
also considers the direction of the facets properly
and therefore the desired rough site is obtained
efficiently. The proposed algorithm consists of the
Figure 7. Roland LPX-600 3D laser scanner

SCANNING OF FRAGMENTS

In order to scan fragments of objects, we have used

Roland LPX-600 3D laser scanner which is found
to create accurately digital information from an
object for 3D computer aided design and computer

707
 Complementary Part Detection and Reassembly of 3D Fragments
following major steps. Firstly, the roughness factor
for each facet both at single granularity and mul-
tiple granularity level is calculated. Secondly, the
facet’s final roughness factor is computed which
can be defined as the maximum value of rough-
ness factors of the facets at different levels. Then,
using a roughness histogram a dynamic roughness
threshold is obtained to mark portions of surface as
rough. In this process due to the presence of noise,
a smooth portion may be incorrectly marked as
rough. Therefore, in order to avoid such wrongly
detected rough portions from the broken fragment,
a roughness consolidation algorithm along with
a noise removal algorithm is used. The detailed
description to calculate roughness factor at each
step is given below.
Measuring Roughness
Roughness can be defined in terms of dihedral
angles (Wu, et. al., 2001), stochastic models (Jafari,
2003, Waechter, et. al., 2004) and fractals (Arrault,
Arnéodo, Davis, & Marshak, 1997) which are all
based on perceptual quality of a mesh or facet.
However, since primary aim of cleavage detection
algorithm is to decrease computations, perfect
roughness modeling is not required. Keeping this
issue in mind, the dihedral angle based approach
to determine roughness factor has been modified
which is found to be very efficient. Figure 8 shows
Figure 8. Cleavage site in a broken object
(Kaushik, et. al., 2007)
708
a cleavage site in a broken object obtained from
an archaeological site.
A rough surface is the surface with high con-
centration of ‘bumps’ of different sizes (Corsini,
Gelasca, Ebrahimi, 2005). It is difficult to define
a suitable metric that can help to differentiate
between roughness due to natural texture and
roughness due to cleavage. Since most fragments
obtained from an archeological site are weathered
and eroded, it becomes difficult to ascertain
whether roughness is due to the natural texture
or not. However, there exist fundamental differ-
ences in roughness generated through natural
texture and roughness due to cleavage. Following
are the observations about rough surfaces obtained
through experiments.
1. Composition of materials in the fragment and
the effect of external agents in the weathered
fragments are the major reasons for natural
roughness of a surface. It can be noted that
weathering agents have a smoothing effect
on surface texture.
2. Bumps of different sizes follow Gaussian
distribution.
3. In natural unbroken surfaces, over a long
measurement range, most of the smaller
bumps average out each other while the
larger bumps are retained. This is shown in
Figure 9.
4. Natural rough surfaces obey fractal property
which is computationally complex and is
difficult to use.
5. In case of a cleaved surface, high roughness
both at lower and at higher granularity levels
is visible.
Computation of Roughness of Single
Facet
Consider a regular triangular mesh M. Let F be
the facet or the triangle under consideration. An
adjacent/ adjoining facet of F is the facet which
Complementary Part Detection and Reassembly of 3D Fragments
Figure 9. Example of single and multilevel roughness (Kaushik, et. al., 2007)
Figure 10. Adjoining Facets (Kaushik, et. al.,
2007)
shares one edge with the facet F and is shown in
Figure 10.
Let θ be the dihedral angle between two facets
Fi and Fi. Finding the value of the angle θ involves
inverse trigonometric function which is compu-
tationally expensive. However, the value of θ can
be found through the angle function which is
given by (1 - cos θ). Let N F . and N F be the
i j
normals of the facets Fi and Fj respectively. Then
ωij is given by
 N N 
 Fi . Fj 
wij = 1 −   (1)
 N Fi . | . | N Fj . 
The value of the dihedral angle ωij is lying be-
tween 0 and 1. Smaller the value of ωij, smoother
is the surface (i.e. roughness of a facet with respect
to its adjoining facets is near zero). One simple
way to estimate the roughness factor of a facet is
by considering areas of the facet and its adjoining
facets along with their corresponding dihedral
angles. Thus, if R(Fi ) is the roughness factor of
the facet Fi, then R(Fi ) can be expressed as
R(Fi ) = fn{ ωij * ( AF . ; AF . )}
i j
where AF . is the area of facet Fi under consider-
i
ation and AF . is the area of facet Fj, ∀ j = 1, 2, 3.
j
It may be observed that the value of ωij can
be very small. Hence, even though there exists
a roughness of a facet with respect to its neigh-
bouring facets, but the difference between the
roughness factors of a smooth facet and a rough
facet may be very small. Thus, to highlight the
cleavage site consisting of such type of rough
facets, we need to use proper multiplicative factor
to upgrade the roughness factor R(Fi).
In this work, we have considered the rough-
ness of facet Fi, R(Fi), as the product of M(Fi)
and V(Fi) as
R(Fi) = M (Fi) ∗V (Fi) (2)
where M(Fi) and V(Fi), can be computed by
1
M(Fi ) = ∑ (ωij ) * (AFi + AFj)
3 Fj ∈A(Fi )
(3)
709
 Complementary Part Detection and Reassembly of 3D Fragments
1 1
V(Fi ) = ∑ (ωij.(AFi +AFj ) − M (Fi))2 R N (Fi) =
Fj ∈A(Fi ) 3
It can be observed that value of R(Fi ) is not
fixed. Thus, to apply a threshold for roughness,
there is a need to normalize the value of R(Fi ) to
the interval [0,1]. This can be done by
R(Fi) − min(R(Fi))
R '(Fi) =
max(R(Fi)) − min(R(Fi))
Multi-Granularity
Roughness Estimation
Corsini et. al. have defined the rough surface as
the surface having high concentration of steep
crests and troughs (Corsini, Gelasca, Ebrahimi,
2005). If one considers multi-scale roughness
i.e. bump size which is exceeding the dimension
of single facet, then the effects of local crest and
an adjacent local trough cancel each other by
making minimum effect on overall bumpiness
of the considered collective facets. Thus, the idea
of roughness estimation can be extended from
roughness of single facet to N levels of rough-
ness estimation which is known as estimation of
multi-granularity roughness.
Let a (N,1)-ring neighborhood of a facet F
comprise of facet F and its adjoining facets. One
can create a (N,2)-ring neighborhood from the
(N,1)-ring by adding all of the facets sharing a
common edge with the (N,1)-ring facets. If the
process is repeated by K times, a set of all facets
within the (N,K)-ring neighborhood can be cre-
ated. The value K is called the clustering radius
of (N,K)-ring. Suppose, S FN is the set of all facets
within the (N,N)-ring neighbourhood where N is
the radius of (N, N)-ring for the facet F. Roughness
of (N,N)-ring or estimation of multi-granularity
roughness at level N is given by
710
N ∑
| S F | Fi ∈SFN
(−1)1+Zij R(Fj ) (6)
(4)
where | S FN | is the number of facets in (N,N)-ring
and the Boolean variable Zij indicates whether the
facet is downward sloping or upward sloping, i.e.,
1 if the facet Fj is upward sloping with respect to Fi
Zij = 
0 if the facet Fj is downward sloping with respect to Fi

(5)
Zij can be computed by
 NFi.(CFj − CFi) 
Zij = 1 + (7)
 | NFi | . | (CFj − CFi) |

where ⌊.⌋ is the floor function, CFi and CFj are the
centroid of facet Fi and Fj respectively. It can be
noted that the complementary (upward and down-
ward sloping) surfaces local bumps, as shown in
Figure 11, may cancel each other’s effect and this
leads to contribute higher scale roughness to the
bump. As a result, it leads to provide larger scale
roughness to the bump.
Also, any surface, which is looking smooth to
the naked eye, may often be found to be rough
under microscope. This type of surface is an ex-
ample of low scale roughness. Again, the surface
of sandstone is an example of high scale rough-
Figure 11. Bumps with different scales (Kaushik,
et. al., 2007)
Complementary Part Detection and Reassembly of 3D Fragments

ness. However, it can be noted that this scale is
also dependent on the resolution of the scanner.
Locating Rough Site at
Multiple Roughness Levels
The values of R(F) which are obtained using
Equation (5) and Equation (6) provide the rough-
ness factors of the single granularity and multi-
granularity level respectively. It can be noted
that the value of R(F) is directly related with the
roughness, i.e. more the value of R(F), more is the
roughness. Thus the cluster of faces which is hav-
ing the values of roughness more than roughness
threshold can be found in the surface and that can
be labeled as rough area. Hence the cleavage site
can be determined which is shown in Figure 12.
Roughness Histogram
minimum value. A facet is marked as rough if its
roughness value is greater than the threshold H.
Figure 14 shows the roughness histogram of
the object considered in Figure 13. Note that the
thick line in the center represents threshold value
H which is used to determine the cleavage site.
It is shown in the highlighted area of Figure 12.
Consolidation of Roughness
Patches
Due to the presence of noise, some small spuri-
ous patches on smooth area can be formed. It is
important to remove such type of noise from the
Figure 13. Source surface used to generate Figure
12 (Kaushik, et. al., 2007)

These roughness values of all facets can be used

to compute its histogram. For a particular rough-
ness granularity, the roughness distribution in
that fragment can be seen from the computed
histogram. It can be noted that roughness value is
always found to be the least at and near the centre
position. The roughness threshold H is selected
from the roughness level corresponding to this
Figure 14. Roughness histogram (Kaushik, et.
al., 2007)
Figure 12. Roughness values of facets and high-
lighting cleavage site using threshold as 0.5914
(Kaushik, et. al., 2007)

711
 Complementary Part Detection and Reassembly of 3D Fragments

cleavage site. Suppose, there are Ψ rough facets in
an (N,K)-ring having T facets. If the value of Ψ is
very small then one can assume that all T facets in
that ring are smooth. In other words, all T facets
of an (N,K)-ring in a fragment are considered to
be smooth if for a predefined threshold value u,
0<u<1, it satisfies the following:
Ψ<T*u.
Again, it can be observed that for any given
threshold value H obtained from roughness histo-
gram, the roughness detection algorithm may not
produce continuous roughness areas in an (N,K)-
ring; i.e. some scattered roughness patches, say Ψ,
can be found in (N,K)-ring through this algorithm
as shown in Figure 10. A continuous rough area
which forms a cleavage site can be obtained by
merging these scattered patches. All T facets of
an (N,K)-ring can be marked as rough if for some
predefined threshold v lying between 0 and 1, it
satisfies the following
Ψ>T*u.
Based on the above discussion the algorithm
for roughness detection of a single facet F of a
broken surface has been given in Algorithm 1.
It can be noted that the time complexity of the
algorithm is O(n) where n is the number of facets
in the cleavage site.
CLEAVAGE SITE DETECTION
During the detection of the rough surface, the
edge facets are also detected as rough because of
their higher dihedral angle and angle function. A
cleavage site from a rough surface of a given 3D
broken object is determined by removing these
edge facets. This approach can be used to find all
cleavage sites of a 3D broken object.
Edge Descriptor
A hypothetical ball of (N,N)-ring diameter is
moved over the entire surface of the object and the
number of facets lying within it is used to extract
edges. The ball is centered at the midpoint of every
facet and the area lying within the ball is used for
edge detection. The edge descriptor, E(Fi) for the
facet Fi can be defined mathematically as follows
E (Fi) = ∑ A(Fj ) ∀j ∋ Dist(Mid (Fj ), Mid (Fi))< R
(8)
where R is the radius of the ball, A(Fj) is the area
of an adjacent facet of Fi, Mid(Fi) is the midpoint
of facet Fi and Dist(X,Y) is the distance between

Algorithm 1. Roughness-Detection (F)

Step 1: Compute the roughness factor at single level R(Fi) using the formula
R(Fi) − min(R(Fi))
R '(Fi) =
max(R(Fi)) − min(R(Fi))
where R(Fi ) is defined in Equation (2)
Step 2: Compute the roughness factor at multiple levels using the following formula
1
R N (Fi) = N ∑
| S F | Fi ∈SFN
(−1)1+Zij R − (Fj )

Step 3: Generate the roughness histogram.

Step 4: Select threshold roughness value H.
Step 5: Mark facets having roughness more than computed threshold.
Step 6: Consolidate all rough patches and remove noise to detect rough surface.

712
Complementary Part Detection and Reassembly of 3D Fragments

point X and Y. If E(Fi) > T, then Fi is an edge facet
where T is the predefined threshold.
The cleavage site is detected by removing all
detected edge facets from its corresponding rough
surface. Figure 15(a) shows the broken 3D object
while Figure 15(b) shows the detected rough sur-
face of applying roughness detection algorithm
and Figure 15(c) gives the rough surface after
removing its edge facets.
Clustering
register the corresponding complementary parts
of the objects in the registration phase.
Let Si be a rough surface and Fj be a facet in Si
then the distance between two rough surfaces can
be defined as the Euclidean distance between their
closest facets. The rough surfaces with distance
less than the predefined threshold form a cluster
Cm. Thus, a cluster Cm can be defined as
Cm = U kn=0Sk : : ∀Si ∃ Sj ∋ Dist(Si, Sj) < threshold
where 0 ≤ i,j ≤ n (9)

Clustering is used to obtain the multiple cleavage

These clusters Cm represent different cleavage
sites (if any) in the broken 3D object. The process
sites in the broken 3D object. Figure 16(a) shows
of grouping a set of physical or abstract objects
the broken object with two cleavage site while
into classes of similar objects is called cluster-
Figure 16(b) shows the result of clustering on a
ing. More clearly, a cluster is a collection of data
broken object.
objects that are similar to each other within the
same cluster and are dissimilar to the objects in
other clusters. That is, a cluster of data objects
COMPLEMENTARY PART
can be treated collectively as one group.
DETECTION
Advantages of clustering are that it is adaptable
to changes and helps to identify useful features
The complementary part detection problem is to
that distinguish different groups. In this case
identify parts of P and Q of a given 3D object that
groups are multiple cleavage sites and the feature
represents the same physical part of the scanned
that distinguishes them is the minimum distance
object. The correspondence descriptor is used to
between them. The Euclidean distance among all
obtain the complementary parts from multiple
closest pairs of rough facets is used to determine
broken objects.
the distance between two cleavage sites. If the
Assume that the objects are broken into two
distance is lying within some predefined threshold,
parts, P and Q, and parts are point-sampled. Let
both the cleavage sites are considered to be a single
parts P and Q consist of n points {p1, p2,.…., pn}
cluster. This process creates a number of clusters
and {q1,q2 …..,qn} respectively. A pair of points
representing each cleavage site in the broken part
from P and Q is in correspondence by assigning to
of the object. These clusters are then processed to
pi a corresponding point qi. This descriptor, termed

Figure 15. Detection of cleavage site from broken 3D object

713
 Complementary Part Detection and Reassembly of 3D Fragments
Figure 16. Detected cleavage sites after clustering
as correspondence descriptor, is the mean of the
Euclidean distances of facets from the center of
the cleavage site which is given by
Cd =
∑E F
n N =
where n is the total number of facets in the cleavage
site and EF is the Euclidean distance of all facets
from the center of a given cleavage site.
REGISTRATION
After detection of complementary cleavage sites,
these complementary parts need to be aligned so
that they can be registered to regenerate the original
object. This phase registers the cleavage sites of
the detected complementary parts by realignment
and translation of the broken parts. The steps for
registration are discussed here.
Changing to Common Orientation
The initial orientation of the complementary
cleavage sites is random; hence they need to be
transformed to a common orientation. The regis-
tration phase first aligns the cleavage sites of the
broken fragments parallel to the ZX-plane. This
is done by fitting a plane on the cleavage site as
shown in Figure 17. The plane is fitted by taking
weighted average of the normals of the rough
714
facets with their corresponding area. Let there
be n facets in the cleavage site. Then average of
normals is given by
n
∑N A k k
k =1
(11)
n
∑A k
k =1
where Nk and Ak are the normal and the area of
the kth facet in the cleavage site. Once the plane
is fitted, it needs to be aligned parallel to a refer-
ence plane which in this case is the ZX-plane, as
shown in Figure 18.
Aligning Cleavage Sites
Cleavage sites which are in a common orienta-
tion, i.e. parallel to ZX-plane, further needs to
be aligned such that one cleavage site is above
the ZX-plane while other one is below it. This is
done by considering two points having maximum
and minimum values of the Y-coordinate on each
cleavage site. Figure 19 shows these coordinates.
The cleavage sites are aligned first by translating
the minimum Y-coordinate values of both the
cleavage sites with respect to the origin and then
by matching the corresponding maximum values
of Y-coordinate so that these two aligned cleavage
sites can be registered. Aligned cleavage sites are
shown in Figure 20.
Complementary Part Detection and Reassembly of 3D Fragments
Figure 17. Initial orientation of fragments
Figure 18. Two cleavage sites in common orientation with respect to ZX-plane
Registration of Cleavage Sites
To register the aligned complementary cleavage
sites, the corresponding maximum Y-coordinates
are matched. This is achieved by rotating one of
the objects about Y-axis with its axis perpendicular
to the plane fitted on the cleavage site passing
through the minimum Y-coordinates.
Consider a vector for each cleavage site which
is formed by joining the maximum and minimum
Y-coordinates. Then the angle between the pro-
jections of the two vectors of the corresponding
cleavage sites on the ZX-plane (by the dot prod-
uct method) is found. Further, one cleavage site
is rotated by that angle about the Y-axis so that
the corresponding maximum Y-coordinates are
matched with each other. This process makes the
registration of two complementary cleavage sites
as shown in Figure 21.
Based on the above discussion the algorithm
for registration of complementary cleavage sites
of broken 3D fragments surface has been given
in Algorithm 2. It can be noted that the time
complexity of the algorithm is O(n) where n is
the number of facets in the cleavage site.
EXPERIMENTAL RESULTS
The proposed roughness detection algorithm has
been tested on more than 50 scanned fragments.
All the fragments are scanned using 3D laser LPX
600 Roland scanner. The results have shown that
the proposed algorithms are well suited for detect-
715
 Complementary Part Detection and Reassembly of 3D Fragments

Figure 19. Cleavage sites with maximum and minimum Y-coordinate

Figure 20. Aligned cleavage sites

Figure 21. Registered cleavage sites

716
Complementary Part Detection and Reassembly of 3D Fragments

Algorithm 2. Registration

Step 1: Detect the cleavage sites of broken 3D fragments following the steps given below
          Step 1.1: Use Roughness Detection algorithm as discussed above

          Step 1.2: Determine cleavage site from a rough surface by removing these edge facets as discussed above

          Step 1.3: Use clustering to obtain all multiple cleavage sites in the broken 3D object using the method discussed

Step 2: Determine the complementary parts of the detected cleavage sites using the correspondence descriptor given by

Cd =
∑E F

n
Step 3: Bring both complementary cleavage sites to common orientation by using the following steps

          Step 3.1: Fit a plane on each cleavage site by taking weighted average of the normals of the rough facets with their corresponding
area which is given by
n

∑N A k k

N = k =1
n

∑A k
k =1

          Step 3.2: Using these planes fitted on the cleavage sites, align them parallel to the ZX-plane

Step 4: Align the cleavage sites such that one cleavage site is above the ZX-plane while other one is below it using

          Step 4.1: Consider two points having maximum and minimum values of the Y-coordinate on each cleavage site

          Step 4.2: Translate the minimum Y-coordinate values of both the cleavage sites with respect to the origin

Step 5: Register the aligned complementary cleavage sites by following the steps given below

          Step 5.1: Form a vector for each cleavage site by joining the maximum and minimum Y-coordinates

          Step 5.2: Compute the angle between the projections of the two vectors of the corresponding cleavage sites on the ZX-plane

          Step 5.3: Rotate one cleavage site by the angle computed in the above step about the Y-axis in order to match the corresponding
maximum Y-coordinates

ing cleavage sites in broken objects in a much
accurate and efficient manner. To illustrate the
performance of the algorithm, the original frag-
ment’s surface as marked in Figure 13 has been
considered. The calculated roughness values of a
single facet as well as multiple facets are shown
in Figure 12. Using these roughness values of all
the facets, a roughness histogram is constructed
as shown in Figure 14. The minimum value near
the center is chosen as the threshold value with
H = 0.5914. Now all the facets having roughness
values more than H are highlighted which form
a rough site as shown in Figure 12.
In order to illustrate the consolidation of
roughness patches, another example is considered.
First the value of clustering radius K is chosen
according to the size of model and the values
of u and v have been experimentally estimated
for consolidation and noise removal as shown
in Table 1. The values of u and v depend on the
type of material. Table 1 depicts these values for
half baked clay fragments of different sizes. Since
values for a particular material are applicable for

717
 Complementary Part Detection and Reassembly of 3D Fragments

Table 1. Values of u and v for varying clustering

The roughness is calculated at (N, N)-ring radius
radii (K)
of 2.
K V U From each of the above two earthen-ware
3 0.44 0.22
fragments, only one cleavage site is detected.
4 0.50 0.16
There may be more than one cleavage site in a
broken fragment. As shown in Figure 24 there are
5 0.60 0.10
two cleavage sites in the scanned fragment. The
detected cleavage sites are shown in Figure 25.
To detect the complementary parts, each of 50
all materials having similar texture, so these values
scanned 3D objects has been broken into two parts
need not be found each time.
and the cleavage site of each of these fragments
Figure 22 shows two original earthen-ware
is identified. Sample of broken objects is shown
fragments. The roughness values at single facet
in Figure 26. Table 2 shows some results of the
level and multiple facet level are computed. Cor-
broken objects after applying the proposed algo-
responding fragments after roughness consolida-
rithm. The algorithm could detect all the comple-
tion and facet de-noising are shown in Figure 23.
mentary parts among the objects successfully.

Figure 22. Two original fragments

Figure 23. Roughness detected fragments

718
Complementary Part Detection and Reassembly of 3D Fragments

Figure 24. Fragment with two cleavage sites

Figure 25. Corresponding cleavage sites of fragments shown in Figure 24

Registration of these detected complementary

parts is done in order to reassemble them at their
Table 2. Broken objects and their complementary
cleavage sites. In order to illustrate the results of
parts
registration, four experiments are given.
3D Broken Object Complementary Part
Experiment 1

The first object considered for our experiment is

an earthenware rectangular object having height
2.0”, width 0.5” and length 1” approximately.

Figure 26. Multiple broken fragments

719
 Complementary Part Detection and Reassembly of 3D Fragments
Figure 27. First fragment with its cleavage site
Figure 28. Second fragment with its cleavage site
The object is broken into two fragments. Figure
27 and Figure 28 show the fragments with their
corresponding detected cleavage site. Figure 29
depicts both the fragments after registration.
Experiment 2
Next object considered in our experiment is a clay
pottery which has the size of approximately 1.5
inches in height and 0.5 inches in diameter. This
Figure 29. Registration of both fragments
720
object is broken into two fragments. Figure 30
shows the first fragment with its detected cleavage
site while the other fragment with its cleavage site
is given in Figure 31. Both the fragments with their
respective detected cleavage sites are registered
and its result is shown in Figure 32.
Experiment 3
In this experiment, we have also considered a clay
pottery. It is approximately 2.5’’ in height and
1’’ diameter. This object is also broken into two
fragments. Both the fragments along with their
respective cleavage sites are shown in Figure 33
and Figure 34. Registered fragments are shown
in Figure 35.
Experiment 4
The object considered in this experiment is a piece
of chalk. The chalk is broken into two fragments.
It can be noted that this chalk piece is a good
Complementary Part Detection and Reassembly of 3D Fragments

Figure 30. First fragment with its cleavage site

Figure 31. Second fragment with its cleavage site

Figure 32. Result after registration example of multi-granularity roughness. Frag-

ments with their respective detected cleavage sites
are shown in Figure 36 and Figure 37 while the
object obtained from fragments after registration
is depicted in Figure 38.

FUTURE WORK

The work discussed in this chapter opens up some

new directions of research which are ranging
from theoretical issues to the latest cutting edge

Figure 33. First fragment with its cleavage site

721
 Complementary Part Detection and Reassembly of 3D Fragments
Figure 34. Second fragment with its cleavage site
Figure 35. Registration of fragments
technologies. The problem studied is found to be
very useful for archeologists who obtain broken
fragments especially in case of potteries at the
time of excavations. Automatic reassembly of
these fragments can be done through the algo-
rithm proposed in this chapter. The algorithm is
found to be very effective for ceramic materials
and earthen-ware. However, it is not well suited
Figure 36. First fragment with its cleavage site
722
for metallic objects and crystalline solids because
the roughness in case of metallic broken objects
is difficult to obtain.
Cleavage site detection algorithm is found
to reduce the area to be matched drastically by
identifying the portions where we should attempt
to fit complementary fragments. This algorithm
can further be enhanced by taking into account
the sharp edges found on cleavage site boundar-
ies. But it requires careful evaluation because all
fragments may not have sharp cleavage bound-
aries. Further, the algorithm for registration can
be improved by taking into consideration the
boundaries of the obtained cleavage sites and ap-
plying the 3D image correlation on it. This may
further improve the accuracy of the registration
of the broken complementary pieces.
Complementary Part Detection and Reassembly of 3D Fragments

Figure 37. Second fragment with its cleavage site

Figure 38. Object after registration

reassembly of 3D thin walled objects. However, the
problem described in this chapter can efficiently
handle thick walled objects as well. In this work,
not only few existing algorithms for roughness
detection were enhanced, but the generated data
was consolidated and cleaned to make it suit-
able for entirely new application of cleavage site
detection. As such 3D fragment reassembly is a
complex problem, involving matching and regis-
tration of complementary surfaces. Cleavage site
detection algorithm reduces the area to be matched
drastically by identifying the portions where we
CONCLUSION
should attempt to fit complementary fragments.
To validate and optimize, the proposed rough-
The chapter has explored a problem for deter-
ness detection algorithm has been tested on more
mining the complementary part of a fragment of
than 50 scanned fragments. All the fragments are
an object and of reassembling them to form the
scanned using 3D laser LPX 600 Roland scanner.
object. It has proposed an efficient surface inspec-
The results have shown that the proposed algo-
tion algorithm which detects the corresponding
rithms are well suited for detecting cleavage sites
cleavage sites of fragments and registers them
in broken objects in a much accurate and efficient
so that the object can be formed from the given
manner. To detect the complementary parts, each
fragments. The algorithm captures differences in
of 50 scanned 3D objects has been broken into
the geometric structure in different parts of the 3D
two parts and the cleavage site of each of these
object represented by polygon meshes. These dif-
fragments is identified. Further, results obtained by
ferences are used to find most probable cleavage/
applying the proposed algorithm for complemen-
fracture sites. It is applicable to all triangulated
tary part detection well suited for identification
surface data even in presence of noise. The algo-
of complementary parts from the multiple broken
rithm is found to be very effective on objects of
objects in a much accurate and efficient manner.
ceramic material and archeological artifacts. The
After the detection of complementary parts, final
work already done in the field of reconstruction
registration of these complementary parts is done
of ceramic material or archeological artifacts is
in order to reassemble them at their cleavage sites.
through reassembly is either on 2D reassembly or

723
 Complementary Part Detection and Reassembly of 3D Fragments
REFERENCES
Arrault, J., Arnéodo, A., Davis, A., & Marshak,
A. (1997). Wavelet based multifractal analysis of
rough surfaces: Application to cloud models and
satellite data. Physical Review Letters, 79(1),
75–78. doi:10.1103/PhysRevLett.79.75
Barequet, G., & Sharir, M. (1997). Partial
surface and volume matching in three dimen-
sions. IEEE Transactions on Pattern Analysis
and Machine Intelligence, 19(9), 929–948.
doi:10.1109/34.615444
Besl, P. J., & McKay, N. D. (1992). A method for
registration of 3D shapes. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 14(2),
239–256. doi:10.1109/34.121791
Boehler, W., & Marbs, A. (2002). 3D scanning
instruments. In CIPA WG6 International Workshop
on Scanning for Cultural Heritage Recording
(CIPA2002) (pp. 9-12).
Bolin, M. R., & Meyer, G. W. (1998). A perceptu-
ally based adaptive sampling algorithm. In ACM
SIGGRAPH (pp. 299–309). ACM Press.
Corsini, M., Gelasca, E. D., & Ebrahimi, T.
(2005). A multi-scale roughness metric for 3D
watermarking quality assessment. In Workshop
on Image Analysis for Multimedia Interactive
Services, (pp. 13-15).
Jafari, G. R. (2003). Stochastic analysis and regen-
eration of rough surfaces. Physical Review Letters,
91(22), 1–4. doi:10.1103/PhysRevLett.91.226101
Kaushik, V. D., Singh, A. K., Varshney, J., Pandey,
P., Rao, K. P., & Pathak, V. K. (2007). Cleavage
site detection in broken 3D objects. In Interna-
tional Conference on Advanced Computing and
Communications (ADCOM 2007) (pp. 339–344).
Levoy, M., Pulli, K., Curless, B., Rusinkiewicz,
S., Koller, D., & Pereira, L. … Fulk, D. (2000).
The digital Michelangelo project: 3D scanning of
large statues. In ACM SIGGRAPH (pp.131-144).
ACM Press.
724
Lindstrom, P., & Turk, G. (2000). Image-driven
simplification. ACM Transactions on Graphics,
19(3), 204–241. doi:10.1145/353981.353995
Masuda, T., & Yokoya, N. (1995). A robust method
for registration and segmentation of multiple range
images. Computer Vision and Image Understand-
ing, 61(3), 295–307. doi:10.1006/cviu.1995.1024
Nicolas, A., Santa-Cruz, D., & Ebrahimi, T. (2002).
Mesh: Measuring error between surfaces using
the Hausdorff distance. In IEEE International
Conference on Multimedia and Expo (ICME):
Vol. I (pp. 705–708). IEEE Press.
Rogowitz, B., & Rushmeier, H. (2001). Are image
quality metrics adequate to evaluate the quality
of geometric objects? In Rogowitz, B. E., &
Pappas, T. N. (Eds.), SPIE on Human Vision and
Electronic Imaging VI (Vol. 4299, pp. 340–348).
SPIE. doi:10.1117/12.429504
Scopigno, R., Cignoni, P., & Rocchini, C. (1998).
Metro: Measuring error on simplified surfaces.
Computer Graphics Forum, 17(2), 167–174.
doi:10.1111/1467-8659.00236
Terry, W. (1995). The challenge of 3D digitizing.
Computer Graphics World, 18(11), 21–25.
Waechter, M., Riess, F., Schimmel, T., Wendt,
U., & Peinke, J. (2004). Stochastic analysis of
different rough surfaces. The European Physi-
cal Journal B- Condensed Matter and Complex
Systems, 41(2), 259-277.
Williams, N., Luebke, D., Cohen, J. D., Kelley,
M., & Schubert, B. (2003). Perceptually guided
simplification of lit, textured meshes. In Sympo-
sium on Interactive 3D Graphics (pp. 113-121).
ACM Press.
Wu, J., Hu, S., Tai, C., & Sun, J. (2001). An ef-
fective feature-preserving mesh simplification
scheme based on face constriction. In 9th Pacific
Conference on Computer Graphics and Applica-
tions (pp. 12 - 21). IEEE Press.
Complementary Part Detection and Reassembly of 3D Fragments

KEY TERMS AND DEFINITIONS Fragment: Each broken piece of a 3D object

Cleavage Site: It consists of all rough facets
in a fragment.
Complementary Part Detection: The method
of obtaining the corresponding fragments belong-
ing to the same 3D object.
Facet: If a surface is represented by triangular
meshes, then each triangle is it is termed as facet.
is known as fragment.
Multi-Level Granularity: It is the roughness
having more than one level in a fragment.
Registration: Alignment of complimentary
parts of fragments.
Roughness Factor: A metric to measure the
roughness of a surface of a 3D object.

This work was previously published in 3-D Surface Geometry and Reconstruction: Developing Concepts and Applications,
edited by Umesh Chandra Pati, pp. 314-337, copyright 2012 by Information Science Reference (an imprint of IGI Global).

725
726

Chapter 44
Biometric Authentication
in Broadband Networks for
Location-Based Services
Stelios C. A. Thomopoulos
National Center of Scientific Research “Demokritos,” Greece

Nikolaos Argyreas
National Center of Scientific Research “Demokritos,” Greece

ABSTRACT
Broadband communication networks have begun to spread rapidly over fixed networks, with wireless
networks following at close distance. The excess capacity allows the offering of broadband services
at competitive rates. Location-based services (LBS) over wireless broadband networks are becoming
mainstream in an emerging ambient intelligence society. For LBS over broadband and, in particular,
peer-to-peer networks, such as ad hoc networks, unambiguous user authentication is of paramount im-
portance to user trust and safety, thus ultimately to the success of such service. Biometric authentication
is an approach to providing irrefutable identity verification of a user, thus providing the highest level
of security. This chapter addresses some of the issues associated with the use of biometric ID for user
and apparatus authentication over broadband wireless networks (e.g., GPRS, UMTS, WiFi, LANs) and
narrow band local networks (e.g., bluetooth, Zigbee, PANs, BANs).

INTRODUCTION enhanced GPS infrastructures, such as EGNOS,

The spreading of broadband networks stimulates
a wealth of Internet services over fixed and wire-
less networks with stationary and mobile devices.
Combining accurate location information from
DOI: 10.4018/978-1-4666-2038-4.ch044
Galileo …, with broadband wireless networks,
provide the necessary infrastructure for delivering
high quality and versatile location-based services
(LBSs) ranging from travel information to enter-
tainment, to crisis and incident management, to
services on demand, to health care and peer-to-peer
communications, to mention just a few.

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Biometric Authentication in Broadband Networks for Location-Based Services
In all these services, the common thread is the
ability to unambiguously identify and authenticate
the mobile user and customer to the LBS provider.
Different LB services may have different authenti-
cation requirements. However, no matter what the
application is all such services, the unambiguous
authentication of the user is paramount to gaining
the trust of the end user and thus achieving the
success of the services. Unambiguous user au-
thentication is paramount to the parties involved
in an LB service and the trust upon which the
service is built. If for example the LBS refers to
the provision of transport services on demand,
the ability to correctly identify and authenticate
both parties involved in the transaction, that is
the passenger (i.e., the user) and the driver (i.e.,
the service provider) build mutual trust and can
be proved life-saving in the case of a car-jacking,
criminal activity, or fraud.
User identification and authentication can be
performed by a variety of means, ranging from a
simple alphanumerical password to a more secure
digital signature, to the ultimate in security bio-
metric ID. Although a digital signature produced
by an electronic device provides the convenience
of a self-contained identification instrument, it
does not prevent fraudulent use of a user ID.
Since there is no unique and inherited connection
between the user and the digital ID, any holder
of the electronic device that produces the digital
ID can produce a fraudulent authentication. The
only means to eliminate such possibility is the
use of biometric ID.
Biometric ID is a digital signature generated
from the measurement of some bodily human
characteristics that are unique, or different enough
to be considered unique, from user to user. This
Biometric ID, encoded properly, constitutes a
unique signature for each user that cannot repli-
cated by an impostor. This biometric ID can be
used to meet the stringent requirements imposed
by LB services and the necessary trust required
by users and operators of such services alike.
Examples of biometrics commonly used for user
identification and verification include fingerprint
identification, iris scan, face and voice recogni-
tion, signature recognition, hand geometry, and
combinations thereof (Reisman & Thomopoulos,
1998; Thomopoulos & Reisman, 1993).
The use of biometric ID imposes certain re-
strictions and technological challenges that need
to be addressed before biometric authentication
becomes widely used as an enabling technology
for irrefutable user authentication in LBS and
other broadband services.
BIOMETRIC ID
A. Requirements for
Biometric ID Usage
Biometric ID is the mathematical encoding of
certain bodily features that are considered unique
for each human being and differ enough from
person to person so that this difference can be
used safely enough to tell apart one person from
another. For example, in the case of fingerprints,
biometric features are the characteristic points
that are formed from the endings or bifurcations
of the finger ridges and/or the pattern of the ridges
themselves. In the case of the iris scan, biometric
features are the radial patterns of the iris. In the
case of the face, biometric features are the relative
location of the eyes, mouth, nose, and so forth.
The mathematical encoding of the biometric
features constitutes what is referred to as biometric
“template” or biometric ID (Thomopoulos, Reis-
man, & Papelis, 1996). No matter which biometric
is used, the biometric features constitute unique
and personal human characteristic and as such,
they are protected by the Personal Information
Protection Act (PIPA) (Personal Information
Protection Act, S.A. 2003, c. P-6.5). This protec-
tion imposes a number of issues, concerns, and
restrictions with the extraction (or capturing)
of biometric features, their (electronic) stor-
age, encoding into a biometric template and the
727
 Biometric Authentication in Broadband Networks for Location-Based Services
subsequent retrieval and use of this template for
user authentication. In addition to privacy, other
concerns with the use of biometric features relate
to the medical information that may be contained
in and revealed by biometric features, the safety
of the process and devices used in the extraction
and verification process, and the protection of
the user’s privacy (ANSI:The American National
Standards Institute, 2005; also ISO: International
Organization for Standardization).
A biometric “template” constitutes a non-
reversible mathematical transformation of the
extracted biometric features (Thomopoulos, Reis-
man, & Papelis, 1996). In that respect, it cannot
be used outside the specific device that is used to
extract it (and subsequently verify it) as a means to
identify one’s identity without his or her consent,
thus violating the PICA. However, it can be used
indirectly through the use of the specific device
and process used to extract it to coerce evidence
about one’s identity without his or her consent.
In that regard, even the biometric template, albeit
its mathematical irreversibility, constitutes private
and personal data and as such it is mandatory that
it is protected by the Private Information Protec-
tion Act (PIPA, S.A. 2003, c. P-6.5).
The restrictions imposed by the PICA reflect on
the technological specifications for the processes
and devices used for biometric ID verification.
Hence, the requirement that any biometric infor-
mation, including the biometric template, consti-
tutes private data and requires that the biometric
data be stored in a memory device that is only
accessible by the owner of this data, either in the
form of raw biometric data or a processed bio-
metric template (Stapleton, 2003). Furthermore,
any retrieval of such information from the storage
device (memory) must be done over secure and
cryptographically protected communication links,
and any subsequent processing of this informa-
tion must be done by a secure processor. These
restrictions lead to almost uniquely defined archi-
tectures for biometric verification (The BioAPI™
Consortium, 2005). The most readily available and
728
seemingly universal architecture involves the use
of a smart card as a memory device for storing
the user biometric ID template and a local secure
processor for capturing the biometric information,
processing it to extract the biometric features, and
comparing it against the stored template. Alterna-
tive storage technologies include memory chips or
tokens. In this case, the biometrics data processor
is part of the memory device, in a size that is not
larger than a conventional memory stick. Using
either technology, the PICA requirements can
be met. The biometric template is always at the
possession of the user, any processing is done
locally, and over links and processors that are
securely isolated from the WWW. However, in
the process of enforcing PICA, one of the ma-
jor advantages of biometrics and biometric ID,
namely the convenience not to carry any type of
ID other than one’s self, is being lost. This is the
trade-off between the convenience offered by the
card-less biometric identification, and the restric-
tions imposed by PICA that translate in the need
to carry and use a plastic card in order to store the
biometrics template in a user-controlled memory.
The cartoon displays in Figures 1 and 2 show
the typical registration and authentication pro-
cesses for biometric ID. These processes have
been implemented in the VeriEasy biometrics
access control system that was developed in the
context of the Bioathletics1 project. The project
aims at implementing a distributed biometrics ID
system to allow secure access to spectators, VIPs
and the press in athletic installations and events
using their biometric ID only.
BIOMETRICS IN ATHLETICS
Athletic events represent an ideal field of appli-
cation and showcase test beds for biometrics as
they combine a usually large crowd, the need for
speedy but secure authentication, electronic tick-
eting, and overall security concerns by potential
troublemakers. VeriEasy™ was developed in the
Biometric Authentication in Broadband Networks for Location-Based Services

Figure 1. Registration process of the PICA-compliant smart card based biometrics (fingerprint) VeriEasy
access control system for athletic events and installations

context of the funded research program Bioathlet-
ics to provide a unified BioAPI compliant platform
for addressing the concerns of the athletic events
industry in secure physical access to stadium for
athletes, spectators, and VIPs. In the VeriEasy™
system, the biometrics (fingerprint) template is
stored directly from the card Verification and
Access Process as implemented in the VeriEasy
access control system for athletic events and in-
stallations reader in a smart card that stays with
the template’s owner at all times. This way the
PICA requirements are fulfilled in terms of both
privacy and protection against eavesdropping,
since the template stays with his or her owner at
all times and processing is done locally over a
trusted link. The user smart card serial number is
then linked to a back office legacy system data-
base that contains demographic (non-private and

Figure 2. PICA-compliant smart card based biometrics (fingerprint)

729
 Biometric Authentication in Broadband Networks for Location-Based Services
sensitive) information about the user and access
rights and athletic event attendance information.
During an athletic event, the user accesses
the stadium or the equivalent athletic installation
using his or her biometric ID smart card. The
stored user’s biometric ID is compared against
his or her fingerprint live scan locally using the
same secure local processor and communication
link. If the authentication is positive, the system
is connected via a broadband connection to the
back office athletic events database to acquire the
appropriate access permissions for the specific
event the user is attempting to gain access to. If
the user has purchased the appropriate ticket for
the event or has access rights to it, an authenti-
cation acknowledgment signal is returned to the
device to enable access to the athletic premise.
Alternative, a positive acknowledgment message
can be displayed on a monitor, if the system oper-
ates with a human attendant. The verification and
access process for VeriEasy is shown in a cartoon
way in Figure 2.
Another restriction that is imposed on bio-
metrics ID systems is the encryption of any bio-
metrics data. Encryption in the VeriEasy system
is achieved using a private key that is encrypted
in the smart card. Other encryption methods are
possible and can be used in conjunction with the
smart card encryption. Another link that must be
secured in the Verification and Access process
is the physical electric connection from the door
controller to the electric door latch.
This link can be secured by using an encryption-
decryption hardware pair to encrypt the electric
signal that commands the opening and closing of
the gate. This way the entire local biometric au-
thentication-access process is secured. It remains,
however, to secure all long haul communication
network links.
In order to make the biometrics ID registra-
tion-verification-access process entirely secure
globally, all long hauls fixed or wireless commu-
nication network links must be secured as well.
One way to achieve this is use VPN tunneling for
730
data communication from the distributed access
points to the centralized back offices database or
databases that hold the athletic event schedules,
ticket information, event schedules, athletic instal-
lations layout, access rights, and authorizations.
Finally, if the biometric authentication devices
communicate with the local Access Rights Con-
troller via a Bluetooth or zigbee wireless link,
this wireless link must be secured as well. Both
bluetooth and zigbee protocols provide a security
layer but the level of security of these layers is
still not widely tested.
Hence, if a Bluetooth or zigbee wireless link
is used in the biometrics ID system, it may be
required to use encryption on top of their security
layer until their security is extensively tested and
proven.
Figure 3 depicts the VeriEasy verification
process as a generic secure verification process
applicable to any biometrics ID authentication
system operating over any broadband communi-
cation network. Hence, the biometrics authenti-
cation system architecture of VeriEasy, and thus
Bioathletics, provides: a trusted and secure means
of biometrics ID extraction and processing at the
local level, compliant with the PICA specifica-
tions for personal data protection and privacy;
a globally secure network architecture for data
exchange and information communication; and a
locally secure wireless network for the intercon-
nection and networking of devices and signaling.
Any biometrics authentication system since
the biometrics field is continually evolving and
no standard has been reached yet.
The Bioathletics paradigm can be general-
ized to other broadband applications that require
biometrics ID authentication. A number of useful
conclusions drawn from the Bioathletics project
are summarized in Table 1. These conclusions are
relevant and applicable to the implementation of
any PICA-compliant biometrics ID authentication
system that may be required for the unambiguous
and trusted user authentication in broadband ap-
plications, with LBS being one of them.
Biometric Authentication in Broadband Networks for Location-Based Services

Figure 3. Securing all communication links, fixed and wireless, either via a VPN, bluetooth, or zibgee,
the security layer provides a globally secure and trusted, PICA-compliant smart card based biometrics
(fingerprint) authentication for any broadband application with emphasis to LBS applications

In addition to the security requirements im- late all manufacturer dependent biometric and
posed on a biometrics ID authentication system, algorithm processes in callable DLLs. In this way,
the system must also be Bio-API compliant. the core structure of the biometrics ID authentica-
Compliance with the Bio-API allows to encapsu- tion system remains fixed, whereas new biomet-

Table 1. Recommendations for the implementation of a biometrics ID authentication system

PICA
Process Device Communication Link Security level PICA requirements
compliant
Biometrics feature Biometrics reader Local and secure. Data Secure and Personal data pro-
extraction and with local processor encryption if necessary trusted tection Yes
template creation
Biometrics tem- Local with the Directly in a smart card or Private and secure Privacy and protec-
plate storage user: Smart Card or memory stick tion of personal data
Memory Stick with Yes
or w/out processor
and built-in memory
Biometrics ID Biometrics reader Local. Encryption of data Secure and Privacy and protec-
authentication with local proces- may be required from smart trusted tion of personal data
Yes
sor & Local Card or card or memory stick to local
Memory Stick reader processor.
Access of back Secure Server VPN with or without ad- Secure N/A unless it carries
office permission ditional data encryption private data info – to N/A
rights db be avoided
Local devices and Biometric readers, Wireless local links: blue- bluetooth or Secure personal data
command signal- controllers, actuators tooth or zigbee zigbee Security transfer if inevitable
ing via wireless protocol with or Yes
links w/out additional
data encryption

731
 Biometric Authentication in Broadband Networks for Location-Based Services
ric feature extraction, encoding, and verification
modules can be easily integrated in the system
with minor modifications. This modularity is
extremely important as it allows adaptation to
new biometrics as time progresses. This adapta-
tion is keen to the success of any biometrics au-
thentication system since the biometrics field is
continually evolving and standards have yet to be
defined.
BioAPI™ and BSP Implementation
The BioAPI™ Specification version 1.1 0 is an
ANSI 0 (INCITS 358-2002) standard and cur-
rently under ISO standardization process (ISO/IEC
JTC1/SC37) specifying an Application Program-
ming Interface (API), which was introduced to
facilitate the implementation of Biometric applica-
tions (The BioAPI Consortium; and The BioAPI
Consortium: BioAPI™ Specification version
1.1). BioAPI is a standard biometric technology
interface, which is intended to provide a high-level
generic biometric authentication model; one suited
for any form of biometric technology. BioAPI
supports, among other lower level functions, the
basic functions of Enrollment, Verification, and
Identification, and includes a database interface
to allow a biometric service provider (BSP) to
manage the Identification population for opti-
mum performance in an application independent
manner. There is a reference implementation
for Microsoft® Windows® (early alpha) publicly
available by the BioAPI™ Consortium 0, as well
as two internally released beta implementations;
one for Java (under Windows®, via JNI) by (Gens
Software Ltd); and one for Unix™ / Linux™ by
SAFLINK. There are several commercial imple-
mentations to date—mainly by members of the
BioAPI™ Consortium—either in the form of
BioAPI™ framework SDK or in the form of BSP
Biometric Service Provider (BSP) SDK or specific
BSP implementations for the vendors’ Biometric
Devices. However, market acceptance is still
immature and market penetration of BioAPI™
732
compliant applications and biometric devices is
speculated to significantly increase in the next two
years (The BioAPI Consortium; and The BioAPI
Consortium: BioAPI™ Specification version 1.1).
A BioAPI™ compliant application consists
of at least two different modules, the application
module, which could be seen as the “top” level
application (Level H) and the BSP (Biometric
Service Provider) module, which could be seen
as the “bottom” level application (Level L).
The application module would be the biometric
application and the BSP module would be the
service provider module, which should control
the biometric hardware. The BSP could be seen
as a black box from the side of the application
developer, since the technical details of a given
device will not concern him.
Access to the biometric mechanisms is through
a set of standard interfaces defined by BioAPI™.
The approach taken by the BioAPI™ is to hide,
to the degree possible, the unique aspects of in-
dividual biometric technologies, and particular
vendor implementations, products, and devices,
while providing a high-level abstraction that can
be used within a number of potential software
applications.
One of the most important terms of the Bio-
API™ standard is the BIR. The term Biometric
Identification Record (BIR) refers to any biometric
data that is returned to the application; including
raw data, intermediate data, and processed sample
ready for verification or identification, as well as
enrollment data.
Theoretically, BSPs that are supplied by
vendors and conform to the BioAPI™ interface
specification, can be used by any application
developed using the BioAPI™ framework.
The BioAPI™ Consortium claims, among
other things, that the BioAPI™ will enable rapid
development of application, flexible deployment
between platforms, and simple application inter-
faces modularity of biometric applications.
For an application to be compliant with the
BioAPI™ specification, a software application
Biometric Authentication in Broadband Networks for Location-Based Services
must perform that operation consistent with the
BioAPI™ specification for each BioAPI™ func-
tion call made. That is, all input parameters must
be present and valid. There is no minimum set of
functions that must be called.
For a BSP to claim compliance to the BioAPI™
specification, it must implement mandatory func-
tions for their category. They are categorized as
either a Verification or Identification BSP. BSPs
must accept all valid input parameters and return
valid outputs. Additionally, they must provide all
required module registry entries. Entries to the
module registry must be performed upon BSP
installation. Biometric data generated by the
BSP must conform to the data structures defined
in BioAPI™ Specification. A BioAPI compliant
BSP must support all Module Management and
Handle operations.
When an application is connected to a given
BSP, the process initializes the BioAPI™ Frame-
work, loads the BSP desired, and then attaches
a module of the given BSP to the application.
Once the application is connected to a BSP, it
can perform BioAPI™ calls for several biometric
operations as Capture, Enrollment, Verification,
and Identification.
The BioAPI™ concept is very interesting, but
some things in the implementation may get con-
fusing. For instance, data exchange between the
BSP and the application can be an issue. There is
no asynchronous way to pass data from the BSP
to the application. Also when it comes to com-
munication of the application and the BSP, the
rules of communication can be quite restraining.
In the given implementation of BioAPI™,
when it comes to the BSP wanting to notify the
application that some event has occurred, the API
defines that the notifications that can be made are
“on insertion” or “on removal” of a biometric
device, “on fault” of a device and “on presence”
or “on removal” of a biometric source (e.g., pres-
ence of a finger on a fingerprint device). In other
words an application can only be notified if:
a. A new device is plugged in the system.
b. An existing device is unplugged from the
system.
c. An existing device doesn’t work properly.
d. A biometric source is presented to one of
the biometric devices in the system, or is
removed from it.
The application can command the BSP with
commands such as BioAPI_Enroll(…) or BioA-
PI_Verify(…). The idea of BioAPI™ includes the
idea that any attached device should be handled
from the BSP so that different devices can be
used if they provide a suitable BSP. However,
as existing biometric devices may not be fully
compliant to BioAPI™, there might be cases
that it may be deemed necessary to command the
device to perform something that is not foreseen
by the BioAPI™ without violating the BioAPI™
standard. Otherwise, the application will no longer
be BioAPI™ compliant, since no other de vice
is expected to work with the same command and
the idea of the BSP module as a black box is lost.
Another problem exists with the data exchange
between the application and the BSP. The API gives
the application the right to use a payload as input
on Enrollment and Create Template, or as output
on Verification and Verify Match (Identification).
In any other case the API does not provide the
application the right to exchange data with the
BSP. Two questions are raised then:
a. What happens in the case that the application
needs to send data to the BSP asynchronously
and vise-versa?
b. What should the payload data be, if the data
that is exchanged through payload are not
specified by the API?
The data exchange through payload might be
anything, but since the API does not define what
this data should be, then it is almost impossible
to find two different BSPs passing the same data
through payload and on the same form. This
733
 Biometric Authentication in Broadband Networks for Location-Based Services
means that the application is highly impossible
to be able to function with the BSP of a different
device, thus a need for BSPs for devices from
different manufacturers and vendors.
Expanding the Capabilities
of VeriEasy™
In many applications, in particular in high security/
high authentication fidelity cases, more than one
biometrics technology is required to exist and
operate under different authentication scenarios
in the same environment. Such a heterogeneous
biometric devices network may consist of fin-
gerprint scanners, voice recognition devices,
and face and iris scanners, all working in various
combinations for ID verification under various
authentication scenarios. Using the VeriEasy™
client-server architecture, we can build BSPs
for each type of biometrics device we intend
to use in the security network and control them
via a single application! This way the highest
security standards are achieved without losing
any flexibility. For example, in a heterogeneous
biometrics network, the Server BSP can grab the
proper fingerprint template from the smartcard or
request to the application for the right one when a
centralized Template Database is in place.
The obvious advantage of this architecture is
that it defines a middle layer between the applica-
tion and the native BSPs, so that the native BSP
programmer needs to know about his device and
nothing more, the middle layer programmer (if
any) needs to know about the biometric network
and nothing more, and the application programmer
needs to know nothing at all about biometrics!
Furthermore we can use plug and play BSPs
(e.g., have a BSP installed, as a client one), and
let the Server BSP to reconfigure itself in order
to handle the new device.
734
MULTIFUNCTIONAL BSPs
Using more than one biometric technology may
be a highly unprofitable venture, because of the
cost of different kinds of devices coming from
different hardware distributors. But even more, the
maintenance of such a system could be proved a
constant headache for our administration person-
nel (different contracts of maintenance). It is here
where VeriEasy™ comes to provide the winning
solution--the multifunctional BSP. That means
that the same device can operate in many different
ways so it is up to us to choose the proper way for
the specific device, whenever we want to.
Hence, setting up the most complex biometric
devices network to meet the security needs of an
organization is only one click away! Moreover, if
for security concerns changes are required in the
security network design at run time (!), VeriEasy™
allows on the fly reconfiguration of the security
network design without service interruption!
Biometrics Authentication
in Broadband Networks and
Location-Based Services
Advances in broadband, and in particular wire-
less broadband networks, enable the provision
of personalized multimedia services, as well as
location-based services (LBS’s). Those services
extend from electronic maps and navigation, to
location sensitive tourist guidance, transportation,
car-pooling, ad hoc networking, and ubiquitous
computing, to incident management, emergency
services, to make reference to just a few. In all
LB services, the sensitive information is the user’s
location and the user’s identity. Both these two
pieces of information must be protected by the
provided service or infrastructure the service is
running on, and authenticated when necessary
by means directly controlled by the user. Simply
imagine how critical it is to authenticate the user’s
ID unambiguously when the LBS is a car pooling
service provided over an ad hoc (peer-to-peer)
Biometric Authentication in Broadband Networks for Location-Based Services
network. In such an application where verification
of the true identity of the user may be a life and
death situation, it becomes imperative to have a
bullet-proof biometrics authentication system to
provide unambiguous user ID verification.
Such a system should protect the user ID from
theft, forgery, and misuse.
CONCLUSION
The architecture and design of a distributed bio-
metric ID verification system has been presented.
The VeriEasy™ system provides protection
of the user privacy and ID, compliant with the
sensitive personal data protection guidelines and
legislations, and is BioAPI compliant using BSP
compliant interfaces.
ACKNOWLEDGMENT
The research for the development of a BioAPI™/
BSP compliant biometrics authentication system
for ID verification in athletics events have been
sponsored by the General Secretariat of Research
& Technology, of the Greek Ministry of Devel-
opment, under Contract No. ΣΠ/ΑΘ/17+32, Op-
erational Program “Competitiveness” Oct. 2003.
REFERENCES
ANSI. American National Standards Institute.
Retrieved from http://www.ansi.org
Gens Software Ltd. (n.d.). Retrieved from http://
www.gensoft.com
ISO. International Organization for Standardiza-
tion. Retrieved from http://www.iso.ch
This work was previously published in Secure E-Government Web Services, edited by A. Mitrakas, P. Hengeveld, D. Polemi,
and J. Gamper, pp. 84-96, copyright 2007 by IGI Publishing (an imprint of IGI Global).
Personal Information Protection Act, S.A. 2003, c.
P-6.5. (n.d.). Retrieved from http://www.psp.gov.
ab.ca/index.cfm?page=legislation/act/index.html
Reisman, J. G., & Thomopoulos, S. C. A. (1998).
Data fusion architecture for automated finger-
print identification for very large databases. In
Proceedings SPIE (Vol. SPIE-3374). SAFLINK
Corporation. (n.d.). Retrieved from http://www.
saflink.com
Stapleton, J. (2003, June 23-26). KPMG, state of
biometric standards. Presentation at the Biomet-
ricTech Conference, New York.
The BioAPI™ Consortium. (2005). Retrieved
from http://www.bioapi.org
The BioAPI™ Consortium: BioAPI™ Specifica-
tion version 1.1. (n.d.). Retrieved from http://
www.bioapi.org
Thomopoulos, S. C. A., & Reisman, J. G. (1993).
Fusion-based, high volume Automatic Fingerprint
Identification System (AFIS). In Proceedings of
SPIE 93, Innsbruck, Austria (Vol. SPIE-2093).
Thomopoulos, S. C. A., Reisman, J. G., & Papelis,
Y. E. (1996). Ver-i-Fus: An integrated access con-
trol and information monitoring and management
system. Proceedings of the Society for Photo-
Instrumentation Engineers, SPIE-2934, 191–200.
ENDNOTES
1
The project “Bioathletics” has been funded
by the General Secretariat of Research &
Technology (GSRT) of the Greek Ministry of
Development under the Contract AΘ 17+32
/ 2-10-2003.
735
736

Chapter 45
Multi-Criteria Decision
Analysis for Identifying
a Suitable Location for
Groundwater Pumping Wells
D. Mimoun
Ecole Nationale Supérieure des Mines, France

S. Gaur
Ecole Nationale Supérieure des Mines, France

D. Graillot
Ecole Nationale Supérieure des Mines, France

ABSTRACT
The paper presents the methodology for the combined use of GIS-based multi-criteria analysis and
simulation-optimisation modelling for management of the groundwater resources of the Dore river basin
in France. The study identifies the suitable location and maximum discharge for the new groundwater
pumping wells. The multi-criteria analysis (MCA), with the help of GIS-based geospatial analysis,
was performed to identify those areas suitable for pumping wells by considering different criteria,
such as hydraulic conductivity, land use, river aquifer exchange, depth to water, and geomorphology.
Different criteria were selected with the help of regional experts and stakeholders. For the study area,
the groundwater flow model was developed. Further, new pumping wells in the suitable zones, those
identified by MCA, were considered and a simulation-optimisation technique was used to identify the
maximum discharge from those wells. Finally, the results obtained from both the methods were to finalise
the potential zone. The developed methodology proves to be a more realistic approach to identifying
new locations for pumping wells.

DOI: 10.4018/978-1-4666-2038-4.ch045

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
1. INTRODUCTION
The demand for water at the industrial, agricultural
and domestic levels is continuously increasing and,
correspondingly, limited fresh water resources are
shrinking. The European Union has established a
community framework for water protection and
management. In this framework, member states
must analyse the characteristics of each river basin
district, the impact of human activity on water
resources and the economic analysis of water
use. The management plans of recommendations
should be executed in order to protect and enhance
groundwater resources, protect them from pollu-
tion and ensure a balance between installing new
pumping wells and designing protected areas. The
above-mentioned objectives must be achieved no
later than 2015.
Selection of a suitable location for the devel-
opment of new pumping wells is a challenging
task when taking of pollution and potentiality of
the area into consideration. With the advance-
ment of GIS technology, a multi-criteria analysis
method is becoming a powerful tool for decision
makers to analyse and solve the problems which
are based on multiple criteria. MCA methods are
used to assist decision makers in either ranking a
known set of alternatives for a problem or mak-
ing a choice among this set while considering
the conflicting criteria (Sumathi, 2007; Keeney
& Raiffa, 1976; Zeleny, 1982; Figueira et al.,
2005).The integration of MCA techniques with
GIS has significantly advanced the map overlay
analysis for site suitability analysis (Carver, 1991;
Banai, 1993 R. Banai, Fuzziness in geographic
information systems: contributions from the ana-
lytic hierarchy process, International Journal of
Geographical Information Systems 7 (1993), pp.
315–329. Full Text via CrossRef | View Record
in Scopus | Cited By in Scopus (41)Eastman,
1997; Malczewski, 1999). An integrated approach
incorporating the application of GIS and MCA
methods have been employed for the suitability
analysis of landfills in an urban matrix by Sumathi
(2007), Lin and Kao (1998), Allen et al. (2002),
and Kontos et al. (2005). In a preliminary, GIS was
normally involved for employing a set of criteria
in order to catalogue an area into defined classes
by creating buffer zones around the geographic
features to be protected. All map layers were then
intersected so that the resulting composite map
contains two suitable and unsuitable areas. As per
the literature review, the authors did not find the
study on the combined use of the MCA approach
with the genetic algorithm model in conjunction
with groundwater model. Moreover, the present
work augments the simulation-optimisation ap-
proach with the use of MCA method coupled with
geographical information system.
In this paper, first the detail of the study area
is presented. Then, the development of methodol-
ogy is explained with a detailed explanation of
multi-criteria analysis, groundwater simulation-
optimisation modelling, objective function and
constraints. Finally, the results and conclusion
of the study are presented.
2. METHODOLOGY
2.1. Location of Groundwater
Extraction Wells
Figure 1 shows the flow chart for the proposed
methodology for the evaluation of a suitable
location for groundwater pumping. In the first
step, the Intrinsic Potentiality Index (IPI) was
calculated using five parameters that control the
groundwater potentiality, i.e., depth to water,
hydraulic conductivity, aquifer river exchange,
geomorphology and land use. The name of the
criteria and their weight and rating are summarised
in Table 1. Two exclusion criteria were also con-
sidered to keep the suitable location away from
a potential polluted area. These criteria include
the distance from the treatment site and the vi-
cinity to an industrial centre. Various buffers to
these areas were developed and used as a mask
737
 Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells

Figure 1. Methodological scheme developed for identify the suitable location for groundwater pumping
wells

layer. All described parameters were developed
in the raster format with a 20 m grid resolution
and map algebra was performed using ArcMap
GIS (ESRI, 99).
Finally, simulation-optimisation modelling
was performed to identify the maximum discharge
from the given number of wells at the locations
identified in the first step. The groundwater flow
model was run by considering the 2 and 3 pump-
ing wells in the zones and the maximum dis-
charges from those wells were computed. The
MGO module of Visual MODFLOW 4.2 were
used to perform the optimisation. The most suit-
able location was identified with the help of results

Table 1. Patterns and grading values of no-exclusionary criteria used to evaluate relative suitability of
areas selected by GIS analysis

Criterion Pattern of Rat- Grading Value

ing Curve
Hydraulic Geometric 0% = 0.06*10-3 m/s; 10% = 0.1*10-3 m/s; 20%=0.5*10-3 m/s; 50%=10-3 m/s; 70%=3*10-3
Conductivity Geometric m/s; 90%=6*10-3 m/s; 100%=8*10-3 m/s
Geomorphology Geometric 0%= outside the zone of recent lateral migration of a meander; 40%=inside the zone of
River Aquifer ex- -- historic lateral migration of a meander; 60%=in the abandoned meander; 100%= inside the
change Linear abandoned meander and near the growing meander
Land Use 0%=low channel dynamics in bed incised river; 20%=low channel dynamics in bed and
Depth to Water riprap bank of the river; 50%=low channel dynamics in bed river; 70%=low channel dy-
namics in bed river and zone of erosion; 80%=moderate channel dynamics; 100%=moder-
ate channel dynamics and zone of erosion
100%=Forest and semi-natural zones; 50%=Aquatic area (swamps) or Pastures and agro-
forest zones; 30%=Discontinuous urban zones or Permanent cultures; 25%=Artificial
covered zones, green zones, continuous urban zones; 20%=gravel pit; 10%=Irrigated and
non irrigated annual culture ; 0%=Industrial discharge, landfill
100%= >9m; 90%=7.5-9m; 80%=6-7.5m; 70%=4.5-6m; 40%=3-4.5m; 10%=1.5-3m;
0%=<1.5m

738
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
achieved by multi-criteria analysis and ground-
water modelling.
2.2. Simulation-Optimisation in
Groundwater Management
Groundwater management problems are typically
solved by researchers through the simulation-
optimisation approach. During the last two de-
cades, several studies have been carried out to
address groundwater management problems using
simulation-optimisation approaches (Lefkoff &
Gorelick, 1986; Ahlfeld et al., 1986; Zheng &
Wang, 2002; Shamir et al., 1984; Willis & Finney,
1988; Finney & Samsuhadi, 1992; Emch & Yeh,
1998). Extensive literature reviews on this topic
have been done by Gorelick (1983), Das and
Datta (2001), Willis and Yeh (1987), and Qin et
al. (2009). In recent years, applications of global
search methods have been growing and these
methods have been applied to solve different kinds
of groundwater quantity and quality management
problems. The global search methods are efficient
at finding a global solution to the problem. The
constraints of the global optimisation method
can be incorporated into the formulation and do
not require derivatives with respect to decision
variables as in non-linear programming (McKen-
ney & Lin, 1994). A genetic algorithm has been
applied by Wang and Zheng (1998), Goldberg
(1989), Wang (1991), Rogers and Dowla (1992),
and McKinney and Lin (1993, 1994). Ritzel et al.
(1994), Cieniawski et al. (1995), and Johnson and
Rogers (1995) used global search methods in their
studies and found that these global search methods
are more efficient and robust in comparison to
other methods like linear, non-linear and mixed
integer. McKenney and Lin (1994) developed
genetic algorithms (GA) based on groundwater
simulation models to solve three groundwater
management problems: (1) maximum pumping
from an aquifer, (2) minimum cost water supply
development and (3) minimum cost aquifer reme-
diation. They found that the GA could effectively
and efficiently be used to obtain globally (or, at
least near globally) optimal solutions to these
groundwater management problems. Mantoglou
and Papantoniou (2008) developed a method
for the optimal design of pumping networks in
coastal aquifers using genetic algorithms. Two
models were developed: first, for the simultane-
ous optimisation of wells’ location and pumping
rates using GA; the second model followed two
stages where the first stage was concerned with
the optimisation of wells’ locations and the sec-
ond stage was concerned with the optimisation
of pumping rates. The last method used GA for
searching the well positions and at each generation
of GA, a simpler optimisation with respect to the
pumping rates was performed using non-linear
programming. Finally, the model was applied
effectively to determine the optimal design of a
pumping network and assess pumping rates in a
coastal aquifer.
3. STUDY AREA
The model has been applied to fulfill the water
demand of the city of Thiers. This is one of the
major cities in the Loire region of France. The total
population of Thiers is 20,000 and the population
in the vicinity of the region is 23,500. Due to an
increasing population and a decreasing quality of
surface water, groundwater is the best alternative
to satisfy the area’s demand for water. Therefore,
water authority officials (municipality, water agen-
cies) are trying to identify potential locations for
new pumping wells with the required discharge.
The Dore river catchment, which is situated in
the eastern part of the Massif-Central in France
(Figure 2), has been considered for establishing
the pumping wells.
The catchment has a moderate climate with
four seasons. The low flow period in the river
occurs in summer but can be extended up to No-
vember. The average annual rainfall recorded is
780 mm. The Dore River is an important tributary
of the Allier River. The average annual flow of
739
 Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells

Figure 2. Location of Dore basin in Massif Central (France)

the river Dore is 20.2 m3 s-1. In the high flow
period, the average flow is 50 m3 s-1 and de-
creases to 5 m3 s-1 in the dry period. The study
area exhibits solid geology where bedrocks are
exposed on the surface in the form of structural
hills. Fluvial quaternary sediments underlain by
marls and clay cover the major part of the area.
The quaternary alluvium is composed of gravels,
sand and pebbles with silt. The impervious sub-
stratum is composed of clay and sand (Oligocene
period). Hydraulic conductivity varies from 1×10-3
to 8×10-3 m s-1 in different parts of the aquifer.
The thickness of the aquifer varies from 12 to 15
m. Elevation of the impervious bottom layer of
aquifer varies from 240 to 258 m, which is mea-
sured from mean sea level (MSL). The location
of different hydrological features and other re-
quired data was extracted from the geological
maps 1/50 000 provided by BRGM (National
Service for Geological Survey). A total of 12
piezometric measurements are available in the
study area, which are used to calibrate the ground-
water flow model. The water level in the river
was observed at 11 different locations and was
used to develop the model.
4. MULTI-CRITERIA ANALYSIS
The suitable location map for the groundwater
pumping wells was produced using a modified
version of the SINTACS model, applied in a GIS

740
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
environment. It is an adaptation of the DRASTIC
method for the hydrogeological, climatic and
impact settings typical for the Italian territory
and the Mediterranean basin. This methodology
includes five non-exclusion criteria i.e., hydraulic
conductivity of the aquifer, geomorphology, river
aquifer exchange, land use and depth of water table.
Recharge by precipitation and topography were
suppressed since they were considered uniform in
the study area. The weighting scheme presented
here is based on local expertise and a preliminary
sensitivity analysis was performed.
The Intrinsic Potentiality concept is used,
which is related to two important controlling
features that determine the accessibility of ground-
water to accommodate the pumping wells, i.e.,
aquifer properties evaluated at the thickness of
the saturated layer and protection against pollut-
ants. The first attribute is related to the hydraulic
conductivity, the river-aquifer exchange and the
geomorphology. In assessing groundwater poten-
tiality based on an index approach, the physical
and conceptual mechanisms take place along the
attempt made to identify the groundwater poten-
tial zones. Therefore, the dependent parameters
for this method (Table 1) were carefully chosen.
For example, hydraulic conductivity is the
most important hydrogeological feature, since it
provides a great chance to extract groundwater
through pumping wells. A high value of hydrau-
lic conductivity helps extract a higher amount
of discharge through pumping wells, whereas a
lower hydraulic conductivity value reduces the
amount of discharge. Available pumping test data
at several locations in the study area were accused
and interpolated to calculate the hydraulic con-
ductivity value on each cell of the grid. Hydraulic
conductivity data was obtained from five aquifer
pumping test databases. Geomorphology takes into
account the evolution of the river. Former chan-
nels make aquifers; therefore, it represents areas
of high aquifer potential. River aquifer exchange
is also an important hydrologic feature. The Dore
River plays an important role in the groundwater
flow: drainage of water during low flow periods
and recharge during high flow periods. Therefore,
it is necessary to take into account the water ex-
change between the existing water table and the
river in the evaluation of a suitable location for
groundwater pumping wells. These exchanges
are governed by the nature and condition of the
banks of streams and by fluvial dynamics. The
phenomenon of river bed incision limits the water
exchange.
The Land use map and the depth to water
table are a sensitive parameter that influences the
vulnerability of groundwater, since it provides a
greater chance for natural attenuation to occur. As
the study area has a dense agriculture activity, a
shallow groundwater depth can increase the risk
of contamination propagation to the water table.
The Land use map was developed on the basis
of CORINE Land Cover. The classification con-
siders the nature and the property of top surface.
Other factors such as aspect, elevation and soil
type were not considered. A total of 7 land use
classes were considered, according to the nature
of their top surface. The land use factor, ranging
from 0 to 100, is assigned to each class of land.
Therefore, the dimension values vary from 0 to
100 to represent the most favourable to the least
favourable class. The weights were assigned to
the classes by the SINTACS method. SINTACS
is an intrinsic vertical vulnerability method
developed by Civita (2001). Data related to the
depth of the water table was obtained from data
on 180 borehole logs. The use of the geostatistical
method, implemented in ARC Info, was neces-
sary to characterise the spatial structure of these
regionalised variables.
Exclusion criteria were defined by creating
the buffer zone around the potential contaminant
location. Further, these buffer zones are used to
prepare the mask layer. Buffer zone values for
different area are as follows: domestic waste treat-
ment site = 300 m buffer zone; industrial centre =
500 m buffer zone and hazardous waste landfill
= 500 m buffer zone.
741
 Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
4.1. Aggregation Analysis
The Analytic Hierarchy Process (AHP) method is
most suitable for the establishment of a platform
organisation and covers the topics necessary to
identify sites suitable for the location of wells.
This method allows for data to be gathered and
organised in a single model considering the system
as a whole, thereby establishing levels of priority
between the criteria. The multi-criteria decision
analysis methods ELECTRE (ELimination and
Choice Expressing REality), also widely used
for solving problems with respect to decision
support environment (ROY 85), does not apply
to this problem. This does not choose the best site
from a given set of sites, but rather identifies all
potential sites.
Appropriate weights have been calculated by
AHP method (Satty, 1980) where pair-wise com-
parisons are made between the criteria, acSaaty,
T.L., 1980. The Analytic Hierarchy Process,
McGraw-Hill, New York. Saaty, T.L., 1980. The
Analytic Hierarchy Process, McGraw-Hill, New
York. according to preference and importance
of the considered criteria. For every new factor,
weights were assigned according to the importance
of each factor. A pair-wise comparison was ap-
plied in order to determine the relative importance
of each new criterion. To make the processes of
assigning factor weights more realistic, criteria
considered one at a time were carried out by differ-
ent experts working on the project. The technique
of pair-wise comparison was developed by Saaty
(1977) in the context of decision-making process
known as the AHP. AHP is characterised by two
fundamental aspects: hierarchical decomposition
of the problem of comparison of complex objects
and estimation of utility functions values from
paired comparisons. In this case, the hierarchy
consisted of two levels (objects and criteria).
A pair-wise comparison was done in order to
determine the relative importance of each cri-
terion. Weights were evaluated with the help of
stakeholder and experts working on the project.
742
In the procedure of multi-criteria evaluation, it
was necessary for the assigned factor weights to
add up to 1. In the study, weight was defined as a
value assigned to an evaluation, which indicates
the importance of one criterion related to another.
In order to perform a pair-wise comparison, the
relative importance of the row variable to its
corresponding column variable was considered
on the basis of following values:
9: the criterion is extremely more important; 7: the
criterion is very strongly more important; 5: the
criterion is strongly important; 3: the criterion is
moderately more important; 1: equal importance
of two criteria; 1/3: the criterion is moderately
less important; 1/5: the criterion is strongly less
important; 1/7: the criterion is very strongly less
important; 1/9: the criterion is extremely less
important.
The result of this pair-wise comparison for 5
non-exclusion criteria is shown in Table 2. The
calculated coherence ratio was 9.14%, which was
within the permissible limit, i.e., 10%. The rat-
ing curve was developed for each criteria, using
a series of arbitrary relationships between the
quantity of the criterion and a suitability score
(0 -100). The mathematical formulation for the
assignment of the five suitability scores required
integration of the weights and scores. This ag-
gregation consisted of multiplying the weight and
score obtained from the rating curves and then
adding up the product to calculate the suitability
score SI so that:
5
SI = ∑Wi X i (1)
i =1
where Wi is the weight of the considered criteria
calculated by pair-wise comparison and Xi is the
normalised score obtained from rating curves. SI
designates the suitability score of the suitable
location for groundwater discharge. Table 2 shows
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells

Table 2. Relative importance of non-exclusionary

the relative importance of non-exclusion criteria
criteria based on the 5-point rating system. These
based on the 5-point rating system. Figure 3 il-
value obtained based on a pair-wise comparison
lustrates the location of suitable zones for ground-
of all criteria using the weighted linear combina-
water pumping after considering the different
tion described by Saaty
parameters explained above. Potentialities of
Criterion Weight suitable zones are defined in four ranges, i.e., low,
Hydraulic conductivity of the 0.463
moderate, high and very high.
aquifer 0.256
Geomorphology 0.195
River aquifer exchange 0.043
Land use 0.041 5. GROUNDWATER MODELLING
Depth to water

A regional groundwater flow model of the Dore

basin, constructed and calibrated to steady-state

Figure 3. Suitable location map for groundwater pumping wells

743
 Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
time-averaged conditions for the period 1976,
developed in a previous study (CETE, 1976), was
used in this study. Groundwater modelling was
performed using the “modular three dimensional
finite-difference groundwater flow model” MOD-
FLOW developed by The United States Geological
Survey (McDonald & Harbaugh, 1988). The active
cells of the model grid cover 60 km² area of the
Dore basin. The y-axis of the model is oriented to
the north to align with the predominant direction
of groundwater flow. The finite-difference model
of the basin was constructed by dividing the al-
luvial plain into tree models. Each model was
subdivided by a rectilinear grid, which consists
of 2,400 active cells that have a uniform grid-cell
spacing of 50 m (a cell area of 2,500 m²) and a
variable thickness. Grid size was taken as uniform
as available data are evenly distributed over the
model domain and each part of the domain was
considered as equal interest.
5.1. Hydrological Units
The overviews of the geology and hydrogeology
of the Dore basin, defined by Mimoun (2009),
was used in this model. The groundwater flow
system of the basin, the thickness of the aquifer,
including its boundary, hydraulic characteristics
and components of recharge and discharge were
also described in Mimoun (2009). The Dore al-
luvial plain is filled with quaternary alluvium and
surrounded by tertiary sedimentary rocks. The fill
consists of sandy gravel, sand, silt and clay. The
alluvium forms a single unconfined aquifer where
groundwater flows from south to north.
The hydraulic characteristics of sediments
that form the aquifers control the direction and
the velocity of groundwater movement within the
system. The horizontal hydraulic conductivity has
been estimated from the pumping well test data
and the published data. Thus the obtained data
have been used as initial distributions that were
subsequently modified during calibration of the
numerical model to achieve the best fit between
744
the simulated and the observed data. The median
values of hydraulic conductivity range from 0.003
m/s for the older sediment to 0.008 m/s for the
actual sediment. Many assumptions are neces-
sary to simplify a real hydrogeological system
represented by a mathematical model. A relatively
simple modelling concept is developed on the
basis of two types of external model boundaries:
physical and artificial. The contact between the
alluvial aquifer and the Dore River is simulated
by a head-dependent flux boundary. The rate of
flow is calculated from the difference of hydraulic
head between the boundary cell and the adjacent
aquifer cells (McDonald & Harbaugh, 1988). The
lateral boundary at the eastern and western parts
of the model between the alluvial and the first
terrace was represented by equipotential lines
with known hydraulic head. These boundaries
were chosen far enough from the area of interest
to avoid the influence on the future flow pattern
created by a proposed suitable location of ground-
water pumping wells.
Some assumptions were made to design a
groundwater model that can limit the scope of
application of the model and the hydrological
questions that can reasonably be addressed. The
model uses a steady-state simulation for the dry
seasons, including climate, land use and water
use. As the model was not calibrated for transient
conditions, the model cannot be used to predict
the transient response of the system. Also, the
boundary conditions involve considerable sim-
plification of the hydrological system and can
generate substantial effects on model results. The
lateral boundary of the Dore river basin model has
been specified as a “constant head boundary” so
that the cells represent an inexhaustible source
of water regardless of changes at the boundary
and in the aquifer. For example, a cell boundary
with known fixed head may dry out because of
the nearby pumping well, and the constant head
cells simulating it will continue to supply water to
the aquifer. For many purposes, these assumptions
do not impose serious limitations on the use of
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
the model. However, interpretation of the results
should be done carefully, as changes in any of these
conditions can influence the suitable location of
the groundwater pumping wells and the discharge.
5.2. Model Run
The model was calibrated by changing the initial
values of recharge and hydraulic conductivity.
The recharge values of different zones were
increased from 15% to 20% and the influence
was observed on the groundwater head. Values
of hydraulic conductivity were also modified for
the domain. The values of groundwater head,
calculated by the model, and observed head values
were analysed at a 95% confidence level. The
groundwater head in the middle part of the basin
was found very close to the observed values of
head. The observation points close to the bound-
ary of the model domain were not found within
the range of the 90% confidence level due to the
complex geological formation, presence of hills
and inselbergs in that area. Therefore, there is a
scope to refine the values of these parameters by
taking more field samples.
5.3. Optimisation Model
and Objective Function
Mathematically, a typical groundwater manage-
ment problem can be defined using three main
components: decision variables, objective function
and constraints (Ahlefeld & Mulligan, 2000). Opti-
misation models are used to maximise or minimise
an objective function without violating the given
set of constraints. The objective function is typi-
cally defined by one or more decision variables
and the optimisation models determine the values
of these decision variables. Most of the traditional
optimization techniques are gradient-based and
the solutions from those methods produce local
optimal values rather than global optimal solution.
Therefore, application of gradient based optimi-
zation techniques may be difficult in realistic
problems related to wells optimization. It is not
always guaranteed that gradient-based methods
will find out the global optimal solution in the
case of highly non-linear optimization problems
(Huang & Mayer, 1997). On the other side, global
search methods are efficient to find out the global
solution of the problem. Constraints in global
optimization method can be incorporated into
the formulation and do not require derivatives
with respect to decision variables as in nonlinear
programming (McKenney & Lin, 1994). In this
study, the genetic algorithm (GA) has been used
as an optimisation model. First pioneered by John
Holland in the 1960s, genetic algorithms have
been widely studied, experimented with and ap-
plied in different fields of engineering. The GA
is a search technique used in computing to find
exact or approximate solutions to optimisation
and search problems. Genetic algorithms are a
particular class of evolutionary algorithms (EA)
and are categorised as global search heuristics. The
GA techniques are inspired by such evolutionary
biology elements as inheritance, mutation, selec-
tion, and crossover.
The basic working steps of GA are as follows:
1. Randomly generate an initial population
IP(0)
2. Compute and save the fitness u(m) for each
individual m in the current population IP(t)
3. Define selection probabilities p(m) for each
individual m in IP(t) so that p(m) is propor-
tional to u(m)
4. Generate IP(t+1) by probabilistically se-
lecting individuals from IP(t) to produce
offspring via genetic operators i.e., crossover
and mutation.
5. Repeat step 2 until satisfying solution is
obtained.
To evaluate the value of fitness function, pen-
alty function approach has been adopted. In this
process, specific penalty is used after the violation
of constraints. This penalty varies linearly accord-
745
 Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
ing to the magnitude of constraint violation. The
next generation population of chromosomes is
selected from the current generation with selec-
tion probability based on the evaluation score of
each chromosome.
In this study, the objective function is to
determine the maximum discharge, which can
be withdrawn from the aquifer through a given
number of wells. Therefore,
 Nw

Maximize ∑ Qi − β1P (h ) (2)
 i =1 
 
Subject to
Qi,min<Qi<Qi, max (2a)
hi>hi,min (2b)
(xi,yi)≠Ai (2c)
(x i − x j )2 + (yi − i j )2 ≥ Sw ,min (2d)
hi,min − hi if hi < hi, min
P(h)=  (2e)
0 if hi ≥ hi,min

where Qi,min and Qi,max are the minimum and maxi-
mum discharge limit for ith well, hi,min = minimum
allowable head of groundwater at ith well, Sw,min
= minimum distance between any pair of wells;
xj and yj are co-ordinates of remaining well, i.e.,
i ≠ j; P(h) = penalty term, which varies linearly
with the magnitude of constraint violation, and
β1 = weighting factor, which can be selected ac-
cording to the problem.
Constraints of the problems were finalised
with the help of water authority’s officials and
stakeholders. The penalty function approach was
used to incorporate the constraints in the problem
solution. The following three constraints were
considered in the optimisation problem.
746
The first constraint was taken to limit the
drawdown of groundwater under the permissible
limit. After the discussion with technical people
from (the) water agencies, minimum 2 m water
column from the bottom impervious layer is
proposed, thus
hi > 258 m (3)
where the average level of the aquifer bottom is
assumed at 256 m and hi is the minimum water
head on the periphery of the ith well and i = 1,
2…., Nw.
The second constraint is defined to ensure the
minimum water discharge from all wells to fulfil
the minimum water demand of the area, and hence
Qtot > 10 000 m3 day (4)
The third constraint prescribes the maximum
and minimum discharge limit of a single well.
On the basis of aquifer properties and the avail-
ability of pumps, the discharge limit was selected
as follows
2000 m3 day < Qi < 4000 m3 day (5)
6. RESULTS AND DISCUSSION
The suitable map values vary from 0 to 100 in
the case study and it falls into four classes cor-
responding to four degrees of vulnerability. The
generic suitable map which has been established
shows four potentiality classes: low, moderate,
high and very high. The weakly potential zones
occupy 39% of the total area and the moderately
potential zones occupy 25%. In the MCA, two dif-
ferent locations in the study area have been found
suitable for establishing the new pumping wells.
After calibration, the model was used to identify
the maximum discharge from the wells at the suit-
able location derived by the multi-criteria analysis.
The set of 3 and 4 wells was used to identify the
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
maximum discharge at this location. Results show
that maximum groundwater can be withdrawn
from the set of 3 wells with a discharge of 4522
m3/sec and from set of 4 wells with a discharge
of 5941 m3/sec. Groundwater modelling results
depict that this location is able to give a greater
amount of groundwater and is less vulnerable to
contamination. It also shows that extracting more
groundwater from this zone lowers the stress
on the aquifer, thereby resulting in an increased
extraction of groundwater in the future.
7. CONCLUSION
The methodology developed aids in the evalua-
tion and management of groundwater resources.
The combined use of GIS-based mapping and
groundwater modelling acts as a powerful tool
for understanding groundwater behaviour and
identifying the best management practices at
watershed level, which can be used judiciously
for agricultural and drinking water purposes. The
multi-criteria analysis is also relevant to the evalu-
ation of relative suitability of the areas selected
by GIS analysis for new pumping wells.
This study also suggests that zones falling
in the potential zones can give a good amount
of groundwater for fulfilling the water demand
of the area. The overall stress on the watershed
can be reduced by using these wells. Hence, the
developed methodology can be applied in other
regions for the evaluation and management of
groundwater resources.
The methodology for identifying the ground-
water potential zones is based on an overlapping
of criteria and not on an analysis of pollutant mo-
tion. These methods ignore the dynamics of the
groundwater flow system, as the results depend
on the weights assigned to each criterion and do
not take the flow into account. The identification
of recharge is also one of the most important
elements of a vulnerability assessment. The
groundwater flow model in conjunction with the
particle-tracking programme will be used to map
recharge areas of the groundwater flow system.
Future works could focus on complementary
sensitivity analysis. In this study criterion weights
are based on the preferences given by experts with
the help of stakeholders. As they can change in time
and space, the methodology could include these
variations. Fuzzyfication of the criteria valuation
could also bring more subjectivity to interpret the
results with buffer around boundaries of the zones
which are recommended to locate pumping wells.
ACKNOWLEDGMENT
The authors thank anonymous reviewers and the
editor for their critical comments which substan-
tially helped to improve the original manuscript
for publication.
REFERENCES
Ahlfeld, D. P., Mulvey, J. M., & Pinder, G. F.
(1986). Designing optimal strategies for contami-
nated groundwater remediation. Adventures in
Water Resources, 9(2), 72–84.
Allen, A., Brito, G., Caetano, P., Costa, C., Cum-
mins, V., Donelly, J., et al. (2002). Procedure for
the location of landfill sites using a GIS model.
In Proceedings of the 9th Congress of the Inter-
national Association of Engineering Geology and
the Environment, Durban, South Africa (p. 100).
Carver, S. J. (1991). Integrating multi-criteria
evaluation with geographical information
systems. International Journal of Geo-
graphical Information Systems, 5(3), 321–339.
doi:10.1080/02693799108927858
Cieniawski, S. E., Wayland, J., & Ranjithan, S.
(1995). Using genetic algorithms to solve mul-
tiobjective groundwater monitoring problem.
Water Resources Research, 31(2), 399–409.
doi:10.1029/94WR02039
747
 Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
Civita, M. (2001). Groundwater vulnerability
mapping in the Lauria Mountains (Southern Italy)
using the SINTACS method. In Sciences et tech-
niques de l’environnement. Bensançon, France:
University of Franche Comté.
Das, A., & Datta, B. (2001). Application of opti-
mization techniques in groundwater quantity and
quality management. Sadhana: Academy Pro-
ceedings in Engineering Sciences, 26(4), 293–316.
Emch, P. G., & Yeh, W. G. (1998). Management
model for conjunctive use of coastal and surface
water and groundwater. Journal of Water Re-
sources Planning and Management, 124, 129–139.
doi:10.1061/(ASCE)0733-9496(1998)124:3(129)
Figueira, J., Greco, S., & Ehrogott, M. (2005).
Multiple criteria decision analysis: State of the
art surveys international series in operations
research and management science. New York,
NY: Springer.
Finney, B. A., & Samsuhadi, W. R. (1992).
Quasi-three-dimensional optimisation model for
J-kata basin. Journal of Water Resources Plan-
ning and Management, 118, 18–31. doi:10.1061/
(ASCE)0733-9496(1992)118:1(18)
Goldberg, D. E. (1989). Genetic algorithms in
search, optimization and machine learning. Read-
ing, MA: Addison-Wesley.
Gorelick, S. M. (1983). A review of distributed
parameter groundwater management modelling
methods. Water Resources Research, 19(2),
305–319. doi:10.1029/WR019i002p00305
Huang, C., & Mayer, A. S. (1997). Pump-and-
treat optimization using well locations and
pumping rates as a decision variable. Wa-
ter Resources Research, 33(5), 1001–1012.
doi:10.1029/97WR00366
Johnson, V. M., & Rogers, L. L. (1995). Loca-
tion analysis in ground-water remediation using
neural networks. Ground Water, 33(5), 749–758.
doi:10.1111/j.1745-6584.1995.tb00021.x
748
Keeney, R. L., & Raiffa, H. (1976). Decisions with
multiple objectives. Cambridge, UK: Cambridge
University Press.
Kontos, T. D., Komilis, D. P., & Halvadakis, C.
P. (2005). Sitting MSW landfills with a spatial
multiple criteria analysis methodology. Waste
Management (New York, N.Y.), 25, 818–832.
doi:10.1016/j.wasman.2005.04.002
Lefkoff, L. J., & Gorelick, S. M. (1986). Design
and cost analysis of rapid aquifer restoration
systems using flow simulation and quadratic
programming. Ground Water, 24, 777–790.
doi:10.1111/j.1745-6584.1986.tb01694.x
Lin, H., & Kao, J. (1998). Enhanced spatial
model for landfill sifting analysis. Journal of
Environmental Engineering, 125(9), 845–851.
doi:10.1061/(ASCE)0733-9372(1999)125:9(845)
Malczewski, J. (1999). GIS and multicriteria deci-
sion analysis. New York, NY: John Wiley & Sons.
Mantoglou, A., & Papantoniou, M. (2008).
Optimal design of pumping networks in coastal
aquafers using sharp interface models. Jour-
nal of Hydrology (Amsterdam), 361, 52–63.
doi:10.1016/j.jhydrol.2008.07.022
McKinney, D. C., & Lin, M. D. (1993). Ground-
water optimization using genetic algorithms. Paper
presented at the SIAM Conference on Computa-
tional Issues in the Geosciences, Houston, TX.
McKinney, D. C., & Lin, M. D. (1994). Genetic
algorithm solution of groundwater management
models. Water Resources Research, 30, 1897–
1906. doi:10.1029/94WR00554
Mimoun, D. (2009). Connaissance et préserva-
tion des ressources en eau de la nappe alluviale
dans la basse vallée de la Dore: Rapport d’étude
pour le Parc Naturel Régional Livradois Forez
(p. 118). Saint-Etienne, France: Ecoles Nationale
Supérieure des Mines de Saint-Etienne.
Multi-Criteria Decision Analysis for Identifying a Suitable Location for Groundwater Pumping Wells
Qin, X. S., Huang, G. H., & He, L. (2009). Simula-
tion and optimization technologies for petroleum
waste management and remediation process
control. Journal of Environmental Management,
90(1), 54–76. doi:10.1016/j.jenvman.2008.07.002
Ritzel, B. J., Eheart, J. W., & Ranjithan, S. (1994).
Using genetic algorithms to solve a multiple objec-
tive groundwater pollution containment problem.
Water Resources Research, 30(5), 1589–1603.
doi:10.1029/93WR03511
Rogers, L. L., & Dowla, F. U. (1992). Groundwater
remediation optimization with artificial neural
networks and the genetic algorithm [abstract].
Eos, Transactions, American Geophysical Union,
73(43), 186.
Saaty, T. L. (1980). The analytic hierarchy process.
New York, NY: McGraw-Hill.
Shamir, U., Bear, J., & Gamliel, A. (1984). Opti-
mal annual operation of a coastal aquifer. Water
Resources Research, 20, 435–444. doi:10.1029/
WR020i004p00435
Sumathi, V. R., Natesan, U., & Sarkar, C. (2008).
GIS-based approach for optimized sifting of
municipal solid waste landfill. Waste Manage-
ment (New York, N.Y.), 28(11), 2146–2160.
doi:10.1016/j.wasman.2007.09.032
This work was previously published in the International Journal of Agricultural and Environmental Information Systems
(IJAEIS), Volume 3, Issue 1, edited by Petraq Papajorgji and François Pinet, pp. 72-85, copyright 2012 by IGI Publishing (an
imprint of IGI Global).
Wang, M., & Zheng, C. (1998). Ground water man-
agement optimization using genetic algorithms
and simulated annealing: Formulation and com-
parison. Journal of American Water Resources,
34(3), 519–530. doi:10.1111/j.1752-1688.1998.
tb00951.x
Wang, Q. J. (1991). The genetic algorithm and
its application to calibrating conceptual rainfall
runoff models. Water Resources Research, 27(9),
2467–2471. doi:10.1029/91WR01305
Willis, R., & Finney, B. A. (1988). Planning model
for optimal control of saltwater intrusion. Jour-
nal of Water Resources Planning and Manage-
ment, 114, 163–178. doi:10.1061/(ASCE)0733-
9496(1988)114:2(163)
Willis, R., & Yeh, W. W. G. (1987). Groundwater
systems planning and management. Upper Saddle
River, NJ: Prentice Hall.
Zeleny, M. (1982). Multiple criteria decision
making. New York, NY: McGraw-Hill.
Zheng, C., & Wang, P. P. (2002). A field demon-
stration of the simulation optimization approach
for remediation system design. Ground Water,
40(3), 258–266. doi:10.1111/j.1745-6584.2002.
tb02653.x
749
750

Chapter 46
Last Online Deposits
Spatial Data in the Web
Mohamed-Khireddine Kholladi
University Mentouri of Constantine, Algeria

ABSTRACT
The posting of interactive mapping is essential for the dissemination of information to the general pub-
lic and in all areas. All the elements constituting a spatial object are represented by agreements with
symbols. Each element is represented at the level of reality. The posting of maps on the Internet can take
many forms. It can be static maps, as a picture. Interactions with the user can be included on the maps
produced. These are movements and functions of the zoom presentation (display information, change
of scale, global view). At this level, it is also possible to examine, by selection of the objects represented
on the map (common facilities). This level is commonly known as Web mapping. The dynamic mapping
is used when information is to be renewed or if the geographical extent of the area is large. In this case,
a server handles in real-time updated database to provide users answers to their complaints. The func-
tions proposed in this case are close to those of GIS software (acquisition, manipulation, management
and processing of geographical data). In this chapter we will explore the possibility of integrating a
dynamic mapping on the Web.

INTRODUCTION picture. As a result, the site or any part thereof

For over thirty years, the cards are offered as a
faithful reproduction of a geographical space ob-
ject. All its components are represented by agree-
ments with symbols. Each is represented on the
scale of reality. The maps posting on the Internet
can take many forms. It can be fixed cards, as a
is not connected to a database. The pages are
already built and are displayed (e.g. PDF) on the
user as they were originally created. Interactions
with potential users can be included on the maps
produced. These are movements, and zoom func-
tions as the display presentation of information,
the change of scale, the global view, and so on.

DOI: 10.4018/978-1-4666-2038-4.ch046

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Last Online Deposits Spatial Data in the Web
It is also possible to examine, by selection
of geographical space on the card, such as mu-
nicipalities, schools, etc. This level is commonly
known as Web-mapping. The dynamic mapping is
applied when the information is to be constantly
updated or if the coverage of the area is great.
In this case, allows a server to query real-time
database updates to give users answers to their
complaints (De Blomac F, 2009), (Dupasque X,
2005), (Dupont G, 2004) and (Mauvière E, 1999).
The functions proposed in this case approximate
those of a geographical information system (GIS)
for the acquisition, handling, management and
processing of geographic data. The publication
of the geographical information as numeric
cards takes more and more interest because of
the progress reached in the domain of the new
technologies of information and the communica-
tion and the arrival to maturity of the software of
spatial representation of the data or geographical
information system (GIS) [De Blomac F, 2001].
In this chapter, we will focus on the possibility
of integrating a dynamic mapping on the Web. Our
aim is to provide an overview of the potential of
the Internet for GIS and to give what he wishes
to put their GIS dynamic web to choose between
different opportunities (AFIGEO, 1998).
STATE OF CARTOGRAPHY ART
The mapping allows the use of techniques for
achieving the cards. It is used to better understand
and define the space, territories and landscapes.
It is also used in related sciences, demography
and economy in order to propose a reading of
the geo-phenomena. How can we be a part of the
irregular shape of the spherical earth on a globe?
To be able to give a reliable description of a ter-
ritory and to locate objects in space, it is essential
to know the exact shape of the earth. That is the
purpose of surveying and mapping work. The
ellipsoid is the reference surface to represent the
best shape of the earth. It is useful to perform the
necessary calculations, if you want to perform the
topographic location of objects by using measures.
The physical figure which best represents the shape
of the earth called geoids. This is the form of the
mean sea level. After the location of each point
in space measurements of angles and distances,
calculations of the projections means from the
three-dimensional reality to the representation
plane (Dupont G, 2004).
ONE POINT ON THE TERRESTRIAL
SURFACE LOCALIZATION
The altimetry permits to determine the height of
one point that is expressed by the difference be-
tween the point and a surface of reference as the
sea level. The geoids are the surface of reference
for the observations of altimeter. The measure of
the heights called levelling makes itself by lines,
whose intermediate points are marked with the help
of ankles implanted in the facades of the buildings.
The set of these reference marks constitutes the
network of levelling of the studied geographical
zone. All information on these reference marks
is listed by the direction of the land-registry of
the wilaya.
The plane coordinates of one point are cal-
culated near to the centimetre with the help of
triangles, where we measured the angles and the
distances with tachometers. To orient their obser-
vations, the geometers determined the coordinates
of the mosques minarets or the water castles that
represent very visible reference marks from afar.
The administration of the land-registry and the
topography manages the geodesic basis network
composed of score thousand points. The system of
positioning by GPS satellites (Global Positioning
System) permits the precise determination of the
coordinates of one point quickly on earth. To be
able to produce a card or a plan from the measured
data, we conduct calculations of a mathematical
projection. Several geodesic systems exist. An
751

important system is the WGS4 that is the basis of
the GPS system (Laurini R. & Thompson D, 1992).
THE CURVES OF LEVEL
The level curve is a polyline that passes by all
points of the land situated to the same altitude.
The earth relief is represented on a plane card by
level curves and points sides. If one represents a
mountain cut in horizontal plates of equal thick-
ness, the contours of these plates, seen from top,
draw the curves of level that follow the relief
shape. The disposition of the level curves on the
card informs on the shape of the relief (stiff slope
or soft slope). The value of the heights curve is
registered on the level curve. The orientation in-
forms of slope direction. The numbers are always
orienting upstream (Kholladi M. K, 1999b).
PRODUCTION OF A CARD
The photogrammetry is a cartographic technique
that permits to measure all visible elements
from aerial photos of the territory. From a plane
equipped with a special camera, some aerial
photos are taken to very short intervals, so that
two successive photos cover themselves to about
60% in direction of the skimming. Thanks to the
principle of the stereoscopy, we can measure and
reconstitute all elements situated inside the 60% of
cover from a couple of aerial photos. The height
of the skimming depends on the scale of the card
that we want to produce. Today the planes of pho-
togrammetry are equipped with the GPS system.
A relation between the earth and the aerial
photo is established by the signalling of the points
knew on the land. These points can be detected
on the photo, what permits to situate all points of
the photo in the system of coordinates wanted. To
re-establish the three-dimensional relief from the
aerial photos, the cartographer uses a grant natural
of the man: the faculty to see in three measure-
752
Last Online Deposits Spatial Data in the Web
ments. At the time of the restitution photogram-
metric, the aerial photos are installed by couples
in a device named restitutor, which permits to
re-establish a model of the relief and to seize all
visible elements numerically to stock them in a
data base. The restitution represents an important
phase in the chain of production of a card.
Our main challenges are the seizure of all
sought-after elements, the seized interpretation
correct of the objects and the guarantee of the
geometric precision required. The result of the
restitution must be controlled therefore and must
be completed by information of descriptive,
nominative and administrative nature, where the
nature of a building (of dwelling, agricultural,
commercial or industrial), its attributes (school,
town hall, reservoir, etc.), the names of the streets
and the toponyms (localities, inhabited places) and
the administrative and other limits. The scale of a
card expresses the report of reduction between the
reality and the representation on the card. A big
scale of card permits to represent a lot of details,
when a card has a small scale. It gives us a better
general preview on a whole region. Different ap-
plications require some cards to different scales
(Kholladi M. K, 1999a; [Kholladi M. K, 1999b;
and 9Laurini R. & Thompson D, 1992).
The topographic card is a faithful representa-
tion of the reality. To arrive there, the cartographer
restores the information in a legible and compre-
hensible cartographic picture. Its instruments are
the symbols, the signatures, the colours and the
texts. The graphic elements used in the card are
regrouped and are explained in the legend of the
card. The choice of the symbols and colours is
guided by the principle to bring closer for example
to best the graphic representation to the reality,
the use of the bruise for the rivers, of the green
for vegetation and symbols for the camping, the
sports, etc. An analysis of the different editions
of a card permits to retrace so much in detail the
evolution of the regional development spatial
such economic and social. The information on the
date and the shape of the update of the edition are
Last Online Deposits Spatial Data in the Web

Figure 1. Example of thematic layers

indicated in the legend of the card, what denotes
the actuality of the card. Available until now as
cards and plans, the geographical information
presents itself henceforth as computer data base,
which permits exchanges of more efficient data,
as well as exploitation and a lot more rational
update (AFIGEO, 1998).

GEOGRAPHICAL INFORMATION
SYSTEM (GIS)

A Geographical information system is a computer

tool permitting to represent and to analyze all
geographical objects as well as all events that
produce themselves of it [Demegre J, 2005]. The
GIS offer all possibilities of the data bases (as ask simple questions as (Kholladi M. K, 2004a)
requests and statistical analyses) and that, through and (Laurini R. & Thompson D, 1992):
a unique visualization and geographical analysis
clean to the cards. • To whom does this parcel of land belong?
Many other domains as research and the de- • Which of these two zones is the most
velopment of new markets, the survey of a con- distant?
struction impact, the regional development, the • Where are the exploitable lands for a given
management of the networks of water, electricity, economic activity?
gas or transportation, the follow-up in real time
of vehicles, the civil protection, the follow-up of Some of the questions integrating an analysis,
the evolution of the pollution of a territory, the as for example:
follow-up of the desertification, etc. are concerned
as directly by the power of the GIS to create
some cards, to integrate all type of information,
Figure 2. The two models of the reality representa-
to visualize the different scripts better, to present
tion: vector and raster
the ideas better and to measure the impact of the
possible solutions better.
A GIS permits the storage of the information
concerning the world under the shape of bound
thematic layers the some to the other by geogra-
phy (as on the Figure 1) (Kholladi M. K, 2004a;
Kholladi M. K, 2004b; Laurini R. & Thompson D,
1992 and Laurini R. & Millert-Raffort F, 1993).
The Systems of Geographical information
exploit two different types of geographical mod-
els: the vector model and the model raster as on
the Figure 2. If one has a GIS and data, one can

753

• What are the virgin lands to construct new
houses?
• What are soils adapted to the plantation of
a type of tree?
• If I construct a freeway here, what will be
the density of the traffic of it?
The GIS arranges many and powerful tools
of analysis, but two among them appear as par-
ticularly essential: the proximity analysis and the
spatial analysis.
PROXIMITY ANALYSIS
How much does existing houses in a zone of 80
meters on both sides of a road to big circula-
tion? (See Figure 3). What is the total number of
customer in a ray of 5Km around a big surface?
Figure 3. A road neighbourhood request
Figure 4. The thematic layers superposition
754
Last Online Deposits Spatial Data in the Web
To answer these questions, the GIS has calcu-
lations algorithms named “buffering” in order to
determine the relations of proximity between the
objects.
SPATIAL ANALYSIS
The integration of data by different information
layers permits to do a spatial analysis. This analysis
by crossing of information, if it can take place
visually (to the tracings identical superimposed the
some to the other) requires the crossing often with
alphanumeric information. The crossing of the soil
nature, of its declivity, of the present vegetation
with the owners and the taxes paid is an example
of sophisticated analysis that permits the GIS use
(as on the Figure 4) (Kholladi M. K, 2004a) and
(Laurini R. & Milleret-Raffort F, 1993).
The card is a much better tool of synthesis and
presentation of information. The cards created
with a GIS can integrate the reports, the three-
dimensional views, the photographic pictures and
all sorts of elements multimedia henceforth eas-
ily. The capacity of the GIS concerning the data
research and the geographical requests execution
allowed numerous enterprises or administrations
to save the hundreds of millions of Algerian Dinars.
A realtor can use a GIS to find houses of a certain
surface, with three rooms and including a garage.
It gets a list of the houses answering these criteria
Last Online Deposits Spatial Data in the Web
Figure 5. Request of a house research with criteria
as well as their geographical localization thus
simultaneously as on the Figure 5.
This request can be refined thereafter with the
help of two new criteria: a renting lower to 5000
DA per month and situated unless about hundred
meters of a school. The result of this new request
will be displayed in the same conditions as on the
Figure 6 (Kholladi M. K, 1999a]) and (Sohm J,
2003).
WEB MAPPING
The needs to communicate some information in
the domains of management of the risks, of the
regional development, or more environment,
the publication of cartographic data on Internet
is means of indispensable communication (De
Blomac F, 2001), (De Blomac F, 2005) and (Sous-
sain G, 2003). The Web Mapping or the diffusion
Figure 6. Composed spatial requests
of the cards via the Internet network is a domain
in full expansion thanks to the development of
the solutions Open Source. Following the GNU
philosophy that allows the copy, the diffusion of the
software and the modification of the code source,
these generally free programs and of emergent
free use at a rhythm sustained.
The concept of the Web cartography consists
in three components: geography, information
and the Web. The approach of the Web Mapping
is adapted to a large public no specialist: simple
interface, oriented “cartographic atlas” and on line
GIS, consultation by connection to weak debit,
fast and independent solution of the navigators,
without Plug in (De Blomac F, 2009) and (Soussain
G, 2001). It permits a complete integration in an
existing site and the possibilities of interactions:
the dual brightness of the entities, the comple-
mentary information, the display of the themes,
etc. The cartographic functionalities are:
• Navigation: inside a same document.
• Global view: seen permitting to localize
the zone of visualization on a general pre-
view of the document.
• Zoom: the user has the possibility to zoom
in the size of the visualization zone.
• Icons: presence of interactive symbols in
the visualization.
• Zones of click: interactive zones in the vi-
sualization triggering some events (pop up,
info bubbles, etc.).
• Management of layers: possibility to vi-
sualize several layers (tracings).
• Requests user: Semantics and/or geomet-
ric, thematic, predefined or no.
• Update: interface of update of the
spatial data and/or attributes without
programming.
755

GENERAL PRINCIPLE
OF A CARTOGRAPHIC
WORKING SERVER
The cartographic server is the automatic ticket
window where we make a call to display some
cards on its station of work. The architecture is
of type client-server. The user, from his terminal,
spear of the requests to ask for a particular card
display. The cartographic server the interprets and
send back the card under the shape of a matrix
picture (png, jpg, etc.) or vectorial (svg, swf,
etc.). The cartographic motor can be controlled
by languages of script as PHP, java script, Python
or Perl that allow him to generate a card dynami-
cally in answer to the user’s request (Thomazeau
J., 1997) and (Soussain G., 2003).
The cartographic server can look for the infor-
mation necessary to the realization of the card in
its own resources, but also on servers of distant
data. The diffusion of the on line data requires an
installation side server with software as Apache
(project Open Source) or IIS (Internet Information
Services, of Microsoft) that turn in task of bottom
and give access to the cards servers to the intranet
and to the Internet. These servers often see their
functions spread by interpreters of scripts as PHP
Figure 7. The cartographic server
756
Last Online Deposits Spatial Data in the Web
or ASP. The cartographic server leans on these
elements to receive some requests and to send
back the pictures and data. Client side, a navigator
Web is sufficient, accompanied by a viewer, to
display the card (cf. Figure 7) (Durand H, 2003)
and (Pewe B, 1997).
The data can be managed by specific software,
the SGBDR, such PostgreSQLs, MySQL, Oracle,
among others, that can be installed directly on the
server containing the cartographic server or on
another server, distant. PostgreSQL and its exten-
sion spatial PostGIS is the management system
of relational data bases the more succeeded in the
domain of the free software. PostGIS is included
henceforth in the official distribution of Post-
GreSQL. It offers functionalities deepened to
stock and to treat geographical reference of the
spatial objects.
The different software bricks have been tested
with the systems of exploitations Windows XP
and Linux Mandrake 10 to value their interoper-
ability. To simulate the situation of exchange of
data via Internet, MapServer 4.4, PHP/MapScript,
the SGBDR extended PostgreSQL 7.5 of PostGIS
0.9.1 with its interface graphic phpPgAdmin has
been compiled on a dedicated server Linux. The
user will have permissions of reading and writing
Last Online Deposits Spatial Data in the Web
in the arborescence of the indexes of the machine
server, for the storage of the solutions side server
while the solutions client side will be placed on
the local machine in a Windows XP pro environ-
ment (De Blomac F, 2001).
SOLUTIONS CLIENT SIDE
The software tested client Open Source, QGIS,
UDIG, THUBAN and JUMP, are software GIS
in full evolution having functionalities advanced
from afar for the consultation and the geographical
data edition (see Figure 8).
In practice, these applications are not always
located entirely side server because they can re-
quire the downloading of a light plug-in by the
customer (for example the Rosa applet). Of basis,
the solutions side server Open Source brings the
possibility from a classic navigator Internet to
visualize the generated dynamically geographical
layers. Currently, new more advanced function-
alities appear as the possibility to add or to
modify on line information.
Figure 8. Solutions server side
THE DM SOLUTIONS PRODUCTS
These are MAPLAB, CHAMELEON, KA-MAPS
and CARTOWEB3. Figure 9 summarizes the
features of each of them.
Other Solutions
On the Web, we can find a big number of other
available solutions in downloading. They are
counted on the sites of FreeGIS and Maptools,
and are of easiness of access and use variables.
In general the simple and easily adaptable solu-
tions for its own geographical data are viewers as
Veremap or Pmapper. This last proposes a set of
tools for example interesting Java scripts as the
zoom by cursor, the legend in hierarchical tree
and the impression in pdf. Then comes the more
complex solutions and evolved like Mapbender or
MapLink. On-line, it is possible to select a polygon
and to modify its geometry (see on the Figure 9)
(Marshall J, 2001) and (Mauvière E, 1999).
Actually the solutions client-server can be
completely complementary, one (server) permit-
ting to do some treatments clean to the problematic
of the GIS (crossings of layers, requests, analysis
thematic), the other (customer) through the inter-
mediary of the mapfile displaying the card in an
757

Figure 9. Web mapping solutions
accessible interface on Internet. Otherwise, we see
the modules Open Source that transplants them on
the big commercial GIS as Amein maintained by
Territories for ArCGIS 8 and Wortomap developed
by usable Tydac with MapInfo 7.8 that allows him
to export the mapfile of the GIS to display the
same card in an interface Map Server.
Web Carto as Solution of
Web Dynamic GIS
The Web Carto product is an interactive Web-GIS,
based on the motor cartographic Open Mapserver
source. It permits the presentation, but also the
acquirement of geographical data through the
Internet. It is a tool permitting to generate some
cards from various spatial data, according to
the instructions contained in what we call some
Mapfile, of the files “.map”, regrouping the pa-
rameters describing the way whose cards must
be presented. We can beside use a model html
(commonly named a template), that will manage
the layout of the display of the data generated.
The most often, it is used on a server Internet
to generate some pictures in Web pages, and so
to permit the display but also the questioning,
the modification, of cartographic pictures on an
Internet site. We can also use MapServer in local,
on its computer, to generate the cards, to do the
758
Last Online Deposits Spatial Data in the Web
requests and analyses. The diagram of the present
Figure 10 how MapServer integrates in a system
server Internet (Gifford F, 1999).
The main libraries used by MapServer are:
• GDAL and OGR: Interpretation and con-
version of the spatial data.
• Proj.4: Reproduction, transfers and distor-
tion of the spatial data.
• GD: Generation and modification of pic-
tures to the formats bitmap of the Web
(Herself will possibly have need of the
JPEG libraries and PNG).
• FreeType: Management of the fonts.
We can also mention the libraries that are going
to allow the generation of cards particular formats:
• MING for the format flash.
• PDFlib for the PDF format.
MapServer accepts a big number of formats
data raster, there again thanks to the library of
GDAL functions. Without this library only some
formats geographical reference rasters is sup-
ported: GeoTIF, hair *, gif *, png *, jpeg *, and
Erdas Lan. The possibilities raster of your instal-
lation of MapServer goes therefore essentially to
depend on the capacities of GDAL that have been
Last Online Deposits Spatial Data in the Web

Figure 10. MapServer in a server Internet

specified at the time of its compilation. To add wms_srs EPSG:42304 EPSG:42101

a layer raster in a Mapfile, here is the syntax of EPSG:4269 EPSG:4326 EPSG:42304
the LAYER block: wms_name
prov_bound
LAYER wms_server_version 1.1.0
NAME Scan wms_formatlist image/gif,image/
DATES Constantine.tif png,image/jpeg,image/wbmp
RASTER TYPE wms_format image/gif
STATUS ONE END
END

THE CARTOWE3 PRESENTATION

SERVERS TO THE NORM WEB MAP
SERVICE (WMS) OF THE OGC
The use of a WMS source requires the presence
of a block PROJECTION in the MAP block, as
well as the use of a METADATA block that is
going to specify the request made to the server.
We have for example:
CONNECTION http://www2.dmsolutions.
ca/cgi-bin/mswms_gmap?
CONNECTIONTYPE WMS
METADATA
Most solutions of interactive cartography proposed
on the market are descended of the world of the
GIS and don’t often answer fully the needs of the
dynamic Internet sites. Camptocamp its developed
CartoWeb thus, a Web application of Geographi-
cal information System (GIS) interactive, while
especially respecting the specific needs of the
web. It is also conceived especially to integrate
easily in nearly all types of sites dynamic internet.
It permits to put easily in evidence the spatial as-
pects of the data presented notably, even though
these are not part of a GIS (Jongler D. & Lestrat

759

I., 2001), (Jongler D. & Lestrat I., 2002) and
(Kholladi M. K., 2008).
The proposed solution is especially interest-
ing since it permits the visualization (diffusion)
and the seizure (collection) of the geographical
objects (lines, points, surfaces) through different
Web sites. It permits to benefit from a solution
of web-GIS to very attractive costs, since this
solution doesn’t require the purchase of licenses
commercial GIS. However, CartoWeb integrates
without any problem in an environment existing
GIS, since the formats supported by this system
are compatible with most formats of the GIS ap-
plications of the market.
SPECIFICATIONS
The Web Carto product is an interactive Web-GIS,
not only permitting the presentation, but also the
acquirement and the geographical data manage-
ment through the Internet and more especially the
Web. The Web Carto has been developed while
respecting the specific needs of the Web in the
GIS domain, to know (De Blomac F, 2005):
• Spatial visualization of the data in the Web
sites (Webmapping).
• Edition and seizure of data through the
Web (interactive Web-GIS).
• Integration of the interactive Web-GIS in a
site web, existing or in creation, simplified
to the maximum.
• Interaction facilitated of the geographical
functionalities with the relational data bas-
es (out GIS) of the site web.
• Possibility to insert the cartography direct-
ly in any page of a site in the content (har-
mony between the site and the GIS tools).
• Possibility to share and/or to rent geograph-
ical information (cartographic funds).
• Weak cost of acquirement and
implementation.
760
Last Online Deposits Spatial Data in the Web
• Compatibility with the applications GIS
standards of the market.
• Compatibility with the GPS and the PDA.
• Respect of the norms and standards (OGC
WMS, OGC SFS, GML, XML, XHTML,
and SOAP).
ADVANTAGES OF USE
We will mention in some points the advantages
of the CartoWeb system:
• Enhancement of the spatial information in
a Web site.
• Easy setting up, little programming.
• Weak cost of implementation (not of ex-
pensive license - federation of the costs of
acquirement of the cartographic data).
• Possibility to seize and to modify geo-
graphical information via Internet.
• Compatibility and integration with the GIS
customer server of the market.
• Possibility to rent the access to functional-
ities (ASP fashion).
CartoWeb will be used in particular for [De
Blomac F, 2009]:
• To distribute the cards or plans on the Web
with specific thematic information and all
functionalities of navigation (zoom, dis-
placement, etc.).
• To bind spatial information to the informa-
tional content of the data bases.
• To seize and to modify the on line geo-
graphical information, by the edition of
objects of type points, lines or surfaces.
ARCHITECTURE OF WEB CARTO
The Web Carto is an efficient solution to the needs
concerning treatment and diffusion by Internet
Last Online Deposits Spatial Data in the Web
of managed data or produced by a GIS classic
client-server. A strong point of Web Carto is the
easiness of interaction between present data in
external data bases and the geographical data. It
permits the presentation, but also the acquirement
of geographical data through internet.
It has been developed while having to the mind
the specific needs of the Web in the domain of the
GIS. It is generally used as application standalone.
However Web Carto is constructed on an archi-
tecture multi third that permit, if in the need be,
to establish it as Web Service. It is then possible
for several Web sites to use the same geographi-
cal server. It possesses two fashions of working
in direct or in SOAP (sees Figures 11 and 12).
Functionalities of the CartoClient are activated
on simple passage of parameters (see Figure 13).
The development of CartoWeb respects the main
condition for a perennial computer application, to
know an efficient separation between the data, the
treatment, the presentation and the language. There
is not a code HTML mixed to the code of PHP
treatment. The texts of the interface also sepa-
rated. It permits to put a multilingual site in place
therefore easily. It assures the application evolutiv-
ity, since the central functionalities are not touched
at the time of a change of graphic charter or the
addition of a peripheral functionality.
Figure 11. Direct Mode
Figure 12. SOAP Mode
CREATION OF A PROJECT OF WEB
MAPPING
The Gait Describes the Project Configuration.
Project Installation
In a line of order the message of solicitation (DOS,
shell), to throw the manuscript of installation of
cw3setup.php.
<PHP-INTERPRETER>cw3setup.php to in-
stall
--base-URL http://localhost/car-
toweb3/htdocs /
--development of profile
--<nameOfProject> of project
So no mistake occurred we must be able to
reach the project by http://localhost/ cartoweb3/
htdocs/client.php of typing for the navigator and
to choose the project from the slim drop-down
of the projects.
DATA OF LOADING
As we can see it, the project that we created is
empty. Thereafter, we will load the spatial data
of the project. To add the next one in the Mapfile
(project_name.map): To PUT THE TYPE THE
COLOR NAMED 240 of region of NAME of
MODEL of class region of ttt of CALIBER of
reg_algerie of POLYGON DATA 240 240 OUT-
LINECOLOR 255 EXTREMITY of extremity of
CODE id_attribute_string of 165 96 of MÉTA-
GIVE extremity of END
And in the file of layers.ini (server_conf/
project_name /) to add what follows:
• layers.region.className = layers.region.
label layer = Regions
• layers.region.msLayer = region
761

Figure 13. The architecture Composition
• And to add the new identification of layer
in the list of children of layer of the root:
• layers.root.children = region
• To publish besides the file of project_name.
ini in a publisher of text in order to place
some initial mapstates by default as chosen
layers.
• To add what follows:
◦◦ mapInfo.initialMapStates.default.
layers.region.selected = rectify
◦◦ In your navigator, the click on the
button of reset_session and one must
see the regions of Algeria that one
wants to display.
• To add our data, one must simply proceed
from the following manner:
◦◦ to get our available data, by default in
the shirt of data
◦◦ to modify the Mapfile size,
◦◦ to modify the size defined in the ini-
tial Mapstateses (project_name.ini),
◦◦ to add the definitions of layers in the
mapfile (project_name.map),
◦◦ to add the definitions of layers in the
layers.ini.
◦◦ to hand to zero the session in the nav-
igator, and to get our data to display.
762
Last Online Deposits Spatial Data in the Web
CONCLUSION
The realization of dynamic cards on Internet is
essential when we want to distribute geographi-
cal information. Often used in the localization of
the places by their addresses, in the calculation
of itineraries or for the geographical marketing,
its uses are increasing and enlarge toward new
sectors. Descended of the data processing applied
to geography, this tool stands to the crossing of
techniques allying the relational data manage-
ment, the representation and the spatial analysis,
the programming in script language (Kholladi
M. K., 1999a).
Actually, we find either of the reusable
components (Chameleon), of the interfaces web
(CartoWeb, Ka-Map), either of the more complete
solutions of help to the management and to the
realization of Web applications (MapLab). As
bending on the construction of a card on line,
only QGIS in customer and MapLab thanks to
its MapEdit tool side server permits to publish
graphically and dynamically a mapfile. In this
domain many progress remain last one has to be
done because this last is not flexible enough in
its syntax and in its adaptability to the different
environments (Jongler D. & Lestrat I., 2001).
Last Online Deposits Spatial Data in the Web
REFERENCES
AFIGEO, (1998, May). L’information
géographique française dans la société de
l’information: état des lieux et propositions
d’action, p. 35.
De Blomac, F. (2001). L’open source, véritable
alternative pour les SIG? p. 4.
De Blomac F, (2005). Les systèmes d’information
géographique territoriaux-La mise en ligne des
données géographiques: principes et expériences,
Cybergeo: European Journal of Geography, n°318,
p. 123.
De Blomac F, (2009, June 01). Publier des cartes
sur Internet: ces solutions qui nous viennent du
graphique. SIG La Lettre, n°108, p. 3.
Demegre, J., & Salge, F. (2004). Que sais-je?
Les SIGs, p. 84.
Dupasque X, (2004, April). Etats des lieux des
solutions Internet de diffusions des données
géographiques, Géo événement.
Dupont G, (2004, January 23). L’IGN cartographie
la France en numérique, Journal: Le Monde, p. 2.
Durand H, (2003, April). Du SIG à la donnée sur
internet: Quel mode de diffusion pour quel public.
Journée professionnelle SIG-LR.
Gifford, F. (1999). Internet GIS Architecture -
Which Side Is Right for You? p. 11.
Jongler D. & Lestrat L, (2001, February). Dossier
technique: les SIG et Internet, Signature n°20,
CERTU, p. 11.
Jongler, D., & Lestrat, L. (2002). Cartographie
sur Internet, Fiche technique du CNIG (Aide à
la Maîtrise d’Ouvrage) n°64, p. 8.
Kholladi M. K, (1999a, November 15-17). Mé-
canismes de raisonnement spatiaux sur des objets
géographiques, Séminaire International sur les
SIGs, « AL-SIG’99 » organisé par le Conseil
National de l’Information Géographique « CNIG
», Alger, p. 9.
Kholladi M. K, (1999b, November 15-17).
Modèle Numérique de Terrain par la Méthode
de Triangulation de Delaunay avec Contraintes,
Séminaire International sur les SIGs, « AL-SIG’99
» organisé par le Conseil National de l’Information
Géographique « CNIG », Alger, p. 21.
Kholladi M. K, (2004a, September). Raisonnement
et Manipulation Spatial sur des Objets Urbains «
RMSOU », CISC’2004: Conférence Internationale
des Systèmes Complexes, Jijel, p. 13.
Kholladi, M. K. (2004b, December). Geographical
Information System for Urban Spatial Analysis,
ACIT’2004: The International Arab Conference
on Information Technology, Mentouri Univer-
sity of Constantine, Algeria, pp. 188-193, ISSN:
1812-0857.
Kholladi, M. K. (2008, October 15-16). Carto
Web Solution for the Web Mapping, IACET: Inter-
national Arab Conference of e-Technology, Arab
Open University, Amman, Jordan, pp. 105-113.
Laurini, R., & Milleret-Raffort, F. (1993). Les
bases de données en géomatique. Paris, France:
Editions Hermès.
Laurini, R., & Thompson, D. (1992). Fundamen-
tals of Spatial Information Systems, the APIC.
Series Academic Press.
Marshall, J. (2001). Developing Internet-Based
GIS Applications, p. 11.
Mauvière, E. (1999). La cartographie sur le web.
p. 40.
763
 Last Online Deposits Spatial Data in the Web

Pewe, B. (1997). GIS online, p. 125. WEB SITES TO VIEW

Pottier, P. (2000). La sémiologie et communica- http://www.camptocamp.com/
tion cartographique, Fiche technique du CNIG
n°40, p. 6. http://www.cartoweb.org/

Sohm, J. C. (2003). Image et graphique pour le http://www.commentcamarche.net

web, p. 5. http://www.esrifrance.fr/
Soussain, G. (2001). Publier des cartes sur inter- http://www.geoscopies.net
net: Les repères. p. 64.
http://www.ieti.fr/Panorama_SIG_WEB.pdf
Soussain, G. (2003). Cartographie sur Internet:
synthèse à l’usage d’une maîtrise d’œuvre. p. 7. http://www.mysql.com

Thomazeau, J. (1997). Le Web support http://www.opensource.org

d’applications client/serveur, une voie suivie par
la DSI pour le système d’information de gestion
du CNRS, p. 12.
http://www.ors-idf.org/etudes/pdf/carto-web.pdf
http://www.portailsig.org
http://www.sitewww.ch/dossier_solution_car-
tographie.php
http://www.webrankinfo.com/actualites/tags/
cartographie-sur-internet.htm

This work was previously published in Handbook of Research on E-Services in the Public Sector: E-Government Strategies
and Advancements, edited by Abid Thyab Al Ajeeli and Yousif A. Latif Al-Bastaki, pp. 182-196, copyright 2011 by Information
Science Reference (an imprint of IGI Global).

764
 765

Chapter 47
Spatial Modeling of Natural
Phenomena and Events
with Artificial Neural
Networks and GIS
Silke Noack Alexey Irkhin
Beak Consultants GmbH, Germany Beak Consultants GmbH, Germany

Andreas Barth Evelyn Bennewitz

Beak Consultants GmbH, Germany Beak Consultants GmbH, Germany
Frank Schmidt
Beak Consultants GmbH, Germany

ABSTRACT
Artificial neural networks (ANN) are used for statistical modeling of spatial events in geosciences. The
advantage of this method is the ability of neural networks to represent complex interrelations and to
be “able to learn” from known (spatial) events. The software advangeo® was developed to enable GIS
users to apply neural network methods on raster geodata. This statistic modeling can be displayed in
a user-friendly way within the ESRI ArcGIS environment. The complete workflow is documented by the
software. This paper presents three pilot studies conducted to illustrate the possibilities of spatial pre-
dictions with the use of existing raster datasets, which described influencing factors and the selection
of known events of the phenomenon to be modeled. These applications included (1) the prognosis of soil
erosion patterns, (2) the prediction of mineral resources, and (3) vulnerability analysis for forest pests.

INTRODUCTION dence of a phenomenon on several main influenc-

Natural phenomena and events are usually caused
by a complex of interacting factors. An exact
mathematical formulation of a geo-scientific task,
however, (with equations describing the depen-
DOI: 10.4018/978-1-4666-2038-4.ch047
ing factors) is rarely feasible. Typically, a model
only refers to some aspects of the phenomenon in
question. Due to the lack of data and knowledge
of details of many geo-processes, mathematical
models cannot be successfully defined and applied
with reasonable amounts of effort. As a result,

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
the exploration of artificial intelligence (artificial
neural networks, ANN, e.g., Haykins, 1998) as a
means to provide a reliable tool to analyze causal
relationships and to make the knowledge available
for predictive tasks was initiated. This approach
differs from traditional methods in that viable
results may be obtained with reasonable efforts
invested in data processing, model design and
computational time.
Results for three different case studies are
discussed in this paper, following an introduc-
tion on the theoretical background of the neural
network approach. The outcomes, predictive maps
illustrating the probabilities of occurrence for a
given event of interest, form an important basis
for planning activities. The case studies deal with
various tasks in applied earth sciences and dem-
onstrate the rather universal applicability of the
coupled ANN/GIS approach that is presented in
this paper. In particular, the following objectives
applied to the case studies presented here:
1. Spatial prediction of soil erosion channels to
localize damage on arable land and to infra-
structure and to model effects of mitigation
measures;
2. Prediction of the most promising locations
for mineral exploration, based on available
geological data within a gis environment;
3. Spatial prediction of the most vulnerable
forest stands in case of spreading bark beetle
infections.
Although the value of the application of neural
network technologies in GIS environments was
recognized in the past, actual application of this
remained a challenge for standard users due to the
lack of user-friendly tools. Thus, the overall goal
of the three case studies was to study the quality
of model outputs and the general applicability of
the ANN/GIS approach included in the software
advangeo®.
766
RATIONALE AND OBJECTIVES
OF THE STUDY: SELECTION OF
AN APPROPRIATE MODELLING
METHOD FOR SPATIAL EVENTS
The location of a natural event is determined by a
complex network of influential causes and subse-
quent effects. Hence, the relationships between the
parameters are usually characterized by qualitative
descriptors rather than quantifiable means, if they
can be described at all in a reasonable amount of
time. This hampers the (mathematical) modeling
and spatial prediction of natural phenomena. In
general, spatial pattern of events can be modeled
by two different approaches:
1. By conducting detailed studies of physical,
chemical and other relations to establish an
accurate quantitative model of the processes
with mathematical-analytical methods (e.g.,
using finite elements to model slope stabil-
ity). With this approach, equations are used
to parameterize and model natural processes.
Equation calibration is usually accomplished
by adjusting “constants”, based on the com-
parison of modeled results and measured
data.
2. By a statistical approach, whereby poten-
tially influential factors are evaluated by
multivariate methods, based on statistical
correlations. The influence of several inde-
pendent factors on the dependent variables
(e.g., the development of a geo-hazard such
as slope failure) is investigated.
The calibration of analytical models is usually
connected with time-consuming and cost-intensive
field studies. As a result, it is not uncommon that
complex natural processes cannot be adequately
described with derived physical equations for more
extensive areas of interest, even given reasonable
efforts to do so.
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
The statistical approach has the advantage of
being able engender usable results with much
lower amounts of effort in terms of data capture
and calibration. Well-known and widely used
multivariate methods include regression, cluster
and discriminant analyses.
The ANN approach is a multivariate analysis
method. With this method, it is possible for users
to model complex, non-linear relations without
prior or initial knowledge of the relationships
amongst the input layers. These relations are
investigated by the ANN software during a
“learning process” in which a sizable amount
of input data can be analyzed at various scales.
However, type of data layers selected and the
manner in which they are prepared by the user
will influence the quality of outputs generated.
Therefore, a general understanding of the baseline
processes to assess the quality of the modeling
results produced.
Various authors have successfully applied
ANN to forecast the occurrence and extent of
spatial events. Examples of applications have
included the accurate identification of slope
failure processes (Fernandez-Steeger, 2002), the
modeling of spatial air pollution patterns (Lin et
al., 2004), the propagation of flood waves (Pe-
ters et al., 2006) and the prediction of mineral
deposits (Redford et al., 2004). However, none
of the previous studies involved the integration
of ANN into a geographic information system,
whereby data processing and the presentation of
results could be streamlined to better facilitate
the storage management and display of generated
information. The goal of this project was initi-
ated to address this with the development of a
software product. This product can be applied by
a GIS user within the standard GIS environment,
to facilitate the production of prediction maps
for planning purposes, following the investiga-
tion of influential factors identified for a given
assessment.
METHOD: ARTIFICIAL
NEURAL NETWORKS
Artificial Neural Networks (ANN) (Haykins,
1998) can be used for the statistical modeling
of spatial events. This method has proven to be
capable of representing complex interactions. In
addition, it can “learn” by analyzing the combina-
tion of input data at a given location with known
events. Following training with such known
events, ANN applies the “knowledge” to more
extensive areas under similar boundary conditions.
ANN help to detect the controlling factors for
analyzed events and can also return a probability
value, which describes the likelihood of event
occurrence for a particular location. These results
are of vital importance, as they can inform about
the nature of geo-hazards and aid in the planning
and implementation of mitigation measures. In
the case of mineral exploration, the analysis can
help prospectors to minimize their efforts and
expenditures by focusing resources and work on
the most promising locations.
It should be noted that all forms of approaches
have their own sets of precautions and shortcom-
ings. The ANN approach is no exception; however,
identification of these challenges and restrictions
allows users to pre-process input data more me-
ticulously and to gain a better understanding of
possible inconsistencies in the generated results.
One such challenge relates to the quality of the
predicted results with respect to the way the ANN
learns: if the training data, which represent the
known events, is only available as incidental points
on the raster layer instead of complete coverage,
the ANN may return an incomplete statistical
model of the phenomenon.
The main principle of the ANN approach is
built on the concept of how networks of biological
neurons are organized and function by imitating the
properties of a real nervous system. These systems
are composed of many chemically interconnected
neurons (nerve cells), which receive and process
signals from other neurons. When these signals
767
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
reach a certain threshold, sets of nerve cells (closest
to the originating signals) are activated and forward
the information to other connected neurons. Learn-
ing processes result from the constant adaptation
of the interconnections amongst neurons, where
the growth and strength of connections are directly
related to their frequency of use and degenerate
when rarely accessed.
The simulation of these biochemical processes
in an ANN is done by artificial neurons, known
as processing units. The connections are realized
by (directed) interconnection weights wij (e.g.,
Backhaus et al., 2003; Kriesel, 2008). In general,
an ANN may be described as an adaptive system,
which changes its structure based on the flow of
external or internal signals through the network.
Artificial neurons are usually organized in
layers, comprised of:
• An input layer that receives input values
(i.e., independent variables and crucial
controlling factors related to the event to
be modeled);
• Hidden layer(s) that process the signals
and calculate the resultant value(s) by
transforming them with the application of
an activating function; and
• An output layer, created from the depen-
dent variable to be modeled.
Figure 1. General scheme of a feed forward network
768
The network topology describes the way the
neurons are interconnected. The direction of sig-
nal propagation (forward or backward), the level
and the type of connection (e.g., completely con-
nected or with shortcuts) are thus defined. Figure
1 shows the scheme of a completely connected,
feed-forward network with 4-3-1 topology. On
the right-hand side, a processing unit (neuron) is
sketched with its output function.
The weights increase or decrease the strength
of the incoming signal and are adaptable during
the learning process. By repeating the input of
the training pattern, the intensity of the connec-
tions is modified, such that the deviation (or error)
between expected and actual outputs (i.e., between
modeled and real data) are minimized.
There are different network paradigms, such
as the multi-layer perceptron (MLP), the radial
basis function networks (RBF), Hopfield networks
and self-organizing maps (SOM). The MLP
paradigm is of particular interest for applications
that involve supervised learning. The information
processing that occurs is characterized as “feed
forward”. This network paradigm is an extension
of earlier developments on the perceptron, which
incorporates higher requirements for the model-
ing of complex relationships. A MLP consists of
an input layer, at least one hidden layer and an
output layer (Figure 1).
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
MLP are usually trained with a “back propa-
gation” algorithm (e.g., Rumelhart et al., 1986;
Rios, 2010). “Back propagation of error” is a
training algorithm, which adapts the weights (wij)
by considering the degree of error from the output
signal so that it may be minimized.
Another important network parameter is the
activation function, which is used to calculate the
output status of the neuron. Differentiable and
non-linear functions are required for this purpose.
Generally, sigmoid functions are applied.
INTEGRATION OF THE
METHOD INTO GIS
ANN has already been used for some time, as
an alternative to other multivariate, statistical
methods and has been proven to produce adequate
results. Although some ANN-based commercial
products and class libraries are currently available,
they usually require the manual generation of input
data in a proprietary format. Some products have
import interfaces, for instance, Microsoft Excel
spreadsheet structures.
GIS users aiming to analyze numerous input
data layers are faced with the organization and
assessment of a sizable number of individual
records. The identification of the spatial relation
of the records (i.e., data layers) enables the GIS
to derive information at any defined point and,
thus, to define training data for the ANN across
all data layers for the same location.
Up to now, the use of ANN to analyze spatial
data has not been widespread in geosciences, due
to the lack of user-friendly software that runs
together with a standard GIS environment. Based
on this observation, a software application was
developed and used in this study which provides
users with tools to evaluate their spatial data with
ANN within the ESRI ArcGIS environment.
This software, called advangeo®, is based on a
client/server architecture and is comprised of two
components, namely “data management” and “cli-
ent application“. The “application meta database”
manages the project data, which includes calcula-
tions, embedded data, parameterized models and
scenarios. Metadata are stored in a Microsoft SQL
Server database (Figure 2).
The “project database” contains all of the
geodata and feature data for the build-up of the
statistical models.
The Data and Model Explorer (a stand-alone
solution) provides support to the user as new
projects are set up and assists in the systematic
parameterization of the models. The workflow is
fully documented. The additional ArcGIS exten-
sion provides all of the necessary GIS function-
alities in a user-friendly interface to facilitate the
geodata with the ANN. Besides the availability
of various functions to pre-process geodata, this
extension also supports the automatic creation of
output map layouts.
The setup of a consistent and reproducible
statistical model for geodata analysis consists of
a number of procedures, which formed the basis
for the design and development of this software.
This included multiple stages, which involve:
1. The exact formulation of the modeling objec-
tive and the delineation of the study area;
2. Describing and compiling base data, which
represent the potentially influential factors
with which predictions will be made in
consideration of,
3. Technical data processing (from the sci-
entific and it point of view) and statistical
pre-processing;
4. The setup and calculation of model scenarios
for training, testing and applications;
5. Statistical post-processing and presentation
of prediction results.
The Data and Model Explorer is the primary
user interface, where the main object types may
be defined for modeling, organized with their
respective metadata sets, modified and visualized.
The processing of geodata and the presentation of
769
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
Figure 2. System Architecture of advangeo ®
subsequent results are supported by a GIS exten-
sion with a special toolbar and a project tree view
option. The system architecture is illustrated in
Figure 2.
Project definition is the first step of the mod-
eling process, and appears at the top of the Data
Explorer hierarchy. The levels are displayed in a
tree view structure for easy navigation. By defin-
ing a project, the user sets the technical boundary
of the intended task. The study, training and test
areas are delineated subsequently.
The basis of every modeling task is founded
upon the ability to use various source data types.
However, data are generally unusable without prior
processing involving the conversion to a standard
or recognizable format for a given application.
The original data are therefore unlikely to be used
directly for calculations, but can be included in
the project for documentation or reference pur-
poses. Necessary processing procedures prior
to modeling can include the calculation of field
values, error identification and elimination, code
770
adaption, scale transformations or data partition-
ing. The “processed source data” are transformed
and saved as spatially compatible raster layers.
Taking the coverage extents of the project area
into consideration, the data are transformed to
“model input data” layers. This process is required
to ensure a spatially consistent set of base data.
Following the data processing procedures, a
parameterized model is set up and used for the
calculation. Model input data, training and model
parameters are then determined in a training sce-
nario (Figure 3). Currently, the MLP approach is
used in advangeo®, along with various adaptable
training algorithms and parameters. This proce-
dure was realized with the open source library
FANN (Nissen, 2009).
The user may opt to retain the standard default
values at the beginning of the model parameter-
ization stage. The model calculation results in a
trained network within the defined training area.
The initial run returns a preliminary estimation
of the model quality. The GIS extension provides
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
Figure 3. Training scenario with advangeo ®
further functions for the statistical validation of
model results.
The trained network can then be applied to
other study areas. The calculations may also be
displayed in a semi-automatically generated pre-
diction map within the GIS environment.
The special new and innovative aspects ad-
vangeo® introduces in ANN processing are the
integrated GIS data management in conjunction
within the standard GIS environment of ESRI
ArcGIS and various tools to pre-process data,
tools to assess the model quality and a customized
map printing function (Figure 9).
CASE STUDY 1: EROSION
PATTERN PREDICTION
The prediction of erosion caused by rainfall was
conducted to evaluate the quality of modeling
results on two different sites, the Limpopo region
in South Africa and the Weisseritz catchment in
the South of Eastern Germany.
In both areas, precipitation events character-
ized by long durations and/or high intensities lead
to the dislocation of soil material, especially in
agricultural fields. Topsoil, which contains the
greatest amount of nutrients, is most vulnerable
to dislocation; seeds may also be carried away,
thereby significantly reducing the potential yield
of the land. These types of adverse effects are
known as “on-site damages”. “Offsite-damages”,
on the contrary, occur where soil is deposited or
accumulated and can spoil water bodies and cause
damages to human infrastructure.
The prediction of the probabilities of erosion
event occurrences, along with a quantitative es-
timation of soil dislocation, forms an important
basis for the planning of preventative or mitiga-
tive measures. Analytical equations to predict the
amount of dislocated soil have been derived and
used for many years (e.g., Wischmeier & Smith,
1978; de Vries et al., 1998). Other soil parameters
have already been predicted successfully with
ANN and the multiplayer perceptron in particu-
lar recently by Behrens et al. (2005), Alavi et al.
771
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS

(2010), and, in case of soil contamination pattern,
by Hertwig et al. (2010).
The goal of this case study was to test the ac-
curacy of the prediction of rill and gully erosion
as a result of intense rainfall with ANN.
Source Data and Pre-Processing
The first project area was located in the Weis-
seritz catchment in Saxony, Germany. The main
controlling factors governing the soil erosion
prediction are well known for the study area, due
to a number of existing algorithms developed for
the parameterization of mathematical models.
Factors include:
• Terrain attributes (slope and flow accumu-
lation; to a lesser extent aspect, profile and
plan curvature and flow length);
• Percent composition of top soil (i.e., soil
particle size – fine, medium, coarse); and
• Land use (e.g., arable land, pastures, for-
ests, urban areas).
Data sources comprised of a digital elevation
model (DEM) with 20 m raster size (DGM 25),
the soil concept map of the Free State of Saxony
(scale 1:25,000) and ATKIS land use data in the
same scale. These data layers were pre-processed
with selected GIS tools to create consistent raster
layers. In order to include the influence of linear
elements such as roads and hedgerows, which
function as barriers to erosional processes, an
“apparent enhancement” of the spatial resolution
was realized by resampling the data to a 5 m grid.
Metric data layers were then normalized to fall
between ranges of [0…1] or [-1…1]. Nominal
scales for layers containing data on the presences
of forests, roads and geological units were reclas-
sified as binary layers.
Training data were manually digitized from
scanned orthophotos, which were taken after the
severe rainfalls of August 2002. The accuracy
of the data was also partially verified during site
visits (Figure 4).
The combination of slope and flow accumula-
tion explained most of the observed erosion pat-
terns. Other terrain attributes (derived from the
DEM) did not enhance the results significantly
whereas top soil composition and land use (arable
land, pastures, forest, urban areas) contributed to
enhanced results.
The second study site was located in the
northeastern region of South Africa. There exist
severe problems for agriculture and transport
infrastructures due to the formation of deep ero-
sion gullies. Comparable data sources to the ones
used in the first study site were also used in this
second analysis of causal relationships, especially
amongst multiple datasets, namely geological data,
field data on existing gullies, remote sensing and
flow accumulation data (derived from the DEM).
The land cover datasets were determined to be
unsuitable for modeling upon visual inspection
as the raster was only available as a 30 m grid.
The DEM was interpolated from a 90 m SRTM
(Shuttle Radar Topography Mission) raster. Vari-
ous terrain attributes were calculated and tested.
For an overview of standard DEM derivations,
please see Wilson and Gallant (2000). The data
sources, which were selected for the (definition
of the) final network, included a geological map,
Figure 4. Training area, field validation

772
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
elevation, slope angle and flow accumulation.
Figure 5 illustrates the types of input data used
and the neural network calculation scheme.
Network Parameterization
The pre-processed data were used as input data
for the model calculation. The models for both
areas were defined with the procedures described
previously and trained with existing influencing
factors. Training was completed in a part of the
defined study area. For modeling, a MLP approach
was used. It was trained with various learning
algorithms, activation functions and parameters.
Acceptable results were attained with the Resil-
ient Back Propagation (RPROP) algorithm, in
combination with a sigmoid activation function.
Validation
The training was repeated several times with the
same parameters. Comparable results (i.e., with
only slight deviations, acquired following a simi-
larly defined training process) were observed. This
confirmed the reproducibility and robustness of
Figure 5. Neural network calculation scheme for erosion gully prediction
the network and is an important parameter for the
assessment of the modeling result. The trained net-
work was then applied on a test site and validated.
The network accurately identified approximately
90% of the test records at Weisseritz catchment.
The threshold for the decision “erosion/no ero-
sion” was set to p > 0.75 and was derived from
statistical calculations of confirmed erosion sites.
This result was confirmed in the field. Figure 6
shows water logging after high precipitation as
well as erosion channels at locations of dark cell
values (high erosion probability).
For the Limpopo study site (Figure 8), visual
validation was conducted, based on satellite im-
ages which clearly showed existing erosion gullies.
The satellite data were not included as a part of
the training data. Relevant gullies were still iden-
tified at the correct location despite the rather low
spatial resolution of 30 m.
Planning of Preventative Measures
and Presentation of Results
The trained network was also used to plan for
the introduction and implementation of erosion
773
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
Figure 6. Visual field inspection of predicted results
prevention measures, especially at Weisseritz
study site. Additional barriers (landscape elements
such as hedgerows) were added in the model on
slopes that were especially susceptible to erosional
processes. This was accomplished by modifying
the input layers. Figure 7 shows that the erosion
probability (dark cell values, dominated by the
factors flow accumulation and slope on arable
land) is significantly reduced by introducing three
barriers to cut of the major flow lines parallel to
Figure 7. Mitigation of erosion processes by landscape elements (barriers) on a southwest exposed slope
774
the elevation contours. The subsequent network
run confirmed the successful mitigation of the
predicted erosion processes. With this approach,
prevention measures could be optimized (number
and location of barriers), thereby facilitating the
conservation of funds.
The integrated function for automatic map
generation helped to display the prediction results
in a user-friendly way. Figure 9 is an example of
a prediction map with the following content: ero-
sion channel probability (in color) on top of the
DEM, hillshade display (top left), and on top of
the aerial image (top right). The bottom line rep-
resents detailed maps on the same background
data (as indicated in the overview map). The
right-hand side diagrams explain the used factors
of influence and display the model calibration
error (training function). The overall erosion
probability is given in a color scale and values
from 0 – 1.
CASE STUDY 2: PREDICTION OF
MINERAL RESOURCES
The ability of ANN to analyze complex, non-linear
combinations of multiple (raster) input data was
tested for the interpretation of mineral occurrences,
in order to generate high quality prediction maps
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS

Figure 8. Visual comparison of erosion gullies on satellite imagery and with ANN predicted gully loca-
tion (image authorized by Council of Geoscience, South Africa)

of mineral resources. Such maps are crucial tools,
especially for developing countries, where positive
results may be used to attract investment interest
in their respective mineral sectors. The identifi-
cation of exploration targets is one of the most
ambitious tasks in geology, as it is instrumental
to the strategic development of companies and
sometimes even countries and is generally tied
to significant amounts of financial resources. The
successful prediction of mineral occurrences plays
a key role in the selection of further exploration
tasks and defines a basis for which investment
decisions can be made in many countries.
The calibration of quantitative analytical
models generally involves substantial temporal re-
sources and cost-intensive field or laboratory work.

Figure 9. Prediction Map for soil erosion probability (original in color, indicating high erosion prob-
ability in red lines/areas)

775
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
Currently, analytical models cannot adequately
describe the complex interactions of geological,
tectonic, geochemical, geo-mechanical and other
factors in the timescale that is required to form
mineral deposits. Only single aspects of it can
be modeled with acceptable levels of precision
when there are fixed boundary conditions (e.g.,
how ore minerals attach to geochemical barriers).
In this case study, input data layers were
selected based on existing knowledge about in-
fluencing factors of resource genesis, which was
derived from geological maps. The study area of
11.000 km² was located in Kosovo in southeastern
Europe. Promising targets for in-depth exploration
could be identified at this scale of investigation.
In addition to the known deposits of lead, zinc,
bauxite, chromium, nickel and coal, the results
also indicated significant potential for gold and
platinum metals.
Source Data and Pre-Processing
Between 2003 and 2007, existing geo-scientific
data were compiled, reviewed, completed and
transformed into a consistent set of digital maps,
under the supervision of the “Independent Com-
mission for Mines and Minerals” (ICMM). All
necessary data layers for the prediction of mineral
resources (especially geology, tectonics, known
deposits, metallogeny, aero-geophysics) were
made to be available, in the scales of 1:100,000
Figure 10. Pre-processing of tectonic data – calculation of distances to potentially relevant lineaments/
faults
776
and 1:200,000, respectively. For modelling pur-
poses, a raster cell size of 50 m was estimated to
be sufficient, resulting in approximately 4,400,000
cell values within the Kosovo administrative
boundaries.
The structural and litho logic relationships
within the lead/zinc deposits were presented by
Anković et al. (2003) and further explored in a
study by Legler et al. (2008). The main controlling
factors for the localization of ore deposits included:
• The striking tectonic elements in the NNW
– SSE direction;
• The proximity of Cenozoic vulcanite
rocks; and
• The presence of highly reactive host rocks,
such as marble and calcareous schist.
Selected influencing factors were derived from
the geological map and transformed to a 50 m grid
(see example of tectonic data preparation in Figure
10). In the cases involving the consideration of
lineaments/faults and Cenozoic vulcanite rocks,
the distance from each cell to the nearest relevant
structure was calculated, which functions as a
viable potential indicator of deposits.
To include high-resolution aero-geophysical
data into the assessment, the following base da-
tasets were identified to be crucial for the gen-
eration of the prediction map:
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS

• Known deposits and their occurrences,
• Tectonic data (lineaments and faults),
• Aero-geophysical data (gamma spectrom-
etry of elements u, th and k, magnetics, and
electromagnetics), and
• Lithological data.
Network Parameterization
The training stage was completed for the test area
illustrated in Figure 11. For the cross-validation
stage, a subset of the known deposits within the
test area was used for a second training scenario,
while the remaining known deposit locations,
which had been removed beforehand, were used
to validate the trained results.
The training results were further assessed by
an iterative process that involved the consideration
of a greater number of input layers. In this case,
further investigation began with 3 layers (aero-
radiometric layers Uranium (U), Thorium (Th)
and Potassium (K)) and increased to the use of 6
layers, which included volcanic, tectonic and
electromagnetic data. For modeling, the MLP
approach was applied.
Validation
Training was repeated several times with the same
parameters, which produced almost identical
results to previous training runs. These results
gave insight into how robust and reproducible
the network-based training runs were. The trained

Figure 11. Training area, used layers (aero-radiometric surveys of Uranium, Thorium, Potassium;
distance to volcanic rocks) and known deposits (dots)

777
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
network was then applied to the full extent of
Kosovo, with the exception of the northern and
northwestern border where no geophysical data
were available. Some known deposits, identified
on the prediction map with probabilities of less
than 0.5, were not used for the training exercise.
Presentation of Results
The results engendered from the modeling stage
were semi-automatically transformed to the GIS,
where a prediction map could be made. This kind
of map can be used to display the modeled results
in a manner that can be easily used by potential
investors in the mineral sector or other interested
parties (Mining Journal, 2009).
CASE STUDY 3: FOREST PEST
VULNERABILITY ANALYSIS
Mass reproduction of the European Spruce Bark
Beetle (Ips typographus) can lead to severe for-
est dieback, since their larvae feed on the inner
bark of trees at different life stages and in various
degrees of deterioration. Knowledge of control-
ling factors pertaining to the reproductive cycle
of I. typographus is crucial for successful forest
management and for the planning of mitigation
measures. The beetles mainly affect Norway
Spruce (Picea abies), and especially older ho-
mogenous stands. Significant influencing fac-
tors for the (mass) dispersion of this bark beetle
include climate parameters such as temperature
sequences along with wind strength and direction,
as well as site parameters such as the percentage
of spruce present, stand age and density. In addi-
tion, topographical attributes such as terrain and
soil conditions can influence the development of
I. typographus populations. The goal of this study
was to test, if the ANN approach with advangeo® is
suitable for the spatial prediction of how probable
a specific site is susceptible to future infections
778
and the mass reproduction of the aforementioned
beetle species. Various approaches and models
have been available for the analysis and predic-
tion of the spread of I. typographus, as described
in Baier et al. (2007), Seidl et al. (2007), and
Netherer and Nopp-Mayr (2005).
Source Data and Pre-Processing
The study site was located in Saxony, Germany.
Source data comprised of a digital elevation model
(DEM) in 20 m resolution and ecological forest
site map at 1:10,000 with soil and soil water data,
and forest stand data on species composition,
stand age, and height. For the ANN training,
known locations of I. typographus infections
from the years 2003 and 2008 were available as
field data for the site of interest. The data layers
were pre-processed based on the knowledge on
influential factors (e.g., elevation and terrain at-
tributes such as slope, aspect, curvature and the
topographic wetness index) and were subjected to
a scale transformation. Soil data were processed to
generate maps differentiating between 1) fine and
coarse soil particle sizes and 2) wetness classes.
A statistical evaluation of 16 site parameters
and three forest stand parameters was conducted
prior to the processing. Histograms were produced
at the end of the evaluation, which helped to de-
lineate data ranges and were used to estimate the
potential influence of aforementioned factors, with
the consideration of previous stand infections by
I. typographus in 2003 and 2008.
Highly correlated and potentially influential
factors were identified to reduce the number of fac-
tors for subsequent training stages. For example,
flow length and flow accumulation were found
to be directly correlated (r=0.92). Therefore, they
were not used simultaneously as input layers, in
order to avoid distortions or bias in the results.
Furthermore, a logarithmic function was applied to
the flow accumulation, which was notably skewed
to the left, to approximate a standard distribution
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
of the range of its values. Data normalization
to ranges [0..1] and [-1..1] was carried out for
most layers. Ordinal and nominal scales were
transformed to binary grids. This is exemplified
in the extraction of soil type data from maps and
their conversion into single binary layers, where
values of 1 or 0 were assigned for the presence
or the non-existence of certain attributes (e.g.,
gley: yes/no or podsol: yes/no). This means of
reclassifying the data facilitated the evaluation of
their significance as potential factors of influence.
Network Parameterization
The training algorithm was executed for the
“Tharandter Wald” located in Saxony, as well
as within parts of the study area. The model was
systematically set up, with various input layers
representing factors of influence by using MLP
coupled with a RPROP algorithm and a sigmoid
activation function.
Validation
To validate the results, several parameters were
evaluated. First, the rate of matches was calculated
(i.e., the percentage of predicted sites to known
sites of infection not used for training). In addition,
the “prediction error sum of squares” (PRESS)
was calculated. The determination of the quadratic
difference of the values (i.e., known infections and
predicted results) at infected locations described
the quality of the predictions.
In addition, a fourfold cross-validation as-
sessment was conducted for the model. For this
purpose, the training data were separated into two
exclusionary subsets (training and test subsets)
and used as input data on a rotational basis, until
each element was a part of the test subset for at
least one time. The results showed that the model
outputs were notably similar, which indicates that
a successful learning process occurred during the
network training stage.
The correlation between predicted and known
values at infected locations raised some questions
about the modelling quality. It was observed
that 60-70% of the infected locations retained
an infection probability at 0.5. There were few
probabilities > 0.8 at all. However, by visual
inspection, the general spatial pattern showed a
more accurate prediction trend. Results for 2003
(a rather dry year) were significantly better than
those for 2008.
The low rates of matches between the two
datasets were caused by the validity and ac-
cessibility of data on the positional accuracy of
the training points (i.e., infected stands). This
in turn, was primarily influenced by different
scales at which data were collected and made to
be available for use. Additional factors that could
significantly enhance the model quality include
information about previous infections and forest
management measures, more detailed forest stand
parameters, information about neighboring stands
and wetness. Another aspect was the absence of
important climatic parameters such as duration
of insolation, wind direction and precipitation,
which may explain the differences of the results
obtained between 2003 and 2008.
A common parameter set (in terms of potential
factors of influence) for the 2003 and 2008 data
turned out to be inefficient. The different envi-
ronmental conditions, particularly differences in
the climate, notably influenced the behavior of the
beetle. It is hypothesized that the period of drought
in 2003 extended the range of beetle dispersion
into previously healthy forest stands (i.e., stress
caused by droughts led to higher susceptibility
to infections).
In the beginning of this study, a grid with an
overall resolution of 5 m was used. Later on, a
resolution of 20 m was determined to be sufficient,
considering the quality of data that was available.
This change in grid size was reflected in the cal-
culation time and the output quality.
779
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
Presentation of Results
In order to compare the model output, the same
classes were used to describe the probability of
infection for each model run, where:
• No susceptibility: 0 – 0.5
• Low susceptibility: >= 0.5 – 0.6
• Average susceptibility: >= 0.6 – 0.7
• High susceptibility: >= 0.7 – 0.8
• Very high susceptibility: >= 0.8 (dark val-
ues on the color scale)
Figure 12 shows one the model output with the
susceptibility (probability of infection) as a mean
value of 15 model runs. Dark values indicate the
highest of 5 probability classes (>= 0.8) and al-
low a quick overview to target actions in case of
starting mass reproduction of the beetles. White
areas represent other land use classes (villages,
pastures, arable land, streams and forest roads.
Finally, a quantitative prediction was calcu-
lated (in m³/infected location). Using the input
layers for 2008, the following parameters were
Figure 12. Mean values of 15 model runs with 5 probability classes (from very high to no susceptibility
to infections by I. typographus)
780
selected as the best (most representative) input
datasets:
• Percentage of spruce (species composition
of the forest stand)
• Forest stand age
• Soil composition
◦◦ Percentage of coarse soil
◦◦ Percentage of silt percentage of sand
• Appearance of gley soils
• Soil moisture levels (soil type)
• Slope angle
• Slope position
• Aspect
CONCLUSION
The case studies demonstrated that the multilayer-
perceptron approach (MLP) is suitable for the
analysis and prediction of various applications
of spatially-dependent natural phenomenon,
including the assessment of soil erosion, the
identification of potential mineral occurrences
and the spread of forest pests. The stepwise
Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
generation of model scenarios with advangeo®
supported the derivation of information while
taking the sensitivity of controlling factors into
consideration. However, one disadvantage to this
approach is its inadequate ability to explain results
(i.e., comparable to a “black box” system), where
the specific influences of single parameters cannot
be directly retraced.
All interrelations amongst factors of influence
and the final result were revealed through the
course of the training process. With this process,
“erroneous learning” cannot be excluded. As a
result, it remains a challenge to define the bound-
aries of neural network applications. To evaluate
the quality of individual components of training
goals and ability to generalize the input events,
“expert knowledge” and the means to critically
assess the outputs of the modelling process are
required, in addition to the generation of statisti-
cal figures on the proximity of predicted and
measured values alone.
Overall, the ANN defined within the advan-
geo® software seemed to be capable of predict-
ing the occurrence of real spatial phenomena.
Further potential fields of applications, which the
developers intend to test in near future, include
the assessment of slope stability, the prediction
of market trends, and the evaluation of traffic
accidents, in addition to the mapping of spatial
events in forestry, agriculture and mining.
The most significant advantage of the ANN is
the relatively small amount of effort required for
data pre-processing, model calibration and short
model run times. As such, this new tool enables
standard GIS users to apply ANN approaches that
are fully integrated with their already familiar
work environment.
ACKNOWLEDGMENT
The authors sincerely thank the Independent
Commission for Mines and Minerals (ICMM) in
Kosovo, the Council of Geoscience (CGS) in South
Africa and the Staatsbetrieb Sachsenforst (State
Agency for Forestry of Saxony) in Germany for
their expressed authorization to present proprietary
project data and figures in this article. The research
project was supported by the Federal Ministry of
Economics and Technology, Germany (project
identification number IW072061).
REFERENCES
Alavi, A. H., Gandomi, A. H., Mollahassani, A.,
Heshmati, A. A., & Rashed, A. (2010). Modeling
of maximum dry density and optimum moisture
content of stabilized soil using artificial neural
networks. Journal of Plant Nutrition and Soil Sci-
ence, 173, 368–379. doi:10.1002/jpln.200800233
Anković, S., Jelenković, R., & Vujić, S. (2003).
Mineral resources and potential prognosis of
metallic and nonmetallic mineral raw materials
in Serbia and Montenegro at the end of the 20th
century. Belgrade, Serbia: Engineering Academy
of Serbia and Montenegro.
Backhaus, K., Erichson, B., Plinke, W., & Weiber,
R. (2003). Multivariate analysemethoden – eine
anwendungsorientierte Einführung. Berlin, Ger-
many: Springer-Verlag.
Baier, P., Pennerstorfer, J., & Schopf, A. (2007).
PHENIPS – A comprehensive phenology model
of Ips typographus (L.) (Col., Scolytinae) as a tool
for hazard rating of bark beetle infestation. For-
est Ecology and Management, 249(3), 171–186.
doi:10.1016/j.foreco.2007.05.020
Behrens, T., Förster, H., Scholten, T., Steinrücken,
U., Spies, E., & Goldschmitt, M. (2005). Digital
soil mapping using artificial neural networks.
Journal of Plant Nutrition and Soil Science, 168,
21–33. doi:10.1002/jpln.200421414
781
 Spatial Modeling of Natural Phenomena and Events with Artificial Neural Networks and GIS
Fernandez-Steeger, T. M. (2002). Erkennung von
hangrutschungssystemen mit neuronalen netzen
als grundlage für georisikoanalysen. Unpublished
doctoral dissertation, Universität Karlsruhe, Karl-
sruhe, Germany.
Haykins, S. (1998). Neural networks: A compre-
hensive foundation (2nd ed.). Upper Saddle River,
NJ: Prentice Hall.
Hertwig, T., Müller, I., & Zeißler, K.-O. (2010).
Management of contaminated soils in urban
areas in the ore mountains (Germany). Paper
presented at the 11th International Conference on
Management of Soil, Groundwater and Sediment,
Salzburg, Austria.
Journal, M. (2009). Kosovo ready to roll. Mining
Journal, 17(7), 14–15.
Kriesel, D. (2009). A brief introduction to neural
networks. Retrieved from http://www.dkriesel.
com
Legler, C., Knobloch, A., & Barth, A. (2008). Map
of minerals – Metallogenic / minerogenic map
(1: 200,000) - Final report – Map description.
Freiberg, Germany: Beak Consultants GmbH,
Prishtinë.
Lin, S.-W., Sun, C.-H., & Chen, C.-H. (2004).
Temporal data mining using genetic algorithm
and neural network: A case study of air pollutant
forecasts. Geospatial Information Science, 7(1),
31–38. doi:10.1007/BF02826674
Netherer, S., & Nopp-Mayr, U. (2005). Predis-
position assessment systems (PAS) as supportive
tools in forest management – Rating of site and
stand-related hazards of bark beetle infestation
in the High Tatra Mountains as an example for
system application and verification. Forest Ecol-
ogy and Management, 207, 99–107. doi:10.1016/j.
foreco.2004.10.020
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 1,
edited by Donald Patrick Albert, pp. 1-20, copyright 2012 by IGI Publishing (an imprint of IGI Global).
782
Nissen, S. (2009). Fast artificial neural network
library (FANN). Retrieved from http://leenissen.
dk/fann/
Penning de Vries, F. W. T., Agus, F., & Kerr, J.
(Eds.). (1998). Soil erosion at multiple scales:
Principles and methods for assessing causes and
impacts. Wallingford, UK: CABI Publishing.
Peters, R., Schmitz, G., & Cullmann, J. (2006).
Flood routing modelling with artificial neural
networks. Advances in Geosciences, 131–136.
doi:10.5194/adgeo-9-131-2006
Redford, S., Lipton, G., & Ugalde, H. (2004). Pre-
dictive ore deposit targeting using neural network
analysis. Society of Exploration Geophysicists
Expanded Abstracts, 23, 1198.
Rios, D. (2010). Neural networks: A requirement
for intelligent systems. Retrieved from http://www.
learnartificialneuralnetworks.com/backpropaga-
tion.html
Rumelhart, D. E., Hinton, G. E., & Williams, R.
J. (1986). Learning internal representations by
error propagation. In D. E. Rumelhart & J. L. Mc-
Clelland (Eds.), Parallel distributed processing:
Explorations in the microstructure of cognition,
volume 1: Foundations (pp. 318-362). Cambridge,
MA: MIT Press.
Seidl, R., Baier, P., Rammer, W., Schopf, A.,
& Lexer, M. J. (2007). Modelling tree mortal-
ity by bark beetle infestation in Norway spruce
forests. Ecological Modelling, 206, 383–399.
doi:10.1016/j.ecolmodel.2007.04.002
Wilson, J. P., & Gallant, J. C. (Eds.). (2000). Ter-
rain analysis: Principles and applications. New
York, NY: John Wiley & Sons.
Wischmeier, W. H., & Smith, D. D. (1978). Pre-
dicting rainfall erosion losses – a guide to conser-
vation planning. USDA Agriculture Handbook,
537, 1–58.
 783

Chapter 48
Stereo-Vision-Based Fire
Detection and Suppression
Robot for Buildings
Chao-Ching Ho
National Yunlin University of Science and Technology, Taiwan

ABSTRACT
A stereo-vision-based fire detection and suppression robot with an intelligent processing algorithm for
use in large spaces is proposed in this chapter. The successive processing steps of our real-time algorithm
use the motion segmentation algorithm to register the possible position of a fire flame in a video; the
real-time algorithm then analyzes the spectral, spatial, and motion orientation characteristics of the fire
flame regions from the image sequences of the video. The characterization of a fire flame was carried
out by using a heuristic method to determine the potential fire flame candidate region. The fire-fighting
robot uses stereo vision generated by means of two calibrated cameras to acquire images of the fire flame
and applies the continuously adaptive mean shift (CAMSHIFT) vision-tracking algorithm to provide
feedback on the real-time position of the fire flame with a high frame rate. Experimental results showed
that the stereo-vision-based mobile robot was able to successfully complete a fire-extinguishing task.

INTRODUCTION tional point-type detectors are installed on walls

Fire incidents can cause loss of lives and damage
to property. Damage due to fire has always been a
major area of concern for museums, warehouses,
and residential buildings. Conventional fire de-
tection sensors (e.g., ionization and photoelectric
detectors) and fire sprinkler systems monitor only
particular points in space. In most cases, conven-
DOI: 10.4018/978-1-4666-2038-4.ch048
or on a ceiling. The delays in the activation of fire
detection sensors and sprinklers in large spaces
are a major problem. Hence, the monitoring ca-
pabilities of point-type sensor devices are limited
to a certain distance, and they are ineffective for
monitoring large areas. These devices are not suf-
ficiently flexible to detect fire incidents, and many
fire-detection sensors and sprinklers are required
to be installed very close to the monitoring areas.
Comparatively, the video camera is a volume

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
sensor, and potentially monitors a larger area
and has a much higher probability of successful
early detection of fire flames. Video surveillance
technology is suitable for early detection of fires
due to its low detection delay, good resolution,
and high localization accuracy. Early detection of
fires can certainly expedite fire-fighting efforts,
and consequently, fires can be extinguished be-
fore they spread to other areas. To monitor large
spaces, the use of a mobile fire-fighting robot is
a more flexible alternative than installing a large
number of detectors and sprinklers. When a fire
is detected, the fire-fighting robot can move to the
position of the fire flame and safely evacuate an
object from the fire area. Stereo vision systems
can provide the robot with precise depth informa-
tion about a target. Hence, the use of two cameras
instead of one increases the suppression efficiency
and adaptability of the robot while detecting and
evacuating a burning object.
BACKGROUND
In recent times, research on the detection of fire
flames using surveillance cameras with machine
vision has gained momentum. The image process-
ing approach involves the extraction of the fire
flame pixels from a background by using frame
difference technologies. Healey et al. (1993)
presented a fire detection algorithm using a color
video input with a pre-partitioning scheme under
some restricted conditions, without rejecting the
similar fire-like alias. Phillips et al. (2002) and
Celik et al. (2007) conducted studies on computer
vision by using spectral analysis and the flicker-
ing property of fire flame pixels to recognize the
existence of fires at a scene. Hue and saturation
are adopted as feature vectors to extract the fire
pixels from the visual images (Chen, 2003). Fire
flame features based on the HSI(hue, saturation,
intensity) color model are extracted, and regions
with fire-like colors are roughly separated from
the image by the color separation method (Horng,
784
2005). Then, the image difference method based
on chromatics is used to remove spurious fire-like
regions such as objects with similar fire colors or
areas reflected from fire flames. A fuzzy-based
dominant flame color lookup table is created, and
fire regions are automatically selected (Wang,
2006). However, either the solution does not
consider the temporal variation of flames or the
approach is too complicated to process in real time.
Fire suppression systems usually use water to
extinguish fires due to its good ability to suppress
fire. Chen et al. (2004) developed a water-spraying-
based fire suppression system. The fire searching
method is realized based on computer vision theory
using one CCD camera that is installed at the end
of a fire monitor chamber. However, it is necessary
to calculate the changes in the space coordinates
of the fire with displacement and the pivot angle
of the CCD camera in the fire searching process.
Ho (2009) proposes a fire-tracking scheme based
on CAMSHIFT. The CAMSHIFT algorithm is
applied to track the trajectory and compute the
2D positions of the specified moving fire-fighting
robot in real time with the aid of a vision system.
Yuan (2010) adopted the computer vision tech-
niques to extract color and motion characteristics
for real-time fire detection. However, the system
was designed to move a water gun along a fixed
path using computer-based control. Hence, the
monitor ranges of the scene are limited and not
sufficiently flexible.
FIRE DETECTION AND
SUPPRESSION ROBOT
Issues, Controversies, Problems
The basic problems associated with conventional
fire detection sensors are that they are not suffi-
ciently reliable due to the time delay between the
start of the fire and nonzero input in the detection
sensor (Podržaj, 2008). A fire-fighting robot,
equipped with vision-based fire detection technol-
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
ogy, is capable of eliminating the conventional
sensor delay factor and providing fire detection
coverage for a larger area with minimized cost. The
overall objective of the mobile robot system is to
develop an autonomous fire detection technology
that is capable of identifying fires at early stages,
alerting fire staff, and extinguishing the fire.
To enable fire detection by the robot, the main
objective here is to develop a color vision-based
fire detection system that is accurate and per-
forms efficiently. The image processing approach
involves the extraction of the fire flame pixels
from a background by using frame difference
technologies. To achieve the segmentation of fire
features, color processing is a better alternative
than gray-scale processing. Color processing
can avoid the generation of false alarms due to
the variations in lighting conditions (e.g., natural
background illumination) better than gray-scale
processing. However, vision-based fire flame
detection still has great technical challenges. Fire
flames are non-rigid objects and do not exhibit
primitive image features and variability in density,
lighting, etc. (Ho, 2009).
For robot navigation and fire extinguishing,
robots with machine vision have been attracting
substantial attention recently. A visual feedback
mechanism commonly used in a robot system is
the machine vision measuring system capable
of non-contacting measurement; this type of
measurement is advantageous because the actual
environment does not always allow contact with
the surface at which the measurement is performed.
One common problem with the acquired images
is the accompanying noises, namely, disturbances
caused by irrelevant objects in the background or
foreground or by substandard illumination. As a
visual tracking system must be capable of record-
ing image features in real time, another problem
with most computer-based vision systems is their
limited performance due to inadequate comput-
ing power for motion tracking after processing
resources, including the memory, have been allo-
cated to the tasks of feature extraction and template
pattern matching, which accounts for the handling
of the disturbances mentioned above (Sumi, 1995;
Di Stefano, 2003). Due to the high cost of im-
age capturing equipment and image processing
components, the need for substantial computing
resources by visual servoing algorithms is not
often taken into consideration. Despite expected
improvements on the performance of visual servo
with image processing and control components
operating on separate processors, the additional
cost involved with the extra components presents
yet another problem.
SOLUTIONS AND
RECOMMENDATIONS
Image Capturing
Image capturing and processing are the two major
challenges in the construction of a stereo visual
tracking system. The main goals of these respec-
tive processes are to provide visual feedback
and identify the objects one wants to track (in
this case a burning target), while simultaneously
determining their 3D positions. In the proposed
stereo tracking system, the video-signal-capturing
process is conducted by stereo sensors, which are
two low-cost CMOS cameras with pixel resolu-
tions of 320 x 240. The captured synchronized
concurrent video frames are then transmitted
via USB to a PC, the image processor, and then
buffered in the PC’s system memory. The frames
are eventually displayed in the capture windows
using the PC’s windows driver model (WDM)
functions. The windows driver model can reach
a high video frame rate, while the two CMOS
cameras can also achieve a frame rate of 30 im-
ages per second; therefore, real-time tracking can
be achieved.
785
 Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
Calibration of Image-Based System
It might be expected that after the raw images are
obtained, the next step would be to process them.
However, camera calibration and robot hand-eye
transformation are two fundamental steps that need
to be taken into consideration before any reliable
image processing or even further visual servoing
can be performed. To obtain the 3D position of an
object portrayed in 2D images, the relationship
between the coordinates of the points in a 3D
frame, in this paper the tank, and the coordinates
of their corresponding points projected onto the 2D
imagery plains must be found. The first step toward
this goal is recognizing the basic characteristics
of the cameras used to retrieve the images. These
characteristics, generally grouped into intrinsic
and extrinsic parameters, are shown in Figure 1.
Camera calibration can be used to obtain the
intrinsic and extrinsic parameters of the camera.
The intrinsic parameters, which are independent
of a camera’s position in the physical environment,
describe the camera’s focal length (fx, fy), princi-
pal point (Cx, Cy), and distortion coefficients. On
the other hand, the extrinsic parameters offer
information on the transformation between the
Figure 1. The robot is calibrated to recognize the intrinsic and extrinsic parameters of the camera
786
coordinate systems of the camera and the concrete
world, including a 3D translation vector t that
provides the translational components and a rota-
tion matrix R that provides the rotational compo-
nents. For stereo camera systems, it is also neces-
sary to obtain the relative extrinsic parameters
between the two cameras. The combination of a
camera’s, or an array of cameras’, intrinsic and
extrinsic parameters presents the full set of data
needed to locate, in the outside world, the cor-
responding position of a point in images taken by
the camera(s). M.-C. Villa-Uriol (2004) and
Hutchinson (2006) provided a complete review
on the fundamentals and techniques of camera
calibration.
In this work, static pattern calibration is ap-
plied to attain the cameras’ two sets of parameters
and empower automatic edge detection. The
key to this calibration process is the use of a flat
checkerboard with a known geometric pattern.
The dimensions of each square on the board are
predetermined. Points on the board model plane
and their projections in digital images taken by the
to-be-calibrated cameras are passed as parameters
to the intrinsic calibration routine.
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
The extrinsic calibration methods used here
also employ digital checker-patterned board im-
ages. Assume that for every point M in the physical
world reference frame we have m = RM + t, where
m is the 3D coordinates of the point’s counterpart
in the camera reference frame, while R and t are
the rotational and translational matrices, respec-
tively, for the coordinate transformation between
the physical world and camera reference frames.
The extrinsic parameters of the to-be-calibrated
cameras can be calculated linearly with the help
of one array that stores the coordinates of chosen
reference points in the digital checkerboard images
and another array that contains the correspond-
ing points in the physical world. The collection
of reference points is selected on the checker-
patterned model plane for the two cameras used
in this chapter.
Vision-Based Fire Detection
The real-time fire-fighting robot is guided by a
vision-based CAMSHIFT tracking surveillance
system and is equipped with a water gun to ex-
tinguish the burning targets. The vision-based fire
flame detection algorithm consists of five steps:
(1) moving pixels or regions in the current frame
of a video are determined with the motion history
image (MHI); (2) the HSI colors of moving pixels
are checked; (3) if the histogram of moving pixels
is correlated with the fire flame color histogram,
then the disordered measurement and temporal
analysis are performed to determine if fire flame
colored pixels flicker or not; and (4) Back-Pro-
jection and CAMSHIFT are applied to track the
fire flame region; (5) the distance between the
robot and the burning target is calculated through
binocular stereo.
Moving Motion Segmentation with
Motion History Image (MHI)
The MHI is a scalar-valued image where intensity
is a function of the recency of motion (Davis, 1999;
Bobick, 2001; Bradski, 2002). This moving history
representation can be used to determine the cur-
rent movement of the object and to segment and
measure the motions induced by the object (e.g.,
fire flame) in a video scene. MHI representations
have the following advantages: a range of times
from frame to frame to several seconds may be
encoded in a single image, direct recognition of
the motion itself is possible, motion recognition
is not computationally taxing and real-time imple-
mentation is possible, and the motion within the
detecting scene can be monitored. An MHI is used
to represent how the fire flame is moving, since
the outward boundaries of the fire flame are less
prone to misdetection than the source regions of
fire flame. In an MHI, the pixel intensity is a func-
tion of the motion history at that location; in the
MHI, brighter values correspond to a more recent
motion. It should be noted that the final motion
locations appear brighter in the MHI.
Correlation of Spectral Characteristics
The first step in detecting possible fire flame pixel
candidates is to transform the color space into HSI
color space and then carry out analysis. The HSI
color system projects the standard red-green-blue
(RGB color model) color space along its principle
diagonal in terms of white to black shades to avoid
the influence of lighting changes (Castleman,
1996). Hue is the dominant color (red, green, and
blue) of an area, and saturation is the colorfulness
of an area in proportion to its brightness. Intensity
is related to the color luminance, e.g., human skin
occupies a small portion of the H and S spaces. The
advantages of the HSI space are the intuitiveness
of the components and the explicit discrimination
between luminance and chrominance. The hue,
saturation, and intensity components of the HSI
model are normalized into the following ranges:
0° ≤ hue ≤ 360°, 0 ≤ saturation ≤ 255, and 0 ≤
intensity ≤ 255. The computed fire flame spectral
histogram correlation coefficient is measured by
the compare correlation analysis. The template
787
 Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
of fire flame spectral histogram, which is based
on empirical analysis results for the fire flames
with colors from red to yellow, was created to
detect the flame-colored pixels (Horng, 2005).
Hue is an attribute of the pure color of the image
scene, and it was demonstrated that it can be used
in assessing the prospect of numerical labelling
of the flame colors (Huang, 2008). Hence, the
detection of flame pixels is carried out using the
hue channel histogram correlation analysis with
the fire flame template, which maps the hue value
of general flames to be distributed from 0° to 60°.
Chaotic Spatial Structure Analysis
The moving object regions with disordered ratios
of perimeter to area for the extracted fire flame
region Ω are defined as: P/A, where P represents
the perimeter of the region and A represents the
area of the region. As the complexity of a shape
increases (i.e., the perimeter increases with re-
spect to the area), the value associated with the
disordered ratio Ω increases. The chaotic and
turbulent nature of a region can be detected by
relating the extracted spatial features to the fire
flame likelihood region and the smoke likelihood
region (Chen, 2004). The likelihood that a flame-
like region is a flame region is highly correlated
with the parameter Ω.
Temporal Analysis
It is not always sufficient to detect fire flame
correctly based on color information. There are
many objects, with similar color properties as the
fire flame spectrum. The key to distinguishing
between the flame and flame-colored objects is the
nature of their motion. The flames in a fire dance
around, so any particular pixel will only see fire
for a fraction of the time. This kind of temporal
periodicity is commonly known as flickering. The
flicker of fire flame causes the spectral values in the
fire flame region to fluctuate in time. The flicker
in fire is also used as additional information. The
788
candidate regions are checked to see whether they
continuously appear and disappear over time. The
level crossing rate LCR is utilized for validating
these extracted fire flickering regions. Temporal
variation for each pixel is computed by finding
the level crossing rate of the most likely fire flame
candidate region above the heuristic threshold
value among consecutive frames. The heuristic
threshold is determined based on the fire flame
models in recorded video sequences.
Back-Projection
If a visual servoing system is applied to natural
backgrounds, color data usually provides more
reliable and flexible information than mono-
chrome data (Kim, 1996; LeGrand, 1996). The
CAMSHIFT tracking engine is based on the his-
togram projection algorithm (Swain, 1990), which
is a useful technique for color object recognition,
especially for object identification in complex
background surroundings. Histogram back-pro-
jection is a primitive operation that finds and
identifies the association between pixel values in
a grabbed image and the values in a particular
histogram bin. Histogram and back-projection
performed on any consecutive frame would gen-
erate a probability image on which the value of
each pixel represents the probability of the exact
same pixel from the input belonging to the target
histogram that was used. Given that m histogram
bins are used, we can define n image pixel loca-
tions. Thus, we have histograms {ŷu}, u=1,…, m
and pixel locations {xi}, i=1,…, n. Let us also
define a function c:R2→{1,…, m} that associates
a pixel at location xi* with a histogram bin index
c(xi*). Then, the histograms can be computed with
n
 ( )
the equation ŷu = ∑ δ c x i∗ − u  . In all cases,

i =1
the values in the histogram bins are rescaled to
fit within the discrete pixel range of the possible
output 2D probability distribution image with the
function
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings

   derived from color histograms. The center and

  UPPER 
pu = min  yˆu ,UPPER . size of the targeted object region is computed
  max (yˆ) 

 u =1...m and used as settings for the search window on the
following frame of the video sequence. Figure
That is, the values in the histogram bins, which 2 shows the images of the tracked object in the
originally lie in the range [0, max(ŷu)], now lie in digital pictures that have been recognized and
the new range [0, UPPER]. In the end, the input processed, and the green bounding ellipse presents
pixels with the highest probability of being in the fire flame region tracked by the CAMSHIFT
the sample histogram will be mapped onto a 2D algorithm. The calculation of the color probability
histogram back-projection image with the highest distribution is not performed on the entire image,
visible intensities. but only on limited regions surrounding the cur-
rent CAMSHIFT window, which includes images
CAMSHIFT Tracking of the specified object that are transformed into
a discrete probability image. This tends to result
The CAMSHIFT algorithm is a non-parametric in a large reduction in the computational costs.
technique that can track a specified target’s 2D The CAMSHIFT algorithm can be summarized
position efficiently across a series of images. When by the following steps:
tracking the 2D position of a colored object, in our
case a fire flame colored region, the CAMSHIFT • Step 1: A region of interest (ROI) window
operates on a color probability distribution image is selected to be the sample image for fu-

Figure 2. The image at the 169th frame is detected as a flame by searching the candidate regions, which
are above the LCR threshold. The CAMSHIFT tracking algorithm is employed to track the movement
of flame pixels, and the flame region is bounded by the green ellipse.

789
 Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
ture color probability distribution compu-
tation. In tracking procedures, this window
is placed over the targeted object.
• Step 2: A mean shift search window is ini-
tially centered at the first frame’s data point
position.
• Step 3: The color distribution of the region
centered at the mean shift search window
is calculated, producing a discrete prob-
ability image. The mean location (the cen-
troid) of the discrete probability image can
be found within the search window by first
obtaining moment values. Given that I(x,
y) is the pixel value function for the inten-
sity of the discrete probability image at
point (x, y) in the search window, one can
compute the zeroth moment for that point.
• Step 4: The mean shift algorithm (step 3)
is iterated, replacing (x, y) with the corre-
sponding (xc, yc), until the centroid of the
search window region’s generated proba-
bility image converges to a constant point.
This point should be at the center of the
tracked target. The zeroth moment (distri-
bution area) and the mean location (the
centroid) are stored.
• Step 5: The size of the search windows
is set as a function of the zeroth moment
found in Step 4 to match the size of the
tracked object, and the center of the search
window is placed on the following frame at
the mean location found in Step 4. The pro-
cess is then repeated, beginning at Step 3.
Depth Calculated through
Binocular Stereo
The purpose of using two cameras instead of one
to track a burning target provides data on the
burning target’s position in a third dimension.
The calculation of depth via binocular stereo is
a common way to extend one’s knowledge of a
scene from 2D to 3D (Jain, 1995). First, feature
points are grouped into teams of two—one each in
790
the two images obtained by each of the two stereo
cameras—to create a set of stereo pairs. Then, for
every pair, each point is envisioned as a ray. The
two rays would intersect in the actual 3D world.
Now suppose the 3D coordinates of the two points
in a stereo pair are P1 and P2; the rotation matrices
and translation vectors for the transformation
between the two camera coordinate systems and
the stereo coordinate system are R1, R2 and t1, t2.
Therefore, given a point M in the stereo coordinate
system, we have P1 = R1M + t1 and P2 = R2M + t2.
Finally, the point where the two rays intersect is
found to be the point M that produces the P1 and
P2 closest to the rays (represented by straight lines)
in their respective camera coordinate systems.
This approach transforms the coordinate systems
of the two cameras into absolute coordinates. In
this work, the real-time 3D depths of the burning
target, tracked with the presently discussed 3D
point measuring method involving the intersecting
rays, are rendered using Open Graphics Library
(Wright, 2000). The specified corner positions of
the chessboard are processed and analyzed with
the calibrated intrinsic and extrinsic parameters
and the absolute coordinates are calculated via
binocular stereo.
a. LCR probability distribution image of flame.
b. The moving flame pixels are separated.
c. The CAMSHIFT tracking algorithm is
employed.
STEREO-VISION-BASED FIRE
EXTINGUISHING MOBILE ROBOT
The real-time stereo-vision-based wheeled mobile
robot (WMR) is composed of a stereo-vision-
based CAMSHIFT tracking algorithm, two CMOS
cameras (with stereo vision) for capturing images,
and a water gun for extinguishing the burning tar-
get. To equip a mobile robot with 3-dimensional
range detection capability, the calibration of the
binocular vision system should be carried out.
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
The inside geometry information of and spatial
relationship between the two cameras used can be
acquired by calibrating their intrinsic and extrinsic
parameters. The binocular image processing sys-
tem first finds the same feature points of objects,
then compares them to see if they are the same
based on the epipolar geometry constraint. The
next step is to calculate the 3-dimensional coordi-
nate of the object using the triangular perspective
theory, and to compute the distance between the
robot and the object.
A fuzzy controller is used to control the robot’s
rotating direction to track the target on the front
side. The robot follows the commands of the
fuzzy reasoning module to manipulate the mobile
rotation. As soon as the tracked target is followed
without colliding with obstacles, the fuzzy reason-
ing visual system orders the mobile robot to cease
rotating. In this work, the two main input variables
for the fuzzy controller are the sum of the target’s
horizontal offset position from the stereo image
pairs Xs = Xl + Xr, as shown in Figure 3, and the
followed target’s depth position relative to the
robot Td. The values of Xs and Td are rescaled to
fit the range [–1, 1]. The value of Xs consists of
Figure 3. Two main input variables for the fuzzy controller are Xs (Xl + Xr) and Td
five fuzzy regions: left far (LF), left near (LN),
zero (ZE), right near (RN), and right far (RF). For
simplicity, these five standard triangular member-
ship functions for the fuzzy region variables {LF,
LN, ZE, RN, and RF} are used. The value of Td
consists of three fuzzy regions: near (N), middle
(M), and far (F), which represent the target’s depth
distance relative to the mobile robot.
The rotating signal θr at the fuzzy controller
output consists of five singletons: left large (LL),
left small (LS), zero (ZE), right small (RS), and
right large (RL). With the values of the four
singletons in hand, the controller output u can
then be calculated using the defuzzifier formula.
The fuzzy rules are listed in Table 1, which also
represent the fuzzy associative matrix. The lines
and columns correspond to the target’s relative
depth distances and horizontal offset position
values, respectively (inputs to the fuzzy reasoning
system), while the values of the matrix correspond
to a robot-rotating signal (output of the fuzzy
reasoning system). Overall, the fuzzy reasoning
system is governed by the min/max inference
technique and the center of gravity for the de-
fuzzification step. The two cameras of the proposed
791
 Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings

Table 1. Fuzzy associative matrix for mobile

the mobile robot for the best effective view of the
robot steering
targeted object. As shown in Figure 4, the robot
Td follows the commands of the fuzzy reasoning
θr
N M F
module to track the burning target with the fire
LF LS LS LL
detection function enabled and hence, the robot
LN
rotates to approach the fire flame on the front
LS LS LS
side. Demo video clips of the experiment are
Xs ZE ZE ZE ZE
available on the Web site (http://www.youtube.
RN RS RS RS
com/watch?v=dgTSsX2ezMU and http://www.
RF RS RS RL
youtube.com/watch?v=JkgC-Q7ogE8).

a. Frame 100, first camera

visual servoing system track the target indepen-
b. Frame 100, second camera
dently of each other with their CAMSHIFT track-
ing engine.
FUTURE RESEARCH DIRECTIONS
EXPERIMENTAL RESULTS Designing a visual tracking system to avoid the
in-path obstacles is a complex task because a
The main purpose of this work is to design a
large amount of video data must be transmitted
wheeled mobile robot with stereo machine vision
and processed in real time. The main task for the
to extinguish a burning target. The stereo visual
target tracking without hitting the obstacles is
system is used to analyze the images to get the
obstacle detection, which is essential for a safe
3D coordinate position of the burning target on
autonomous mobile robot. Detecting obstacles
the ground, mark the region and its gravity center,
requires an active perception of the surround-
and then the mobile robot is driven to extinguish
ings. Laser scanners have the great advantage
the burning object with a water gun. The position
of providing accurate depth information that has
and orientation of the stereo cameras with respect
to be computed from calibrated stereo images if
to the real world’s coordinate system could be
cameras are used for the same task (Ho, 2009). The
obtained through extrinsic calibration routines.
real-time object tracking and collision avoidance
Both cameras are mounted at the front part of

Figure 4. Position deviations between the robot and the fire and viewpoint from the robot’s first and
second camera after applying the visual servoing rules

792
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
method for mobile robot navigation in indoor and
outdoor environments using stereo vision fused
with laser sensors is an emerging trend.
Multi-sensor fusion is necessary to cut down
on the number of false alarms, since it can reduce
the effects of errors in measurements. A variety
of multi-sensors were fused together in sensor
packages and evaluated based on a set of cost
and performance criteria. Vision-based real-time
detection for early fire flame detection can be fused
with the multiple sensors in order to have a more
robust video-based fire detection system. The
development of a more sophisticated algorithm,
versus the simple threshold rule, for multi-sensor
detectors is currently under investigation.
In addition, multiple fires may break out si-
multaneously, and hence the conducting research
on tracking multiple fire regions concurrently and
extinguishing dynamically by the behavior of the
flame is great technical challenges.
Significant progress has been made in visual
servoing during the last few years. Several robust
tracking algorithms have been developed, which
can track objects in real time in simple scenarios.
The proposed framework can find further applica-
tions in versatile fields like automated surveillance,
human computer interaction, video retrieval, traf-
fic monitoring, and vehicle navigation. Further,
motion estimation is a very active area of research
in which new solutions are continuously being
developed. One challenge in tracking is to develop
robust algorithms suitable for tracking objects via
hardware logic IP (system-on-chip technology).
The combination of different sensors employed in
visual servoing (e.g., audio and force sensors) is
also a new direction for further development. Such
hybrid sensor-based servoing provides additional
information that can be used in conjunction with a
video-based tracker to solve problems like severe
occlusion or estimating tracking more robustly.
CONCLUSION
This chapter proposes a stereo-vision-based
wheeled mobile fire detecting and fighting system
that has been successfully implemented and shown
to work in non-ideal real-world residential build-
ings. Spectral, spatial, and temporal motion fea-
tures and a heuristic-based classifier are employed
to extract real fire flame data and are adopted for
helping the validation of that fire flame. Stereo
vision tracking can be achieved by applying the
CAMSHIFT algorithm and using two low-cost,
calibrated USB cameras, which enable high-speed
image capturing. A computationally efficient and
robust implementation of the visual measurement
and servo mobile robot can be used to obtain reli-
able real-time online 3D positioning of a particular
burning object. Moreover, the fire-fighting system
provides more safety in fire fighting and is very
economical when incorporated with other fire
alarm systems for use in large spaces. Intelligent
and automatic control of the fire-fighting robot
improves its detection efficiency and suppres-
sion adaptability. Experimental results show that
real-time fire flame detection and suppression is
achieved even under non-ideal lighting conditions.
ACKNOWLEDGMENT
This research was partially supported by the
National Science Council in Taiwan under Grant
NSC 98-2218-E-224-003, NSC 100-2221-E-224-
016 and Industrial Technology Research Institute.
Special thanks are given to anonymous reviewers
for their helpful suggestions.
793
 Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
REFERENCES
Bobick, A., & Davis, J. W. (2001). The represen-
tation and recognition of action using temporal
templates. IEEE Transactions on Pattern Analy-
sis and Machine Intelligence, 23(3), 257–267.
doi:10.1109/34.910878
Bradski, G. R., & Davis, J. W. (2002). Motion seg-
mentation and pose recognition with motion his-
tory gradients. Machine Vision and Applications,
13(3), 174–184. doi:10.1007/s001380100064
Castleman, K. R. (1996). Digital image process-
ing. Upper Saddle River, NJ: Prentice Hall Press.
Celik, T., Özkaramanlı, H., & Demirel, H. (2007).
Fire and smoke detection without sensors: Image
processing based approach. Proceedings of 15th
European Signal Processing Conference (pp.
1794-1798).
Chen, T., Yuan, H., Su, G., & Fan, W. (2004). An
automatic fire searching and suppression system
for large spaces. Fire Safety Journal, 39(4),
297–307. doi:10.1016/j.firesaf.2003.11.007
Chen, T. H., Kao, C. L., & Chang, S. M. (2003).
An intelligent real-time fire-detection method
based on video processing. Paper presented at the
IEEE 37th Annual 2003 International Conference
on Carnahan.
Chen, T. H., Wu, P. H., & Chiou, Y. C. (2004).
An early fire-detection method based on image
processing. Paper presented at the International
Conference on Image Processing.
Davis, J. (1999). Recognizing movement using
motion histograms.
Di Stefano, L., Mattoccia, S., & Mola, M. (2003).
An efficient algorithm for exhaustive template
matching based on normalized cross correlation.
Paper presented at the 12th International Confer-
ence on Image Analysis and Processing.
794
Ho, C.-C. (2009, 16-19, August). Machine vision
based 3D scanning system. Paper presented at the
International Conference on Electronic Measure-
ment & Instruments, Beijing.
Ho, C.-C. (2009). Machine vision-based real-time
early flame and smoke detection. Measurement Sci-
ence and Technology, 20(4), 045502(045513pp).
Huang, H. W., & Zhang, Y. (2008). Flame colour
characterization in the visible and infrared spec-
trum using a digital camera and image process-
ing. Measurement Science & Technology, 19(8),
085406. doi:10.1088/0957-0233/19/8/085406
Hutchinson, T. C., Kuester, F., Doerr, K. U., &
Lim, D. (2006). Optimal hardware and software
design of an image-based system for capturing
dynamic movements. IEEE Transactions on In-
strumentation and Measurement, 55(1), 164–175.
doi:10.1109/TIM.2005.860872
Jain, R., Kasturi, R., & Schunck, B. G. (1995).
Machine vision. McGraw-Hill.
Kim, K. I., Oh, S. Y., Kim, S. W., Jeong, H., Lee,
C. N., Kim, B. S., et al. (1996). An autonomous
land vehicle PRV III. Paper presented at the IEEE
Intelligent Vehicles Symposium.
LeGrand, R., & Luo, R. C. (1996, 22-28 April).
Position estimation of selected targets. Paper pre-
sented at the International Conference on Robotics
and Automation, Minneapolis, MN
Podr aj, P., & Hashimoto, H. (2008). Intelligent
space as a framework for fire detection and evacu-
ation. Fire Technology, 44(1), 65-76.
Sumi, K., Hashimoto, M., & Okuda, H. (1995).
Three-level broad-edge matching based real-time
robot vision. Paper presented at the IEEE Interna-
tional Conference on Robotics and Automation.
Swain, M. J., & Ballard, D. H. (1990). Indexing
via color histograms. Paper presented at the Third
International Conference on Computer Vision.
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
Villa-Uriol, M. C., Chaudhary, G., Kuester, F.,
Hutchinson, T., & Bagherzadeh, N. (2004). Ex-
tracting 3D from 2D: Selection basis for camera
calibration. Paper presented at the 7th Interna-
tional Conference on Computer Graphics and
Imaging (CGIM).
Wang, S. J., Tsai, M. T., Ho, Y. K., & Chiang, C.
C. (2006). Video-based early flame detection for
vessels by using the fuzzy color clustering algo-
rithm. Proceedings of International Computer
Symposium, 3, 1179–1184.
Wright, R. S. Jr, & Sweet, M. R. (2000). OpenGL
superBible (2nd ed.). Waite Group Press.
Yuan, F. (2010). An integrated fire detection and
suppression system based on widely available vid-
eo surveillance. Machine Vision and Applications,
21(6), 941–948. doi:10.1007/s00138-010-0276-x
ADDITIONAL READING
The references to materials directly cited in the
chapter have been included in the reference sec-
tion. However, there are many additional readings
relevant to stereo vision, some of which are listed
in this section.
Bradski, G. R. (1998). Computer vision face
tracking for use in a perceptual user interface.
Intel Technology Journal, 2(2), 12–21.
Bradski, G. R., & Davis, J. W. (2002). Motion seg-
mentation and pose recognition with motion his-
tory gradients. Machine Vision and Applications,
13(3), 174–184. doi:10.1007/s001380100064
Chen, T. H., Yin, Y. H., Huang, S. F., & Ye, Y. T.
(2006, Dec. 2006). The smoke detection for early
fire-alarming system base on video processing.
Paper presented at the International Conference
on Intelligent Information Hiding and Multimedia
Signal Processing.
Davis, J. W., & Bobick, A. (1997, June). The
representation and recognition of action using
temporal templates. Paper presented at the Pro-
ceedings of IEEE Conference on Computer Vision
and Pattern Recognition, Puerto Rico.
Fang, J., Jie, J., Hong-Yong, Y., & Yong-Ming, Z.
(2006). Early fire smoke movements and detec-
tion in high large volume spaces. Building and
Environment, 41(11), 1482–1493. doi:10.1016/j.
buildenv.2005.05.036
Ho, C.-C. (2010). Image processing algorithm for
tracking and picking an object for an open plat-
form mobile robot. Key Engineering Materials,
419, 569–572. doi:10.4028/www.scientific.net/
KEM.419-420.569
Ho, C.-C. (2010). A stereo vision based target
tracking and obstacle avoidance. Key Engineer-
ing Materials, 419, 565–568. doi:10.4028/www.
scientific.net/KEM.419-420.565
Ho, C.-C., Chen, M.-C., & Lien, C.-H. (2011).
Machine vision-based intelligent fire fighting
robot. Key Engineering Materials, 450, 312–315.
doi:10.4028/www.scientific.net/KEM.450.312
Ho, C.-C., & Kuo, T.-H. (2009). Real-time video-
based fire smoke detection system. Paper presented
at the IEEE/ASME Conference on Advanced
Intelligent Mechatronics.
Ho, C.-C., & Shih, C.-L. (2007). Real-time track-
ing and stereo vision-based control of a goldfish-
catching system. Paper presented at the Tthe 35th
International MATADOR Conference.
Ho, C.-C., & Shih, C.-L. (2008). Machine vision
based tracking control of a ball-beam system.
Key Engineering Materials, 381-382, 301–304.
doi:10.4028/www.scientific.net/KEM.381-
382.301
Ho, C.-C., & Shih, C.-L. (2008). A real-time visual
servo architecture for a golden fish catching sys-
tem. Measurement and Control, 41(5), 151–154.
795
 Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings
Ho, C.-C., & Shih, C.-L. (2009). A real-time
fuzzy reasoning based control system for catch-
ing a moving goldfish. International Journal of
Control. Automation and Systems, 7(5), 755–763.
doi:10.1007/s12555-009-0508-x
Horng, W.-b., Peng, J.-w., & Chen, C.-y. (2005,
March 19-22). A new image-based real-time
flame detection method using color analysis.
Paper presented at the International Conference
on Networking, Sensing and Control, Tucson,
Arizona, USA.
Kang, K. (2007). A smoke model and its applica-
tion for smoke management in an underground
mass transit station. Fire Safety Journal, 42(3),
218–231. doi:10.1016/j.firesaf.2006.10.003
Ko, B. C., Cheong, K. H., & Nam, J. Y. (2008).
Fire detection based on vision sensor and support
vector machines. Fire Safety Journal.
Kopilovic, I., Vagvolgyi, B., & Sziranyi, T. (2000,
09/03 - 09/07). Application of panoramic annular
lens for motion analysis tasks: surveillance and
smoke detection. Paper presented at the 15th In-
ternational Conference on Pattern Recognition,
Barcelona, Spain.
Lai, C. L., Yang, J. C., & Chen, Y. H. (2007, May).
A real time video processing based surveillance
system for early fire and flood detection. Paper
presented at the The 24th IEEE Instrumentation
and Measurement Technology Conference.
Liu, C.-B., & Ahuja, N. (2004). Vision based fire
detection. Paper presented at the The 17th Interna-
tional Conference on Pattern Recognition, Urbana.
Phillips, W. III, Shah, M., & da Vitoria Lobo, N.
(2002). Flame recognition in video. Pattern Rec-
ognition Letters, 23(1-3), 319–327. doi:10.1016/
S0167-8655(01)00135-0
796
Schroeder, D. (2004). Evaluation of three wildfire
smoke detection systems. [FERIC]. Advantage-
Forest Engineering Research Institute of Canada,
5(24), 8.
Toreyin, B. U., Dedeoglu, Y., & Cetin, A. E.
(2005). Flame detection in video using hidden
markov models. Paper presented at the Interna-
tional Conference on Image Processing.
Toreyin, B. U., Dedeoglu, Y., & Cetin, A. E. (2005).
Wavelet based real-time smoke detection in video.
Paper presented at the The 13th European Signal
Processing Conference, Antalya, Turkey.
Toreyin, B. U., Dedeoglu, Y., & Cetin, A. E. (2006).
Contour based smoke detection in video using
wavelets. Paper presented at the The 14th European
Signal Processing Conference, Florance, Italy.
Toreyin, B. U., Dedeogˇlu, Y., Gudukbay, U., &
Cetin, A. E. (2006). Computer vision based method
for real-time fire and flame detection. Pattern
Recognition Letters, 27(1), 49–58. doi:10.1016/j.
patrec.2005.06.015
Wieser, D., & Brupbacher, T. (2001, March 25-28).
Smoke detection in tunnels using video images. Pa-
per presented at the 12th international conference
on automatic fire detection, Gaithersburg, USA.
Xiong, Z., Caballero, R., Wang, H., Finn, A. M.,
Lelic, M. A., & Peng, P.-Y. (2007). Video-based
smoke detection: possibilities, techniques, and
challenges. Paper presented at the Fire Suppres-
sion and Detection Research and Applications,
Orlando, FL.
Yuan, F. (2008). A fast accumulative motion ori-
entation model based on integral image for video
smoke detection. Pattern Recognition Letters,
29(7), 925–932. doi:10.1016/j.patrec.2008.01.013
Stereo-Vision-Based Fire Detection and Suppression Robot for Buildings

KEY TERMS AND DEFINITIONS
3D Tracking: Position based tracking algo-
rithm to provide feedback on the real-time posi-
tion of the specified target with a high frame rate.
Fire-Fighting Robot: The mobile robot is able
to successfully complete a fire-extinguishing task.
Fire Surveillance System: Video based fire
surveillance technology is suitable for early detec-
tion of fires due to its low detection delay, good
resolution, and high localization accuracy.
Stereo Vision: 3D position can be generated
by means of two calibrated cameras.
Video Based Fire Flame Detection: Analyze
the spectral, spatial, and motion orientation char-
acteristics of the fire flame regions from the image
sequences of the video to detect fire.
Video Based Smoke Pattern Recognition:
Analyze the spectral, spatial, and motion orien-
tation characteristics of the fire smoke regions
from the image sequences of the video to detect
fire smoke.
Visual Servoing: A visual tracking system to
track an object and video data must be transmitted
and processed in real time in order to feedback
to the system.

This work was previously published in Depth Map and 3D Imaging Applications: Algorithms and Technologies, edited by
Aamir Saeed Malik, Tae Sun Choi and Humaira Nisar, pp. 418-432, copyright 2012 by Information Science Reference (an
imprint of IGI Global).

797
 Section 4
Utilization and Application

This section discusses a variety of applications and opportunities available that can be considered by practitioners
in developing viable and effective Geographic Information Systems programs and processes. This section includes
16 chapters that review topics from case studies in Africa to best practices in Asia and ongoing research in the
United States. Further chapters discuss Geographic Information Systems in a variety of settings (environment,
city planning, disease/disaster management, etc.). Contributions included in this section provide excellent cover-
age of today’s IT community and how research into Geographic Information Systems is impacting the social
fabric of our present-day global village.
 799

Chapter 49
Children’s Maps in GIS:
A Tool for Communicating Outdoor
Experiences in Urban Planning

Kerstin Nordin
Swedish University of Agricultural Sciences, Sweden

Ulla Berglund
Swedish University of Agricultural Sciences, Sweden

ABSTRACT
Since 2002 the authors have successively developed “Children’s Maps in GIS”, a method for children’s
participation in spatial planning. Their studies show that 10-15 year-olds are capable of reading maps
and using a GIS-application for communicating their interests in a stable and useful manner. The purpose
of this article is to discuss the first stages of implementation in a real world project, in relation to ICT.
The authors report experiences from a Swedish municipality using Children’s Maps in GIS in a survey
with over 600 children as part of a comprehensive planning process and give examples of how data can
be visualized. A significant digital divide between different parts of the administration is noted. In the
ongoing development into an Internet version of the method the authors aim to increase the access to
the GIS-application and develop standard procedures for categorizing and analyzing data.

INTRODUCTION in effect implemented, by most countries around

It is considered that children (0-18 years) should be
afforded the opportunity to influence those matters
that affect or concern them, as outlined in Article
12:1 of the UN Convention of The Rights of the
Child (UNICEF, 1990), a convention signed, and
the World. The local environment is such an issue
concerning children, this is identified in the UN
action plan Agenda 21 (United Nations, 1993)
where it is stated that young people’s participation
is vital to the realization of a sustainable society.
Therefore there is an expectation for young

DOI: 10.4018/978-1-4666-2038-4.ch049

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

people’s experiences and wishes to be included
in local-authority planning.
In the background section of this article we
report on our theoretical framework and describe
the method Children’s Maps that we have devel-
oped through our research. In the result section
we document how this method was used outside
the research-context in a municipal as a part of a
comprehensive planning process. In the discus-
sion section we comment on the results and make
comparisons with experiences from previous
research-pilots. Finally we draw conclusions on
further development of the method.
BACKGROUND
This paper reports on research carried out within
the discipline of Landscape architecture although
the theoretical framework shows an interdis-
ciplinary approach including planning theory,
social theory as well as theory of environmental
psychology, geography and GIScience.
Children, Planning and Participation
In today’s construction of childhood, children are
acknowledged as active social and cultural actors
(Holloway & Valentine, 2003; Christensen &
Prout, 2002). According to Christensen & O’Brian
(2003, p. 2) they are also recognized as informants
and participants in research, having “emerged
as key source for understanding their everyday
life”. Christensen further stresses the relevance
for planning of children’s “emplaced knowledge
[…] full with personal and social meaning, built
up through their everyday encounters” with their
local environment” (Christensen, 2003, p.16). The
overall idea of governance using communicative
planning (e.g., Healey, 1997; Healey, 1999) with
reference to Habermas’ theories on communica-
tive rationality and communicative action (e.g.,
Habermas, 1984) today is widely practiced in
developed countries. However, this practice is
800
Children’s Maps in GIS
criticized for lacking strategies for handling of
biased power relations. This means that less
powerful stakeholders, and especially those whose
perspectives deviate from existing policies, are at
risk of being unable to achieve the influence that
their arguments call for (Flyvbjerg, 2001; Sager,
1994; Sager, 2006).Children and youth tend to
fall into this category, putting their participation
at risk of being reduced to pseudo-democratic
practices, such as manipulation, decoration or
tokenism, the lower steps in the “Ladder of Young
People’s Participation” (Hart, 1997). Case studies
in various European countries highlight tendencies
towards superficial participation resulting in little
real influence (Rogers, 2006; Tonucci, Prisco,
& Horelli, 2004) and unwillingness or inability
from decision-makers to take children’s interests
into account has been noted by many researchers
(Chawla, 2002a; Lynch, 1977; Matthews, 1998;
Wilhjelm, 1999; Woolley et al., 1999).
Conversely, literature shows a strong sup-
port for the idea that planning and design of the
physical environment attracts the interest of young
people and it is therefore possible to include them
in society’s democratic processes (Horelli, 1998;
Percy-Smith & Malone, 2001; Urban Green Spaces
Taskforce, 2002; Norsk form, 2005.) Furthermore,
following our personal experiences, there is a de-
sire among many planners to engage children and
youth to a larger extent. We, as well as Freeman
& Aitken-Rose (2005) have noticed a growing
interest toward child-focused and youth-focused
methodologies for working with adolescents.
There is an obvious need to find solutions to
the dilemma of how to incorporate young peoples’
perspectives into local planning processes (e.g.,
Christensen & O’Brian, 2003; Driskell, 2002).
A model for enhancing children’s participation
in decision-making is the “Pathways to par-
ticipation”, developed by Shiers (2001). Shiers,
drawing on Hart (1997), identifies five levels of
participation. The model can be used as a tool for
planning for participation. Level 3, “Children’s
views are taken into account” is the minimum
Children’s Maps in GIS
you must achieve if you endorse the UN Con-
vention on the Rights of the Child. The higher
levels of participation require that children are
involved in decision making processes (level 4)
and children share power and responsibility for
decision-making (level 5). A method aiming to
enhance children’s participation should be useful
at least at level 3.
Several initiatives concerning the empower-
ment of children within spatial planning and related
issues have been influenced by the UN Convention
of the Right of the Child. Among these are the
“Town of Children Project” started in 1991 which
has engaged more than 200 Italian municipalities
(Baraldi, 2003), the UNESCO Project “Growing
up in Cities” (Chawla, 2002b; Driskell, 2002),
and the “Children as Community Researchers”,
initiated by Roger Hart and launched by UNI-
CEF (UNICEF, 2001).These varied approaches
highlight the diversity of methodologies available
for children’s participation in planning of cities.
The changes brought about by globalization
and the diffusion of information and communica-
tion technology (ICT) is gradually opening up new
opportunities for the involvement of young people
in planning and development of cities (Horelli &
Kaaja, 2002). One such example is a tool in the
form of an Internet-based design game “The Ad-
venture Forest” invented by Finnish researchers
(Kyttä, Kaaja, & Horelli, 2004); other example is
the sharing of knowledge on the Internet site “The
FreeChild Project” (The FreeChild Project, 2008);
There are several reports concerning the insti-
tutional constraints of adapting new methods and
technologies within the planning practice (Horelli
& Kaaja, 2002; Vonk, Geertman, & Shot, 2005).
An assumed reason for this is that most current
tools are too inflexible and incompatible for most
planning tasks, being oriented towards techno-
logical problems rather than planning problems
(Vonk et al., 2005)
A method for communicating children’s
outdoor experiences into every-day planning
practices should be an easy-to-handle method that
can collect, store, analyze and visualize data that
is relevant and trustworthy to all participants in
a planning process. The first step is then to make
data concerning children’s use and perceptions
of the outdoor environment available in a plan-
ning decision system in a way that is true to the
child’s own views.
Geographic Information
Geographic information, often represented in the
form of maps, plays an important role in planning
practice; today almost all of this information is
digital and part of a geographic information sys-
tem, GIS. Although the use of GIS has increased
in Sweden in the public sector since 2003, there is
still a gap between those who can and those who
cannot use GIS for more advanced applications.
To look at static digital maps is still the most
common use of GIS (ULI, 2008).
MacEachren (2004, pp. iii-iv) points out that
the map is evolving as an interactive interface to
geospatial information, not just a static representa-
tion of information. He states that understanding
the processes of map-making and the perceptual-
cognitive processes through which these repre-
sentations are understood are of importance in
decision making and the enhancement of the role
of maps in collaborative activities.
Maps and cartography has been critically dis-
cussed since the 1990´s with relation to linking
geographic knowledge with power (Crampton
& Krygier, 2006). They note drawing on Pickles
(1995) that “Maps are active: they actively con-
struct knowledge, they exercise power and they can
be powerful means of promoting social change”
(Crampton & Krygier, 2006, p.15).
Geovisualization of data enhance visual
thinking in exploring data and in collaboration
processes (Jiang, Huang, & Vasek, 2003). Children
tend to develop a visual competence in an early
age. Of special interest is specifically children’s
understanding of a map as a representation of the
real world. Plester, Blades, and Spencer (2006),
801

describe how children during the first school years
learn to master correspondence and to interpret
direction and distance, i.e. skills needed to under-
stand the representation of an environment in an
aerial photograph or map.
If we focus on GIS as a data collecting tool, is
it useful when collecting information from chil-
dren? When it comes to drawing on digital maps
the competence of children above 10 years of age
concerning interpretation of symbols, scale etc
can be expected to be as good (or bad) as that of
adults (Plester et al., 2006). Positive experiences of
the use of digital maps with children are reported
by for example Kyttä et al. (2004) and Berglund
(2008). Holloway and Valentine (2003) in their
study of children and ICT comment that children
on the whole tend to be more technically compe-
tent in the use of computers than their parents are.
Elwood (2006) points out that GIS has a his-
tory of being an expert tool in a rational, expert-
driven planning process. Information that can be
displayed visually is privileged as are the use of
quantitative techniques for spatial analysis. This
may implicate that other forms of knowledge may
be excluded from processes in which GIS is used.
The widespread use of Internet and mobile GIS
makes it important to discuss the digital divide
with aspect to access to GIS and to spatial data
(Elwood, 2006). So why use GIS? “Elwood an-
swers (2006, p. 693) “In spite of its limitations
and challenges, GIS is tremendously important
because it is such a powerful mediator of spatial
knowledge, social and political power …” The
use of such an expert tool as GIS, can give the
information from the children some legitimacy
(Elwood, 2002).
Research with Children in School
There is a significant amount of literature regard-
ing pupils of 11-12 years of age. In our judgment
this is due not only to their experience of place but
also to their writing and reading abilities and in
some cases the fact that technical skills of children
802
Children’s Maps in GIS
from this age group are supposedly developed,
providing the researcher with multiple options
concerning methodology.
School is a frequently used environment for
research with children. In school you find an
organization that can assist with information and
the permissions needed for research that concerns
children. If the aim and the methods of the research
can make it fit into the curriculum at school, a
win-win situation could be created. If this is not
possible it might be regarded unethical to ask for
the time of teachers and pupils.
According to the report “Benchmarking Ac-
cess and Use of ICT” (Korte & Hüsing, 2006),
all Swedish schools have access to computers
and Internet, but the access is inconsistent across
schools, with students and teachers at independent
schools having more access to ICT than students
and teachers in municipal schools. This includes
hardware as well as connection to broadband and
software (Skolverket, 2009). Furthermore, most
teachers in Sweden do not feel that that they have
sufficient knowledge of ICT, and roughly half of
the teachers do not believe that ICT encourages
learning. That puts Sweden almost at the bot-
tom of the list for ICT use in schools in Europe
(Skolverket, 2009).
The Method Children’s Maps
Since 2002 we have in our research successively
developed “Children’s Maps in GIS”, a method
for communicating young people’s outdoor experi-
ences into urban planning. The method Children’s
Maps consist of several activities: contact with
school and parents, information to children, the
use of a GIS-application and feed-back to the
children and teachers (Berglund, 2008; Berglund,
Nordin, & Eriksson, 2009). In the following we
refer to the method as “Children’s Maps” and
depending on the context, we use young people as
well as children and youth, students and children
for labeling our target group.
Children’s Maps in GIS
We aimed at developing a method that was easy
to manage within the school context, and stimulat-
ing and fun for children to participate with. We
tested the method with and adapted it for children
from 10 to 15 years of age. As we were aware of
that GIS technology may empower or disempower
actors and institutions as described in Elwood
(2002) we wanted to build a GIS-application that
should not demand technical skills or expensive
software or databases. The result should be trust-
worthy and relevant to a town district level in the
planning practice.
We have in several research pilots (Berglund
& Nordin, 2007; Berglund, 2008; Berglund et al.,
2009) tried to answer the following questions:
Can we attain location bound information that is
relevant to planning and suitably reliable, with
the help of children? Will this information be ac-
cepted by planning authorities? Are children able
and willing to document their movements using
GIS on digital maps that are used within today’s
planning? Can the children’s school become a
mediator between the children and the planners;
can it be the place for developing and implement-
ing the method?
During the period 2003-2006, the method
was tested with children and teachers in several
research-pilots in different locations in Sweden.
Several tests were carried out in Stockholm; in a
high-status inner-city district as well as in a low-
status outer-city district with a high proportion
of immigrant families within an interdisciplinary
research project “Children and Open spaces in the
City – Accessibility, Use and Influence”. In 2006
we got an opportunity to test the method in a rural
context (a village, Örbyhus) as we got funding
from the Road Administration. In that project
we adjusted the method to better catch questions
concerning traffic safety. In all our studies, we
supplied computer, software and supervision for
the children participating, and did not ask for
any IT-qualifications from the teacher. We have
through diverse means, evaluated the function
and the trustworthiness of the method. We have
found no special complications regarding the use
of the method in the different research-contexts.
Our studies show that 10-15 year-olds are capable
of using a GIS-application for communicating
their experiences and interests into the local urban
planning process in a stable and useful manner.
We found it easy to fit the project into schools and
to get some teachers to run a special teacher’s ap-
plication. The development and tests are described
in detail in Berglund (2008) and supplemented in
Berglund et al. (2009).
The GIS-application can be described as a
computerized questionnaire. Each student an-
swers seven questions individually. The student
answers the question by making a sketch map
directly on the screen, and when required, the
student includes text information as an attribute
to the geographical object created in the sketch
map. An overview of the procedure is presented
in Table 1. The exercise generally takes about 15
minutes and is carried out during class hours. The
software is an application built in ESRI ArcView
3. Laptops are used to store the application as well
as the data collected.
The teachers participating in a study are using
an application that works in the same way as the
children’s application. They indicate routes and
places that are used for teaching by themselves
or used by other teachers of their school.
APPLICATION OF THE METHOD
IN REAL-WORLD USE IN TÄBY
We have in the text above described the method
Children’s Maps through our research-pilots.
What was still lacking was the evaluation of the
implementation into planning practice. In 2008,
the municipal of Täby wanted to use the method
“Children’s Maps” to get information about how
children and young people perceive and use the
outdoor environment in Täby. The project was
called “Place to grow” (our translation) and a part
of the comprehensive planning process. This gave
803
 Children’s Maps in GIS

Table 1. Questions included in the survey and data inputs that are possible to make. The questions are
directly translated from Swedish and not adapted to English speaking children

Questions on the screen Geographic data Attribute

A call to register a new student by
writing a unique number
1.Where do you live?
2.What routes do you use to school?
3.What routes do you use in your
free-time?
4.Where do you go when you are
out-doors?
5.Are there places that are hazardous
or make you feel uncomfortable?
6.Do you have a favorite place?
7.Do you have any suggestions for
improvements?
School, grade, number and sex
Point object(s) No attributes
Poly-line object(s) Means of travel: walk, cycle, walk/cycle to public transporta-
tion
Comments: Free wording
Poly-line object(s) Movements: walk, run, walk the dog, walk/cycle to public
transportation
Destination: friends, activity or else
Comments: Free wording
Polygon(s) Common activities all year around: meet friends, play, play
boll, training, sun-bath, barbeque, skate
Common activities in winter time:
Ice-skating, skiing, play, sliding
Comments: Free wording
Point object (s) Comments: Free wording
Point object (s), max 3 objects Comments: Free wording
Point object (s) Comments: Free wording

us an opportunity to test if the method Children’s
Maps could be transferred from research to prac-
tical use. We decided to use the version that we
had developed and tested in 2006 with special
focus on traffic safety (Berglund et al., 2009).
Our conclusions from the research-pilots are that
this version works well as an over all tool and that
traffic safety has proved to be of high interest in
local planning with a child perspective.
By participating in the project we wanted to
get some answers on the following questions:
• How can the method work outside the re-
search context?
• How can the information be accessed, han-
dled and valued within a local authority?
Täby has about 62 000 inhabitants and is situ-
ated 15 km north of Stockholm city. The residential
areas consist partly of block of flats and partly
of areas with private homes. In the centre of the
municipal are a big shopping mall and some sports
arenas. The topography is varied, and green areas
with woods and lakes are situated close to the
residential areas. For an overview, please, look
at http://maps.google.se/maps, search for Täby.
The initiative for a co-operation came from
Technical Office at the municipal administration
in Täby. Technical Office is divided into five
units: Building Permits Unit, Surveying Division,
Street and Park Unit, Water and Sewage Unit and
Urban Planning Unit. The project leader and the
project members came from the Street and Park
Unit. The unit works with developing and main-
tenance of streets, roads, parks and forests. The
aims of the project “Place to Grow” were (Täby
kommun, 2009b):
• To make the voices of children and youth
heard in urban planning in Täby,
• To gather the results into a planning-friend-
ly tool for all who plan and design the pub-
lic places and out-door environments in the
municipality,

804
Children’s Maps in GIS
• To get concrete maps concerning places
and routes that children and youth use, and
in what way they use them,
• To get some general information about
what children and youth see as good or bad
about the neighborhoods of Täby,
• To draw conclusions from the answers
about which places and routes ought to be
protected, developed or changed.
During 2008, surveys were carried out in 28
out of 42 schools in Täby that have students in the
age between 11-18 years. The schools were chosen
to give a good geographical representation of the
municipality. In all a total of 635 children from
34 classes participated. 54% of the children were
11-12 years, 37% were 15-16 years and 9% were
17-18 years old (Täby kommun, 2009b). In Table
2 there is an overview of the schools participating
in the project. Schools with children with special
needs did not participate.
The Surveying Division of Täby provided the
digital maps with detailed and updated informa-
tion. As data were in MapInfo–format (tab) they
had to be exported to ESRI ArcView format (shape)
and then loaded into the Children’s Maps GIS
application on laptop-computers.
The Use of the Method
Children’s Maps in Täby
The method used in Täby is very much like the
method described earlier in this article, with some
Table 2. Numbers of schools participating sorted on level and school form
Total number of schools municipal
Compulsory schools 24 participating of 37 16 participating of 19
Age: 11-16 years possible possible
Upper secondary 4 participating of 7 pos- 1 participating of 2
schools sible possible
Age: 16-18
years
exceptions. Besides the spatial information the
project “Place to grow” wanted to get as much
verbal comments from the children as possible
about their perceptions of dangerous or unpleas-
ant places (question 5). The supervisor clarified
that boring or weary places could be marked as
well. The supervisors also reminded the children
to think about the winter aspect (part of question
4) (Täby kommun, 2009b).
This project did not prioritize the participation
of the teachers. However, 16 teachers from 11
schools were given the opportunity to take part.
All schools, teachers and children participated
voluntarily.
The data was transferred from the lap-tops into
the data server administrated by the IT-department
of the local government. Before storing the data,
some harmful comments on individuals were
taken away by the project team. A copy of the
data set was sent to us for research purposes and
back-up. The administration of data showed to
be a time-consuming activity when using the ap-
plication developed for small-scale research in a
large scale survey.
REPRESENTATION OF DATA
The project team had full access to data through
an internal digital support system. The team
represented data in several maps and charts pre-
sented it in a report (Täby kommun, 2009b). The
project also produced additional information as
independent Not participating
8 participating of 18 7 schools with children
possible with special needs, 6 of
them are independent
schools
3 participating of 5 1 school for students
possible with special needs, 2
schools with vocational
education
805

field notes, analysis and presentations that have
not been published.
A summary of the results is a part of the com-
prehensive plan (exhibited for consultation until
September, 15, 2009), under the heading Public
Interests – important out-door places for children
and youth. The comprehensive plan highlights the
following conclusion:
“[…] many popular places have some qualities
in common – space, variation, people and vicin-
ity. […] The biggest problem for children in the
local environment is traffic, followed by lack of
illumination and worries about other people’s
behavior. Bad maintenance is another factor that
makes places difficult to use.” (Täby kommun,
2009a, p. 39).
The results were also presented on a seminar
April 14, 2009. Participants were politicians and
employees representing the Technical Office, the
Child Care and Elementary School and Culture and
Leisure. In the seminar and in the report “Place
to Grow” (Täby kommun, 2009b), the following
aspects were brought forward.
In general there are no big differences noted
between how boys and girls are moving around
except in the 9th grade (15-16 years), where there
are some notable differences. The boys report more
physical activities than the girls. The girls report
that they “hang around”, looking at activities,
talking to each other. When it comes to safety/
security, girls 15 years and older, have made
more comments than the boys on places that they
perceive as insecure. Such places can be where
people are gathering at nights, empty places or
badly illuminated places like woods and parks.
Places to meet friends are a favorite in all ages.
Many children have marked places indoors as the
shopping mall, training grounds or at a friend’s
home as a favorite place. Most comments in all
ages about insecure places concerns traffic, lack of
illumination and suspicious and unsavory people.
Places that are often used also have most marks
806
Children’s Maps in GIS
as dangerous or insecure places. The explanation
given is that these are places the children know
well, and therefore have a lot of information about.
An example is the central shopping mall that has
most marks for favorite place and as a place that
can be perceived as insecure. A closer look at the
comments shows that the indoor environment,
with a lot of people and activities is perceived as
secure, and the parking lot outside is perceived
as empty no mans land and insecure. Also, the
children have made some comments on rebuild-
ing of parks that in one case has been successful
according to the comments, in another case the
comment from the children are “it was better
before”. Some new places for skating and biking
are asked for in the survey, and that information
is also highlighted in the report.
The participants of the seminar expressed their
appreciation of the information of children’s use
and perceptions on the local environment and saw
connections to prioritized issues like on school-
road safety and security issues in the outdoor
environment, as well as a basis for prioritizing
maintenance of streets and parks.
A follow-up meeting was held on May 18,
2009 with one researcher and people in the local
administration that were most likely to use the
information from the children (the GIS-expert
from Surveying Division and employees from the
Street and Park Unit and from the Urban Planning
Unit together with Child Care and Elementary
School and Culture and Leisure departments).
The question was: what part of the dataset is the
most interesting, and how do you want to access
it? The following alternatives were discussed: a)
plain thematic maps (Figure 1), b) map visualizing
some attributes e.g. places for meeting friends
(Figure 2), c) summarized maps (Figure 3), and
d) text information connected to geographical
object (Figure 4).
The participants from the administration an-
ticipated to use the information in urban planning
and in the detailed planning of projects concern-
ing traffic safety, green-structure planning and
Children’s Maps in GIS

Figure 1. A thematic map showing all the places marked in a part of Täby without editing afterwards.
The attributes are not visualized. Symbolization of the basic map is different from what the children saw
on the screen when drawing the sketch map.

Figure 2. A map showing places for meeting friends. It is one of the multiple-choice options that are
given in question number 4. Each bar represents one annotation from one child.

Figure 3. A representation using the centroids of each polygon to construct a density map. When there
are many centroids of the polygons close to each other, the color gets darker.

807
 Children’s Maps in GIS

Figure 4. A map showing places that are perceived as hazardous or dangerous. Connected to the symbol
are text annotations made by the children. The examples on the right are directly translated from Swedish.

maintenance of open spaces. The Technical Office
had access to data from the survey, through the
internal digital support system while the Child
Care and Elementary School unit did not have
access to the digital information. Because of the
character of the information, officials from the
Child Care and Elementary School unit expressed
their interest in having access as well.
In the report “Place to Grow” there is no
presentation of the information given by the
teachers although there is some information in the
dataset. Figure 5 shows the difference between
the information given by the children and the
information given by the teachers of one of the
schools. It shows the same pattern that we have
seen in our research pilots; schools tend to use
other areas than the students do themselves, and
with less concentration towards central parts of
the neighborhoods.
DISCUSSION
By reporting on this initial test of the method
Children’s Maps in a real-world planning project
we want to discuss problems and possibilities with
implementation into a local government context.
We also reflect on empowerment of children and
school and on the digital divide within the local
authority administration.

Figure 5. Focus in “Place to grow” has been on children’s mapping. Some teachers have contributed
with information as in this example.

808
Children’s Maps in GIS
On a general level we could ask if the project
“Place to Grow” shows that Children’s Maps can
function as a vehicle to facilitate young people’s
influence on the local environment. One conclu-
sion is that the project allows the opportunity for
children to participate to a degree that corresponds
with level 3 “Children’s views are taken into ac-
count, and given due weight in decision making”
(Shier, 2001, p. 111). A weakness of the method
is that young people are delivering data, yet have
neither ownership nor control over it. The maps
produced are adapted to the conventional planning
process, which we judge crucial for the information
to be taken into account in real-world planning.
The power of the children is embedded in the fact
that their knowledge and perceptions are visual-
ized, and have the same technical significance
as other forms of planning information. At this
point, the deliberation between different aspects
becomes political issue.
Lessons Learned by Stepping
Out of the Research Context
The method has shown to function outside the
research context as a tool for collecting and storing
information. Information from over 600 children
concerning their use and perceptions of the local
environment has been collected within the local
administration. On the whole the results are con-
sistent with the results from the research–pilots.
The seven questions (see Table 1) in our research
studies are tested in different social contexts and
in different age-groups. The development and tests
are described in detail in Berglund (2008) and
supplemented in Berglund et al. (2009). By using
the same pre-formulated questions in every survey
we assert that the questions are not biased by any
ongoing planning debate or by personal agendas.
The “Place to Grow” project had no procedures
to communicate back to the children (and teachers)
on the impact of their participation. Even if their
information is taken into account, they may not
be aware of it and the project did not provide for
further communication within the planning pro-
cess. This must be regarded as a weakness when
considering the empowerment of children, making
it less likely to positively influence their attitude
to future contributions to society. Feed-back is,
however, prescribed by us in the steps for the use
of Children’s Maps. This is based on numerous
studies of participation projects with children stat-
ing that being listened to and getting feed-back
are cornerstones needed for a fair involvement
of children in a participation process (Chawla,
2002b). In this sense the “Place to Grow” project
is incomplete, as well as our own research-pilots.
As reported, some of the results from the
project “Place to grow” have been integrated in
the consultation for the comprehensive plan yet
it is too early to assess the real impact on the lo-
cal environment or on the empowerment of the
children. If the administration can not assimilate
the information in its different activities, it may
be another case of tokenism (Hart, 1997).
The supervisors came from the Street and Park
Unit of Täby Municipality. The result indicates
that the supervisors have encouraged the children
to make comments on favorite places (question 6)
and suggestions for improvements (question 7),
besides the expressed aim of gathering as much
verbal information as possible about unpleasant
and dangerous places. The extent of the comments
from Täby is larger than from the research-pilots.
The influence from the supervisors can imply
that the children are used as a transmitter for the
supervisors own views, which could lead others to
assess the information as less trustworthy. On the
other hand, the Children’s Maps can be seen as a
tool for getting on speaking terms with children
and youth, and thus gaining information that is
otherwise difficult to attain. This is important to
consider when choosing persons to instruct the
children.
One bias that the Täby case highlights is the
choice of schools participating in mapping. If the
case is just a small project, concerning one small
village this might not be an issue. When the method
809

is used for gaining information about a whole mu-
nicipality, there needs to be some consideration on
how the schools and classes are chosen. In Täby,
schools and classes with students with special
needs were excluded. We have not reflected or
tested how the method can or cannot be used by
children with different kind of disabilities. But to
exclude these students from the use of Children’s
Maps is certainly not our intention.
In the report “Place to grow”, it is noted that
most of the children were able to orientate on
the map quite well, except from one district with
a very hilly topography, were the slopes are not
shown in the basic orientation map (Täby kom-
mun, 2009b, p. 16). This is an aspect that we
have not encountered earlier and, of course, it
is important that the basic map used for orienta-
tion represents features in the landscape that are
important for children.
Another point to observe is the low number
of teachers participating in the mapping. The
project leader was reluctant to include teachers’
mapping in the project as the focus was on how
children used and perceived the outdoor environ-
ment. However, we know from our own pilots that
teachers might be reluctant to use the computer
themselves to input information. Some have will-
ingly done so while others have preferred to be
helped by the supervisor. This observation is in
line with the reporting from the Swedish National
Agency for Education (Skolverket, 2009). We
have not engaged ourselves in developing the
teacher’s graphical interface to the same degree
as the children’s version. The interface that we
have used has still to prove that it functions. As
the teacher does not report a personal use but
functions as a representative for the school’s use
of the outdoor environment, it is not considered
to be a significant problem that not all teachers
are involved. The school’s use and need for good
environments and safe routes for transportation
can be reported by those teachers who are willing
and able to do so.
810
Children’s Maps in GIS
The project “Place to grow” shows that there
is a great demand for information from children
and from various administrations. In Täby there
were demands of information from different units
within the Technical Office as well as from the
Child Care and Elementary School and from Cul-
ture and Leisure. The most obvious demands came
from the park unit, from the traffic department
and from people involved in “Safe in Täby” (our
translation), a project within the local government.
As we have stated earlier in this article, maps
(and GIS) represent power and this means that
parts of the administration that cannot express
their interests in maps (or GIS) have a weaker
position when it comes to maintaining their
interests in urban planning. Child Care and El-
ementary School Unit have until now had little
interest in geographic information. This may be
a result of different approaches toward different
sources of information and technologies within
an administration.
The seminars showed that the written com-
ments were of high interest. As our focus is on
the spatial information, we have not considered
how to analyze or categorize the written infor-
mation. The participants wanted the information
to be categorized for easy access. Qualitative
analyzes of the text information calls for time
and competence to carry it out in a qualified way.
One solution is to have standard categories in the
questionnaire, yet if we restrict the children to
placing information into predefined categories,
we are afraid that this will produce biased and
incomplete results. By this we mean that it is
better to let children express themselves in their
own words in relation to personal feelings such
as favorite places, dangerous places and wishes
for change. It is then possible to make different
interpretations while the source material is kept
intact. This ambition may lead to a somewhat
less administration-friendly representation, yet
the strength of the written text is that it provides
an easier medium to hear the voice of the child,
than by simple markings on the map.
Children’s Maps in GIS
Digital Divide
The geographical databases produced by local
government in Täby are of high quality regarding
scope, scale of details and actuality. It is mainly
used as a background map for the internal digital
system, containing a large amount of different
information used in administration. Geographic
information was not accessible for Child Care
and Elementary School units within the internal
digital support system. The preliminary finding
from Täby tells us that the use of GIS is still in-
consistent across the administration, and that it is
still a technique that demands training to master.
In Täby, as in many municipalities in Sweden,
there are just a couple of people engaged in the
technical support of the data system that have the
mandate to set up and decide the functionalities
that can be obtained by users in administration.
The GIS-expert plays a crucial role in deciding
what is possible to do or not.
The full potential of the internal digital system
with interactive maps seems not to have been used
or even requested. The same situation is reported
from research on bottlenecks with implementation
of planning support system (Vonk et al., 2005). If
the information stored in digital information sys-
tem is not relevant for you, why use the systems?
Interactive maps and multimedia representation
of data helps the user to construct knowledge. If
important information is stored in files, charts and
archives or unreadable static maps they have very
little influence. If relevant information, such as
children’s experiences of the outdoor environment,
is accessed through an interactive information
system, this will enhance the use of the informa-
tion, and maybe even transform it into knowledge.
Need for Improvements
on Children’s Maps
Problems with the GIS-application in its cur-
rent stage are; access to the GIS application, the
adaptation of the maps in the GIS-application to
each case and the administration of data. From the
very beginning we aimed at an Internet solution,
we now have funding for transferring the GIS-
application into Internet so we will address issues
of unequal access to hardware between different
schools as mentioned by the Swedish National
Agency for Education (Skolverket, 2009). We
will also develop standard procedures for catego-
rizing and analyzing data, making the threshold
as small as possible for the user to interactively
explore the data.
Parallel to this, we will write a manual de-
scribing, among other things, how the supervi-
sors should conduct themselves when using the
method and how we consider it could be combined
with other means for gaining deeper information
and higher levels of participation. We will also
develop guidelines concerning feed-back to chil-
dren and teachers and further explain the role of
the teacher’s map.
CONCLUSION
The information that young people and teachers
can provide is asked for in many contexts, not
only in urban planning. The use of GIS as an
interactive tool for analyses and presentations is
not well established in Swedish local authorities.
In Täby it is still an expert tool, and especially
the social administration has not integrated the
technique in the daily work. Here the representa-
tion on static maps still plays an important role in
communication processes between the municipal
administrations as well as with the public.
We look forward to transferring Children’s
Maps to Internet. We do not aim for the application
to be open for use by everyone, only accessible
within projects, to secure that the information
given by children and youth is not misused. Be-
sides, local authorities, a school, or an organization
with children’s wellbeing on the agenda could get
access to the tool. As the use of Google maps and
Web 2.0 is spreading, we think it is more important
811

than ever that the knowledge and the interests of
children and youth are made visible and accessible.
We believe that the presence of Children’s Maps
on Internet will contribute to this and in the end
hopefully be helpful in giving young people the
influence on local planning and maintenance of
their neighborhoods that they have the right to.
REFERENCES
Baraldi, C. (2003). Children’s social participa-
tion in the town of adults. In Christensen, P., &
O’Brian, M. (Eds.), Children in the city: Home
neighbourhood and community (pp. 184–205).
London: RoutledgeFalmer.
Berglund, U. (2008). Using children’s GIS maps
to influence town Planning. Children. Youth and
Environments, 18(2), 110–132.
Berglund, U., & Nordin, K. (2007). Using GIS
to make young people’s voices heard in urban
planning. Built Environment, 33(4), 469–481.
doi:10.2148/benv.33.4.469
Berglund, U., Nordin, K., & Eriksson, M. (2009).
Barnkartor i GIS och trafiksäkerhet (Children’s
maps in GIS and traffic safety) (Tech. Rep.). Up-
psala, Sweden: Rapporter Institutionen för stad
och land.
Chawla, L. (2002a). Toward better cities for
children and youth. In Chawla, L. (Ed.), Growing
up in an Urbanizing World (pp. 15–34). London:
Earthscan.
Chawla, L. (Ed.). (2002b). Growing up in an
urbanising world. London: Earthscan.
Christensen, P. (2003). Place, space and knowl-
edge: Children in the village and the city. In
Christensen, P., & O’Brian, M. (Eds.), Children
in the City: Home neighbourhood and com-
munity (pp. 13–28). London: RoutledgeFalmer.
doi:10.4324/9780203167236
812
Children’s Maps in GIS
Christensen, P., & O’Brian, M. (2003). Children
in the city: Introducing new perspectives. In
Christensen, P., & O’Brian, M. (Eds.), Children
in the city: Home neighbourhood and com-
munity (pp. 1–12). London: RoutledgeFalmer.
doi:10.4324/9780203167236
Christensen, P., & Prout, A. (2002). Work-
ing with ethical symmetry in social research
with children. Childhood, 9(4), 447–497.
doi:10.1177/0907568202009004007
Crampton, J. W., & Krygier, J. (2006). An in-
troduction to critical cartography. ACME An
International E-Journal for Critical Geographies,
4(1), 11-33. Retrieved Mach 29, 2009 from http://
www.acme-journal.org
Driskell, D. (2002). Creating better cities with
children and youth. London: Earthscan.
Elwood, S. (2002). GIS use in community plan-
ning: A multidimensional analysis of empower-
ment. Environment & Planning A, 34, 905–922.
doi:10.1068/a34117
Elwood, S. (2006). Critical issues in participatory
GIS: Deconstructions, reconstructions, and new
research directions. Transactions in GIS, 10(5),
693–708. doi:10.1111/j.1467-9671.2006.01023.x
Flyvbjerg, B. (2001). Making social science
matter: Why social inquiry fails and how it can
succeed again. Cambridge, UK: Cambridge
University Press.
Freeman, C., & Aitken-Rose, E. (2005). Future
shapers: Children, young people, and planning in
New Zealand local government. Environment and
Planning. C, Government & Policy, 23, 227–256.
doi:10.1068/c0433
Habermas, J. (1984). The theory of communica-
tive action (Vol. 1). Boston, MA: Beacon Press.
Children’s Maps in GIS
Hart, R. A. (1997). Children’s participation: The
theory and practice of involving young citizens in
community development and environmental care.
London: Earthscan.
Healey, P. (1997). Collaborative planning: Shap-
ing places in fragmented societies. Basingstoke,
UK: Palgrave Macmillan.
Healey, P. (1999). Institutionalist analysis, com-
municative planning, and shaping places. Jour-
nal of Planning Education and Research, 19(2),
111–121. doi:10.1177/0739456X9901900201
Holloway, S., & Valentine, G. (2003). Cyberkids:
Children in the information age. London: Rout-
ledgeFalmer.
Horelli, L. (1998). Creating child-friendly envi-
ronments: Case studies on children’s participation
in three European countries. Childhood, 5(2),
225–239. doi:10.1177/0907568298005002008
Horelli, L., & Kaaja, M. (2002). Opportunities
and constraints of internet-assisted urban plan-
ning with young people. Journal of Environ-
mental Psychology, 22, 191–200. doi:10.1006/
jevp.2001.0246
Jiang, B., Huang, B., & Vasek, V. (2003). Geovi-
sualisation for planning support systems. In Geert-
man, S., & Stillwell, J. (Eds.), Planning support
system in practice (pp. 177–191). Berlin: Springer.
Korte, W. B., & Hüsing, T. (2006). Benchmark-
ing Access and Use of ICT in European Schools
2006. Retrieved from http://empirica.com/pub-
likationen/documents/No08-2006_learnInd.pdf
Kyttä, M., Kaaja, M., & Horelli, L. (2004).
An internet-based design game as a media-
tor of children’s environmental visions. En-
vironment and Behavior, 36(1), 127–151.
doi:10.1177/0013916503254839
Lynch, K. (Ed.). (1977). Growing up in cities:
Studies of the spatial environment of adolescence
in Cracow, Melbourne, Mexico City, Salta, Toluca,
and Warszawa. Cambridge, MA: MIT Press.
MacEachren, A. (2004). 1995). How maps work:
Representation, visualization, and design (pp.
iii–iv). New York: Guilford.
Matthews, H., Limb, M., & Taylor, M. (1999).
Young people’s participation and representation
in society. Geoforum, 30, 135–144. doi:10.1016/
S0016-7185(98)00025-6
Norsk Form. (2005). Barnetråkk: Barns tilgjen-
gelige uteareal (Children’s tracks: Children’s
accessible outdoor environments). Retrived
June 20, 2009 from http://www.norskform.
no/?V_ITEM_ID=1282
Percy-Smith, B., & Malone, K. (2001). Making
children’s participation in neighbourhood settings
relevant to the everyday lives of young People.
PLA Notes, 42, 18–22.
Pickles, J. (Ed.). (1995). Ground truth: The social
implications of geographic information systems.
New York: Guilford.
Plester, B., Blades, M., & Spencer, C. (2006).
Children’s understanding of environmental repre-
sentations: Aerial photographs and model towns.
In Spencer, C., & Blades, M. (Eds.), Children and
their environments: Learning, using and designing
spaces (pp. 42–56). New York: Cambridge Univer-
sity Press. doi:10.1017/CBO9780511521232.004
Rogers, P. (2006). Young people’s participation in
the renaissance of public space: A case study of
Newcastle upon Tyne, UK. Children. Youth and
Environments, 16(2), 105–130.
Sager, T. (1994). Communicative Planning Theory.
Avebury. UK: Avebury Press.
813

Sager, T. (2006). The logic of critical com-
municative planning: Transaction cost al-
teration. Planning Theory, 5(3), 223–254.
doi:10.1177/1473095206068629
Shiers, H. (2001). Pathways to participation: Open-
ings, opportunities and obligations. Children &
Society, 15, 107–117. doi:10.1002/chi.617
Skolverket. (2009). IT-användning och IT-kom-
petens i förskola, skola och vuxenutbildning (IT
usage and IT skills in preschool, school and adult
education) (Dnr 75-2007:3775). Retrieved from
http://www.skolverket.se/publikationer?id=2192
Täby kommun. (2009a). Det nya Täby: Översikt-
splan 2010-2030, utställningshandling maj 2009,
allmänna intressen (The new Täby: Comprehen-
sive plan 2010-2030, consultation May 2009,
public interests). Retrieved June 20, 2009 from
http://www.taby.se
Täby kommun. (2009b). Plats att växa (Place to
grow). Retrieved June 13, 2009 from http://www.
taby.se/Miljo-natur-halsa/Planeringsdokument/
Plats-att-vaxa/
The FreeChild Project. (2008). Retrieved No-
vember 15, 2009 from http://www.freechild.org
Tonucci, F., Prisco, A., & Horelli, L. (2004). A
comparison of the models for children’s participa-
tion in decision-making, in Rome and Helsinki.
In L. Horelli & M. Prezza (Eds.), Child-Friendly
Environments: Approaches and lessons (pp.
83-93). Espoo, Finland: Helsinki University of
Technology.
This work was previously published in the International Journal of Information Communication Technologies and Human
Development (IJICTHD), Volume 2, Issue 2, edited by Susheel Chhabra and Hakikur Rahman, pp. 1-16, copyright 2010 by
IGI Publishing (an imprint of IGI Global).
814
Children’s Maps in GIS
ULI. (2008). Lägesbild GI Sverige: en lägesbild
av användandet av geografiska informations-
sytem i Sverige, resultat från delundersökningen
offentliga sektorn (Situational awareness GI
Sweden: Study on use of geographic informa-
tion and geographic information technology in
Sweden, public sector). Retrieved from http://
www.geoforum.se/page/158/332
UNICEF. (1990). Convention on the rights of the
child. In First Call for Children (pp. 43–79). New
York: UNICEF.
UNICEF. (2001, March). The children as com-
munity researchers. Retrieved November 15, 2009
from http://www.unicef.org/teachers/researchers/
United Nations. (1993). [Earth summit - the United
Nations programme of action from Rio. New York:
United Nations Dept. of Public Information.].
Agenda (Durban, South Africa), 21.
Urban Green Spaces Taskforce. (2002). Green
spaces, better place: Final report. London: De-
partment for transport, local government, and
the regions.
Vonk, G., Geertman, S., & Shot, P. (2005). Bottle-
necks blocking widespread usage of planning
support systems. Environment & Planning A,
37, 909–924. doi:10.1068/a3712
Wilhjelm, H. (1999). Hvor har du vært? - ingen
steder: Miljøtilknyttende infrastruktur og barns
hverdagsliv - en kunnskapsoversikt (Where have
you been? nowhere: Environmental infrastructure
and children’s everyday life- a review). Trond-
heim, Norway: Norsk senter for barneforskning.
Woolley, H., Spencer, C., Dunn, J., & Rowley, G.
(1999).Thechildascitizen. JournalofUrbanDesign,
4(3), 255–282. doi:10.1080/13574809908724451
 815

Chapter 50
Landscape Biodiversity
Characterization in Ecoregion
29 Using MODIS
Nitin Kumar Tripathi
Asian Institute of Technology, Thailand

Aung Phey Khant

Asian Institute of Technology, Thailand

ABSTRACT
Biodiversity conservation is a challenging task due to ever growing impact of global warming and cli-
mate change. The chapter discusses various aspects of biodiversity parameters that can be estimated
using remote sensing data. Moderate resolution satellite (MODIS) data was used to demonstrate the
biodiversity characterization of Ecoregion 29. Forest type map linked to density of the study area was
also developed by MODIS data. The outcome states that remote sensing and geographic information
systems can be used in combination to derive various parameters related to biodiversity surveillance
at a regional scale.

1. INTRODUCTION of natural selection to fit in the constellation of

A natural environment is self-renewing, self-
perpetuating and stable one, in which every or-
ganism contributes in some way, however, small
to the overall stability. In natural ecosystems,
the plants and animals have evolved at their own
pace and in their own way under the influence
DOI: 10.4018/978-1-4666-2038-4.ch050
certain environmental factors or niches. In the
process, they help to sustain others, each species
controlling its own population growth and at the
same time limiting of other species, so that a
reasonable ecological balance may be achieved
and maintained for hundreds of years The satellite
remote sensing can identify the important param-
eters for biodiversity characterization like size,
fragmentation, porosity, patchiness, interspersion

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
and juxtaposition at the landscape level. The role
of remote sensing is emphasized in quick ap-
praisal of regional biodiversity surveillance. This
becomes of high importance in present context of
biodiversity loss due to climate change.
There is an urgent need to inventory and
monitor indicators of biological diversity such as
species richness and habitats. Remotely sensed
data provide a means to accomplish part of this
task, but there has been no comprehensive scien-
tific framework to guide its effective application
(Stoms & Estes, 1993). Most of the discussion
concerning potential roles for remote sensing in
biodiversity assessment has come from conserva-
tion biologist and ecologists (Soule & Kohm, 1989;
Noss, 1990; Lubchenco et al., 1991). The remote
sensing community has had little involvement to
date in supporting biodiversity research, largely
concentrating instead in the global change domain
(Stoms & Estes, 1992). Very little quantitative
analysis has been accomplished to determine the
actual value of remote sensing and geographic
information systems in biological research.
Remote sensing provides spatial data, which
are less used but they are the powerful source to
acquire accurate, up-to-date information essential
for conserving biodiversity and wildlife habitat
mapping. Although technically complex, the
remote sensing techniques have revolutionized
the process of data gathering and map making.
The combinations of Remote Sensing (RS) and
Table 1.­
Ecoregion Name
Bioregion
Major Habitat Type
Ecoregion Number
Political Unit(s)
Ecoregion Size
Biological Distinctiveness
Conservation Status
Conservation Assessment
816
Geographical Information System (GIS) have
proven to be very effective tools to analyze the
landscape patterns for biodiversity characteriza-
tion at various levels.
2. STUDY AREA
Kayah-Kayin and Tenisserim ecoregion (Ecore-
gion 29) are the richest in species in mainland
Southeast Asia, for this area is the cross road to
exchange species among different geographic
regions of Holartic, Oriental and Greater Sundas
Island. On the other hand, this ecoregion forming
a juncture of the Indo-Chinese, Indo-Burmese, and
Malaysian floral and faunal elements. Formerly,
it is divided into the Kayah-Karen Montane
Rain Forests (ecoregion 51: Table 1) and the
Tenasserim-South Thailand Semi-evergreen Rain
Forests (ecoregion 53: Table 2). In the year 2000,
World Wide Fund For Nature (WWF) scientists
team combined these two regions and defined
as ecoregion (29). The flora and fauna in this
region is distinct and includes several endemic
species. Among the ecoregions of Indochina, this
ecoregion contains some of the highest diversity
of both bird and mammal species found in the
Indo-Pacific region
This ecoregion encompasses the mountainous,
semi-evergreen rain forests of Thailand, Myanmar
and Malaysia, and includes the extensive lowland
: Kayah-Karen Montane Rain Forests
: Indochina
: Tropical and subtropical moist broadleaf forest
: 51
: Myanmar, Thailand
: 119,200 km2
: Globally outstanding
: Relatively intact
: III
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS

Table 2.­

Ecoregion Name : Tenasserim-South Thailand Semi-evergreen Rain Forests

Bioregion : Indochina
Major Habitat Type : Tropical and subtropical moist broadleaf forest
Ecoregion Number : 53
Political Unit(s) : Myanmar, Thailand, Malaysia
Ecoregion Size : 96,900 km2
Biological Distinctiveness : Globally outstanding
Conservation Status : Relatively intact
Conservation Assessment : III

plain that lie between the peninsular mountains
and which until recent decades supported exten-
sive lowland forest. The southern margin of this
ecoregion is defined by the Kangar-Pattani flo-
ristic boundary (Whitmore & Sayer, 1992). Figure
1 shows the geographic extent of the study area.
2.1. Vegetation
The vegetation of this ecoregion includes both
tropical and subtropical moist broadleaf forest,
montane forests, lowland rainforest; and a higher
proportion of evergreen broad-leaved species. This
ecoregion represents the semi-evergreen forests of
the Kayah-Karen Mountains in the broad transition
zone between the subtropical broadleaf evergreen
forests in the north and the southern tropical and
dry deciduous forests in Tenesserim and southern
Thailand (Figure 2). Tropical hardwood trees in
the family Dipterocarpaceae dominate forests
throughout the ecoregion. Forests to the east are
dominated, especially at the lower elevations,
trees that have a drought-deciduous phenology,
while the west-facing slopes are a mixture of de-
ciduous and evergreen species. At low elevation
(<= 1000 m) on the east side of the Tenasserim
Hills, the potential vegetation consists of drought
deciduous forest or savanna woodland. Higher
elevations support much richer broad-leaved
forest communities with a mixture of evergreen

Figure 1. Map of Ecoregion 29

817
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
Figure 2. Land use map of study area
and deciduous species. Forests of teak Tectona
grandis represent the climax vegetation at low
elevation in the absence of fire, but today the
teak forests are nearly extirpated in Thailand
and also declining rapidly in Myanmar. At 800-
1200m, a well-developed undestroyed shrub
grows beneath a tall, closed forest canopy that
includes some very large, buttressed trees that
share an affinity with tropical Asia together with
temperate tree taxa in the families Magnoliaceae
and Lauraceae (IUCN, 1991). Although fire is
frequent today, there is little consensus as to the
historical frequency of fire or its importance in
this ecosystem. An important unresolved question
is whether fire (mostly anthropogenic) or pre-
monsoon drought stress (non-anthropogenic) is
primarily responsible for limiting species diversity
in these places. This ecoregion remains relatively
unexplored scientifically, especially those parts
that lie in Myanmar, it will very likely yield more
818
biological surprises. However, after Thailand
banned timber exploitation in its forests in 1988,
Myanmar granted large logging concessions to
Thai companies, and illegal timber extraction in
Myanmar by Thai loggers has become common
in recent years (WWF & IUCN 1995).
2.2. Fauna
This ecoregion contains one of the most intact
vertebrate faunas of Indochina, including one of
the richest assemblages of mammals in Asia. The
fauna is also distinctive, with characteristics of
the islands of the Malay Archipelago as well as
the mountains of China and India. The relatively
intact and contiguous hill and montane habitat
has potential for conserving large landscapes
that will provide adequate habitat to maintain
a viable population of Asia’s largest carnivore,
the tiger (Panthera tigris), and Asian elephant
(Elephas maximus). This ecoregion lies within a
high priority, Level (I) Tiger Conservation Unit
(Dinerstein et al. 1997). This range of forests in
conjunction with the Kayah-Karen Mountains
represents some of the best landscapes for Asian
elephant conservation in Indochina. Numerous
other mammals are of conservation significance,
primarily the elusive and endemic Fea’s muntjac
(Muntiacus feae). The population of the Malayan
tapir (Tapirus indicus), the only Old World tapir
representative, has been drastically reduced. It
survives in the hill and montane protected areas
of this ecoregion and scattered pockets throughout
peninsular Malaysia and Sumatra. More than 25
pairs have been found in some of the last remain-
ing forest, and that forest is now contained within
Bang Kram wildlife sanctuary (Stewart-Cox,
1995). Several primate species are found in these
forests, and include the threatened banded langur
(Trachypithecus melalophus) and slow loris (Loris
nycticebus), a small, nocturnal prosimian. Other
species of conservation concern include the Dyak
fruit bat (Dyacopterus spadiceus), the endangered
clouded leopard (Pardofelis nebulosa), common
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
leopard (Panthera pardus), sun bare, binturong
(Arctictis binturong), gaur (Bos gaurus), and
banteng (Bos javanicus) (Stewart-Cox, 1995).
The wide diversity of habitats within this
ecoregion, from deciduous forests in the north to
seasonal evergreen forests in the south, habitats
lowland to montane, make it one of the richest in
bird species for the entire Indo-Pacific. A total of
560 bird species have been recorded there. The
ecoregion is the fourth richest in the Indo-Pacific
region for mammals with 168 known species.
These include one ecoregional endemic species,
the tiny Kitti’s hog-nosed bat (Craseonycteris
thonglongyai). Some of the other mammals of con-
servation importance include several threatened
species such as the tiger (Panthera tigris), Asian
elephant (Elephas maximus), gaur (Bos gaurus),
banteng (Bos javanicus), wild water buffalo (Buba-
lus arnee), southern serow (Naemorhedus suma-
traensis), clouded leopard (Pardofelis nebulosa),
Malayan tapir (Tapirus indicus), wild dog (Cuon
alpinus), Asiatic black bear (Ursus thibetanus),
Assamese macaque (Macaca assamensis), stump-
tailed macaque (Macaca arctoides), smooth-
coated otter (Lutrogale perspicillata), great Indian
civet (Viverra zibetha), and Particoloured flying
squirrel (Hylopetes alboniger). Sumatran rhinoc-
eros is believed to have inhabited remote regions
of the Tenasserim Hills in recent years, but this
critically endangered species is now thought to
have been extirpated from this ecoregion. Hunt-
ing has decimated most of the large mammal
populations, such as elephant, banteng, gaur and
tiger (IUCN, 1991). On the other hands, many
species are declining to unsafe population levels,
important habitats are destroyed, fragmented and
degraded, and the ecosystems are destabilized
through climate change, population, invasive
species, and human impact.
The relatively intact, contiguous habitat has
potential to conserve large landscape that will
provide adequate habitat to maintain a viable
population of Asia’s largest carnivore, the tiger,
as well as other species of critical conservation
significance. Several of Myanmar and Thailand’s
largest and most intact wildlife reserves lie within
this ecoregion, including Myinmo-let-khat Na-
tional Park (proposed) and Huai Kha Khaeng
Wildlife Sanctuary (2,575 km2) and several other
protected areas with which it forms a contigu-
ous network. Huai Kha Khaeng is prized for the
high diversity of cat species it supports, and it’s
relatively intact vertebrate communities and intact
lowland dipterocarp forests. Moister habitats on
the Myanmar side of the Tenasserim ranges also
include significant amounts of intact habitat,
probably still in better condition overall than the
forest on the eastern (Thailand) side of the range.
However it is difficult to assess ecological condi-
tions in the forests of eastern Myanmar at this time.
The existing (50) protected areas that cover
32% (35,030 km2) of the ecoregion and most of
these protected areas are located in Thailand. Large
blocks of intact seasonal evergreen forest habitats
are still remaining in Myanmar, but these are
mostly not protected. Some protected areas have
been designated in the portion of this ecoregion
that lies within Myanmar, but their effectiveness
is difficult to assess at this time due to the political
instability of the region.
2.3. Agriculture
Shifting cultivation is the main cause of defor-
estation throughout the region. But in areas like
northern Thailand, where previously nomadic
tribal peoples have been settled, pressure exists to
convert forest into more intensive agricultural land
devoted to cash crops like cabbage, coffee, and
lychee. Opium replacement efforts, in Thailand,
have compelled local people to grow alternative
crops that require more cultivated land area and
higher pesticide inputs. The land requirements
of an increasing population have forced itinerant
farmers to reduce the cycle of cultivation-fallow
periods, and have pushed them deeper into the
forest and into more marginal areas.
819
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
Lowland area in Peninsular Thailand (Bang
Kram) still supports a significant amount of late
successional forest. Other areas support vast
tracts of rubber plantation, monocultures of a fast-
growing, short-lived tree species native to South
America, and plantations of oil palm. Pineapple
is grown there as a rotation crop following the
removal of senescent rubber trees. Paddy is also
grown in some lowland areas. Unfortunately,
none of these crops with the possible exception of
paddy provide any significant support for natural
biological diversity. Hill slopes support more
native forest than the lowland areas, and the hill
forests of Southern Thailand are relatively intact,
although swidden (slash and burn) agriculture is
still practiced in hill areas in the northern part
of the ecoregion. Mature forest cut for swid-
den agriculture is generally succeeded in this
ecoregion by a grassy sub-climax that supports
far fewer species than the mature forest (IUCN,
1991). General land-use map of the study area is
shown in Figure 2.
2.4. Geology
Ecoregion (29) consists of hills of Paleozoic lime-
stone that have been much dissected by chemical
weathering. The overhanging cliffs, sinkholes, and
caverns characteristic of tropical karst landscapes
are all-present in this ecoregion.
2.5. Landscape Structure
Terrain throughout much of this ecoregion is
rugged and intricately folded. Hillsides tend to
be steep and ridges exceed 2,500 m elevation.
Valley bottoms are narrow but fertile and tend to
lie at about 300 m elevation. The western slopes
drain into the Salween River that flows through
Myanmar and entered Gulf of Martaban in the
Indian Ocean. The eastern slopes drain into the
Chao Phraya River that drains into the gulf of
Thailand.
820
2.6. Climate
Annual precipitation increases southward as
the length of the dry season and the magnitude
of pre-monsoon drought stress declines. The
southern mountain ranges receive rain from both
the northeast and southwest monsoons so that,
unlike mountain ranges further north, there is
no significant rainshadow. The entire region has
a monsoon climate with warm, moist summers
and mild winters that tend to be dry. Overall an-
nual rainfall average is 1,500 mm to 2,000 mm.
Although this ecoregion lies within the Tropic of
Cancer, winter temperatures can be quite cool,
especially at the higher elevations where frost
has been recorded from the northern part of the
ecoregion. West-facing slopes (Myanmar side)
face the Bay of Bengal and receive more precipi-
tation; and East-facing slopes (Thailand side) lie
within a partial rainshadow and tend to be drier.
This climatic difference is clearly reflected in the
vegetation. The Köppen climate system placed
this ecoregion in the Tropical Wet Climate Zone
(Townsend & Cohoon, 1999).
3. DATA USED
Fortunately, enough data is available with institu-
tions and agree to share the information. These
are Biogeographical zone map, protected areas
networks and species list from WWF-Thailand,
topographic/ vegetation/ LU & LC/ geology/ soil
maps from Department of Forestry (Myanmar and
Thailand); and annual rainfall and temperature for
ten years from Department of Geography. Lists of
data used for this study are as follows:
• MODIS satellite data for study area with
the spatial resolution of 250 m.
• Digital map of entire ecoregion with the
scale of 1: 250000
• Topographic/ Vegetation/ LU&LC/
Geology/ Soil map
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS

Figure 3. Kayah-Kayin and Tenisserim ecoregion

• Annual rainfall and temperature for 10
(Ecoregion 29) in MODIS image
years
• Other ancillary data

MODIS image in Figure 3 shows the overview

of the ecoregion 29 and surrounding. This image
was utilized to extract the biodiversity character-
istic of the ecoregion.

4. BIODIVERSITY RICHNESS
ESTIMATION MODEL

Growing concerns over the loss of biodiversity

has spurred land managers to seek better ways of
managing landscapes at a variety of spatial and
temporal scales. A number of developments have
made possible the ability to analyze and manage
entire landscapes to meet multi-resource objec-
tives. The developing field of landscape ecology
has provided a strong conceptual and theoretical
basis for understanding landscape structure, func-
tion, and change (Forman & Godron 1986; Urban
et al., 1987; Turner, 1989). Growing evidence that
habitat fragmentation is detrimental to many spe-
cies and may contribute substantially to the loss
of regional and global biodiversity (Saunders et
al., 1991; Harris, 1984) has provided empirical
justification for the need to manage entire land-
scapes, not just the components.
As growing numbers of researchers and re-
source managers rely on digital geographic data,
and look to remotely sensed imagery as a source
of data for their GIS, issues of geographic data
models and remote sensing scene models (Good-
child, 1992, 1994; Strahler et al., 1986), as well
as image processing algorithm development, are
central to research on the integration of remote
sensing and GIS.
A rotational approach to the management of
biological resources requires advanced technol-
ogy in many areas of information processing.
The conceptual framework is quite complex, with
many interactions between species and environ-
ments. Advanced computation and massive data
archives would produce quantitative improvement
in analysis of biodiversity through more accurate
and efficient solution. Visualization and interac-
tivity would produce greater understanding and
clarity on the status of the environment and effort
to maintain it.
A landscape is composed typically of several
types of landscape elements. Of these, the matrix is
the most extensive and most connected landscape
element type, and therefore plays the dominant role
in the functioning of the landscape. As landscape
patterns are complex and heterogeneous, variety
of matrices are needed to measure different aspect
of patterns to characterize biodiversity. Digital
satellite data provides a consistent and complete
spatial data for a very large area and may be con-
sidered very useful tool for measuring landscape

821
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS

patterns (patch, fragmentation, terrain complexity, 4.1 Indices for Biodiversity

porosity, interspersion and juxtaposition) to esti- Characterization
mate biodiversity richness together with species
diversity index (Shannon and Simpson). And also, Species Diversity Index
the landscape spatial configurations influence
ecological processes such as biodiversity, habitat Species diversity can be described as the number
or animal population dispersal and abundance. of species in a sample or habitat per unit area. The
Landscapes are distinguished by spatial rela- indices are as follow:
tionships among component parts. A landscape
can be characterized by both its composition Shannon’s diversity index (log base 10 and natu-
and configuration, and these two aspects of a ral log)
landscape can independently or in combination
affect the ecological processes and organisms. H = -Σ(Pilog[Pi])
The difference between landscape composition
and configuration is analogous to the difference H = -Σ(Piln[Pi])
between floristic and vegetation structure that are
commonly considered in biodiversity studies at Simpson’s diversity index (D)
the within-patch scale.
Spatial data layers are generated using digital D = Σ(Pi2)
satellite data through digital image processing
and classification. In addition to that spatial and D = 1-Σ(Pi2)
non-spatial data from other ancillaries data sources
are collected. Database is generated using GIS Where,
software and analyzed. Then biodiversity char-
acterization was done analyzing the landscape Pi = the total number of species
parameters of patch, fragmentation, porosity, H = Shannon’s index value
interspersion and juxtaposition, and using ground D = Simpson’s index value
base data on species richness, terrain complexity
and diversity index. Fragmentation
First, and more importantly, subsets of par-
ticular landscape have to be analyzed in order to Forest fragmentation occurs when large, continu-
evaluate the distribution of spatial characteristics ous forests are divided into smaller blocks, either
(to detect the high diversity or where the habitat by roads, clearing for agriculture, urbanization,
is more endangered because of the loss of con- or other human development.
nectivity).
N
Therefore, it is necessary to know how the F = ∑ Di
extent of analyzed pattern is going to influence the i=1
results of estimation and whether the obtained val-
ues are reliable to the entire region under analysis. Where,

F = Index value of Fragmentation

Di = dissimilarity value for the ith boundary be-
tween adjacent cells

822
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
N = number of boundaries between the adjacent
cell
Generally, forest type map is classified into
two classes such as forest and non-forest. A grid
cell (n * n) is convolved with the spatial data layer
with a deriving number of forest patches within
the grid cell. This will be repeated by moving
the grid cell through the entire spatial layer. The
output layer with patches was derived.
Patchiness
Patchiness is a measure of the density of patches
of all types or number of clusters in given mask.
It is a measurement of number of polygons over
a particular area.
Pj = ((∑Di) / Nj ) * 100
Where,
N = Nj is the number of boundaries between
adjacent intervals along Transect j
Di = Dissimilarity value for the i th boundary
between adjacent Cells
In landscape ecology, patches are spatial
units at the landscape scale. Patches are areas
surrounded by matrix, and may be connected
by corridors. The geomorphology of the land
interacting with climate factors, along with the
other factors such as the establishment of flora
and fauna, soil development, natural disturbances,
and human influences work to determine patch
size, shape, location, and orientation (Forman &
Godron, 1986).
Five major types of patches:
(1) spot disturbance patch
(2) remnant patch
(3) environmental resource patch
(4) introduced patch
(5) ephemeral patch
Porosity
Porosity is the measure of number of patches or
density of patches within a landscape, considers
only number of patches, not size.
n
PO = ∑ Cpi
l=1
Where, Cpi = number of closed patches of ith
cover class
Interspersion
Interspersion is count of dissimilar neighbors with
respect to central pixel or measurement of the
spatial intermixing of the vegetation types. It can
also be used to represent the landscape diversity.
n
I = ∑ (∑ Fi / n)
l=1
Where, Fi is shape of factor
For the determination of interspersion a
‘Convolution’ will be used with the forest type
map to compute the number of dissimilar pixels
in the nearest neighborhood. The computation
is performed in an interactive mode through the
entire spatial layer to derive an output intersper-
sion layer. Calculation of interspersion gives the
magnified view of resistance; the central pixel or
class has with respect to its surroundings.
Juxtaposition
Juxtaposition is defined as measure of proximity
of the vegetation of vegetation. It’s measurement
includes relative weight assigned by the impor-
tance of the adjacency of two cover types for the
species in question. It is species-specific measure-
ment i.e. edge between cover and quality of edge.
J = ∑ Di (Ji) / Jmax
823
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
I=1
Where,
Di = the shape desirability weight for each cover
type combination
Ji = the length of edge between combination of
cover types on either side of an edge
Jmax = the average total weighted edge per habitat
unit of good habitat
A grid cell of size 5 (e.g. 5*5) is convolved
with the derived layer in an interactive manner
by assigning higher weight to natural vegetation
and lower weight to unnatural vegetation. The
juxtaposition helps in characterizing the param-
eter porosity with respect to natural or unnatural
(manmade) vegetation type. Since porosity can be
understood as one of the important factors influ-
encing the disturbance index, the added weightage
through juxtaposition gives the right perspective
in ultimately deriving disturbance index.
4.2 Landscape Configuration Metrics
for Biodiversity Characterization
Landscape Similarity Index (LSIM)
n Mean Nearest Neighbor (MNN), Measure
∑a ij
j=1
LSIM = Pi = (100) distance of an individual patch is the shortest
A
LSIM equals total class area (m2) divided by total
landscape area (m2), multiplied by 100 (to convert
to a percentage); in other words, LSIM equals
the percentage of the landscape comprised of the
corresponding patch type.
Patch Density (PD)
ni
PD = (10, 000)(100)
A remotely sensed data has kept pace with the infor-
824
PD equals the number of patches of the correspond-
ing patch type (NP) divided by total landscape
area, multiplied by 10,000 and 100 (to convert
to 100 hectares).
Landscape Shape Index (LSI)
m
∑e ik
LSI = k-1
2 À_A
When LSI = 1, the landscape consists of a single
patch of the corresponding type and is circular
(vector) or square (raster); LSI increases without
limit as landscape shape becomes more irregular
and/or as the length of edge within the landscape
of the corresponding patch type increases.
There are a number of landscape configuration
metrics to characterize the biodiversity spatially.
The spatial statistics output table created from a
landscape pattern analysis contains all the results
of the analysis (Class area, Total landscape area,
number of patches, Total edges, Edge density,
Mean shape Index, Fractal Dimension, Mean Near-
est Neighbor Distance, Mean Proximity Index,
Interspersion and Juxtaposition Index, Shannon’s
Diversity Index, Shannon’s Evenness Index).
of patch isolation, means the nearest neighbor
distance to a similar patch (edge to edge). The
mean nearest neighbor distance is the average of
these distances (meter) for individual classes at
the class level and the mean of the class nearest
neighbor distances at the landscape level.
5. RESULTS AND DISCUSSIONS
The improvement in spectral, spatial and temporal
resolutions and better digital image processing of
mation needs of the characterizing biodiversity.
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
The geographic information system (GIS) tool has
developed fast with greater facility of large area,
multiple spatial and non-spatial data integration
and analysis.
5.1 Results Discussion
on Biodiversity Richness
Estimation Model
In the biodiversity richness estimation model, the
most significant landscape patterns, which can be
extracted from the digital satellite data were used
for biodiversity characterization.
For meaningful estimation and characterization
of biodiversity, the degree of variation of each
index is related to landscape patterns that are
used for biodiversity richness estimation model
and not to artifacts derived from the methodologi-
cal problems involved in the measurement. This
uncertainty associated with the index estimated
is arguably one of the major limitations of this
kind of quantitative analysis for biodiversity
characterization. And it is also greatly related to
the scale such as spatial resolution of satellite data
and the spatial extent (i.e. total area).
The biodiversity richness estimation model can
provide variety of landscape value and allowing
obtaining wide range of patterns with intermediate
level of spatial differences, in which patchiness
and fragmentation is generated.
The use of these indices and landscape con-
figuration matrices instead of real world landscape
data is preferable in this study because this ap-
proach is possible to generate and isolate different
factors affecting the landscape to characterize the
biodiversity.
The influences of fragmentation, patchiness,
porosity, interspersion and juxtaposition can be
adequately separated, and thus the biodiversity
richness estimation using landscape patterns and
configuration metrics can be specifically analyzed
avoiding confusion with other land cover data
(Table 3).
Patch shape and orientation also play an im-
portant ecological role. An ecologically optimum
patch shape usually has a large core with some
curvilinear boundaries and narrow lobes (Forman,
1995). This shape may allow both interior species
and edge species to flourish. Patch shapes also
determine the edge length.
Porosity is very important parameter for
conservation aspect at landscape and regional
level. The porosity (1) provides an over all clue
to the change of species isolation present and to
the potential genetic variability present within
population of animals and plants in regional
level, (2) sometime indicates the presence of high
conservation value and lower interaction among
landscape elements, homogeneity and low frag-
mented habitats, (3) higher porosity value indicates
the higher interaction among landscape elements,
heterogeneous and high fragmented habitats, and
(4) disturbance in natural landscape increases
patch density and decrease matrix connectivity.
Interspersion Juxtaposition Index, Measure of
patch adjacency, the result will be zero when the
distribution of unique patch adjacencies becomes
uneven and 100 when all patch types are equally
adjacent. At the landscape level it is a measure of
the interspersion of the each patch in the landscape.
The dispersal ability of species depends on
spatial organization of landscape. The fragile
or rare species occur only in highly connected
landscape mosaic. Higher value of interspersion
Table 3. Indices for Biodiversity characterization
of study area
No. Index name Result value
1 Patchiness 60.98
2 Fragmentation 2.009
3 Porosity 2,962
4 Interspersion/ Juxtaposition 49.015
5 Landscape similarity 48.07
6 Patch fractal dimension 33
825
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
means dispersal ability of the central class will
be low or reduce.
Landscape Similarity Index Matrices ap-
proaches 0 when the corresponding patch type
(class) becomes increasingly rare in the landscape.
LSIM equal to 100 when the entire landscape con-
sists of the corresponding patch type; that is, when
the entire image is comprised of a single patch.
Fractal dimension is a measure of shape com-
plexity that can be computed for each patch and
then for landscape, or it can be computed from
the landscape as a whole.
These diversity measures are influenced by
two components: (1) richness and (2) evenness.
Richness refers to the number of patch types
present and evenness refers to the distribution
of area among different types. Richness and
evenness are generally referred to as the compo-
sitional and structural components of diversity,
respectively. Simpson’s diversity index is more
sensitive to richness than evenness. Thus, rare
types have a disproportionately large influence
on the magnitude of the index. This diversity
index has been applied to measure one aspect of
landscape structure and landscape composition
to characterize biodiversity. Thus, the higher the
value the greater the likelihood that any randomly
drawn patches would be different patches that is
greater in diversity.
In many biodiversity applications, the pri-
mary interest is in the amount and distribution
of a particular patch. Forest fragmentation is a
landscape-level process in which forest tracts
are progressively sub-divided into smaller, geo-
metrically more complex, and more isolated forest
fragments as a result of both natural processes and
human land use activities. This process involves
changes in landscape composition, structure, and
function and occurs on a backdrop of a natural
patch mosaic created by changing landforms
and natural disturbances. Forest fragmentation
is the prevalent trajectory of landscape change
in several human dominated forest regions, and
826
is increasingly becoming recognized as a major
cause of declining biodiversity.
The use of diversity measures in community
ecology has been heavily criticized because diver-
sity conveys no information on the actual species
composition of a community. Species diversity is
a community summary measure that does not take
into account the uniqueness or potential ecological,
social, or economical importance of individual
species. A community may have high species
diversity yet be comprised largely of common
or undesirable species. Conversely, a community
may have low species diversity yet be comprised of
especially unique, rare, or highly desired species.
Although these criticisms have not been discussed
explicitly with regards to the landscape ecological
application of diversity measures, these criticisms
are equally valid when diversity measures are
applied to patch types instead of species. In addi-
tion, these diversity indices combine richness and
evenness components into a single measure, even
though it is usually more informative to evaluate
richness and evenness independently (Tables 4
and 5, Figures 4, 5 and 6).
Therefore, a landscape with many habitats will
be richer than a less heterogeneous one. How-
ever, if habitat patches become too fragmented
and disjunctive, as typically results from human-
induced land use conversion, regional richness
declines.
5.2 Forest Type Distribution
The potential of remote sensing and Geographic
Information Systems is displayed in providing
accurate and timely information on forest density
map. The forest types are classified base on the
dominant species composition. As a result, in the
study area various forest types were classified and
mapped with reasonable amount of accuracy. The
vegetation type map of ecoregion 29 is classified
into nine types: (1) Deciduous Broadleaf Forest,
(2) Semi-deciduous Broadleaf Forest, (3) Ever-
green Needleleaf Forest, (4) Lowland Evergreen
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS

Table 5. Overall species richness estimation

No. of plots Estimation

Summarized Index
1 31.6
2 45.4
3 55.4

9.997
0.785
1.179
1.502

0.110
4 59.4
5 61.6
Plot 12

6 64.4
0.703
0.195
1.058
1.505

5.12
7 66.2
8 67.2
Plot 11

0.096
10.45
1.196
1.519
0.788

9 67.8
10 68.4
Plot 10

11 68.6
0.735
0.147
6.826
1.107
1.505

12 69
Plot 9

7.727
0.676
0.129
1.052
1.556

Broadleaf Rain Forest, (5) Semi-evergreen Moist

Plot 8

10.575
0.799
0.095
1.192
1.491

Broadleaf Forest, (6) Upper Mountain Forest,

(7) Lower Mountain Forest, and (8) Mangrove
Plot 7

11.686

Forest (Table 6).

0.808
0.086
1.228
1.519

The degradation activities such as extensive

Table 4. The results of Shannon’s and Simpson’s species richness index.

shifting cultivation, excessive logging, increasing

Plot 6

16.228
1.431
0.887
0.062
1.27

in human population, development of infrastruc-

ture, and conversion of forest habitats to agricul-
Plot 5

10.383
1.415
0.841
0.096

tural land have altered the natural landscape to

1.19

great extent in ecoregion 29. Because of these

Plot 4

10.008
1.232
1.58
0.78
0.1

Table 6. Forest density distribution

Plot 3

14.046
0.848
0.071
1.287
1.519

Percentage
No. Forest Type Area (km2)
(%)
Plot 2

9.577
0.789
0.104
1.229
1.556

Deciduous/semi-
1 deciduous broadleaf 26,886.55 12.45
forest
Plot 1

0.136
7.343
1.101
1.431
0.77

Disturbed natural
2 20,438.47 9.46
forest
3 Evergreen forest 4,597.43 2.13
Simpsons Diversity (1/D)
Simpsons Diversity (D)
Shannon Hmax Log10.

Semi-evergreen moist
4 55,246.34 25.58
Shannon H’ Log 10.

broadleaf forest
Index

5 Lower montane forest 22,093.60 10.29

Shannon J’

6 Upper montane forest 1,327.88 0.67

7 Mangrove 839.82 0.38
8 Others 84,521.71 39.04
Total 215,951.79 100.00

827
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS

Figure 4. Graph of Shannon’s species richness index

Figure 5. Graph of Simpson’s species richness index

Figure 6. Graph of overall species richness estimation

828
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
increased anthropogenic activities, the natural
landscape have become fragmented. In addition
to forest cover lost, fragmentation has a great
impact on biodiversity. Bio-habitat degradation
due to fragmentation is also considered to be one
of the major threats to biodiversity and often re-
duced genetic diversity, consequently.
The fragmentation of forest communities
leads to high patch density and also affected to
various ecological processes such as species dis-
tribution and degradation of habitats of keystone
and indicator species. For core habitat zonation,
as well as mapping of habitat for endangered
species deals with these transformations. These
land transformations have resulted in the altera-
Figure 7. General status of Ecoregion 29
Figure 8. Forest density distribution
tion of natural habitats and have brought in lots
of biodiversity. Figures 7 and 8 show the status
of forest in ecoregion 29.
6. CONCLUSION
Study demonstrates that MODIS data can be
used to characterize biodiversity and help the
understanding of landscape diversity and ter-
restrial ecosystem functions at the regional or
landscape level.
The assessment to calculate the biodiversity
richness using important indices (Fragmentation,
Patch, Terrain complexity, Porosity, Interspersion
829
 Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS
and Juxtaposition, Species richness, etc.) includ-
ing landscape configuration metrics generated
important quantitative information about the status
of different resources and landscape patterns of
Ecoregion 29. The use of these index indicate the
status of entire landscape and, therefore, can be
assumed as essential tools for biodiversity appli-
cation towards conservation strategies to prevent
biodiversity loss.
As known, the biodiversity is a reflection of
several biotic, abiotic and climatic factors, there-
fore the ecological relationships was considered
as the most important underlying factor for their
in situ conservation.
Earlier study lack using moderate resolution
satellite data such as MODIS towards biodiversity
mapping. This data is regularly available and is
well suited for regional level biodiversity surveil-
lance (inventory and monitoring) to extract the
data and information of landscape patterns, land
cover, and vegetation. Conventional methods for
biodiversity mapping are very costly and time
consuming. Since biodiversity loss and ecological
imbalance are the major concern of climate change
impact on biological richness of ecoregion 29, the
surveillance based on MODIS satellite data and
limited ground observations will facilitate the
prioritization for conservation.
The results from biodiversity richness estima-
tion model indicate that the overall biodiversity
richness is still high except the southern part of
Thailand.
REFERENCES
David, M. S., Davis, F. W., & Hollander, A. D.
(1994). Hierarchical representation of species
distribution for biological survey and monitoring.
In Goodchild, M., Parks, B. O., & Steyaert, L.
(Eds.), Environmental modeling: Progress and
research issues (pp. 445–449). Fort Collins, CO:
GIS World Books.
830
Dinerstein, E. (1997). A framework for identifying
high priority areas and actions for the conserva-
tion of tigers in the wild. World Wildlife Fund-US.
Foreman, R. T. T. (1995). Some general prin-
ciples of landscape and regional ecology. Land-
scape Ecology, 10(3), 133–142. doi:10.1007/
BF00133027
Foreman, R. T. T., & Gordon, M. (1986). Land-
scape ecology. New York, NY: John Wiley & Sons.
Goodchild, M. F. (1994). Integrating GIS and
remote sensing for vegetation analysis and model-
ing: Methodological issues. Journal of Vegetation
Science, 5(5), 615–626. doi:10.2307/3235878
Harris, L. D. (1984). The fragmented forest: Is-
land biogeography theory and the preservation
of biotic diversity. Chicago, IL: University of
Chicago Press.
IUCN. (1991). Caring for the Earth. The World
Conservation Union, Gland.
Lubchenco, J., Olson, A. M., Brubaker, L. B.,
Carpenter, S. R., Holland, M. M., & Hubbell, S.
P. (1991). The sustainable biosphere initiative:
An ecological research agenda: A report from
the Ecological Society of America. Journal of
Ecology, 72(2), 371–412. doi:10.2307/2937183
Noss, R. F. (1990). Indicators for monitor-
ing biodiversity: A hierarchical approach.
Conservation Biology, 4(4), 355–364.
doi:10.1111/j.1523-1739.1990.tb00309.x
Saunders, D. A., Hobbs, R. J., & Margules, C. R.
(1991). Biological consequences of ecosystem
fragmentation: A review. Conservation Biol-
ogy, 5(1), 18–32. doi:10.1111/j.1523-1739.1991.
tb00384.x
Soule, M. E., & Kohm, K. A. (Eds.). (1989).
Research priorities for conservation biology.
Washington, DC: Island Press.
Landscape Biodiversity Characterization in Ecoregion 29 Using MODIS

Stewart-Cox, B. (1995). Wild Thailand. Cam-
bridge, MA: MIT Press.
Stoms, D. M., & Estes, J. E. (1993). A re-
mote sensing research agenda for mapping
and monitoring biodiversity. International
Journal of Remote Sensing, 14, 1839–1860.
doi:10.1080/01431169308954007
Strahler, A. H., Woodcok, C. E., & Smith, J. A.
(1986). On the nature of models in remote sensing.
Remote Sensing of Environment, 30(2), 121–139.
doi:10.1016/0034-4257(86)90018-0
Townsend, P. A., & Cohoon, K. P. (1999). Sen-
sitivity of distributional prediction algorithms
to geographic data completeness. Ecological
Modelling, 117(1), 159–164. doi:10.1016/S0304-
3800(99)00023-X
Turner, M. G. (1989). Landscape ecology: The
effect of pattern on process. Annual Review
of Ecology and Systematics, 20(1), 171–197.
doi:10.1146/annurev.es.20.110189.001131
Urban, D. L., O’Neill, R. V., & Shugart, J. R.
(1987). Landscape ecology. Bioscience, 37,
119–127. doi:10.2307/1310366
Whitemore, T. C., & Sayer, J. A. (1992). Tropi-
cal deforestation and species extinction. IUCN,
Gland.

This work was previously published in Land Use, Climate Change and Biodiversity Modeling: Perspectives and Applications,
edited by Yongyut Trisurat, Rajendra P. Shrestha and Rob Alkemade, pp. 133-149, copyright 2011 by Information Science
Reference (an imprint of IGI Global).

831
832

Chapter 51
The Spatially Interactive
Literature Analysis
System Study Tool:
A GIS-Based Approach to Interpreting
History in the Classroom

Alyssa K. Moore
University of Redlands, USA

Lillian I. Larsen
University of Redlands, USA

Diana Stuart Sinton

University of Redlands, USA

ABSTRACT
The integration of spatial technologies with humanistic approaches to scholarship is expanding and
creating new opportunities and challenges for students. Documenting information about source material
is a fundamental step of the historical research process. Traditional geospatial data documentation is
maintained within standardized metadata forms, but these are not well suited for documenting historical
sources and are cumbersome for people new to GIS and metadata. The Spatially Interactive Literature
Analysis System Study Tool (SILAS.ST) addresses this and other usage issues for students in undergradu-
ate humanities courses. Through customized toolbars and map templates designed to be used with Esri
software, SILAS.ST lowers the barriers to understanding the parameters of spatial datasets while at the
same time models the processes of historical inquiry, documentation, and communication of research
results. The use of natural language and focused prompts help students begin to understand complex
mapping topics such as authorship, relevance, purpose, and uncertainty. This prototype tool aids in the
introduction of digital mapping technologies to humanities students.

DOI: 10.4018/978-1-4666-2038-4.ch051

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
The Spatially Interactive Literature Analysis System Study Tool
INTRODUCTION
Maps and mapping have long held the interest of
historians and humanities scholars. The where of
geography and the when of history are intrinsically
linked, making place and location central concepts
for both disciplines (Jessop, 2008; Mostern, 2010;
Staley, 2007). For humanities scholars, maps have
traditionally served as visual texts that effectively
reveal patterns, support arguments, and illustrate
narratives (Ackerman & Karrow, 2007; Black,
1997; Edney, 1997; Sparke, 1995; Staley, 2007).
Maps and globes in digital form also serve as
research tools for exploring historical questions
(Holdsworth, 2003; Schultz et al., 2008; Univer-
sidad Francisco Marroquín, 2009).
The focus of humanistic map use is increas-
ingly shifting towards maps generated through
geographic information systems (GIS). Through
its flexible tools, GIS offers the capacity to or-
ganize, archive, analyze, and display prodigious
amounts of information in novel ways (Bailey &
Schick, 2009; Bodenhamer et al., 2010; Brennan-
Horley et al., 2010; Goodchild & Janelle, 2010;
Gregory & Healey, 2007; Jessop, 2004; Knowles,
2002). Large scale research projects on China (Bol,
2008), the Holocaust (Knowles et al., 2009), the
Salem Witch Trials (Ray, 2002), the Dust Bowl
(Cunfer, 2002), racially-based housing discrimi-
nation (Hillier, 2003), the development of the
railway system (Schwartz, 2010), and the Civil
War (Knowles, 2008) have been undertaken in
efforts to glean new insights from applying spatial
analysis to historical questions.
The growing availability of digital geographic
tools has also spurred increased interest in edu-
cation from a spatial perspective (Brown, 2010;
Sinton & Bednarz, 2007). New generations of
superior mapping tools and data resources have
significantly facilitated spatial analysis and, as
importantly, qualitative learning (Mostern, 2010).
This is due not solely to advances in mapping
technology. Rather the combination of intellectual
and technical skills has helped researchers and
students ask robust and interesting questions when
analyzing spatial patterns, allowing them to more
deeply probe spatial relationships and to investi-
gate causes for observed patterns (Gregory & Ell,
2007; Kwan & Ding, 2008; Schultz et al., 2008).
Humanities-oriented GIS applications have
particularly captured the imagination of teaching
faculty at many undergraduate institutions (Brown,
2008; Lloyd, 2001; Robinson, 2010; Sinton, 2009;
Sinton & Lund, 2007). Instructors in these settings
realize that simply mapping in “layers” can greatly
enhance students’ understanding of history and
place (Mostern, 2010; Staley, 2007). Even the
rudimentary process of manually overlaying clear
plastic sheets, printed with different themes and
map elements, reinforces the point “that places
are complex, and that mapping spatial phenom-
ena can reveal important, sometimes surprising
juxtapositions” (Knowles, 2000, p. 29).
At the same time, the role of GIS in humanities
scholarship and instruction is complex. Because
the tools were not created with historical sources
in mind, their analytical potential is more readily
applicable to data rich fields within the natural and
social sciences, less so in the humanities. Because
historical data are often inherently imprecise and
incomplete, measures of uncertainty and gener-
alization must also be taken into account when
interpreting historically informed spatial trends
and patterns (Gregory & Ell, 2007; Knowles,
2002; Rumsey & Williams, 2002). The quality
of historical data is difficult to evaluate with
respect to many of the categories foundational
to GIS (such as error and accuracy), since the
data are rarely formulated in ways conducive to
the exploration and manipulation typically used
with GIS software (Gregory & Ell, 2007; Pearce
& Louis 2008; Plewe, 2002). Such realities often
thwart attempts to confidently analyze and inte-
grate historical data within GIS (Brown, 2010).
Including historical data on digital maps
compels scholars to address issues of geographic
data quality and completeness. For example, in
an ancient frame, even data sets derived from the
833
 The Spatially Interactive Literature Analysis System Study Tool
most reliable sources, may be deduced from in-
herently nebulous archeological remnants (Reed,
2007). Such sources are susceptible to historical
inaccuracy, and some place locations will never
be known with certainty. This ambiguity is rarely
depicted in published maps. In turn, contemporary
digital maps, whether they are addressing histori-
cal themes or not, often fail to cite their sources
of information. This undermines and stymies
critical inquiry by students who tend to trust the
visual representation of mapped data (Schultz et
al., 2008). Other formats of map publications,
such as atlases, incorporate additional information
that may facilitate the interpretation of ambiguous
historical geographic data. Yet as a design format,
atlas pages are not often used by default in typical
layout templates for map production or within
classrooms exploring GIS-based instruction.
Humanities scholars and students alike also
find that the use of GIS poses initial challenges
(Brown, 2010; Jessop, 2004, 2008; Mostern,
2010). The GIS software most often available on
university campuses, Esri’s ArcGIS, is industrial
strength and designed for professional applica-
tions. It is not ideally suited to casual or educa-
tional use. Acquiring and processing data can be
cumbersome and learning to be an independent
user of the software can be an overwhelming
task. To make GIS an effective scholarly tool
for humanities students, it may be necessary to
guide and simplify its use. Furthermore, inquiry
into the factors behind the spatial patterns is more
consistently the final learning achievement, rather
than software expertise for its own sake (Gregory
& Ell, 2007; Kerski, 2003; Mostern, 2010).
With these challenges in mind, graduate students
and faculty at the University of Redlands (California,
USA) have built a series of GIS-based interfaces and
tools designed to support undergraduate humanities
students as they critically investigate ancient texts
and contexts (Benzek, 2008; King, 2009; Larsen,
2009). The interest to develop these has emerged
from partnerships between graduate students in the
University’s Master’s of GIS program and a faculty
834
member, Lillian Larsen, in the Religious Studies
department. With a focus on history and critical
reading of primary texts, these non-theological
religion courses have proven ideal undergraduate
“laboratories” for humanities students to explore
geospatial technologies. Conceiving of maps not as
“an end product; [but as] a research tool” (Gregory
& Ell, 2007, pp. 10-11), these interfaces have been
designed to allow students to visually overlay
information from various geographic sources and
store their own notes and observations for individual
layers. Default exploratory and drawing tools are
simplified and grouped, making the most common
visualization tasks more accessible. In addition,
customized templates are provided so that students
can generate layouts resembling an atlas page, a
format that accommodates maps, text, images, and
other graphics. This combination allows students
to integrate maps with the other range of media
they more typically use in their scholarship. The
additional explanatory text and documentation as-
sociated with each atlas page provides information
that aids in qualifying and interpreting the history
behind and within the included maps.
MAPPING AS A CLASSROOM
ACTIVITY
Initially, simple paper and pencil exercises were
used to introduce maps into the undergraduate
classroom. These exercises have proved an effective
preliminary step in generating insights. For ex-
ample, in a course on Christian Scripture, religious
studies students were tasked with representing the
geographical information included in the letters of
Paul, and the Acts of the Apostles, on two separate
maps. While there is broad scholarly consensus
that Acts is a well-fashioned literary reconstruction
of events being reported up to a full century after
Paul lived and wrote, and general agreement that
only seven of the thirteen letters ascribed to Paul in
fact derive from Paul’s hand, biblical geographers
have traditionally assigned these sources equal
The Spatially Interactive Literature Analysis System Study Tool
historical truth and weight and woven a seamless
map of the “three missionary journeys of Paul”
(Larsen, 2009). In creating their own maps, students
quickly recognized the disparities in data derived
respectively from the Acts of the Apostles, and the
undisputed letters in Paul. Their student-generated
maps differed significantly from the traditional map
of Paul’s missionary journeys. Creating maps that
visually differentiated between discrete historical
data layers, equipped students to develop their own
criteria for critically assessing the relative historicity
of ancient sources (and, derivatively, the maps used
to elucidate these sources). As importantly, students
gained a new appreciation for the value of asking
historical questions through a geographical lens.
With even more ambitious educational goals in
mind, the Spatially Interactive Literature Analysis
System (SILAS) was constructed as a web ser-
vice that first addressed issues of GIS usability
(Benzek, 2008; Larsen, 2009). Using ArcGIS
Server, the tool allowed students to add data lay-
ers from a pre-existing ArcSDE database and the
user interface supplied simple annotation tools.
Historic roads, populated places, and topography
were provided as an initial collection of layers.
Additionally, a scripture search web application
was accessible through the interface, enabling a
keyword search of biblical texts (Benzek, 2008).
A step up from paper mapping, GIS afforded an
even richer laboratory for facilitating learning.
While the gains were significant for introduc-
tory GIS classroom exercises, the scope of the
initial SILAS project did not extend into other
pedagogically-important areas of historical inter-
est, including the support of deeper data inquiries.
In particular, the complex, fragmentary, incom-
plete, and uncertain nature of historical data, of
great interest to scholars, was not readily explor-
able through the SILAS web service.
To address these issues, a second project phase,
the Spatially Interactive Literature Analysis System
Study Tool (SILAS.ST), focused on three areas:
data documentation, data discovery and exploration,
and cartographic design and layout (King, 2009;
Larsen, 2009). SILAS.ST is comprised of a col-
lection of customized toolbars, built to work with
ArcGIS 9.3, desktop software that was available
to University of Redlands students in computer
labs. One toolbar provides access to a set of forms
within which students can document important
information for individual data layers. A second
set of toolbars facilitates both data exploration and
design of maps as atlas pages. This platform also
supports data creation and editing, processes that
were harder to manage with the original web-based
version of SILAS. During classroom activities
involving GIS, students can add the SILAS.ST
toolbars to an ArcMap document, highlight any
data set or map layer, and display the customized
application for viewing or editing of pertinent map
information (Figure 1). The custom SILAS.ST
metadata-like forms behave as a regular metadata
editor, storing the information along with the dataset
for future access and editing by students working
independently or in groups.
The SILAS Study Tool was designed to encour-
age and support classroom learning objectives:
critical engagement with historical texts and con-
texts and exploration of complex cartographic
topics such as authorship, relevance, purpose,
uncertainty, and accuracy. The prototype was de-
signed during the spring of 2009, concurrent to the
offering of a course on the Historical Search for
Jesus. In this course, students used a variety of
media including digital historical maps, to explore
and construct a conceptual image of the historical
Jesus as recorded and stereotyped through time.
For young researchers and introductory GIS users,
we sought to create a venue for contemplating the
Figure 1. Final SILAS.ST Map Information tool-
bar. Accessed within ArcMap (9.3), this toolbar
gives students easy access to the customized
metadata-like forms while avoiding ArcCatalog.
835
 The Spatially Interactive Literature Analysis System Study Tool
cause of patterns and discrepancies among data
layers as information from multiple sources was
juxtaposed. Students overlaid and critiqued data
layers representing historical and modern biblical
sources, a process that challenged them to evaluate
the presumed authority of printed texts and maps.
Students were also encouraged to interrogate the
causes and generative forces that have contributed
to data being represented as they are. Working with
the Study Tool gave students greater understanding
of the historical relevance and uncertainty inherent
in the maps they create and study. Students simul-
taneously became more critical consumers of the
texts and map layers they encountered while in-
vestigating historical topics. Applying these ap-
proaches of historical inquiry to maps not only
helped student avoid “crude spatial-pattern spot-
ting” (Gregory & Ell, 2007, p. 118), but also cul-
tivated their senses of intellectual curiosity.
SILAS.ST is not a replacement for conven-
tional and standardized geospatial metadata,
which is still available for students as they are
using ArcGIS. However, in our experience, the
categories that populate traditional metadata
standards are opaque and rarely accessible to
students in humanities courses. In designing the
content of the metadata tool, we made careful use
of natural language, avoided disciplinary jargon,
and included gentle “prompts” to direct and guide
student responses to questions. We deliberately
embedded open-ended, subjective, and qualitative
parameters to model the lines of scholarly inquiry
being pursued in broader humanities courses.
Student input and feedback were essential
components of the SILAS.ST development pro-
cess from the beginning. When students did not
understand any particular component, or were
unclear about how they were to use the tools in
class, the design scheme was revisited to ensure
as much transparency as possible. Likewise, while
both SILAS.ST and its predecessor, SILAS, were
initially coupled with Religious Studies curri-
cula, their functionality was designed to support
learning in many other disciplines as well. Each
836
iterative version of the tool was further refined
with continued classroom use.
METHODS OF TAGGING MAP
DATA WITH KNOWLEDGE
Enabling the systematic and thoughtful consider-
ation of historical mapped data was the primary
motivation behind SILAS.ST. Its customized
forms blend portions of the traditional structure
of metadata with a more open-ended documen-
tation approach. The metadata-like forms were
modeled on two sets of form and function: source
documentation standards and common evaluative
methods used in humanistic fields. We considered
several existing schemes for their ability to docu-
ment, elucidate, and evaluate information. The
structure was then organized around different
categories of information evaluation using the
same type of terminology that a librarian might
employ in teaching students to question and as-
sess the sources and information they encounter.
We focused primarily on the identification and
qualification of data, attentive to the nature of
the information we expect students to evaluate.
The organizational structure of SILAS.ST
draws largely from an information literacy docu-
ment known as the CRAAP (Currency, Relevance,
Authority, Accuracy, and Purpose) Test (Meriam
Library, 2010). Designed with students in mind, this
sequence of evaluative questions encourages critical
assessment of the quality and original purpose of a
given body of information, an important component
of the research process. While it was originally
geared towards assessing information found on the
Internet, where many students automatically begin
their data searches, its value extends well beyond
this immediate function. Focusing the students’ at-
tention on the currency of information, the sequence
begins by asking students to note when a source
was published and last updated. Relevance targets
the appropriateness of the source by determining
the intended audience, which may influence the
The Spatially Interactive Literature Analysis System Study Tool
language and scholarly integrity of included infor-
mation. The named authority of the information is
an important qualifier, as the author represents a set
of credentials and qualifications that speak of their
ability to adequately contribute reliable information.
Students are also asked to assess the accuracy of
a source, whether it is supported by a substantive
body of evidence, or compromised by bias. Last
but not least, the purpose of the information sheds
light on the larger contextual picture for why the
information exists. Is the information’s purpose
to persuade? Entertain? Sell? (Meriam Library,
2010). By inviting students to assess a GIS data
layer, such as a scanned historical map, using the
same criteria that would be applied to any other
historical source underscores the textual aspect of
mapped information. It likewise equips students
with a familiar rubric on which they can base their
initial discussion and research.
In adapting the framework of the CRAAP
Test’s sequenced components, other sources were
considered for basic objective metadata categories.
Of these, the metadata standard from the Electronic
Cultural Atlas Initiative (ECAI) and the ArcGIS
Citation Data Model were particularly useful.
The ECAI Metadata Clearinghouse has its own
metadata standard for the purpose of cataloging
historical spatial data. As a collaborative venture,
ECAI’s information base draws on a worldwide
network of scholars working to create a digital atlas
of cultural information, tagged with spatial and
temporal data (University of California, 2007). The
ECAI documentation standard is composed of the
Dublin Core Standard. This widely used metadata
format documents both spatial and temporal ele-
ments in one field, called Coverage. For use in a
spatial database comprised of historical sources, a
single field documenting both time and place may
not be satisfactory, so ECAI extended the Coverage
field to include the bounding latitude and longitude
points, the relevant historical or archaeological
periods, resolution of a map or digital position
recording, and the geography by which the data
was aggregated (Johnson, 1998). By adding their
own contextualizing fields and redefining Dublin
Core’s Coverage to better document spatial and
temporal attributes, ECAI has made their database
searchable by location and time, within the frame
of Dublin Core formatting (Gregory & Ell, 2007;
Lancaster & Bodenhamer, 2002).
Further complicating the objective documenta-
tion of historical sources is how to best recognize
that the recorded location of a feature can be
subject to multiple interpretations. The ArcGIS
Citation Data Model (CDM) was designed to
connect bibliographic information to geographic
features in ArcGIS (Frye, 2008). The CDM enables
scholars to document why a given feature has been
digitally located where it has, and why others may
disagree. The interface links such information to
single geographic features by utilizing a database
schema that systematically joins and relates tabu-
lar data. Each object within a feature class can
reference up to three corroborating sources, as
well as one conflicting source, should one exist.
This referencing system ensures efficient use of
sources and establishes their fidelity.
Confidence in cartographic representations can
be established by contextualizing map sources or
even the cartographers themselves (Frye, 2008;
Harley, 1988). The CDM includes a notes field in
both the source table and feature layer table, inviting
insights and information that might help qualify a
source or feature. By applying the CDM to GIS
research, a variety of data sources can be incor-
porated, validated, easily maintained and shared.
SILAS.ST FOR DATA RESEARCH
AND DOCUMENTATION
By design, SILAS.ST incorporates elements from
each of the aforementioned schemes and standards.
Those most suitable for evaluative documentation
and objective metadata were selected and modified
so as to better align with our pedagogical purposes.
These elements were arranged on digital forms
under five different tabs: Quick Facts, Origin &
837
 The Spatially Interactive Literature Analysis System Study Tool

Authority, Relevance & Purpose, Accuracy &
Reliability, and Notes. The content of each tab
was then organized with a mind to supporting
critical and creative historical inquiry.
The Quick Facts section covers the basic ele-
ments that a student would initially need to know
about a map or data layer. This tab integrates the
ECAI notion of “coverage” of time and space,
but separates location and time into two differ-
ent categories (Figure 2). Instead of identifying
precise bounding coordinates for any particular
location, students are asked to provide a more
general place and time period. This encourages
classification of locations and time using descrip-
tors that are inherently meaningful to the student
user. Likewise, if the details of time or place are
ambiguous or conflicted, students can state this
using qualitative description.
The notions of scale and resolution were com-
plex and important ones to address in designing
SILAS.ST. These are difficult concepts for many
students to grasp, especially as quantitative values
(Goodchild & Janelle, 2010). What is the best way
to inform Google Earth-era students about scale,
when the digital mapping tools with which they
are most familiar have a notable absence of meta-
data? What does it take for students to appreciate
why the coastline of a country will appear different
within a GIS if it was originally digitized from
maps at 1:10,000 or 1:1,000,000 scale? If we
limit students’ ability to view data above a certain
scale, with the scale-dependent display functional-
ity available in GIS, what lessons are they learning?
When one zoomed in very close to the ground with
Google Earth or Google Maps imagery, a message
previously indicated that “imagery is not available
at this resolution.” In a GIS interface, no such
message will appear. At very large scales, Google’s
images now convert from an overhead view to the
“Street View” and one suddenly appears to stand
on the sidewalk (if you have zoomed to an urban
area, for example) and can look down the street.
What lessons about digital image resolution does
this impart to students?

Figure 2. Final SILAS.ST Map Information design for the ‘Quick Facts’ tab. Fields are methodically
grouped to promote comprehension of scale and resolution.

838
The Spatially Interactive Literature Analysis System Study Tool
Thus the growing familiarity with these online
maps and virtual globes, and their ability to zoom
and preserve resolution at each fixed extent level,
undermines their understanding of how digital data
in GIS can differ. For example, during one class
session a student zoomed in on a junction of two
vector roads that, from afar, appeared to intersect.
However, as the student continued to zoom in, the
roads no longer connected. This situation launched
a rich conversation about the limits of generating a
spatial dataset and how errors enter the digitizing
processes. Students gained a new awareness of
how the original scale of a map affects the nature
of the data, knowledge otherwise absent following
independent forays with virtual globes.
Understanding the differences among scale,
resolution and map size is a challenge, but link-
ing the topics may facilitate learning (Frye, per-
sonal communication, 2009; Rumsey & Williams,
2002). Here, the issue lies in maintaining a balance
between mathematically specific definitions and
more open-ended, conceptual parameters. Thus,
we added a field where one could indicate the
dimensions of the original source map that had
been used to generate the GIS layer the student
now observed, so that students could envision
how the data were now being represented in a
dramatically different size. Knowing the method
and purpose of preservation is also relevant in this
case, and we provided a tag to indicate the quality
and resolution of the original scan. Ultimately, we
included a combination of both subjective and
objective tags to reflect these ideas and metrics.
One of the goals of SILAS.ST is to move beyond
the when and where of historical sources and involve
students in thinking about how a source may have
been used and what factors drove its creation in the
first place (including student generated maps and
data layers). Thus the Origin and Authority section
prompts students to document the owner, distribu-
tor, publisher, or cartographer of the maps and data
layers they are studying (Figure 3). By breaking
down the various forms of map ownership, students
begin to understand that not all historic maps were
“published” in the ways that a contemporary map
might be. Through this section students can see
how historic maps may represent a certain school of
thought or tradition of map making. They are also
invited to consider a cartographer’s motivation for
making maps for different purposes. Vernaculars
of geographers and cartographers vary depending
on social and cultural background (Harley, 1989).
Historical maps often have a distinct lineage and
may have undergone many revisions and updates
(Black, 1997; Rumsey & Williams, 2002; Wilford,
2000). When researching the cartographer and the
map, students are prompted to consider how the
information was gathered, and in turn which other
maps might have been known or consulted. They
can assess the map maker’s level of familiarity
with the geography represented, as these elements
affect the relative reliability of a map.
In our experience with visual overlay of maps,
students were eager to single out the differences
they saw in change over time, or differences between
two or more maps produced within a single time
period. When students encounter these discrepan-
cies, SILAS.ST offers prompts in its Purpose and
Relevance section to consider and note what the
differences might mean and why they might exist,
such as the results of different map-making prac-
tices or the intentions of the cartographers (Figure
4). This subjective and evaluative section guides
students to critically evaluate the map layer as an
information source. By design, completing this
section is likely to require further independent
research by the student. Students are encouraged
to leave their name and document the references
they use for this research, so that future students
can add to and accrue information over time. In
this way, the tool retains the potential to serve as
an ongoing evaluative document for students, who
can revisit, revise and rethink its informational
content through consecutive semester courses.
The notions of uncertainty and accuracy have
long been central to mapping and GIS (Coucle-
lis, 2003; MacEachren, 1992; Plewe, 2002; Unwin,
1995; Zhang & Goodchild, 2002). Even the am-
839
 The Spatially Interactive Literature Analysis System Study Tool

Figure 3. Final SILAS.ST Map Information design for the origin and authority

Figure 4. Final SILAS.ST Map Information design for the relevance and purpose

840
The Spatially Interactive Literature Analysis System Study Tool
biguity associated with the words themselves
lends rich complexity to these ideas. Students
bring with them a prior conceptual knowledge of
“uncertainty” and “accuracy” but the depth and
breadth of how those ideas translate into mapping
and GIS is new to them. Moreover, students are
also learning how to apply these concepts as junior
scholars of history. Many enter the classroom
without appreciating the importance of critical
thinking in addressing historical texts, much less
maps (Robinson, 2010). The challenge lies in
introducing both sets of ideas to students so that
they can begin to appreciate complexity but not
be overwhelmed by a perceived burden of trying
to “fix” something. We debated the balance be-
tween quantitative and qualitative measures of
uncertainty and accuracy, designing an approach
to accommodate both in the Accuracy and Reli-
ability section of SILAS.ST (Figure 5).
In historical records, problems of incomplete-
ness, incoherency, and ambiguity are part of
documenting a narrative and interpreting it over
time (Gaddis, 2002; Plewe, 2002). Historical maps
Figure 5. Final SILAS.ST Map Information design for the accuracy and reliability tab
of the Holy Land, in particular, clearly show the
great effort expended in detailing particular fea-
tures relevant to the desired audience (Long, 2002).
Prior to the invention of precision instruments,
maps were simply a mental construct of space,
limited by the observer’s (or primitive instru-
ment’s) ability to measure. Early observances and
records may not be adequately preserved, or fall
victim to bias. Thus a large challenge to overcome
before using historical sources is adequate research
to compensate for the place names, geographic
features and measurement standards that have
changed or have not been properly recorded.
One quantitative measure where accuracy can
be applied is in georeferencing (Rumsey & Wil-
liams, 2002). Georeferencing involves the careful
placement of a digital image, often an historical
map, over the GIS-based representation of that
same geographic extent on the Earth’s surface.
Once the image is “tacked” down, it becomes
a layer within the GIS, aligned with the other
layers in the vertical stack of data. How well the
georeferencing placement process occurs is some-
841
 The Spatially Interactive Literature Analysis System Study Tool
thing that can be measured; the root mean square
error (RMSE) is one calculation of how far the
“tacks” have had to stretch the image to fit where
it has been placed. With explanation and practice,
students can learn not only to georeference maps
(and record their RMSE values within SILAS.
ST) but also to understand when and why some
maps can never be georeferenced with acceptable
measures of accuracy.
Storing information on data lineage and pro-
cessing is also a component of the Accuracy and
Reliability section. Ideally completed by a more
experienced GIS user who may have processed the
data, in class settings this information may be used
to facilitate discussion of GIS data creation and
conversion techniques. Such conversations may,
in turn, shed light for the students on the merits
and limitations of GIS for historical research. At
the same time, it is important that the documenta-
tion of data lineage and processing be written so
that students less familiar with GIS techniques
can understand how the original information has
already been manipulated in order to be viewed
within GIS.
SILAS.ST FOR DATA DISCOVERY
AND REPRESENTATION
In addition to the metadata-like forms avail-
able within SILAS.ST, we also created several
other customized toolbars designed to support
classroom-usage of GIS within undergraduate
Figure 6. Final SILAS.ST Explore Data Toolbar is comprised of common map exploratory functions
Figure 7. Final SILAS.ST Hand Draw Toolbar helps students edit graphic features and attributes
842
courses. This involved reorganizing and renaming
existing ArcMap functionality for the support of
historical inquiry.
To facilitate map navigation, the Explore Data
toolbar contains the common Pan and Zoom func-
tions for initial exploration of a dataset, along with
Bookmarks, so students can save extents they
wish to revisit later (Figure 6). For map overlay
visualizations, the Transparency and Swipe tools
were also included. The Add Layers command
opens to the ArcSDE database, containing the
original SILAS data layers and any others that
students may have contributed. The Search Bible
functionality opens up the SILAS Scripture Search,
in which students can perform a keyword search
of names or places included in the biblical text.
To encourage sketching as a form of inquiry
and data creation, the Hand Draw toolbar contains
common drawing functionality that supports the
use of graphic elements (Figure 7). Drawing al-
lows students to generate point, lines and polygon
features directly on the map, and provide their
own annotation. We have found this particular
functionality very important in our classes as we
often ask students to trace the routes of historic
figures, based on information they are extracting
from the texts as they read. By engaging students
in the tracking of historical journeys, reading
comprehension is improved because students are
required to read for geographical indicators (Oden,
1992). Students have the option to save their
drawings as a layer for future use outside of their
current map document.
The Spatially Interactive Literature Analysis System Study Tool

Lastly, we encourage students to frame their
use of maps within an atlas format, one that blends
the use of maps, text, images, and other graphics
to document and share their historical scholarship.
Not only does this model the use of multiple
sources and formats of information to document
historical inquiry, but it facilitates the creation of
professional, well-designed final research prod-
ucts. The Create Atlas (Figure 8) toolbar links to
customized map templates that include preset base
maps with ArcSDE database connections, map
layout schemes, and customized historical fonts
and symbols.
The final map templates were modeled on
layouts that students themselves had designed in
previous mapping exercises (Figure 9). The or-
ganizational choices reflect student interest in
comparing maps of similar geographic extents,
or using a single map to illustrate a journey.
NEXT STEPS AND CONCLUSION
As a prototype tool, SILAS.ST will undoubtedly
continue to evolve as it is used in more classroom
settings, with new data sets and new learning
objectives. Within this frame, the capacity for
ongoing student involvement in shaping and re-
shaping SILAS.ST is significant. In developing
both SILAS and SILAS.ST, student needs were

Figure 8. Final SILAS.ST Create Atlas Toolbar eases map layout for print

Figure 9. One of several map layout templates available to students

843
 The Spatially Interactive Literature Analysis System Study Tool
anticipated and then directly evaluated during
in-class sessions. Student feedback was system-
atically solicited and incorporated as part of the
work flow of tool development.
Diverse opportunities for student feedback
and lesson integration continue to be integral to all
future development and modifications. Though
the use of SILAS.ST has been limited to courses
in the Religious Studies department to date, and
even with those classes its deployment has been
sporadic, we expect this to expand significantly
in the near future. Our active campus-wide ini-
tiative to promote the use of maps, mapping
and spatial perspectives into our undergraduate
curricula has recently focused on the themes of
mapping people and mapping migrations, topics
that are clearly of broad interest to humanities
and social science faculty (LENS, 2010). In
the future, there are plans to explore the use of
SILAS.ST with multiple courses in history and
archaeology as well. These additional use cases
will undoubtedly provide important feedback to
the on-going development of this prototype tool.
Software upgrades have had to be accom-
modated with each new phase. For example,
accessing the customized SILAS.ST forms
from a toolbar within ArcMap 9.3 required the
use of VB.Net, and this function may behave
differently in more recent versions of ArcGIS.
Software enhancements create new hurdles but
also new opportunities and possibilities. With
new versions, we can now reconsider the idea
of enabling students to edit data and work with
SILAS.ST over web-based maps.
In developing SILAS.ST, we found that
blending mapping technologies with intellectual
contexts and technical skills yields a rich learn-
ing environment. Of the many possible ways of
documenting sources, SILAS.ST’s capacity for
accommodating both factual and more ambigu-
ous information has broken new ground. The
balance it affords will be particularly valuable in
the interdisciplinary and collaborative classroom
settings that are often home to undergraduate
844
classroom GIS use. Objective and subjective
scrutiny of historical sources is helping students
more fully comprehend the complexities of
history. This juxtaposition allows humanities-
oriented researchers and students to ask better
and more nuanced questions, whether students are
analyzing patterns, probing spatial relationships,
investigating causes, or formulating queries
(Gregory & Ell, 2007). As a prototype, SILAS.
ST helps students appreciate the processes of
historical scholarship and encourages critical
thinking and inquiry. As students seek to bring
historical data sources into conversation with
21st century technology platforms, they begin
to view both ancient and modern categories in
new and productive ways.
Within historical GIS, the nature of the project
and data sources must influence the documenta-
tion method. Many projects focus on the pres-
ervation of history, with an eye to utilizing GIS
database management for information efficiency.
These GIS projects share similar challenges to
those met by SILAS.ST. Documenting coverage,
scale and accuracy, and the thoughtful incorpora-
tion of customized metadata standards are issues
that any historical GIS must address. Metadata
and source documentation have been considered
from a technical and scholarly perspective for
many years, but our intention was to design a
tool that prioritized pedagogical objectives. Ul-
timately, by making these topics more relevant
for historians and researchers, we increase the
likelihood that scholars and students alike will
benefit from the powerful resources of GIS.
Interest in and use of these tools will be one
measure of our success. A second measure will
be SILAS.ST’s effectiveness in validating data
sources for exploratory use in the humanities
classroom, while encouraging students to ques-
tion both conventional readings and geographical
representations of history.
The Spatially Interactive Literature Analysis System Study Tool
REFERENCES
Ackerman, J. R., & Karrow, R. W. Jr. (2007).
Maps: Finding our place in the world. Chicago,
IL: University of Chicago Press.
Bailey, T. J., & Schick, J. B. M. (2009). Enabling
the collision of history and geography. Social
Science Computer Review, 27(3), 291–296.
doi:10.1177/0894439308329757
Benzek, S. (2008). Spatially Interactive Literature
Analysis System (SILAS): An interactive tool to
study the narrative landscape of ancient texts
(Unpublished master’s thesis). University of
Redlands, Redlands, CA.
Black, J. (1997). Maps and history: Construct-
ing images of the past. New Haven, CT: Yale
University Press.
Bodenhamer, D. J., Corrigan, J., & Harris, T. M.
(2010). The spatial humanities. Bloomington, IN:
Indiana University Press.
Bol, P. (2008). Creating a GIS for the history
of China. In Knowles, A., & Hillier, A. (Eds.),
Placing history: How maps, spatial data, and GIS
are changing historical scholarship (pp. 27–60).
Redlands, CA: ESRI Press.
Brennan-Horley, C., Luckman, S., Gibson, C., &
Willoughby-Smith, J. (2010). GIS, ethnography,
and cultural research: Putting maps back into
ethnographic mapping. The Information Society,
26(2), 92–103. doi:10.1080/01972240903562712
Brown, J. (2008). Using GIS to drive research
in undergraduate history classes: Five years of
experimentation with the modern world history
survey course. Journal of the Association for
History and Computing, 11(2).
Brown, P. (2010). Exploring historical space and
environments in the history/social studies class-
room: Discussion and reflections. Journal of the
Association for History and Computing, 13(1).
Cunfer, G. (2002). Causes of the Dust Bowl. In
Knowles, A. (Ed.), Past time, past place: GIS for
history (pp. 93–104). Redlands, CA: ESRI Press.
Edney, M. H. (1997). Mapping an empire. Chicago,
IL: University of Chicago Press.
Frye, C. (2008). The ArcGIS citation data model
and method. Paper presented at the Meeting of the
Social Science History Association, Miami, FL.
Gaddis, J. L. (2002). The landscape of history:
How historians map the past. New York, NY:
Oxford University Press.
Goodchild, M., & Janelle, D. (2010). Toward criti-
cal spatial thinking in the social sciences and hu-
manities. GeoJournal, 75(1), 3–13. doi:10.1007/
s10708-010-9340-3
Gregory, I. N., & Ell, P. S. (2007). Historical GIS:
Technologies, methodologies and scholarship.
Cambridge, MA: Cambridge University Press.
doi:10.1017/CBO9780511493645
Gregory, I. N., & Healey, R. G. (2007). Historical
GIS: Structuring, mapping and analysing geogra-
phies of the past. Progress in Human Geography,
31(5), 638–653. doi:10.1177/0309132507081495
Harley, J. B. (1988). Maps, knowledge, and power.
In Cosgrove, D., & Daniels, S. (Eds.), The ico-
nography of landscape (pp. 277–312). Cambridge,
UK: Cambridge University Press.
Harley, J.B. (1989). Deconstructing the map.
Cartographica: The International Journal for
Geographic Information and Geovisualization,
26(2), 1–20. doi:10.3138/E635-7827-1757-9T53
Hillier, A. (2003). Redlining and the home owners
loan corporation. Journal of Urban History, 29(4),
394–420. doi:10.1177/0096144203029004002
Holdsworth, D. W. (2003). Historical geogra-
phy: New ways of imaging and seeing the past.
Progress in Human Geography, 27(4), 486–493.
doi:10.1191/0309132503ph440pr
845
 The Spatially Interactive Literature Analysis System Study Tool
Jessop, M. (2004). The visualization of spatial data
in the humanities. Literary and Linguistic Com-
puting, 19(3), 335–350. doi:10.1093/llc/19.3.335
Jessop, M. (2008). The inhibition of geographi-
cal information in digital humanities scholarship.
Literary and Linguistic Computing, 23(1), 39–50.
doi:10.1093/llc/fqm041
Johnson, I. (1998). ECAI Metadata Manual
(Version 1.1). Retrieved from http://www.ecai.
org/documentation/ecai_metadata_standard.html
Kerski, J. J. (2003). The implementation and ef-
fectiveness of geographic information systems
technology and methods in secondary education.
The Journal of Geography, 102(3), 128–137.
doi:10.1080/00221340308978534
King, A. (2009). The spatially interactive litera-
ture analysis system study tool: A GIS tool for
interpreting history (Unpublished master’s thesis).
University of Redlands, Redlands, CA.
Knowles, A. K. (2000). A case for teaching geo-
graphic visualization without GIS. Cartographic
Perspectives, 36, 24–37.
Knowles, A. K. (Ed.). (2002). Past time, past
place: GIS for history. Redlands, CA: ESRI Press.
Knowles, A. K. (2008). What could Lee see at
Gettysburg? In Knowles, A., & Hillier, A. (Eds.),
Placing history: How maps, spatial data, and GIS
are changing historical scholarship (pp. 235–265).
Redlands, CA: ESRI Press.
Knowles, A. K., Beorn, W., Cole, T., Gigliotti, S.,
Giordano, A., & Holian, A. (2009). Geographies
of the Holocaust. Geographical Review, 99(4),
563–574.
Kwan, M.-P., & Ding, Q. (2008). Geo-narrative:
Extending geographic information systems for
narrative analysis in qualitative and mixed-method
research. The Professional Geographer, 60(4),
443–465. doi:10.1080/00330120802211752
846
Lancaster, L., & Bodenhamer, D. (2002). The
electronic cultural Atlas Initiative and the North
American Religion Atlas. In Knowles, A. (Ed.),
Past time, past place: GIS for history (pp. 163–
177). Redlands, CA: ESRI Press.
Larsen, L. (2009). Implementing SILAS: The fu-
ture of using GIS to study the past. Paper presented
at the Meeting of the Association of American
Geographers, Las Vegas, NV.
LENS. (2010). Learning spatially at the Uni-
versity of Redlands. Retrieved from http://lens.
redlands.edu
Lloyd, W. (2001). Integrating GIS into the
undergraduate learning environment. The
Journal of Geography, 100(5), 158–163.
doi:10.1080/00221340108978443
Long, B. O. (2002). Imagining the Holy Land:
Maps, models, and fantasy travels. Bloomington,
IN: Indiana University Press.
MacEachren, A. M. (1992). Visualizing uncer-
tain information. Cartographic Perspective, 13,
10–19.
Meriam Library, California State University,
Chico. (2010). Evaluating information -- apply-
ing the CRAAP test. Retrieved from http://www.
csuchico.edu/lins/handouts/eval_websites.pdf
Mostern, R. (2010). Putting the world in world
history. Journal of the Association for History
and Computing, 13(1).
Oden, P. (1992). Geography is everywhere in chil-
dren’s literature. The Journal of Geography, 91(4),
151–158. doi:10.1080/00221349208979503
Pearce, M. W., & Louis, R. P. (2008). Mapping
indigenous depth of place. American Indian
Culture and Research Journal, 32(3), 107–126.
Plewe, B. (2002). The nature of uncertainty in
historical geographical information. Transac-
tions in GIS, 6(4), 431–456. doi:10.1111/1467-
9671.00121
The Spatially Interactive Literature Analysis System Study Tool
Ray, B. (2002). Teaching the Salem Witch Trials.
In Knowles, A. (Ed.), Past time, past place: GIS
for history (pp. 19–33). Redlands, CA: ESRI Press.
Reed, J. L. (2007). The HarperCollins visual guide
to the New Testament: What archaeology reveals
about the first christians. New York: HarperOne.
Robinson, B. (2010). The integration of GIS-based
information mapping into an enthnohistorical
seminar. Journal of the Association for History
and Computing, 13(1).
Rumsey, D., & Williams, M. (2002). Historical
maps in GIS. In Knowles, A. (Ed.), Past time,
past place: GIS for history (pp. 1–18). Redlands,
CA: ESRI Press.
Schultz, R. B., Kerski, J. J., & Patterson, T. C.
(2008). The use of virtual globes as a spatial
teaching tool with suggestions for metadata stan-
dards. The Journal of Geography, 107(1), 27–34.
doi:10.1080/00221340802049844
Schwartz, R. (2010). Rail transport, agrarian cri-
sis, and the restructuring of agriculture: France
and Great Britain confront globalization, 1860–
1900. Social Science History, 34(2), 229–255.
doi:10.1215/01455532-2009-026
Sinton, D. (2009). Roles for GIS within higher edu-
cation. Journal of Geography in Higher Education,
33(1), 7–16. doi:10.1080/03098260903034046
Sinton, D., & Bednarz, S. (2007). About that G in
GIS. In Sinton, D., & Lund, J. (Eds.), Understand-
ing place: GIS and mapping across the curriculum
(pp. 19–34). Redlands, CA: ESRI Press.
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 3,
edited by Donald Patrick Albert, pp. 70-85, copyright 2012 by IGI Publishing (an imprint of IGI Global).
Sinton, D., & Lund, J. (2007). Understanding
place: GIS and mapping across the curriculum.
Redlands, CA: ESRI Press.
Sparke, M. (1995). Between demythologizing
and deconstructing the map: Shawnadithit’s
New-found-land and the alienation of Canada.
Cartographica, 31(1), 1–21. doi:10.3138/WW47-
6X0N-475Q-7231
Staley, D. (2007). Finding narratives of time and
space. In Sinton, D., & Lund, J. (Eds.), Under-
standing place: GIS and mapping across the cur-
riculum (pp. 35–48). Redlands, CA: ESRI Press.
Universidad Francisco Marroquín. (2009).
Quauhquechollan, El Lienzo de la Conquista (The
Lienzo). Retrieved from http://www.lienzo.ufm.
edu/cms/en/home
University of California. (2007). Electronic
Cultural Atlas Initiative. Retrieved from http://
www.ecai.org/
Unwin, D. (1995). Geographical information sys-
tems and the problem of ‘error and uncertainty.’
Progress in Human Geography, 19(4), 549–558.
doi:10.1177/030913259501900408
Wilford, J. N. (2000). The mapmakers (Revised
ed.). New York, NY: Knopf.
Zhang, J., & Goodchild, M. (2002). Uncertainty in
geographic information. Boca Raton, FL: Taylor
and Francis. doi:10.4324/9780203471326
847
848

Chapter 52
Geographic Visual
Query Languages and
Ambiguities Treatment
Arianna D’Ulizia
Consiglio Nazionale delle Ricerche, IRPPS, Italy

Fernando Ferri
Consiglio Nazionale delle Ricerche, IRPPS, Italy

Patrizia Grifoni
Consiglio Nazionale delle Ricerche, IRPPS, Italy

ABSTRACT
The main issues of spatial databases and Geographic Information System (GIS), concern the representa-
tion, the management and the manipulation of a large and complex number of spatial objects and spatial
relationships. In these systems many concepts are spatial and, therefore they are intrinsically related
with a visual representation, which makes also easier to formulate queries by non-expert users. The
main problems in visual query languages for spatial databases concern imprecision, spatial integrity and
ambiguities in query formulation. Our concern in this chapter is with the ambiguity of visual geographi-
cal queries. In particular, a review of existing visual query languages for spatial databases and their
classification on the grounds of the methodology adopted to resolve the ambiguity problem are provided.

INTRODUCTION Visual queries for spatial databases can be

Spatial databases and Geographic Information
System (GIS) represent, manage and manipulate
a large and complex number of spatial objects
and spatial relationships.
DOI: 10.4018/978-1-4666-2038-4.ch052
expressed using one of the following four ap-
proaches. The first approach uses predefined
icons to retrieve pictorial information. Examples
of languages that use this approach are: Cigales
(Calcinelli and Mainguenaud, 1994), the language
defined by Lee and Chin in (1995) and the card-
based language proposed by Ju et al. (2003). The
second approach specifies spatial relationships by

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Geographic Visual Query Languages and Ambiguities Treatment
freehand drawing. Sketch (Meyer, 1993) Spatial-
Query-By-Sketch (Egenhofer, 1997; Blaser and
Egenhofer, 2000), VISCO (Wessel and Haarslev,
1998), GeoPQL (Ferri and Rafanelli, 2005; Ferri
et al., 2004) and, finally, the language proposed
by Erwig and Schneider (2003), belong to this
approach. The third approach uses symbolic im-
ages for representing a set of objects and a set
of spatial relations among them. Languages that
belong to this approach are Pictorial Query By
Example (Papadias and Sellis, 1995), SVIQUEL
(Kaushik and Rundensteiner, 1997), and the lan-
guage proposed by Rahman et al. (2005). Finally,
the fourth approach combines text and sketching in
a hybrid solution, such as the language proposed
by Szmurlo et al. (1998).
The main problems in visual query languages
for spatial databases concern imprecision, spatial
integrity (Favetta & Laurini, 2001) and ambi-
guities in query formulation. Some authors have
proposed solutions to resolve ambiguities. For
example, Favetta and Aufaures-Portier (2000)
proposed a taxonomy for classifying different
types of ambiguity during query formulation.
They state that the best solution for ambiguities
is a hybrid language (textual and visual) with a
more intensive dialog between user and system.
Lbath et al. (1997) have proposed to resolve
ambiguities through a standard for semantics using
specific menus. They argue that it is possible to
define the interpretation for the query and suggest
a hybrid visual language named Aigle-Cigales, in
which the system works with default semantics.
Details should be explicitly mentioned through
a specific contextual menu or by textual format.
Since ambiguity can represent a restriction for
visual languages, it is very interesting to analyze
several language proposals and classify them
according to the methodology used to resolve
the problem of ambiguity. A first group of lan-
guages, such as Pictorial Query By Example and
SVIQUEL, faces ambiguity by allowing the use
of a few operators and/or spatial relationships. A
second group disambiguates language through the
use of actions in query formulation by modifying
the query semantics. The iconic language defined
by Lee and Chin (1995) belongs to this group.
A third group of languages tries to increase the
user’s ability to formulate more complex queries
by the use of several operators without facing the
ambiguity problem. Languages that belong to this
group are Cigales and LVIS. A fourth group of
languages proposes approximate solutions as well
as the exact answer to the query enabling the user
to select what he/she requires. Sketch and Spatial
Query By Sketch are part of this category. Finally,
a fifth group of languages, such as GeoPQL,
resolves ambiguities by introducing special new
operators to manage them.
This chapter is structured as follows. Section
2 gives a brief overview of the approaches used
for the definition of visual querying for spatial
databases. Section 3 illustrates problems about
ambiguity treatment in these kinds of visual lan-
guages. In section 4 a classification of different
languages on the grounds of methodology adopting
to resolve the problem of ambiguity is proposed.
Section 5 presents some future perspectives on the
growth of visual languages for spatial databases
and conclusions.
VISUAL QUERY LANGUAGES
FOR SPATIAL DATABASES
Several proposals of visual languages for geo-
graphic data exist in the literature. The following
discussion expressed in detail the classification
of the languages presented in the introduction.
To conceptually represent geographic objects,
different visual query languages consider three
types of symbolic graphical objects (SGO): point,
polyline and polygon.
The first approach uses predefined icons for
retrieving pictorial information. The shortcoming
of this method is that the predefined icons do not
have strong expressive power, and the consequent
query capability is limited.
849
 Geographic Visual Query Languages and Ambiguities Treatment
Among languages based on this approach, there
is Cigales (Calcinelli and Mainguenaud, 1994),
which allows the user to draw a query. It is based
on the idea of expressing a query by drawing the
pattern corresponding to the result desired by
the user. To achieve this it uses a set of icons that
model the geometric objects, polyline and polygon
(point is not considered), and the operations carried
out on these objects (e.g., intersection, inclusion,
adjacency, path and distance). Symbolic graphical
objects and icons that conceptualize the operators
are predefined.
Another language, defined by Lee and Chin
(1995), allows the user to compose a query utilizing
the three symbolic graphical objects: rectangle,
line and point. These SGOs can compose an iconic
sentence whose meaning is due to topological
relations among the icons of the query. In this
language the user draws a new SGO and can set
the state of all of the previously drawn SGOs to
the foreground or background. SGOs for which
the relationships with the new SGO have to be
considered, must be placed in the foreground,
while those SGOs whose relationships have not
been considered are placed in the background.
The card-based language, proposed by Ju et
al. (2003), uses the card iconic metaphor for rep-
resenting both complex spatial objects and spatial
relationships between them. The user can describe
his/her query requirements in a visual environment
by selecting the appropriate cards and putting
them into the proper query boxes. The result of
the query is also displayed in graphical form.
The second approach specifies spatial relation-
ships by a freehand drawing. First of all there is
Sketch! (Meyer, 1993), that allows the user to
draw a visual representation of his/her query, as
if on a blackboard, without explicit references to
operators to be applied to geographical objects
involved in the query. In fact, the spatial query is
expressed through a sketch on the screen, which is
later interpreted by the system. It means that spatial
operators are directly derived from the sketch.
850
The Spatial-Query-By-Sketch language
(Egenhofer, 1997; Blaser and Egenhofer, 2000),
similarly to Sketch!, is based on a formal model for
topological spatial relations and a computational
model for the constraints relaxation. Each query
produces a set of candidate interpretations as result
and the user selects the correct one.
Another language based on the second ap-
proach is VISCO (Wessel and Haarslev, 1998)
which considers geometric as well as topological
constraints as drawn by the user. The system parses
the geometry of query sketches and supports the
annotation of meta information for specifying
relaxations which are additional constraints, or
“don’t cares”, that define the query interpretation.
Among the languages based on the second ap-
proach there is the GeoPQL (Ferri and Rafanelli,
2005; Ferri et al., 2004) that is based on twelve
operators. The nine traditional topological opera-
tors, the distance operator and finally two further
operators, ALIAS and ANY, devoted to solving
ambiguities in query interpretation. Only the last
two operators need to be explicitly expressed,
while topological operators are automatically de-
duced from the visual representation of the query.
Finally, the proposal of Erwig and Schneider
(2003) consists in a language devoted to analyz-
ing two-dimensional traces of moving objects
to infer a temporal development of their mutual
spatial relationships.
The third approach uses symbolic images for
representing a set of objects and a set of spatial
relations among them. For instance, Pictorial
Query-by-Example (PQBE) (Papadias and Sellis,
1995) uses symbolic images to find directional
relationships. This language considers a symbolic
image as an array that could correspond to visual
scenes, geographical maps or other forms of spa-
tial data. The main limitation of PQBE is that it
considers directional relationships only.
An evolution of PQBE is SVIQUEL (Kaushik
and Rundensteiner, 1997) that also includes topo-
logical operators which consider 45 different types
of primitives, allowing to represent topological
Geographic Visual Query Languages and Ambiguities Treatment
and directional relationships between two SGOs
of type polygon.
Finally, the proposal by Rahman et al. (2005),
allows elicit information requirements through
the interactive choice of wireless web services.
To define a visual spatial query language this ap-
proach combines text and sketching in a hybrid
solution. Using this method, users can draw spatial
configurations of the objects they would like to
retrieve from the GIS, while the textual part permits
the specification of the geographical semantics.
An example of this approach is represented by the
language proposed by Szmurlo et al. (1998). This
language allows the user to draw the configura-
tion of the objects he/she is interested in, and thus
defines spatial constraints between these objects.
Any geographic object can be classified as a zone,
a line or a point. Moreover, constraints written in
natural language are collected in labels that are
graphically connected to the object they refer to.
QUERY’S AMBIGUITY PROBLEM
The main problems of visual query languages
for spatial databases concern imprecision, spa-
tial integrity and ambiguities in query formula-
tion. Favetta and Aufaure-Portier (2000) details
problems due to topological imprecision and
integrity in spatial relations. Some researchers
have proposed solutions to resolve ambiguities.
For example, Favetta and Aufaures-Portier (2000)
have suggested a taxonomy for classifying dif-
ferent kinds of ambiguity that can be produced
during the query formulation. They state that the
best solution to resolve ambiguities is a hybrid
language (textual and visual) with a dialog more
intense between user and system.
Another proposal to resolve the ambiguities
has been introduced by Lbath et al (1997). They
proposed standard semantics, through specific
menus, that make interpretation of the query
possible. They propose a hybrid visual language,
Aigle-Cigales, in which the system works with
default semantics and details are mentioned ex-
plicitly through a specific contextual menu or by
textual format.
Moreover, Carpentier and Mainguenaud (2002)
distinguish two ambiguities: visual ambiguity that
appears when a given visual representation of a
query corresponds with several interpretations,
and selection ambiguity that appears when sev-
eral metaphors correspond to a given selection.
To reduce these problems the authors propose
a drawing process with a grammar for interac-
tion: the more powerful the grammar, the more
important the level of ambiguity. A compromise
needs to be found.
Visual languages offer an intuitive and incre-
mental view of spatial queries, but often times
they may offer different interpretations of the
same query. Among the reasons for multiple
query interpretations, the most important is that
the user has a different intention in formulating
his/her query with respect to the analysis that the
system makes of it. Moreover, when the user draws
two icons for representing different objects of a
sentence, he/she can avoid defining one or more
spatial relations between them. Then the system
can formulate different interpretations to obtain
the required result.
For example, suppose the user formulate the
following query: “Find all the regions which pass
through a river and overlap a forest”. The user
is not interested in the relationship between the
river and the forest and the absence, in natural
language (NL) formulation, of explicit relation-
ships between them produces an ambiguity. The
different visual queries of Figure 1 represent the
query in natural language.
To remove the ambiguity, the complete natu-
ral language query, “Find all the regions which
are passed through by a river and overlap a forest,
irrespectively of the topological relationships
between the river and the forest”, could be con-
sidered. However, when the user draws an SGO
representing a forest and another representing a
851
 Geographic Visual Query Languages and Ambiguities Treatment
Figure 1. Visual queries for the same NL query
river he/she cannot avoid representing a topo-
logical relationship between them.
Since ambiguity can constitute a restriction
of visual language, in the following section we
analyze methodologies adopted by different lan-
guages to resolve this problem.
VISUAL SPATIAL QUERY
LANGUAGES AND THE
AMBIGUITY PROBLEM
In section 2 we presented visual query languages
for spatial databases that have been proposed in
the literature. Now we classify these languages
on the grounds of methodologies they adopt to
resolve the problem of ambiguity.
A first group of languages handles the am-
biguity by allowing the use of few operators or
spatial relationships, such as Pictorial Query By
Example and SVIQUEL. Considering only limited
kinds of spatial relations (directional relations)
PQBE avoids multiple interpretations of the query
but reduces the possibility of formulating more
complex queries that involve topological rela-
tionships. The SVIQUEL language also includes
topological operators. However, it avoids multiple
interpretations by limiting the number of objects
involved (to just two) and provides a tool with a
low expressive power for specifying the relative
spatial positions.
A second group disambiguates the language
by the use of actions in query formulation that
modify query semantics. The iconic language
852
defined by Lee and Chin (1995) belongs to this
group. In this language it is possible to remove
undesired relationships among drawn symbolic
graphical objects or impose an a priori restrictive
interpretation using the foreground/background
metaphor. The relationships of a new symbolic
graphical object depend on the state (foreground
or background) of the previously drawn symbolic
graphical objects. To interpret a query the parser
must consider both the visual representation and
the drawing process. In this manner some proce-
dural steps influence the semantics of the query
they do not influence its representation. Queries
having the same representation may have differ-
ent semantics. Another language that belongs to
the second group is VISCO. This language offers
tools that specify meta-information resembling the
user’s idea of the interpretation. This approach
demands more skill from the user but makes the
intended interpretation explicit.
A third group of language tries to increase a
user’s possibility of formulating a more complex
query by the use of several operators without facing
explicitly the ambiguity problem. A language that
belongs to this group is Cigales. In this language the
system is not able to give a unique interpretation
of the visual query representation. Two possible
solutions proposed to reduce ambiguity of Cigales
are: to introduce various interactions (feedback)
with the user and to increase the complexity of the
resolution model. However, different obstructions
may arise so that the semantics of the query are
fully user dependent and complex queries with
numerous basic objects, are not expressible. An-
Geographic Visual Query Languages and Ambiguities Treatment
other language of the third group is that proposed
by Szmurlo et al. (1998). This language allows the
user to draw the configuration of the objects and
thus to define spatial constraints between these
objects and thematic constraints that each object
has to respect. However, as this language allows the
user to have great freedom, many ambiguities may
arise due to the incoherencies between the object
and the thematic constraints, between constraints
for the same object or between constraints for
different objects. A solution to ambiguity consists
of detection of incoherencies and proposal of a
possible solutions to the user.
A fourth group of languages proposes to give
query approximation solutions to the user and the
user then selects what he/she requires. Sketch and
Spatial Query By Sketch are part of this category. In
particular, Spatial-Query-By-Sketch resolves the
ambiguity problem by considering and proposing
to the user both the exact solution of the query, if
possible, and other approximate solutions obtained
by relaxing some relationships. In this manner the
language includes multiple interpretations in the
result, and the user selects the representation that
provides a correct interpretation of his/her query.
Figure 2. The same visual query represented with different languages
Finally, a fifth group of languages resolves the
ambiguity problem by introducing new special
operators that serve to manage the ambiguity.
Among the languages belonging to this group there
is GeoPQL, which allows the user to represent only
the desired relationships. The system interprets
the query considering all relationships between
symbolic graphical objects of the sketch, and it
is possible to remove or modify some undesired
relations using ad hoc operators introduced in the
language. Figure 2 shows the few examples of
visual queries represented by using some of the
languages introduced in this article.
CONCLUSION
The number of applications using spatial or geo-
graphic data has been ever increasing over the
last decade. New small-scale GISs, often called
desktop GIS, are gradually becoming available and
people are becoming familiar with using the web
to access remote information. An important future
direction is to provide the existing system through
the web by using standards created by the Open-
853
 Geographic Visual Query Languages and Ambiguities Treatment
GIS consortium such as WMS (Web Map Server)
and GML (Geographic Markup Language). In
particular, GML is an open standard for encoding
geographic information in an eXtensible Markup
Language (XML). It is not related to any specific
hardware or software platform. Any data encoded
using it can be easily read and understood by any
programming language and software system able
to parse XML streams. Moreover, there are com-
mercial and open source tools that translate GML
data into the Scalable Vector Graphics (SVG)
format in order to display maps.
Consequently, the World Wide Web diffusion
and the increasingly widespread usage of mobile
devices has made possible to query and to access
geographical databases available online from mo-
bile devices. There are mainly two drawbacks in
using GML for mobile devices. First, it is memory
and bandwidth-consuming for storage and transfer
respectively. Moreover, maps described with it
have to be projected and scaled before being plot-
ted. These characteristics make GML not directly
accessible with small devices. It is necessary to
reduce GML size in order to make cartographic
data accessible from mobile devices.
Another future perspective concerns the repre-
sentation of dynamic phenomena that change over
space and time in GISs. While current GIS can
provide snapshot views at discrete time intervals,
they fall short in providing an ability to link the
process models with data from multiple sources
or simulate scenarios of change for users.
REFERENCES
Blaser, A. D., & Egenhofer, M. J. (2000). A Vi-
sual Tool for Querying Geographic Databases.
Advanced Visual Interfaces – AVI 2000. Palermo,
Italy: ACM Press (pp. 211-216).
854
Calcinelli, D., & Mainguenaud, M. (1994).
Cigales, a visual language for geographic infor-
mation system: the user interface. Journal of
Visual Languages and Computing, 5(2), 113–132.
doi:10.1006/jvlc.1994.1006
Carpentier, C., & Mainguenaud, M. (2002).
Classifying ambiguities in visual spatial lan-
guages. GeoInformatica, 6(3), 285–316.
doi:10.1023/A:1019770003896
Egenhofer, M. J. (1997). Query processing in spa-
tial-query-by-sketch. Journal of Visual Languages
and Computing, 8(4), 403–424. doi:10.1006/
jvlc.1997.0054
Erwig, M., & Schneider, M. (2003). A visual
language for the evolution of spatial relation-
ships and its translation into a spatio-temporal
calculus. Journal of Visual Languages and
Computing. Elsevier, 14, 181–211. doi:10.1016/
S1045-926X(02)00057-5
Favetta, F., & Aufaure-Portier, M. (2000). About
Ambiguities in Visual GIS Query Languages: a
Taxonomy and Solutions. Proceedings of the 4th
International Conference on Advances in Visual
Information Systems. LNCS Springer-Verlag Pub-
lications, LNCS 1929, (pp. 154-165).
Favetta, F., & Laurini, R. (2001). About Preci-
sion and Integrity in Visual Query Languages
for Spatial Databases. Proceedings of the 7th
International Conference on Database Systems
for Advanced Applications (DASFAA 2001). IEEE
Computer Society, (pp. 286-293).
Ferri, F., Grifoni, P., & Rafanelli, M. (2004).
XPQL: A pictorial language for querying geo-
graphic data. Databases and Expert Systems
Applications (DEXA 2004). LNCS 3180, Springer-
Verlag Publications (pp. 925-935).
Geographic Visual Query Languages and Ambiguities Treatment
Ferri, F., & Rafanelli, M. (2005). GeoPQL: A Geo-
graphical Pictorial Query Language that resolves
ambiguities in query interpretation. Journal of
Data Semantics. LNCS 3534, 3, 50-80. Springer-
Verlag Publications.
Ju, S., Guo, W., & Hernández, H. J. (2003). A
Card-based Visual Query System for Geographi-
cal Information Systems. Scandinavian Research
Conference on Geographical Information Science
(pp. 62-74).
Kaushik, S., & Rundensteiner, E. (1997).
SVIQUEL: A Spatial Visual Query and Explora-
tion Language, Databases and Expert Systems
Applications (DEXA 1998), LNCS 1460, Springer-
Verlag Publications (pp. 290-299).
Lbath, A., Aufaure-Portier, M., & Laurini, R.
(1997). Using a Visual Language for the Design
and Query in GIS Customization. International
IEEE Conference on Visual Information Systems.
San Diego, CA. (pp. 197-204).
Lee, Y. C., & Chin, F. (1995). An iconic query
language for topological relationship in GIS. In-
ternational Journal of geographical. Information
Systems, 9(1), 25–46.
Meyer, B. (1993). Beyond Icons: Towards New
Metaphors for Visual Query Languages for Spatial
Information Systems. International Workshop
on Interfaces to Database Systems. Glasgow,
Scotland, (pp. 113-135).
Papadias, D., & Sellis, T. (1995). A pictorial query-
by-example language. Journal of Visual Lan-
guages and Computing, 6(1), 53–72. doi:10.1006/
jvlc.1995.1004
Rahman, S. A., Bhalla, S., & Hashimoto, T. (2005).
Query-By-Object Interface for Information Re-
quirement Elicitation Implementation. Fourth
International Conference on Mobile Business
(ICMB2005). IEEE Computer Society, Sydney,
Australia, (pp. 667-670).
Szmurlo, M., Gaio, M., & Madelaine, J. (1998).
The geographical Anteserver: a Client/Server
Architecture for GIS. EOGEO’98 workshop.
Salzburg, Austria, (pp. 78-85).
Wessel, M., & Haarslev, V. (1998). VISCO:
Bringing Visual Spatial Querying to Reality.
Proceedings of the IEEE Symposium on Visual
Languages. IEEE Computer Society, Halifax,
Canada, (pp. 170-179).
KEY TERMS AND DEFINITIONS
eXtensible Markup Language (XML): is
a W3C-recommended general-purpose markup
language for creating special-purpose markup
languages, capable of describing many different
kinds of data. In other words, XML is a way of
describing data.
Geographical Database: A database in which
geographic information is store by x-y coordi-
nates of single points or points which identify the
boundaries of lines (or polylines, which sometimes
represent the boundaries of polygons). Different
attributes characterize the objects stored in these
databases. In general the storing structure consists
of “classes” of objects, each of them implemented
by a layer. Often a geographic databases include
raster, topological vector, image processing, and
graphics production functionality.
Geographical Information System (GIS):
A computerized database system used for the
capture, conversion, storage, retrieval, analysis
and display of spatial objects.
Geography Markup Language (GML): is the
XML grammar defined by the Open Geospatial
Consortium (OGC) to express geographic features.
GML serves as a modeling language for geographic
systems as well as an open interchange format for
geographic transactions on the Internet.
Icon: Small pictures that represent commands,
files, objects or windows.
855
 Geographic Visual Query Languages and Ambiguities Treatment

Metaphor: Figurative language that creates an
analogy between two unlike things.  A metaphor
does not make a comparison, but creates its anal-
ogy by representing one thing as something else.
Scalable Vector Graphics (SVG): SVG is a
language for describing two-dimensional graphics
and graphical applications in XML.
Visual Query language: A language that
allows the user to specify its goals in a two-(or
more)-dimensional way with visual expressions
- spatial arrangements of textual and graphical
symbols.

This work was previously published in Handbook of Research on Geoinformatics, edited by Hassan A. Karimi, pp. 340-348,
copyright 2009 by Information Science Reference (an imprint of IGI Global).

856
 857

Chapter 53
Multi Depot Probabilistic
Vehicle Routing Problems
with a Time Window:
Theory, Solution and Application

Sutapa Samanta
Maryland State Highway Administration, USA

Manoj K. Jha
Morgan State University, USA

ABSTRACT
Vehicle Routing Problems (VRPs) are prevalent in all large pick up and delivery logistics systems and are
critical to city logistics operations. Of notable significance are three key extensions to classical VRPs: (1)
multi-depot scenario; (2) probabilistic demand; and (3) time-window constraints, which are considered
simultaneously with VRPs in this paper. The issue then becomes a Multi Depot Probabilistic Vehicle
Routing Problem with a Time Window (MDPVRPTW). The underlying complexities of MDPVRPTW
are analyzed and a heuristic approach is presented to solve the problem. Genetic algorithms (GAs) are
found to be capable of providing an efficient solution to the so-called MDPVRPTW. Within the GAs, two
modification operators namely, crossover and mutation, are designed specially to solve the MDPVRPTW.
Three numerical examples with 14, 25, and 51 nodes are presented to test the efficiency of the algorithm
as the problem size grows. The proposed algorithms perform satisfactorily and the limiting case solutions
are in agreement with the constraints. Additional work is needed to test the robustness and efficiency of
the algorithms as the problem size grows.

DOI: 10.4018/978-1-4666-2038-4.ch053

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
INTRODUCTION
The Vehicle Routing Problem (VRP) has received
widespread attention in recent years because of
its applicability in many pickup and delivery
systems. Multiple depots become necessary for
systems consisting of larger networks, such as
for the operation of larger logistics systems of
distribution companies. When the vehicle routing
problem involves multiple depots, it is called a
Multi Depot Vehicle Routing Problem (MDVRP).
In such problems the number of vehicles can be
equal to or exceed the number of depots. In most
of the real-life VRPs, demands at the customer
nodes vary due to various factors, such as loca-
tional and temporal seasonal factors. Hence, the
stochasticity has to be considered for the demands
at the customer nodes. The vehicle routing prob-
lem which considers stochastic demands of the
customers, is called Probabilistic Vehicle Routing
Problem (PVRP). Another important criterion
for any logistics system is to provide the service
within a specified time period. Hence, the time
window concept is associated with the VRPs,
which is coined as Vehicle Routing Problem with
Time Window (VRPTW).
The real life vehicle routing problem with a
larger network can have all the issues mentioned
above, such as the nodes can have variable or sto-
chastic demand, the services have to be provided
within fixed time periods at the nodes and multiple
depots are needed to cover the network. Hence all
the three components are important for a vehicle
routing problem The proposed MDPVRPTW is
complex in nature as it combines all the three prob-
lems, such as, VRPTW, MDVRP, and PVRP. There
are many applications of this type of problem like
the pick-up of children by school buses, garbage
collection, delivery of laundry, and delivery by
United Postal Services (UPS). The problem in all
these cases can be described as follows:
A set of customers in an area has to be provided
with a service by a fleet of vehicles. The vehicles
858
will start from a set of depots to cover larger areas.
Each node or point has a demand, which varies
stochastically and must be picked up by any of the
vehicles. The customers specify a period of time
in which the pick up must occur. This particular
time period is called the time window. The vehicles
have specific limited capacities. The total demand
picked up by each vehicle should not exceed its
corresponding capacity. The objective is to find
out a set of routes formed by the fleet of vehicles
in order to cover these customers in such a way
that total travel cost becomes minimum while sat-
isfying the capacity and time window constraints.
Previous researchers have approached vari-
ous components of MDPVRPTW separately.
The stochasticity in demand, the time window
constraint, and the multidepot scenario is either
addressed individually or in the combination
of two in most of the cases. As the number of
variants increase with the introduction of each
component, it becomes difficult to devise an
analytical approach to obtain an optimal solution.
This might be the reason for the MDPVRPTW to
be unsolved, to date. The earlier works in the area
have not addressed all the three components of the
vehicle routing problem together, but in real life
problems, they exist together in most cases. The
attempt in this study is to develop an innovative
algorithm, which is capable of handling these
three components simultaneously and is able to
produce reasonable results. The algorithm can be
modified and improved further in future; neverthe-
less, it provides a good start to handle the different
components of the VRP that make the problem
relatively complex to solve. The new heuristic-
based algorithm to solve the MDPVRPTW is
developed after carefully scrutinizing significant
earlier works and the underlying complexities
associated with the problem.
In the next section, we briefly review the
VRP literature. The third section describes the
formulation and the proposed methodology to
solve the MDPVRPTW. The last section presents
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
three numerical examples of growing complexity.
Sensitivity of the results on the problem domain is
examined by relaxing or restricting the time win-
dow constraint so that the results can be validated
by justifying them within the specified boundaries.
LITERATURE REVIEW
This section reviews various methodologies
developed by the researchers to solve different
components of the MDPVRPTW at different
times. A selective list of literature review has been
presented here. The various studies available in the
literature can be broadly categorized as follows:
• Vehicle routing problem
• Vehicle routing problem with time window
• Vehicle routing problem with stochastic
demand
• Multi depot vehicle routing problem
• Multi depot vehicle routing problem with
stochastic demand
• Vehicle routing problem with time win-
dows and stochastic demand
The vehicle routing problem is a classical
optimization problem. Jaillet and Odoni (1998)
present a comprehensive survey of some advanced
methods used to solve VRPs. The methods include
Integer programming, Lagrangian relaxation
method, genetic algorithm, tabu search, etc. These
methods assume deterministic demand and don’t
consider the time window constraint. Laporte et al.
(2000) presented classical and modern heuristics
to solve vehicle routing problems.
Literature Considering Time Window
Some researchers identify the vehicle routing
problem with time windows as a separate class
of problem. As time window is an important con-
straint for most of the vehicle routing problems,
an emphasis is put on optimizing the routes that
require the time window constraint. Kohl and Mad-
sen (1997) used Lagrangian Relaxation for VRP
with time windows. The demand was considered
deterministic in these problems. They solved some
previously unsolved problems and obtained the
results with competitive computation time. Jung
and Haghani (1999, 2001) solved a vehicle rout-
ing problem and a pickup-delivery problem with
time windows. They used a genetic algorithm to
solve these problems. It worked better than the
exact algorithms for smaller problems. Bent and
Hentenryck (2004) solved VRP with time windows
with a two-stage hybrid local search approach.
They tried to minimize the number of routes in
the first stage using simulated annealing and tried
to minimize the total travel cost by using a large
neighborhood search method in the second stage.
The hybrid algorithm improves or matches with
the solutions of most of the Solomon Benchmark
problems (2009) though the approach to minimize
the number of vehicles does not guarantee better
results for minimizing the total cost. Cordeau et
al. (2001) attempt to solve vehicle routing prob-
lem with time windows using tabu search. Their
method, however, does not seem efficient for all
the cases presented. For example, their approach
does not necessarily minimize the number of
vehicles, and yet they suggest their algorithm to
be an efficient one. Two optimization algorithms,
generalized K-tree algorithm and variable splitting
algorithm were developed by Fisher et al. (1997)
to solve vehicle routing problems with time win-
dows. They were able to solve problems optimally
for up to 100 customers. Desaulniers et al. (2008)
used tabu search, partial elimentarity and k-path
inequalities to solve the vehicle routing problem
with time windows. Braysy and Gendreau (2005,
2005) presented metaheuristics and local search
algorithms for the vehicle routing problems with
time windows. Ropke and Cordeau (2009) solved
the pickup delivery problem with time windows
by Branch and Cut, and Price method.
859
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
Literature Considering
Stochasticity in Demand
Stochastic demand is an important consideration
in real-world VRPs. A number of research works
involving stochastic demand were found in the
literature. Jaillet and Odoni (1998) discussed the
probabilistic vehicle routing problem, its applica-
tions, and some heuristics to solve it. Bertsimas et
al. (1990) solved the vehicle routing problem with
an a-priori optimization strategy. They considered
the demand to be stochastic. They approximated
the solutions as the problem became complex.
Bertsimas et al. (1992) solved a stochastic vehicle
routing problem in the Euclidean plane. The main
objective of that work was to minimize the total
waiting time, not the total travel cost. The problem
involved a queuing phenomena and considered
the demand to follow a Poisson distribution. They
solved a VRP with stochastic demand and proposed
two different strategies: (1) a re-optimization
strategy when the demands are known and, (2)
an a-priori strategy when the demand is known
only when the customer reaches at the node. They
fixed the upper and lower boundaries of the de-
mand based on the general probabilistic demands
and optimized the routes within those limits. The
accuracy of the method depends on making the
boundaries tight, which is not always guaranteed.
Dror et al. (1989) and Dror (1993) modeled
vehicle routing with stochastic demand. They
handled the problem by applying a stochastic
programming model and the Markov Decision
Process model. Gendreau et al. (1995, 1996)
solved vehicle routing problem with stochastic
demands by an exact algorithm and with a tabu
search method, and obtained satisfactory results.
Gendreau et al. (1996) presented a comprehensive
review of stochastic vehicle routing problems and
summarized selected cases. Secomandi (2000)
solved a vehicle routing problem with stochastic
demand by formulating it as a Markov decision
problem. The demand distribution was known
beforehand and it was assumed to be discrete and
860
independent. The author applied Neuro-Dynamic
programming in order to tackle the complexity of
the problem. While all of the above mentioned
models were capable of handling stochastic de-
mand in VRPs, they did not take the multi depot
case and time window constraint into account.
Aykagan and Erera (2007) solved vehicle rout-
ing problem using a tabu search heuristic. The
uniqueness of their strategy was to use vehicles
in pairs instead of a single vehicle for nodes. Liu
and Lai (2009) solved vehicle routing problem
considering uncertain demand at nodes.
Literature Considering
the Multi Depot Case
Nagy and Salhi (2005) solved a multi depot
pickup-delivery problem. They found better results
by using their heuristics over the benchmark pro-
cedures available in the literature. Salhi and Sari
(1997) developed heuristics to solve a multi depot
vehicle fleet mix problem. The heuristics were
also applied on the multi-depot vehicle routing
problem and produced solutions which were about
1% better than the best solutions available in the
literature, with lesser computation time. Renaud
et al. (1996) solved a multi depot vehicle routing
problem with a tabu search algorithm. The algo-
rithm clusters the nodes initially and then improves
the solution by intensification and diversification
methods. They obtained better solutions to all 23
benchmark problems on which the algorithm was
employed. Laporte et al. (1988) solved multi-depot
vehicle routing and location routing problems with
an exact algorithm and could obtain comparable
results within relatively short computing time for
the large multi depot vehicle routing problems of
upto 80 nodes. Tangiah and Salhi (2001) solved a
multi depot vehicle routing problem using genetic
clustering. In the optimal solution they obtained
lesser number of vehicles but larger travel times
for most of the problems on which they employed
the developed algorithm. Crevier et al. (2007)
solved a multi depot vehicle routing problem in
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
which vehicles may be replenished at intermedi-
ate depots, using a tabu search algorithm. The
method is shown to be robust and fast enough for
larger problem sizes. Wu et al. (2002) provided
heuristics solutions for multidepot location routing
problems. They developed an approximate reason-
ing algorithm and improved sweeping algorithm
to solve the problem.
Literature on the Simultaneous
Consideration of Multi
Depot and Stochasticity
Chan et al. (2001) solved the problem of multi
depot vehicle routing and location routing with sto-
chastic demand by formulating it as a binary integer
program. They used space filling curves (SFC)
for the stochastic model and the Clark-Wright
Savings method for the deterministic instances
of the problem. They provided a validated way
to estimate the probable demand using a queuing
network. They did not however, incorporate the
time window constraint in the VRP.
Literature on the Simultaneous
Consideration of Time
Window and Stochasticity
There are limited studies reported that considered
time windows and stochastic demand in VRPs.
Bramel and Simchi-Levi (1996) solved a VRP with
time windows and stochastic demand. Chan et al.
(2001) solved a vehicle routing and scheduling
problem with time windows and stochastic demand
using a tabu search based heuristic algorithm.
Chen et al. (2006) developed a heuristic to solve
the vehicle routing problem with time windows
for a real time and time dependent travel time and
a real time demand scenario.
From the literature review, it is observed that
there are no studies available, which cover all
the components of MDPVRPTW. As most of the
practical vehicle routing problems contain all three
components, multi depot condition, stochastic
demand, and the time window constraint, the
need to develop a heuristic to solve real-world
MDPVRPTWs within a reasonable computation
time arises.
Baker and Ayechew (2003) solved vehicle
routing problems with a GA and proved that the
method is superior to Simulated Annealing or tabu
search methods. Chakroborty and Mandal (2005)
obtained comparatively good results efficiently
while solving single vehicle routing problems
with a GA. Ombuki et al. (2006) solved a vehicle
routing problem with time windows using a multi-
objective GA. They used Pareto-optimal ranking
to determine the number of vehicles to minimize
the cost. The obtained results were superior to the
best solutions reported by other researchers. Tong
et al. (2004) solved vehicle routing problem with
time window and unknown number of vehicles
using a GA and proved that GAs can efficiently
search for the optimal solution. Hwang (2002)
developed a Graphical User Interface (GUI)-
program based on a GA to solve the Traveling
Salesman Problem (TSP) and obtained efficient
results for problems containing small and large
number of nodes. Berger and Barkaoui (2004)
developed a new hybrid genetic algorithm to solve
VRPs with time windows. Though the algorithm
does not seem be any superior to other established
heuristics, the obtained results were comparable to
those obtained from other heuristics. Alvarenga et
al. (2007) proposed a two-phase approach using a
GA and a set partitioning method for a VRP with
time window. The method was proved to perform
better than other established heuristics or exact
algorithms. The above literature review makes the
GA a viable candidate to solve the MDPVRPTW.
MODEL DEVELOPMENT
The MDPVRPTW problem can be formulated as a
Mixed Integer Programming (MIP) and Constraint
Programming (CP) problem. Since MIP and CP
problems are non-convex, they can be solved
861
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
either by exact solution methods or by some kind
of systematic exhaustive search algorithm. For a
MIP, memory and solution time increase exponen-
tially with the increase in integer variables. In this
paper, a genetic algorithm (GA) based heuristic is
proposed for solving the MDPVRPTW problem
as unlike many conventional search methods,
GAs are not restricted by such assumptions as
linearity, convexity, and differentiability (Jong &
Schonfeld, 2003; Kim et al., 2005). GAs search
from a population rather than a single point, reach-
ing to a globally optimum solution eventually.
GAs generate candidate solutions which satisfy
the integer constraints. Initially the solutions are
far from optimal, but they transform the existing
solutions into new candidate solutions through
operators, such as crossover and mutation, which
improve the solution over successive search gen-
erations. Thus, GAs handle far more variants in
a larger and complex search space than the clas-
sical optimization algorithms, thereby reducing
the chance of earlier convergence of the solution.
The objective of developing the GA-based
heuristic is to obtain an optimal solution which
consists of a number of routes traveled by a number
of vehicles within a fixed time window. The problem
includes more than one depot. Each vehicle has a
given capacity. The demands at the customer nodes
are considered to be stochastic in nature.
Mathematical Formulation
of the Problem
As noted earlier, the problem is formulated as a
mixed-integer programming and constraint pro-
gramming problem. The objective function is to
minimize the total travel cost with the constraints
of time windows and vehicle capacity. The nota-
tions are given as follows:
Notations for parameters:
N = {1,2,…..n} set of customers
K = {1,2, ….k} set of depots
862
M = {1,2,….m} set of vehicles
(ai, bi) = Time window for customer node i
qi = Demand of customer i
Qm= Capacity of vehicle m
cij = Cost per unit travel distance
Dmax= Maximum distance which the vehicles may
cover in a tour
Coordinates of the nodes
Notations for decision variables:
1, if arc ij is part of route k
x ijk = 
0, otherwise

dij = Distance between customer i and j
C w = Penalty for the early arriving at the cus-
tomer node i
C d = Penalty for late arriving at the customer
node i
ti = Actual arrival time at node i
wi = Waiting time at the customer node i
di = Delay time at the customer node i
q(m) = Total demand picked up by vehicle m
pi = Probability value of the demand at node i
The total expected cost of the routes formed
from multiple depot is calculated considering
factors, such as stochastic demand, time window,
and shortest routes.
Multiple Depots
It is assumed that the depots are located around
the customer nodes such that the customers should
be efficiently served. Depending on the stochastic
demand and time window constraints for pickups
and deliveries, the natural tendency is to assign
vehicles to customer nodes from the nearest depot.
Hence the nearest neighborhood method has been
used to assign the customer nodes to the depot.
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

But, distance is not the only governing factor for
the assignment; demands of the customer nodes
and the capacities of the vehicles play important
roles during the assignment. The customer nodes
are divided in the following two sets during the
assignment.
demand (Equation 2b below) distributed over the
time of a day is shown in Figure 1. The probability
factor (pi) of demand at any customer node over
the time period of a day is given by
pi = ps ⋅ qi (2)

U: Set of nodes allocated to other depots

where,
V: Sets of available nodes to be allocated to the
kth depot
f (z )
U ∪V ⊆ N ; ps = ; 0 ≤ k <1 (2a)
A k : {q }, q ∈ V , k ∈ K ; Ak represents ∑ f (z )
∀z
the set of nodes allocated to the kth depot.
V where,
B k: k , k ∈ K ; Bk represents the set of
Ak
rest of the nodes available for the allocation u
−(z −T /2)

to the (k+1)th depot. f (z ) = × m; 0 < u < 1 (2b)

Travel Time Cost
The shortest route set is developed using the GA-
based heuristic discussed in Section 3.4. The travel
time cost is calculated by multiplying the total
travel time by the unit travel time cost for each
route formed. Total travel time cost is estimated
for all the routes and for all the depots
Ttotal = ∑∑ ∑ cij ⋅ dij ⋅ x ijk (1)
k ∈M i ∈N ∪K j ∈N ∪K
u −T /2
The demand varies over the days of a week
(Figure 2). The demands for the weekdays are as-
sumed to be more than those on weekends. The
probability factor (pk ) is determined from Figure
2 depending on the day of a week the vehicle ar-
rives at a particular node. The functions for ps and
pk are designed randomly for this study and is
replaceable by another function as specified by the
user. The expected value of the total cost of the
route due to the stochastic demands is given by

Cost Due To Stochastic Demand

Figure 1. Function showing the distribution of
probability factor of demand over the time of a day
We assume that the exact demand at the customer
nodes is known only when the vehicle arrives at
the node. The demand at a customer node varies
stochastically for various reasons, e.g., the time
in a day or the day in a week when the vehicle
is arriving at the customer node, the proximity
of the node to important activity areas, etc. A
vehicle will decide to pick up the delivery from
the customers depending on the stochastic demand
at the node and available capacity of the vehicle.
The function depicting the probability factor of

863
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

Figure 2. Probability factor (pk) for demand over

where,
a week
q(m ) = ∑ piqi (7)
i ∈N

∑ ∑ dij ⋅ x ijk ≤ Dmax (k ∈ M ) (8)

i ∈N ∪K j ∈N ∪K

∑∑ x ijk = 1 (j ∈ N ) (9)
k ∈M i ∈N ∪K

∑∑ x ijk ≥ 1 (j ∈ K ) (10)
C (x s ) =
k ∈M i ∈N ∪K
∑ ∑ pi ⋅ dij ⋅ x ij (3)
i ∈N ∪K j ∈N ∪K

∑ x ijk = ∑ x jik (j ∈ N ∪ K , k ∈ M )
i ∈N ∪K i ∈N ∪K
Penalty Due to the Time Window (11)
Violation
w i ≥ a i − ti (i ∈ N ) (12)
The customer deliveries have to be picked up
from a particular node within a time window.
Penalties are imposed for both the actions of early di ≥ ti − bi (i ∈ N ) (13)
arriving and/or late arriving from the specified
time window. The objective (Equation 5) is to minimize the
The cost incurred due to the early arrival or total travel cost embedded with the penalty for
delay at ith node is given by: the delay and waiting at the node. Constraints (6)
and (7) are the maximum capacity of the vehicles
C (Tx ) = C w × wi or , = C d × di (4) and the maximum distance (Secomandi, 2000),
respectively. Equation (8) represents the total
The objective function and constraints are demand which consists of probabilistic values
given as follows: of the customer demand. Constraints (9) and (10)
impose that every customer belongs to one and
Minimize only one route or vehicle.
Z =∑ ∑ Constraint (11) implies that every customer
∑ C ij ⋅ dij ⋅ x ijk + C w × ∑ wi
is serviced by the same vehicle. Constraints (12)
k ∈M i ∈N ∪K j ∈N ∪K i

+ C d × ∑ di + ∑ ∑ ps ⋅ dij ⋅ x ij and (13) are the time window constraints (Jung
i i ∈N ∪K j ∈N ∪K
& Haghani, 2001).
(5)
Upper and Lower Boundaries
subject to of the Objective Function

q(m ) ≤ Qm (m ∈ M ) (6) The boundaries of the objective function are

defined based on the time windows assigned to

864
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

the customer nodes. The upper boundary of the when

objective function, i.e., the maximum value of
the cost is estimated by relaxing the time window ti ⊂ (ai , bi )
constraint.
Proposition 1: The total expected cost, i.e.,
Proof: The time window is maintained most
the value of the objective function (Z), will be
stringently; the vehicles arrive within the time
maximum when none of the nodes satisfy the
window at all customer nodes. Hence, the delay
time window constraint.
and waiting time become 0, which leads the total
cost to be minimum, other constraints remaining
Z u = Z max
constant.

when (di + wi ) = 0;

ti ⊄ (ai , bi )
as

Proof: As the time window constraint is ti ⊂ (ai , bi ) ;

relaxed, the vehicles may arrive at the customer
nodes anytime in which case, the arrival time
Hence,
might not lie within the upper and lower limit
of the arrival times. Hence, the delay time and
C (Tx ) ≤ C (Tx )* ,
waiting time become maximum which leads the
cost due to delay or early arrival to be maximum.
as
C (Tx ) ≥ C (Tx )* ,
(di + wi ) ≤ (di + wi )* → Z ≤ Z * ,
as
where (⋅) corresponds to the optimum value,
*

(di + wi ) ≥ (di + wi ) → Z ≥ Z ,
* *

which is minimum total cost, in this case. This

particular condition of the arrival time of the
where (⋅) corresponds to the optimum value,
*
vehicle yields the lower boundary ( Zl ) for the
which is minimum total cost, in this case. The total expected cost for MDPVRPTW, expressed
specific case of arrival of the vehicles without by Z * ≥ Zl .
considering the time window constraint yields the
upper boundary ( Z u ) for the total expected cost Assumptions
for MDPVRPTW, expressed by Z * ≤ Z u .
Two important assumptions to the proposed model
Proposition 2: The lower boundary of the
are as follows:
objective function is estimated by satisfying the
time window constraint for all the nodes.
• No individual demand is greater than the
vehicle capacity.
Zl = Z min

865
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
• One node can only be assigned to one
depot.
The GA-Based Heuristic
A GA-based heuristic is developed to solve the
above MDPVRPTW problem. The overview of
the heuristic is shown by a flow-chart in Figure
3. GA is a heuristic search algorithm based on
the principle of natural selection and genetics. It
has been extensively applied in many engineering
and logistics applications involving optimization.
The proposed algorithm consists of the following
steps. First, initial set of solutions are generated.
Then they are modified to produce improved
Figure 3. Overview of the proposed methodology
866
results. In the proposed algorithm the modified
genetic operators used are reproduction, crossover
and mutation. The steps of the algorithm can be
summarized as:
1. Generation of initial set of solutions
2. Evaluation
3. Modification
Two special modification operators, namely
Crossover-mutation and Mutation-crossover are
developed for the modification scheme, to suit
the nature of the problem.
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
1. Generation of initial set of solutions: The
first step presents the procedure for gen-
erating the initial solutions. First a proper
representation scheme is designed for the
solution of the problem. Each solution set
will consist of a number of routes equal to
the number of vehicles. A typical solution
set (nth population) will be represented as
the following:
M X1 D1A B C D1 X2 D2E F D2 X3 D3G H
I….D3 …… XMDM P Q R DM
where,
M = Number of routes or vehicles
K = number of depots
Xi = number of nodes for ith route traveled by ith
vehicle
Dk = kth depot
A, B, C are the customer nodes visited by the first
vehicle from first depot
E, F are the customer nodes visited by the third
vehicle from second depot
G, H, I are the customer nodes visited by the third
vehicle from third depot
P,Q,R are the nodes visited by the Mth vehicle
from Mth depot
For the present study, the number of vehicles
is considered to be the same as the number of
depots, which is M = K.
The steps for constructing the initial feasible
solution are given below:
Step 1: Initial feasible solution set will consist of
the routes formed by the vehicles from the
depots. Vehicles of varying capacities are
assigned to each of the depots randomly.
Different sets of solutions are generated.
Step 2: Each vehicle starts from a depot. It picks
up the nodes one by one following a nearest
neighborhood method.
Nearest Neighborhood Method
The nearest neighborhood method is the simplest
method used to order the nodes. In this method,
once a node is selected, the node which has the
minimum travel time from the previously selected
node is selected as the next node from the set of
unselected nodes.
As for example, if there are 4 nodes and node
2 is selected randomly as the first node to form
a route. The travel times from node 2 to node
1, node 3 and node 4 have to be evaluated. The
node with the least travel time is node 4 and will
be selected as the next node. The ravel time from
node 4 to node 1 and node 3 will be evaluated and
the least time is for mode 3 and will be chosen
to be the next node. Hence the route formed by
Nearest Neighborhood method will be 2-4-3-1.
Step 3: The deterministic value of the demand is
known. The probabilistic demand is calcu-
lated based on the probability distribution
function. The nodes are associated with the
stochastic demands.
Step 4: As a vehicle keeps on adding the nodes,
the capacity of the vehicle gets exhausted
gradually by adding up the demands of the
picked up nodes.
Step 5: Once the capacity of the vehicle is reached,
i.e., it can no longer serve any more nodes,
it returns to the depot. The time window
constraint has to be validated while forming
these routes. If the total time for the route
set does not fall within the time window, the
total delay time or waiting times are calcu-
lated and associated with the evaluation of
the solution set.
Step 6: Repeat steps 1 to 5 for the remaining de-
pots using the method mentioned for multi
depot condition in section 3.1.2.
By following the above steps, one initial set
of solution is generated. Similarly, k initial set of
solutions are generated randomly, given by:
867
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
Si :
{m x i 1 D1Ai BiC i D1x i 2 D2 Ei Fi D2 x i 3 D3 Gi H i I i D3 }
Sj :
{m x j 1 D1Aj B jC j D1x j 2 D2 E j Fj G j D2 x j 3 D3 H j I j D3 }
Sk :
{m x k 1 D1Ak BkC k D1x k 2 D2 Ek Fk D2 x k 3 D3 Gk H k I k D3 }
The initial set of solutions is represented by:
I : {S i , S j ,...Sk }
A sample problem has been presented to
demonstrate the mechanism of different steps of
the algorithm. A problem of 14 nodes is shown
in Figure 4. An initial set of routes (Figure 5) for
the problem may be represented as,
3 5 1 4 6 9 10 1 6 3 11 2 7 8 13 3 312 5 14 14
where, the first route is given by 1 4 6 9 10 1
where, the depot node is 1 and the 1st vehicle
traverses 4 nodes which are 4, 6, 9 and 10 other
than the depot 1. The second route consists of 6
Figure 4. Position of the customer nodes and the
depots shown in the problem
868
nodes, of which, 3 is the depot node and 11, 2, 7,
8 and 13 are the nodes traveled by 2nd vehicle
starting from depot node 3 and returning to the
same depot. The 3rd route consists of 5 nodes of
which 14 is the depot node and the nodes traversed
are 12 and 5.
2. Evaluation: After generation of the initial so-
lution sets, the fitness values of the solutions
are determined. Fitness value is calculated
by the total travel time cost and the penalty
values incurred due to the waiting and delay
time while picking up the deliveries at the
customer nodes. The fitness function is given
by Equation (11).
TC = ∑ ∑ dij × cij + C w × ∑ wi
(11)
+ C d × ∑ di + ∑ ∑ pi ⋅ dij ⋅ qi
3. Modification: The initial set of solutions is
modified to result into improved solutions.
Various modification schemes such as,
Reproduction, Crossover and Mutation are
used for this purpose. Two unique operators
are designed specially in accordance with
the nature of the problem. The operators are
i) Crossover-mutation, which is applied at
Figure 5. An example of initial route set
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

the end of the crossover and ii) Mutation- route of another solution. The
crossover, which is applied during mutation. exchange takes place between
a. Reproduction: Tournament Selection the randomly chosen sub routes
is applied for the reproduction opera- of two route sets.
tion. The intent of this operation is to Step 3: The selected segments are ex-
duplicate good solutions and eliminate changed provided the constraint
bad ones from the population. The steps of vehicle capacity does not get
to apply this operator are given below: violated in either case.
Step 1: The value of the tournament
selection of size is assumed to The process is described below:
be p for this purpose.
Step2: The m better solutions are cho- Sm : {m x m 1 D1Am BmC m D1x m 2 D2 Em Fm D2 x m 3

sen from the initial set of solutions D3Gm H m I m D3 }
based on the fitness Step function
values. The m selected solutions
are copied n times to develop Sn : {m x n 1 D1An BnC n D1x n 2 D2 En Fn D2 x n 3
the mating pool of population
D3 Gn H n I m D3 }
size mxn.
Step 3: From mxn population size, p
solutions are selected.
Step 4: The best of p solutions are The parts to be exchanged are selected ran-
selected and copied p times and domly from the two sets of solutions chosen for
kept in the mating pool. Thus, the the crossover, which are given as:
new mating pool will consist of
 m × n  pm : {BmC m } ⊂ Sm
 
 p  sets of solutions, each
of which contain p number of pn : {BnC n } ⊂ Sn
solution sets.
b. Crossover: Crossover takes place in
After the crossover the new sets of locations
a set of two solutions. New solutions
generated are:
are created by exchanging the segments
of any two solutions from the mating
Sm' : {m x m 1 D1Am BnC n D1x m 2 D2 Em Fm D2 x m 3
pool. So, there will be p number of D3 Gm H m I m D3 }

2
crossover operations in a mating pool
of p solutions. The steps to apply the
crossover operator are given below: Sn' : {m x n 1 D1An BmC m D1x n 2 D2 En Fn D2 x n 3
Step 1: Two solutions are picked up
D3 Gn H n I m D3 }
randomly.
Step 2: A few nodes of one route of
one solution set are selected
randomly and interchanged with Crossover-mutation operator: A new operator
the same number of nodes of a called Crossover-Mutation operator is de-
signed for the modification scheme while

869
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

solving the MDPVRPTW with the GA heu- Q ~ [β ]

ristic. It is a post-crossover operator. After the
nodes of the sub routes of two solutions get and
exchanged in the crossover process, the sub
routes except the participating sub routes in P ∪ [α ] → [{Smp ~ [β ]} ∪ [α ]] ∩ [{Svq ~ [α ]} ∪ [β ]
the crossover get mutated in order to satisfy
the constraint that each node can be visited
only once. The already existing nodes which For example, we consider two populations
become identical with the newly imported and after the crossover the two offspring (Figure
nodes due to crossover are eliminated and 6) obtained are:
the exported nodes are appended to the
same sub route. Thus, a modified sub route Population 1: 3 6 1 11 12 8 9 6 1 5 3 4 10 13 7
is generated due to the sudden change made 3 3 3 14 5 2 14
after the crossover. Population 2: 3 7 1 4 6 9 10 13 8 1 5 3 12 11 7
2 3 2 14 5 14
Offspring 1: 3 6 1 4 10 8 9 6 1 5 3 12 11 13 7 3
T
P : Sp  3 14 5 2 14
 
Offspring 2: 3 7 1 6 9 13 8 12 11 1 5 3 4 10 7 2
3 2 14 5 14
T
Q : S q 
 
c.. Mutation: Application of the mutation op-
erator is an occasional phenomenon where
[α ] : {α1, α2 ,...αm } ⊆ Snq random modification takes place within the
solution. A few of the nodes of one route are
and exchanged with some other nodes of another
route in the same solution set. A completely
[α ] : {α1, α2 ,...αn } ⊆ Snq new route set is developed by this process.

Parent:
[β ] : {β1, β2 ,...βn } ⊆ Svq Sn :
{m x n 1 D1An BnC n D1x n 2 D2 En Fn D2 x n 3 D3Gn H n I m D3 }

and Offspring:
Snnew :
{m x n 1 D1An BnnewC n D1x n 2 D2 En Fn D2 x n 3 D3 Gn H n I m D3 }
[β ] : {β1, β2 ,...βm } ⊆ S
p
u

In the above mentioned example in the first

The crossover is explained as follows:
route node Bn has been replaced by a new node,
P ~ [α ] Bnnew which gives rise to a new route set.

Mutation-Crossover Operator
and
The mutation takes place through intra-solution
P ∪ [β ] → [{Smp ~ [α ]} ∪ [β ]] ∩ [{Sup ~ [β ]} ∪ [α ] crossover, hence the name of the operator is

870
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

Figure 6. Development of new route sets by

Population 1: 3 6 1 8 13 10 4 7 1 5 3 9 6 5 2 3 4
crossover operator
14 12 11 14
Offspring: 3 5 1 8 9 6 7 1 6 3 13 10 4 5 2 3 4
14 12 11 14

The new solution sets are evaluated. The al-

gorithm is run for a number of generations until
the improvement in objective function becomes
negligible. The optimum set of routes which in-
curs the minimum cost is obtained. The program
terminates either when the solution converges or
the algorithm runs for a pre-specified number of
generation gmax.

ANALYTICAL RESULTS

Three artificial instances are generated randomly

and the heuristic is applied on them. The coordi-
nates and the demand of the nodes are generated
proposed as mutation-crossover. The nodes of randomly for the first two instances of 14 and
two sub routes within a solution are exchanged 25 nodes (P14 and P25), respectively. The third
to give rise to two new sub routes. The exchange
of information does not take place between two Figure 7. Development of a new route set by
solutions; rather it remains restricted within the mutation operator
same solution. The assignment of the nodes to
the depots changes within the same solution. So,
the operator can be presented as a special type of
mutation operator.

T
P : S 
 

[α ] : {α1, α2 ,...αm } ⊆ Sm ;

[β ] : {β1, β2 ,...βn } ⊆ Sn

P ' : [{Sm ~ [α ]} ∪ [β ]] ∩ [{Sn ~ [β ]} ∪ [α ] ]

Figure 7 shows an example of mutation and the

representation of the new chromosomes, which
are given as follows:

871
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

instance of 51 nodes (P51) is a modification from Population size for GA=30x2

the problem Eil51 (capacitated vehicle routing
problem) obtained from TSPLIB95 (TSPLIB
2001). The single depot problem has been modi-
fied to a multi depot problem by assigning a few
nodes as depot randomly. The demands of the
depot nodes have been changed to zero for the
problem while the demands of the other nodes
and the coordinates of the other nodes remain
the same. The heuristic is run 10 times on each of
the instances. The description of the problems is
given in Tables 1 through 3, respectively. The time
windows are set based on the near-optimal travel
times obtained for MDPVRP. The input values
used for the three instances are shown in Table 4.
The input parameter values used in the algo-
rithm are given below:
Vehicle speed=60 mph
Penalty coefficient or waiting and delay param-
eters (time window) Cw=0.2 and Cd=0.1
Parameter values for the probability function for
stochastic demand m=10, u=0.85
Number of generations for GA=100
A - indicates data not applicable
The optimal travel times obtained for MDVRP,
MDPVRP and MDPVRPTW for the 3 instances
are given in Table 5. Figures 8a through 8c show
the results obtained for the P14 problem for the
three cases, such as, MDVRP, MDPVRP and
MDPVRPTW. Similarly, Figures 9a through 10c
show the results for the P25 and P51 problems,
respectively.
For all the three instances the lower boundar-
ies of the objective functions are set by establish-
ing the time windows in such a way that all the
nodes satisfy the time window constraints. Hence,
there is no need for adding any penalty to the
total cost due to waiting or delay at the correspond-
ing nodes.
On the other hand, the coefficients of waiting
(C w ) and delay (C d ) are set to very low values
which are 0.002 and 0.001, respectively, to obtain
the upper boundaries of the objective functions.
Setting the coefficients to very low values allow
the vehicles to travel with the violation of the time

Table 1. Input values for P14 (14-node problem)

Node Type of node Coordinates of the node Demand (qi) Time window
(AM)
1 Depot 1.5, 8 - -
2 Customer 2.5, 10 30 9:54-10:02
3 Depot 4.5, 11 - -
4 Customer 8, 10.5 10 9:34-9:39
5 Customer 12, 11.5 00 10:16-10:20
6 Customer 6, 8.5 20 9:42-9:46
7 Customer 4.5, 7 30 9:45-9:52
8 Customer 6.5, 5 20 9:25-9:30
9 Customer 7.5, 7.5 30 9:31-9:36
10 Customer 10, 7.5 20 10:12-10:15
11 Customer 3, 2 10 9:10-9:14
12 Customer 5, 1 20 9:15-9:20
13 Customer 11, 4 10 10:03-10:08
14 Depot 8, 5 - -

872
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

Table 2. Input values for P25 (25-node problem) Table 3. Input values for P51 (51-node problem)

Node Type of Coordinates Demand Time Node Type of Coordinates Demand Time
node of the node (qi) window node of the node (qi) window
(AM) (AM)
1 Customer 2, 2 10 9:07- 1 Customer 37, 52 7 10:32-
9:12 10:36
2 Customer 4, 1 15 9:10- 2 Customer 49, 49 30 9:20-9:25
9:14
3 Customer 52, 64 16 10:02-
3 Customer 6, 2 20 9:00- 10:08
9:05
4 Customer 20, 26 9 10:48-
4 Depot 9, 1 - - 10:55
5 Customer 10, 2 20 9:04- 5 Customer 40, 30 21 9:18-9:23
9:10
6 Customer 21, 47 15 9:48-9:55
6 Customer 1, 4 10 9:03-
7 Customer 17, 63 19 9:25-9:32
9:09
8 Customer 31, 62 23 9:05-9:10
7 Depot 3, 4 - -
9 Customer 52, 33 11 9:44-9:50
8 Customer 5, 4 12 9:05-
9:08 10 Customer 51, 21 5 9:34-9:38
9 Customer 9, 4 15 9:01- 11 Depot 42, 41 - -
9:05
12 Customer 31, 32 29 9:40-9:46
10 Customer 7, 5 10 9:07-
13 Customer 5, 25 23 10:05-
9:10
10:14
11 Customer 11, 5 20 9:05-
14 Customer 12, 42 21 9:19-9:25
9:07
15 Customer 36, 16 10 10:20-
12 Customer 7, 6 16 9:04-
10:28
9:11
16 Customer 52, 41 15 9:35-9:44
13 Depot 8, 6 - -
17 Customer 27, 23 3 10:40-
14 Customer 5, 7 10 9:10-
10:45
9:17
18 Customer 17, 33 41 9:27-9:36
15 Customer 2, 8 15 9:05-
9:12 19 Customer 13, 13 9 10:37-
10:44
16 Customer 10, 8 10 9:17-
9:23 20 Depot 57, 58 - -
17 Depot 4, 10 - - 21 Customer 62, 42 8 9:23-9:30
18 Customer 6, 10 12 9:01- 22 Customer 42, 57 8 10:22-
9:05 10:29
19 Customer 11, 10 20 9:22- 23 Customer 16, 57 16 9:31-9:36
9:27
24 Depot 8, 52 - -
20 Customer 5, 11 10 9:02-
25 Customer 7, 38 28 9:45--9:53
9:06
26 Customer 27, 68 7 9:13-9:18
21 Customer 10, 11 - -
27 Customer 30, 48 15 10:45-
22 Customer 2, 12 00 9:10-
10:52
9:14
28 Customer 43, 67 14 10:10-
23 Customer 8, 12 15 9:05-
10:18
9:10
29 Customer 58, 48 6 9:16-9:24
24 Customer 7, 13 10 9:03-
9:08 30 Customer 58, 27 19 9:40-9:50
25 Customer 11, 13 20 9:18- 31 Depot 37, 69 - -
9:25
continued on following page

873
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

Table 3. Continued Table 4. Input parameter values

Node Type of Coordinates Demand Time Problem Number of vehicles Vehicle

node of the node (qi) window in the fleet capacity
(AM)
P14 3 80
32 Customer 38, 46 12 9:35-9:42
P25 5 80
33 Customer 46, 10 23 10:08-
P51 8 95
10:16
34 Depot 61, 33 - -
35 Customer 62, 63 17 9:55-9:62 Figure 8. a. Near-optimal route sets for MDVRP
36 Customer 63, 69 6 9:35-9:43 for P14 b.Near-optimal route set for MDPVRP for
37 Customer 32, 22 9 10:34- P14 c. Near-optimal route set for MDPVRPTW
10:38
for P14
38 Customer 45, 35 15 10:03-
10:08
39 Customer 59, 15 14 9:55-10:05
40 Customer 5, 6 7 10:02-
10:08
41 Depot 10, 17 - -
42 Customer 21, 10 13 9:44-9:50
43 Customer 5, 64 11 9:04-9:12
44 Customer 30, 15 16 10:27-
10:34
45 Depot 39, 10 - -
46 Customer 32, 39 5 9:56-10:12
47 Customer 25, 32 25 9:33-9:41
48 Customer 25, 55 17 9:35-9:45
49 Customer 48, 28 18 9:49-10:00
50 Customer 56, 37 10 9:27-9:38
51 Depot 30, 40 - -

Table 5. Results for the instances

Problem Number of depot Travel time (min) Travel time Travel time
MDVRP) (min) (min)
(MDPVRP) (MDPVRPTW)
P14 3 57.60 54.64 81.85
P25 5 90.68 86.69 113.69
P51 8 804.85 827.95 903.97

874
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
Figure 9. a. Near-optimal route sets for MDVRP
for P25 b. Near-optimal route set for MDPVRP
for P25 c. Near-optimal route set for MDPVRPTW
for P25
window constraints as much as possible and hence
incurring larger penalty cost, which is evaluated
using the original values of the coefficients after
formation of the final route set. The final objective
function values and the lower limit (LL) and up-
per limit (UL) of the objective function are shown
in Table 6.The final values of the objective func-
tions remain within the boundaries for all the three
instances. Hence, they validate the optimality
conditions for the instances randomly generated
for the study. The heuristic improves the objective
Figure 10. a. Near-optimal route sets for MDVRP
for P51 b.Near-optimal route set for MDPVRP for
P51 c.Near-optimal route set for MDPVRPTW
for P51
function from the initial solution sets by 22.7%
for the P14 problem, 15.7% for the P25 problem,
and 18.8% for the P51 problem. The average
computational time for all three examples was 30
minutes.
CONCLUSION
This paper presented a new heuristic for solv-
ing the multi depot probabilistic vehicle routing
875
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
Table 6. Lower and Upper boundaries of the
objective functions
Problem LL UL Objective
function value
P14 54.64 104.50 81.85
P25 86.69 135.31 113.79
P51 827.95 1117.85 903.97
problem with time window (MDPVRPTW). The
MDPVRPTW represents a real-world situation
and has many applications, such as the pick-up
of children by school buses, garbage collection,
delivery of laundry, and delivery by the United
Postal Service (UPS). Past studies only solved
various components of MDPVRPTW separately
since it was difficult to devise a solution procedure
for a vehicle routing problem by simultaneously
considering multi-depot, probabilistic demand,
and time window constraints. In our work pre-
sented here we were able to develop a customized
genetic algorithm based heuristic to solve the
MDPVRPTW. We think this will open up the pos-
sibility for the implementation of the developed
approach in many real-world applications.
The GA-based heuristic exhibits robust nature
and efficiency in terms of computing time for the
MDPVRPTW problem. The optimal objective
function values lie within the specified limits for
all the three instances. Hence, we can infer that
the proposed heuristic provides a feasible solu-
tion. While there is no rigours proof to show the
optimality of the solutions, it is at least intuitively
evident that the solutions are optimal as they have
been compared in their limiting cases with classical
VRPs by relaxing the time-window and stochastic
restitutions. It is also found that on the average,
there is about 20% improvement from the initial
solution sets for the three instances. Additional
tests need to be performed to test the efficiency
of the heuristic on larger problems.
876
ACKNOWLEDGMENT
This work was completed at the Center for
Advanced Transportation and Infrastructure
Engineering Research (CATIER) at the Morgan
State University.
REFERENCES
Alvarenga, G. B., Mateus, G. R., & Tomi, G.
(2007). A genetic and set partitioning two-phase
approach for the vehicle routing problem with time
windows. Computers & Operations Research,
34, 1561–1584. doi:10.1016/j.cor.2005.07.025
Aykagan, A., & Erera, A. L. (2007). A paired-
vehicle recourse strategy for the vehicle-routing
problem with stochastic demands. Transpor-
tation Science, 41(2), 222–237. doi:10.1287/
trsc.1060.0180
Baker, B. M., & Ayechew, M. A. (2003). A ge-
netic algorithm for the vehicle routing problem.
Computers & Operations Research, 20, 787–800.
doi:10.1016/S0305-0548(02)00051-5
Bent, R., & Hentenryck, P. V. (2004). A two-stage
hybrid local search for the vehicle routing problem
with time window. Transportation Science, 38(4),
515–530. doi:10.1287/trsc.1030.0049
Berger, J., & Barakoui, M. (2004). A parallel
hybrid genetic algorithm for the vehicle routing
problem with time windows. Computers & Op-
erations Research, 31, 2037–2053. doi:10.1016/
S0305-0548(03)00163-1
Bertsimas, D. J. (1992). A vehicle routing problem
with stochastic demand. Operations Research,
40(3), 574–585. doi:10.1287/opre.40.3.574
Bertsimas, D. J., Jaillet, P., & Odoni, A. R. (1990).
A priori optimization. Operations Research, 38(6),
1019–1033. doi:10.1287/opre.38.6.1019
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
Bramel, J., & Simchi-Levi, D. (1996). Probabi-
listic analyses and practical algorithms for the
vehicle routing problem with time windows. Op-
erations Research, 44(3), 501–509. doi:10.1287/
opre.44.3.501
Braysy, O., & Gendreau, M. (2005). Vehicle rout-
ing problem with time windows, part I: Route
construction and local search algorithms. Trans-
portation Science, 39(1), 104–118. doi:10.1287/
trsc.1030.0056
Braysy, O., & Gendreau, M. (2005). Vehicle
routing problem with time windows, part II:
Metaheuristics. Transportation Science, 39(1),
119–139. doi:10.1287/trsc.1030.0057
Chakroborty, P., & Mandal, A. (2005). An asexual
genetic algorithm for the general single vehicle
routing problem. Engineering Optimization,
37(1), 1–27. doi:10.1080/03052150410001721
468
Chan, Y., Carter, W. B., & Burnes, M. D. (2001). A
multiple-depot, multiple vehicle, location routing
problem with stochastically processed demands.
Computers & Operations Research, 28, 803–826.
doi:10.1016/S0305-0548(00)00009-5
Chen, H., Hsueh, C., & Chang, M. (2006). The
real-time time dependent vehicle routing prob-
lem. Transportation Research Part E, Logistics
and Transportation Review, 42(5), 222–237.
doi:10.1016/j.tre.2005.01.003
Cordeau, J.-F., Laporte, G., & Mercier, A. (2001).
A unified tabu search heuristic for vehicle routing
problems with time windows. The Journal of
the Operational Research Society, 52, 928–936.
doi:10.1057/palgrave.jors.2601163
Crevier, B., Cordeau, J.-F., & Laporte, G. (2007).
The multi depot vehicle routing problem with
interdepot routes. European Journal of Op-
erational Research, 176, 756–773. doi:10.1016/j.
ejor.2005.08.015
Desaulniers, G., Francois, L., & Ahmed, H. (2009).
tabu search, partial elimentarity, and generalized
K-path inequalities for the vehicle routing prob-
lem with time windows. Transportation Science,
42(3), 387–404. doi:10.1287/trsc.1070.0223
Dror, M. (2003). Modeling vehicle routing with
uncertain demand as a stochastic program: Prop-
erties of the corresponding solution. European
Journal of Operational Research, 64, 432–441.
doi:10.1016/0377-2217(93)90132-7
Dror, M., Laporte, G., & Trudeau, P. (1989). Ve-
hicle routing with stochastic demands: Properties
and solution framework. Transportation Science,
23(3), 166–175. doi:10.1287/trsc.23.3.166
Fisher, M. L., Jornsten, K. O., & Madsen, O. B. G.
(1997). Vehicle routing with time windows: Two
optimization algorithm. Operations Research,
45(3), 488–492. doi:10.1287/opre.45.3.488
Gendreau, M., Laporte, G., & Seguin, R. (1994).
A tabu search heuristic for the vehicle routing
problem with stochastic demands and custom-
ers. Operations Research, 44(3), 469–477.
doi:10.1287/opre.44.3.469
Gendreau, M., Laporte, G., & Seguin, R. (1996).
An exact algorithm for the vehicle routing problem
with stochastic demands and customers. Trans-
portation Science, 29(2), 143–155. doi:10.1287/
trsc.29.2.143
Gendreau, M., Laporte, G., & Seguin, R.
(1996). Stochastic vehicle routing. European
Journal of Operational Research, 88, 3–12.
doi:10.1016/0377-2217(95)00050-X
Hwang, H. (2002). An improved model for ve-
hicle routing problem with time constraint based
on Genetic Algorithm. Computers & Industrial
Engineering, 42, 361–369. doi:10.1016/S0360-
8352(02)00033-5
877
 Multi Depot Probabilistic Vehicle Routing Problems with a Time Window
Jaillet, P., & Odoni, A. R. (1998). The probabi-
listic vehicle routing problem. Vehicle Routing:
Methods and Studies, 2, 125–247.
Jong, J.-C., & Schonfeld, P. (2003). An evolution-
ary model for simultaneously optimizing 3-di-
mensional highway alignments. Transportation
Research Part B: Methodological, 37(2), 107–128.
doi:10.1016/S0191-2615(01)00047-9
Jung, S., & Haghani, A. (1999). A genetic algo-
rithm for pick-up and delivery problem with time
windows. Transportation Research Record, 35,
123–126.
Jung, S., & Haghani, A. (2001). A genetic al-
gorithm for the time-dependent vehicle routing
problem. Transportation Research Record, 1771,
164–171. doi:10.3141/1771-21
Kim, E., Jha, M. K., & Son, B. (2005). Improving
the computational efficiency of highway alignment
optimization models through a stepwise genetic
algorithms approach. Transportation Research
Part B: Methodological, 39(4), 339–360.
Kohl, N., & Madsen, O. B. G. (1997). An optimi-
zation algorithm for the vehicle routing problem
with time windows based on Lagrangian relax-
ation. Operations Research, 45(3), 395–405.
doi:10.1287/opre.45.3.395
Laporte, G., Gendreau, M., Potvin, J. Y., & Semet,
F. (2000). Classical and modern heuristics for the
vehicle routing problem. International Transac-
tions in Operational Research, 7(4-5), 285–300.
doi:10.1111/j.1475-3995.2000.tb00200.x
Laporte, G., Nobert, Y., & Taillefer, S. (1988).
Solving a family of multi depot vehicle routing
and location routing problems. Transportation Sci-
ence, 22(3), 161–172. doi:10.1287/trsc.22.3.161
878
Liu, S., & Lai. (2009). The vehicle routing problem
with uncertain demand at nodes. Transporta-
tion Research Part E, Logistics and Transpor-
tation Review, 45(4), 517–524. doi:10.1016/j.
tre.2008.11.002
Nagy, G., & Salhi, S. (2005). Heuristic algorithms
for single and multiple depot vehicle routing
problems with pickups and deliveries. European
Journal of Operational Research, 162, 126–141.
doi:10.1016/j.ejor.2002.11.003
Ombuki, B., Ross, B. J., & Hanshar, F. (2006).
Multi-objective genetic algorithm for vehicle
routing problem with time windows. Applied
Intelligence, 24, 17–30. doi:10.1007/s10489-
006-6926-z
Renaud, J., Laporte, G., & Boctor, F. F. (1996). A
tabu search heuristic for the multi-depot vehicle
routing problem. Computers & Operations
Research, 23(3), 229–235. doi:10.1016/0305-
0548(95)O0026-P
Ropke, S., & Cordeau, J. (2009). Branch and Cut
and Price for the pick up and delivery problem with
time windows. Transportation Science, 43(3),
267–286. doi:10.1287/trsc.1090.0272
Salhi, S., & Sari, M. (1997). A multi-level com-
posite heuristic for the multi-depot vehicle fleet
mix problem. Operations Research, 103, 95–112.
Secomandi, N. (2001). Comparing neuro-dynamic
programming algorithms for the vehicle routing
problem with stochastic demands. Comput-
ers & Operations Research, 27, 1201–1225.
doi:10.1016/S0305-0548(99)00146-X
Thangiah, S. R., & Salhi, S. (2001). Ge-
netic clustering: An adaptive heuristic for the
multi depot vehicle routing problem. Ap-
plied Artificial Intelligence, 15, 361–383.
doi:10.1080/08839510151087293
Multi Depot Probabilistic Vehicle Routing Problems with a Time Window

Tong, Z., Ning, L., & Debao, S. (2004). Genetic
algorithm for vehicle routing problem with time
window and uncertain vehicle number. In Pro-
ceedings of the 5th World Congress on Intelligent
Control and Automation, Hongzhou, China.
Wu, T., Low, C., & Bai, J. (2002). Heuristic solu-
tions to multi-depot location-routing problems.
Computers & Operations Research, 29(10), 1393–
1415. doi:10.1016/S0305-0548(01)00038-7

This work was previously published in the International Journal of Operations Research and Information Systems (IJORIS),
Volume 2, Issue 2, edited by John Wang, pp. 40-64, copyright 2011 by IGI Publishing (an imprint of IGI Global).

879
880

Chapter 54
Estimating Potential Woody
Biomass in Communal Savanna
Woodlands from Synthetic
Aperture Radar (SAR)
Charles Paradzayi
University of Johannesburg, South Africa & Midlands State University, Zimbabwe

Harold J. Annegarn
University of Johannesburg, South Africa

ABSTRACT
Recent developments in Synthetic Aperture Radar (SAR) technologies have shown their potential for as-
sessing and quantifying above-ground biomass (AGB) at landscape levels in different biomes. This paper
examines the application of full polarimetric data to retrieve information related to potential woody bio-
mass in sparse communal savanna woodlands in southern Africa using the Advanced Land Observation
Satellite’s Phased Array L-band Synthetic Aperture Radar (ALOS PALSAR). Woody vegetation classes
were obtained from the unsupervised entropy/alpha Wishart classification of the full polarimetric ALOS/
PALSAR data. A combination of Differential GPS and conventional surveying techniques was used for a
field inventory survey to estimate plot-level biomass densities in Welverdiend communal woodlands of
South Africa. Regression analysis was used to derive the logarithmic relationship between the sampled
plot AGB densities and the mean backscatter intensity of the microwave signal, which is transmitted in
the horizontal plane and received in the vertical plane (HV). The AGB density for each woody vegeta-
tion class is estimated by solving the logarithmic equation after extracting the mean HV backscatter
intensity for the particular vegetation class. The potential woody biomass is estimated from the derived
AGB densities and the areal extent of the respective woody vegetation classes.

DOI: 10.4018/978-1-4666-2038-4.ch054

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)
INTRODUCTION
Population growth and persistent poverty are
increasing pressure on forest and woodland
resources, thereby threatening their sustainable
utilisation. In some electrified communities, the
high cost of electricity has generally prevented
a move away from dependency on bio-energy
(Madubansi & Shackleton, 2007). At the same
time, intensive harvesting practices and demands
for arable farming land are contributing to the con-
tinued decrease in the availability of tree species
preferred for fuelwood and/or charcoal. Indicators
of energy poverty such as increased distance of
preferred species from households are beginning
to emerge in certain communal woodlands that
are not subject to rigorous management regimes
(Howells et al., 2003; Madubansi & Shackleton,
2007). This is resulting in the emergence of both
micro- and macro-level woody biomass trading
within and among villages, and from rural to urban
areas respectively. In Mozambique, Tanzania and
Zambia, charcoal from communal woodlands is
exported to urban centres for sale to low-income
urban dwellers. However, there is a general paucity
of quantitative data on available woody biomass
resources to provide critical input into energy
planning models at the landscape level in most
rural economies (Mabowe, 2006). The objective
of this paper is to investigate the potential of using
the intensity of the backscatter signal of polari-
metric synthetic aperture radar to quantify above
ground biomass density and then map the spatial
distribution of available above ground biomass
resources in savanna woodlands.
Utilisation of Woody
Biomass Resources in Low-
Income Communities
Savanna woodlands provide a traditional source
of woody biomass and other non-timber forestry
products for a large percentage of households in
both rural and urban low-income communities in
southern Africa (Scholes & Archer, 1997; Twine,
Moshe, Netshiluvhi, & Siphugu, 2003; Whitlow,
1979). The predominant source of bioenergy in
most low-income rural and urban households is
woody biomass in the form of charcoal and fu-
elwood (Banks, Griffin, Shackleton, Shackleton,
& Mavrandonis, 1996; Wamukonya & Jerkins,
1995). In Mozambique, biomass (charcoal and
firewood) provides up to 80% of energy consump-
tion while 87% of rural households in Zambia use
fuelwood for cooking (Brouwer & Falcão, 2004;
Chidumayo, Masialeti, Ntalasha, & Kalumiana,
2002). Quantitative information about available
biomass resources required to meet domestic
energy demands is lacking in most developing
economies. In addition, quantitative information
on woody biomass resources is becoming critical
for national inventories of woody biomass stocks
for carbon sequestration and accounting purposes
as well as for meeting the reporting requirements
for international conventions and protocols such
as the Kyoto Protocol on Climate Change and the
Reduction of Emissions from Deforestation and
forest Degradation (REDD) initiatives (Gibbs,
Brown, Niles, & Foley, 2007; Woodhouse, Cas-
sells, Mitchard, & Tembo, 2009).
Biomass Estimation in
Savanna Woodlands
Conventional field inventory and optical remote
sensing approaches have been used in the past to
estimate available above ground biomass (Chidu-
mayo, 1997; Ellegård, Chidumayo, Malimbwi,
Pereira, & Voss, 2002; Grundy, 1995). Field
inventory measurements provide essential data
for validating remotely sensed biomass estimates.
However, such approaches are tedious and time-
consuming. They are also constrained by ac-
cessibility problems, thus limiting the sampling
approaches to areas where measuring equipment
can be delivered (JICA, 2005; Mabowe, 2006;
Netshiluvhi & Scholes, 2001). In savanna wood-
lands, allometric equations relating measurable
881
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)
tree parameters to above ground biomass quantities
were established in previous studies (Netshiluvhi
& Scholes, 2001). The total tree biomass (TB) is
estimated using the equation:
TB = 0.035 * DAH2.5 (1)
where DAH is the diameter at ankle height.
Recent developments in the airborne and
spaceborne synthetic aperture radar technologies
have shown great promise in obtaining standing
woody biomass estimates in boreal, temperate
and tropical forests (Mette, 2007; Papathanas-
siou, Lee, Kugler, & Hajnsek, 2008; Prakoso,
2006). Studies in these biomes have shown that
the HV polarization of the L-band is sensitive
to biomass densities up to 200 t ha–1 (Imhoff,
1995; Viergever, Woodhouse, & Stuart, 2007;
Wallington, Viergever, Stuart, Moss, & Wood-
house, 2006; Watanabe et al., 2006). In sparse
and heterogeneous woodlands, the backscatter
intensity from the HH and VV polarisation has
shown positive correlation with increasing woody
biomass densities (Lucas et al., 2006; Saatchi,
Halligan, Despain, & Crabtree, 2007; Santoro,
2003; Santos et al., 2003). Integrated approaches
that combine terrestrial and spaceborne techniques
have been used for assessing biomass resources at
the landscape level (Avitabile et al., 2008; Moss,
Stuart, Viergever, & Wallington, 2006; Narvaes,
Silva, & Santos, 2007).
METHODS
Field Inventory
The field inventory survey involved GPS mea-
surements using the OmniLOGGER Differential
GPS system (from OmniSTAR, South Africa) to
establish control points with known coordinates
for the subsequent detailed topographic surveying
of trees in sampled plots. A total of 24 plots were
sampled, ranging between 0.05 – 0.32 ha. The
882
training plots were characterised by heterogeneous
distribution of tree species and only trees with trunk
circumferences greater than eight centimetres were
sampled. A Total Station (surveying instrument
with integrated direction and distance measuring
devices) was used to measure the geographic
location and heights of sampled trees from the
established control points because GPS degrades
when operated under tree canopies (Paradzayi,
Annegarn, Matsika, & Erasmus, 2008). Tree basal
diameter at ankle height (DAH) and the tree height
parameters were also recorded during the field
inventory surveys. The relatively short vegetation
structure in the savanna woodlands of this region
makes it impractical to measure the conventional
diameter at breast height to use with allometric
equations (Mabowe, 2006).
ALOS/PALSAR Quadpol Dataset
This study analysed the relationship between the
normalised HV backscatter intensity and above-
ground biomass density using full polarimetric
ALOS/PALSAR [ALPSRP148026690] dataset.
The data were acquired over the communal sa-
vanna woodlands around Welverdiend Village
(24° 35’S, 31° 20’E) in South Africa under wet
conditions on 3 November 2008. The data were
obtained at an incidence angle of 23.1°. The
PALSAR data have a pixel spacing of 9.4 m in
azimuth, resulting in a ground range resolution of
23.88 m. The pixel spacing in the range direction
is 3.5 m. The ground range resolution is used to
determine the ratio between the azimuth and range
pixels to produce square pixels in the SAR prod-
ucts. This process is referred to as multi-looking.
A multi-look ratio of 7:1 was adopted for this
work to give square pixels of 25 m x 25 m. The
quadpol data were processed using a combina-
tion of open source Polsarpro® and commercial
GAMMA® SAR software. Polsarpro was used to
decompose the ALOS/PALSAR data into entropy,
anisotropy and alpha angle elements. The terrain
classes were generated by running the unsuper-
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)
vised entropy/alpha (H/α) Wishart classification
algorithm. The commercial SAR software was
used for normalising the backscatter intensity
and also to geocode the products to the Universal
Transverse Mercator projection.
H/A/alpha Decomposition
The polarimetric data were decomposed into
entropy (H), anisotropy (A) and alpha angle (α)
elements using the Cloude and Pottier (eigenvec-
tor-eigenvalue) decomposition approach (Cloude
& Pottier, 1997). The entropy describes the degree
of statistical disorder of the scattering phenom-
enon from the logarithmic sum of the eigenvalues
of the coherency matrix. High entropy values
indicate the presence of more than one dominant
scattering mechanism. The anisotropy measures
the relative importance of the second and third
eigenvalues of the H/A/alpha decomposition
(Harter, Engdahl, & Desnos, 2007). Anisotropy
expresses the relation between the secondary
scattering processes such as between surface and
double bounce if volume is dominant. The alpha
angle is continuous within a range of 0o ≤ α ≤ 90o
and is used together with the entropy to differen-
tiate the scattering mechanisms present in a SAR
decomposition. α = 0o represents isotropic sur-
face scattering produced by a rough surface;
α = 90o represents dihedral (double-bounce)
scattering; α = 45o represents oriented dipole
(volume) scattering typical of canopy scattering
(Hajnsek, 2001; Lee et al., 1999, 2004).
Unsupervised H/α Wishart
Classification
The classification procedure combines the unsu-
pervised classification approach based on polari-
metric target decompositions and the maximum
likelihood classifier based on the complex Wishart
distribution (Harter et al., 2007; Lee et al., 1999).
The classification segmentation modifies the
boundaries in an adaptive way to better fit the
natural distribution of the scattering mechanisms
and also takes into account information related
to the backscattered power. In Figure 1, the en-
tropy/alpha angle plane is segmented into eight
zones representing the scattering mechanism of
each pixel of the PALSAR image after the H/α
decomposition. The boundaries of the segmented
H/α plane are used to discriminate the surface,
volume and double bounce scattering along the
alpha axis and low, medium and high degree of
randomness along the entropy axis. The volume
and dihedral scattering mechanisms are associated
with the presence of woody vegetation.
RESULTS
Sampled Plot Biomass Density
Individual tree biomass was obtained from exist-
ing allometric equations (Equation 1) from the
ground-truthing survey of test plots within the
savanna woodlands. The plot biomass density was
calculated by dividing the total plot biomass by
the respective areal extent of the sampled plot. A
summary of the field inventory results in selected
plots around Welverdiend village are shown in
(Table 1). A GIS layer (shapefile) of the sampled
plots were generated and used to extract the mean
of the normalised HV backscatter intensity over
the sampled plots.
Unsupervised Classification Results
Scattering from buildings is associated with
the complex structures class. Strong reflections
from residential buildings and tree colonies in
open areas result in the strong double bounce
scattering. Surface reflections are classified as
either Bragg or rough surfaces. The scattering
from natural vegetation and some built up areas
is associated with volume and dihedral reflections
corresponding to the dense, mixed and sparse
883
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)

Figure 1. H/α segmented plane

woody vegetation classes. Areas such as roads and double bounce, anisotropic, dihedral and dipole
some built-up areas that are dominated by both scattering mechanisms are closely related to the
surface and double bounce scattering are shown interaction between the microwave signal and tree
in shades of pink. The results of the unsupervised components. Hence, these classes can be used to
H/α Wishart classification is made up of eight infer the amount of above ground woody biomass.
classes representing complex structures, random Aerial photographs were used to validate the
anisotropic scatterers, Bragg surfaces, forestry polarimetric classification results. In Figure 2, the
double bounce scattering, anisotropic (vegetation) classified map was validated by overlying it on
scattering, rough surfaces, dihedral scatterers aerial photographs from 2007 obtained from the
and dipole scattering mechanisms. The forestry Bushbuckridge Municipality (Vincent, personal
communication, June 2009). A visual analysis
of the classified product shows a high level of
Table 1. Summary of training plot biomass den-
correlation between the classes and the expected
sities
scattering mechanisms depicted on the aerial
Training AGB Density Training AGB Density photographs. Relatively new built-up areas show
Plot (tons ha-1) Plot (tons ha-1) dominant surface and double bounce scattering as
N1 1.7 SE2 2.8 indicated by the pinkish colour. Established resi-
N2 1.5 SE3 7.0 dential areas show increased presence of dipoles,
N3 0.8 W1 24.9 dihedrals and vegetation scattering mechanisms
NE1 5.5 W2 5.8 from citrus trees and also some preserved natural
NE2 4.1 W3 14.5 trees and hedges used for providing shade. These
NE3 0.7 W4 7.6 are shown in shades of yellow and green. Bragg
NW1 6.0 W5 80.4 and rough surfaces are also notable in sky and
NW2 3.9 North 2 0.8 navy blue colours respectively.
S1 1.5 North 3 2.8 The south-western part has a good woody
S2 45.3 West 6 9.9
vegetation cover as evidenced by the dominant
S3 4.7 North 1 1.8
anisotropic (vegetation), dipoles and forestry
SE1 11.1 WLFC 17.0

884
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)

Figure 2. Validation of classified map with 2007 aerial photograph

double bounce scattering mechanisms in green,
yellow and red colours respectively.
Relating HV Backscatter Intensities
and Plot Biomass Densities
Analysis of the relationship between AGB density
and SAR polarisation backscatter intensity showed
that HV and VV have better regression coefficients
for biomass assessment in savanna woodlands
compared to the backscatter intensity from the
HH polarisation (Paradzayi, 2010). A 25 m buffer
was introduced around each of the training plots
to account for possible geolocation errors when
extracting the mean backscatter intensity for the
respective training plot. Regression analysis was
performed to test the correlation between HV
backscatter intensities and the plot biomass densi-
ties obtained during the field campaign. Figure 3
shows the results of the regression analysis using
logarithmic fit between the plot biomass densities
and the respective plot HV backscatter intensities.
The mean backscatter intensities showed strong
correlation with increased woody vegetation
densities for the sampled woodlands around
Welverdiend village, with a correlation coefficient
of 0.84. The regression equation derived from the
logarithmic fit between the field inventory above
ground biomass estimates and the mean HV
backscatter intensities (σHV
o
) is given in Equation
2.
o
σHV = 19.084 + 1.019 * ln(AGB ) (2)

Figure 3. HV backscatter intensity and plot AGB density regression analysis

885
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)
Table 2. Above ground estimates in Welverdiend communal woodlands
Vegetation Class Mean HV
4. Forest double bounces -14.68
5. Vegetation -16.65
7. Dihedral scatterers -12.44
8. Dipoles -14.30
The forest double bounce, anisotropic scatter-
ing, dihedral scattering and dipoles classes were
used to calculate the potential woody biomass
within the limits of Welverdiend communal wood-
lands because they are related to components of
above-ground biomass. The total above-ground
biomass within Welverdiend communal areas
was then calculated as the sum of the product of
the areal extent of the vegetation classes and the
mean AGB densities (Table 2).
The areal extent of each vegetation class was
derived from GIS analysis. The potential above
ground biomass is estimated to be over 52 000
tons for the communal woodlands within the
boundaries of Welverdiend village.
DISCUSSION
The results presented in this paper have dem-
onstrated the potential of using full polarimet-
ric ALOS PALSAR data to retrieve potential
woody biomass quantities in communal savanna
woodlands. The microwave imagery was decom-
posed and classified using both open source and
commercial SAR applications. The commercial
software was used to project the POLSAR data
onto conventional coordinate systems such as the
Universal Transverse Mercator. The results of the
unsupervised H/α Wishart classification showed
relatively high qualitative correlation between
the derived vegetation classes and the aerial pho-
tographs and other existing geospatial datasets.
886
AGB Density (tons
/ ha) Area (ha) AGB (tons)
75 126 9472
11 732 8013
678 21 14375
109 192 20924
1071 52783
The results of the regression analysis between
the HV backscatter and above ground biomass
densities enable the estimation of AGB densities
using the mean HV backscatter intensity values
for the woody vegetation related polarimetric
classes. The logarithmic equation applies only
to acquisitions under wet conditions since the
soil moisture has a significant influence on the
interaction of the backscatter intensity with woody
vegetation. The 24 training plots were confined to
woodlands around Welverdiend due to logistical
and time constraints. The methods used for the
field inventory surveys proved to be tedious and
time-consuming but allowed the detailed col-
lection of tree parameters required to determine
the plot-level biomass densities. However, the
regression analysis revealed the need for more
training plots with larger geographic coverage.
Such training plots must transcend the full spec-
trum of savanna vegetation (from sparse to dense
vegetation). Larger training plots will account for
the heterogeneous nature of savanna woodlands
and minimise the effect of residual geolocation
errors, leading to more rigorous logarithmic fitting
and robust regression equations.
CONCLUSION
The logarithmic equation can be used to estimate
the amount of available above ground woody
biomass resources at the landscape scale after
classifying woody vegetation classes from full
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)
polarimetric SAR data. Improved estimates can
be obtained by accounting for restrictions imposed
by land tenure and nature conservation policies
governing the Welverdiend community. Biomass
resources in built-up areas, along river banks
and other restricted areas within the limits of the
communal area jurisdiction must be excluded
from the calculation of the available resources for
firewood and charcoal production. Results from
such analysis will provide decision makers with
more accurate information on available woody
biomass resources for modelling the sustainable
utilisation of such resources in rural communities,
providing a quantitative tool for monitoring and
assessing savanna woodlands.
ACKNOWLEDGMENT
This paper was supported by VW Foundation
Southern Africa Biomodelling Project and a Na-
tional Research Foundation Focus Area Grant:
FA2005040600018 “Sustainability Studies Using
GIS and Remote Sensing” to Harold J Annegarn.
The authors acknowledge European Space Agency
(ESA) and the Japan Aerospace Exploration
Agency (JAXA) for providing ALOS PALSAR
full polarimetric data for research purposes under
project C1P. 6143. The authors also acknowledge
the technical assistance from Andreas Vollrath and
Ralf Knuth (University of Jena, Germany) in pro-
cessing and geocoding the ALOS PALSAR data.
REFERENCES
Avitabile, V., Marchesini, L. B., Balzter, H.,
Bernoux, M., Bombelli, A., & Hall, R. (2008).
Biomass: Assessment of the status of the devel-
opment of standards for the terrestrial essential
climate variables. Rome, Italy: Global Terrestrial
Observing System.
Banks, D. I., Griffin, N. J., Shackleton, C.
M., Shackleton, S. E., & Mavrandonis, J. M.
(1996). Wood supply and demand around two
rural settlements in semi-arid savanna, South
Africa. Biomass and Bioenergy, 11(4), 319–331.
doi:10.1016/0961-9534(96)00031-1
Brouwer, R., & Falcão, M. P. (2004). Wood
consumption in Maputo, Mozambique. Biomass
and Bioenergy, 27, 233–245. doi:10.1016/j.biom-
bioe.2004.01.005
Chidumayo, E. N. (1997). Miombo ecology and
management: An introduction. Stockholm, Swe-
den: Intermediate Technology Publications.
Chidumayo, E. N., Masialeti, I., Ntalasha, H., &
Kalumiana, O. S. (2002). Charcoal potential in
Southern Africa final report for Zambia. Stock-
holm, Sweden: Stockholm Environment Institute.
Cloude, S. R., & Pottier, E. (1997). An entropy
based classification scheme for land applica-
tions of polarimetric SAR. IEEE Transactions
on Geoscience and Remote Sensing, 35, 68–78.
doi:10.1109/36.551935
Ellegård, A., Chidumayo, E., Malimbwi, R.,
Pereira, C., & Voss, A. (2002). Charcoal poten-
tial in Southern Africa (CHAPOSA) final report.
Stockholm, Sweden: Stockholm Environment
Institute.
Gibbs, H. K., Brown, S., Niles, J. O., & Foley,
J. A. (2007). Monitoring and estimating tropical
forest carbon stocks: Making REDD a real-
ity. Environmental Research Letters, 2, 1–13.
doi:10.1088/1748-9326/2/4/045023
Grundy, I. M. (1995). Wood biomass estimation
in dry miombo woodland in Zimbabwe. For-
est Ecology and Management, 72, 109–117.
doi:10.1016/0378-1127(94)03467-B
Hajnsek, I. (2001). Inversion of surface parameters
using polarimetric SAR. Unpublished doctoral
dissertation, University of Jena, Jena, Germany.
887
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)
Harter, M., Engdahl, M., & Desnos, Y. L. (2007,
January 22-26). First assessment of polarimetric
images from ALOS - PALSAR: Check of polari-
metric calibration and assessment of land-cover
classification potential. Paper presented at the
3rd International Workshop on Science and Ap-
plications of SAR Polarimetry and Polarimetric
Interferometry, Frascati, Italy.
Howells, M. I., Alfstad, T., David, G., Victor, D.
G., Goldstein, G., & Remme, U. (2003). An energy
model for a low income rural African village.
Stanford, CA: Stanford University.
Imhoff, M. L. (1995). Radar backscatter and
biomass saturation: Ramifications for global
biomass inventory. IEEE Transactions on
Geoscience and Remote Sensing, 33, 511–518.
doi:10.1109/36.377953
JICA. (2005). Manual of biomass survey and
analysis. Tokyo, Japan: Japanese International
Cooperation Agency (JICA).
Lee, J. S., Boerner, W. M., Schuler, D. L., Ain-
sworth, T. L., Hajnsek, I., & Papathanassiou, K. P.
(2004). A review of polarimetric SAR algorithms
and their applications. Photogrammetry and
Remote Sensing, 9(3), 31–80.
Lee, J. S., Grunes, M. R., Ainsworth, T. L., Du, L.
J., Schuler, D. L., & Cloude, S. R. (1999). Unsu-
pervised classification using polarimetric decom-
position and the complex Wishart classifier. IEEE
Transactions on Geoscience and Remote Sensing,
37(5), 2249–2258. doi:10.1109/36.789621
Lucas, R. M., Cronin, N., Lee, A., Moghaddam,
M., Witte, C., & Tickle, P. (2006). Empirical
relationships between AIRSAR backscatter and
LiDAR-derived forest biomass, Queensland,
Australia. Remote Sensing of Environment, 100,
407–425. doi:10.1016/j.rse.2005.10.019
888
Mabowe, B. R. (2006). Aboveground biomass
assessment in Serowe woodlands, Botswana.
Unpublished doctoral dissertation, ITC, Enschede,
The Netherlands.
Madubansi, M., & Shackleton, C. M. (2007).
Changes in fuelwood use and selection fol-
lowing electrification in the Bushbuckridge
lowveld, South Africa. Journal of Environmental
Management, 83, 416–426. doi:10.1016/j.jenv-
man.2006.03.014
Mette, T. (2007). Forest biomass estimation from
polarimetric SAR interferometry. Unpublished
doctoral dissertation, Technischen Universität
München, Munich, Germany.
Moss, D., Stuart, N., Viergever, K., & Wallington,
E. (2006). SAR remote sensing for natural resource
management - is there a role for ground survey?
Geomatics World, 24-28.
Narvaes, I. S., Silva, A. Q., & Santos, J. R. (2007,
July 23-27). Evaluation of the interaction between
SAR L-band signal and structural parameters of
forest cover. In Proceedings of the IEEE Inter-
national Symposium on Geoscience and Remote
Sensing, Barcelona, Spain (pp. 1607-1610).
Netshiluvhi, T. R., & Scholes, R. J. (2001). Al-
lometry of South Africa woodland trees. Pretoria,
South Africa: CSIR.
Papathanassiou, K. P., Lee, S. K., Kugler, F., &
Hajnsek, I. (2008, July 6-11). Polarimetric SAR
interferometry for forest structure parameter esti-
mation: Potential and limitations. In Proceedings
of the IEEE International Symposium on Geosci-
ence and Remote Sensing, Boston, MA.
Paradzayi, C. (2010). Polarimetric synthetic aper-
ture radar (POLSAR) biomass estimation in com-
munal African savanna woodlands. Unpublished
doctoral dissertation, University of Johannesburg,
Johannesburg, South Africa.
Estimating Potential Woody Biomass in Communal Savanna Woodlands from Synthetic Aperture Radar (SAR)
Paradzayi, C., Annegarn, H. J., Matsika, R., &
Erasmus, B. (2008, July 6-11). Field surveys for
biomass assessment in African savanna wood-
lands. In Proceedings of the IEEE International
Symposium on Geoscience and Remote Sensing,
Boston, MA (pp. 632-635).
Prakoso, K. U. (2006). Tropical forest mapping
using polarimetric and interferometric SAR data
- a case study of Indonesia. Unpublished doctoral
dissertation, Wageningen Agricultural University,
Wageningen, The Netherlands.
Saatchi, S., Halligan, K., Despain, D. G., &
Crabtree, R. L. (2007). Estimation of forest fuel
load from radar remote sensing. IEEE Transac-
tions on Geoscience and Remote Sensing, 45(6),
1726–1740. doi:10.1109/TGRS.2006.887002
Santoro, M. (2003). Estimation of biophysical
parameters in boreal forests from ERS and JERS
SAR interferometry. Unpublished doctoral disser-
tation, University of Jena, Jena, Germany.
Santos, J. R., Freitas, C. C., Araujo, L. S., Dutra,
L. V., Mura, J. C., & Gama, F. F. (2003). Airborne
P-band SAR applied to the aboveground bio-
mass studies in the Brazilian tropical rainforest.
Remote Sensing of Environment, 87, 482–493.
doi:10.1016/j.rse.2002.12.001
Scholes, R. J., & Archer, S. R. (1997). Tree-grass
interaction in savannas. Annual Review of Ecol-
ogy and Systematics, 28, 517–544. doi:10.1146/
annurev.ecolsys.28.1.517
Twine, W., Moshe, D., Netshiluvhi, T., & Siphugu,
V. (2003). Consumption and direct-use values of
savanna bio-resources used by rural households in
Mametja, a semi-arid area of Limpopo Province,
South Africa. South African Journal of Science,
99, 467–473.
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 1,
edited by Donald Patrick Albert, pp. 53-62, copyright 2012 by IGI Publishing (an imprint of IGI Global).
Viergever, K. M., Woodhouse, I. H., & Stuart, N.
(2007, July 23-27). Backscatter and interferometry
for estimating above-ground biomass in tropical
savanna woodland. In Proceedings of the IEEE
International Symposium on Geoscience and Re-
mote Sensing, Barcelona, Spain (pp. 2346-2349).
Wallington, E., Viergever, K. M., Stuart, N., Moss,
D., & Woodhouse, I. (2006). SAR Remote sens-
ing for natural resources management - retrieving
spatial extent and height of savanna vegetation.
Geomatics World, 31-35.
Wamukonya, L., & Jerkins, B. (1995). Durabil-
ity and relaxation of sawdust and wheat-straw
briquettes as possible fuels for Kenya. Biomass
and Energy, 8(3), 175–179. doi:10.1016/0961-
9534(95)00016-Z
Watanabe, W., Shimada, M., Rosenqvist, A.,
Tadono, T., Matsuoka, M., & Romshoo, S. A.
(2006). Forest structure dependency of the relation
between L-Band sigma nought and biophysical
parameters. IEEE Transactions on Geoscience and
Remote Sensing, 44(11), 3154–3165. doi:10.1109/
TGRS.2006.880632
Whitlow, J. R. (1979). The household use of wood-
land resources in rural areas. Harare, Zimbabwe:
National Resources Board.
Woodhouse, I., Cassells, G., Mitchard, E., &
Tembo, M. (2009, July 12-17). The use of ALOS
PALSAR for supporting sustainable forest use
in Southern Africa: A case study in Malawi. In
Proceedings of the IEEE International Symposium
on Geoscience and Remote Sensing, Cape Town,
South Africa (pp. 206-209).
889
890

Chapter 55
OOSTethys/Oceans IE
Service Registry Based on
Catalog Service for Web
Manil Maskey Ken Keiser
The University of Alabama in Huntsville, USA The University of Alabama in Huntsville, USA

Helen Conover Luis Bermudez

The University of Alabama in Huntsville, USA Southeastern Universities Research Association,
USA
Sara Graves
The University of Alabama in Huntsville, USA

ABSTRACT
Service registries can play a big role in helping developers, collaborators and agencies find deployed
resources without difficulty. A service registry is especially useful if it follows a well-known, predefined
specification that allows for automatic machine interactions and interoperability, such as the Open
Geospatial Consortium (OGC) specification for Catalog Services for the Web (CSW). This chapter
discusses a CSW-compliant registry developed as part of an OGC-sponsored interoperability experi-
ment involving the ocean sciences community. The development approach for selecting, adapting and
enhancing an open source implementation of the CSW is described. Implementation goals for the registry
included support for OGC Sensor Observation Services (SOS) and additional functionality to minimize
requirements on service providers and maximize the robustness of the registry. The registry’s role in the
OGC Ocean Science Interoperability Experiment is also discussed.

INTRODUCTION amounts of data available, scientists must address

Geo-scientists today have access to unprecedented
volumes of data collected by the many in-situ and
remote sensing environmental sensors currently
deployed. But in order to make use of the vast
DOI: 10.4018/978-1-4666-2038-4.ch055
issues related to multiple data formats, discovery
and access methods in use. Solutions across these
issues, such as data format interchange and in-
teroperable discovery and access services, would
allow scientists to focus on science tasks rather
than data technology. Data services supporting
common standards specifications, such as com-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
munity-accepted XML definitions, can provide
the required information in a form understandable
by machines, thus promoting data interoperability
and inter-use. Web service specifications have
provided new levels of interoperability, allowing
users and applications to more easily discover and
bind to available services. While interoperability is
one of the main purposes behind the standardiza-
tion of web services, in practice, it is sometimes
difficult to obtain due to loose conformance to,
or interpretation of, these specifications in actual
implementations.
Standards development is a slow and difficult
process that requires collaboration, iteration and
consensus within the target community. Often the
result can be large, complex specifications that
attempt to satisfy everyone. Although the devel-
opment of standards and specifications generally
involves rigorous revision and approval processes,
an accurate assessment of a standard’s usefulness
cannot be made until diverse teams of developers
implement the standard in real systems. Standards
often suffer from ambiguities and scalability
issues. There are, however, tangible benefits to
developing and using a standard, that are realized
when interchangeable tools are developed that
researchers can use and plug into their workflow
as needed.
Hence, it is important that experienced de-
velopers identify such ambiguities as well as
test the reliability and scalability of new or pro-
posed standards in their areas of expertise. The
Open Geospatial Consortium (OGC) conducts
interoperability experiments to address the issue
of interoperability for newly approved geospa-
tial specifications (Percivall, 2009). OGC is an
international organization consisting of various
institutes and agencies with a common goal of
developing standards for geospatial services and
products. Thus, OGC is an ideal coordinator of
the type of interoperability experiments needed.
One such initiative is the Ocean Science Interop-
erability Experiment (Oceans IE http://www.
opengeospatial.org/projects/initiatives/oceansie),
which engages the marine science community to
advance interoperability using OGC standards for
data exchange. As part of Phase I of this experi-
ment, the oceans community investigated the use
of OGC Web Feature Services (WFS) and Sensor
Observation Services (SOS), utilizing the refer-
ence implementations of the services produced by
a related community activity, called OOSTethys.
OOSTethys is a collaboration among software
developers and marine scientists who develop
open source tools in multiple languages, for use in
the evolving Integrated Ocean Observing System
(IOOS). Reference implementations and devel-
oper guides, such as the OOSTethys cookbooks
for SOS standards (http://www.oostethys.org/
best-practices), can be very helpful by providing
implementation examples, usage tips and commu-
nity specification profiles, for use by developers
and system integrators – as was demonstrated in
the Oceans IE project.
As an integral part of the Ocean IE, the authors
developed a service registry compliant with the
Catalog Service for Web (CSW) OGC standard
for registries of data and services. The Oceans
IE registry was created to index the growing
number of data services (both SOS and WFS) that
became difficult to track within the experiment.
This registry supports easy registration of SOS
and WFS, as well as discovery of the services
using the CSW specifications. While CSW was a
natural choice for an OGC service registry, it has
other advantages as well, described in the CSW
subsection later in the chapter.
This chapter describes the concepts and capa-
bilities of an OGC service registry based on the
CSW standard that delivers the value of a service
registry in a heterogeneous environment. First, we
discuss a variety of service metadata specifications
and existing service registries. The next section
describes the OGC CSW in some detail. Then we
describe Oceans IE and its CSW service registry
implementation, followed by a consideration of
future research directions.
891
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
BACKGROUND: SERVICE-
ORIENTED ARCHITECTURE
AND OGC WEB SERVICES
Service Oriented Architecture (SOA) approaches
offer increased interoperability, software reuse
and reduced maintenance, as compared to more
traditional software architectures. SOAs can
transparently tie together services that provide
different functions, are often distributed, and that
communicate via standards. An SOA includes pro-
ducers that offer specific functionality as a service
and consumers, who make use of the services in
the SOA framework. These architectures are typi-
cally built around web services, loosely coupled
and often distributed services that are critical to
creating workflows and problem solutions. Use
of Extensible Markup Language (XML) has been
key to attaining this goal. XML with a predefined
schema (specification) has become the common
way of storing and passing service information,
especially because XML is able to provide precise
machine-readable information about web services.
Being machine-readable allows better automa-
tion, which eventually helps in development of
supporting tools.
Geoscience tools can also take advantage of
the SOA approach to implement flexible and eas-
ily adaptable workflows and processes. In such a
scenario, consumers may be geospatial visualiza-
tion, environmental data analysis or other tools
that interface with the SOA framework using a
standard protocol (usually HTTP) and messages
that follow a certain specification. Embracing
SOA for geoscience applications can enrich an
organization’s service offerings and provide value
in overall efficiency, accuracy, and accessibility to
organizational marketability. A geoscience SOA
is supported by the efforts of geospatial profes-
sionals and environmental data scientists who use
various industry proven software to configure,
serve, and maintain science data and models.
A complete geoscience SOA includes servers
that provide web-accessible services along with
892
supporting tools. OGC advocates use of its suite
of web services along with supporting tools for
almost all types of geospatial applications.
OGC Sensor Data Access Services
The Oceans Interoperability Experiment demon-
strated the SOA approach using OGC data services.
A primary focus of Oceans IE was the OGC Sen-
sor Observation Service (SOS), a service protocol
that provides a standard interface for requesting,
filtering and retrieving sensor system information
and observations. SOSs allow the user or calling
program to select any number of the observation
variables available from the data source, and to
subset the data by spatial and temporal range,
thereby significantly reducing data volume and
transfer time. Sensor data is converted from
its native format to the OGC Observations &
Measurements encoding specification, and may
be embedded directly into the SOS response or
delivered as either ASCII or binary attachments,
in order to foster data/application interoperability
and reuse of multiple sensor data products. The
mandatory operations for SOS are GetCapabili-
ties, GetObservations, and DescribeSensor. Each
of these operations is briefly described below.
• GetCapabilities: This operation returns
the metadata about the SOS including de-
scriptions of offered observations and links
to description of the sensors being used.
• GetObservation: This operation re-
turns observation data from sensors for
the specified temporal and spatial range.
GetObservation uses the Observations &
Measurements (O&M) schema in its re-
sponse. O&M is an OGC standard that
provides models and XML specifications
for encoding sensor observations and mea-
surements. Using O&M, the SOS can de-
fine the data fields, data types, and struc-
tural organization in the data records. This
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
allows for machine-readable responses for
virtually any type of observation.
• DescribeSensor: This operation returns
about a description of the sensor itself.
The response is encoded in Sensor Model
Language (SensorML), an OGC standard
that provides models and XML specifica-
tions for describing sensors and processes
associated with measurements.
Used in conjunction with other OGC specifi-
cations such as CSW, the SOS provides a broad
range of interoperable capability for discovering,
binding to and interrogating individual sensors,
sensor platforms, or network of sensors in real-
time, archived or simulated environments.
Survey of Web Service
Metadata Specifications
While web services and SOA are usually thought
to be synonymous from a technical perspec-
tive, they are not. Web services are only one of
the important aspects of SOA. Another equally
pertinent component of SOA is the service reg-
istry. A service registry handles the management
of service descriptions and serves as a record
keeper of information on how to interface with
the services. Usually the service registry uses ex-
tensive metadata to make services searchable and
discoverable. Standards-based service registries
expose service metadata for discovery by users
or automated tools, fostering interoperability and
promoting service reuse. Additional information
about the functional capabilities and operational
metadata can also be made searchable. Science
data and analysis services have often been regis-
tered in ad hoc registries that are populated and
maintained by the organizations that developed
the services, and often contain only those services
developed by the organization. Obviously, custom
tools and interfaces would need to be developed
to communicate with these proprietary registries.
Thus, compatibility and interoperability are big
issues and human interactions are almost certainly
needed.
A solution is the use of standardized protocols
for registering and discovering information about
the services. The key questions when selecting
a service registry specification should be what
are the functional requirements and whether the
registry is easy to use. By understanding which
functions are needed to support an SOA approach,
an appropriate registry solution can be selected.
A simple service registry lifecycle is shown in
Figure 1. For a viable registry, all three opera-
tions – publish, find, and invoke – follow defined
specifications. Service providers publish their
services into the service registry. Service consum-
ers find services using search filter mechanisms.
Once the service is located, the service consum-
ers can invoke the service at the service provider
using the protocols and descriptions defined in
the service registry.
OGC web services serve geospatial content
and capabilities. OGC has defined standard pro-
tocols for these services, including the required
GetCapabilities operation for accessing service
metadata. Hence, these services should be easily
discovered and published using standard inter-
faces. Before we discuss CSW, we describe a few
alternate approaches that have been used to
tackle the service registry problem. We also dis-
cuss the GEOSS interoperability registry that has
the ability to interconnect the services using
various service registries.
Universal Description, Discovery,
and Integration (UDDI)
Universal Description, Discovery, and Integration
(UDDI) is a specification for online directories
to be used by both service providers and service
consumers (OASIS, 2010). The UDDI specifica-
tion complies with the Web Services Interoper-
ability Organization (WS-I) standards of technical
interoperability as a communication protocol for
exchanging data. Service providers register their
893
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
Figure 1. Simple standard based service registry lifecycle
web services in a UDDI directory and service
consumers search the service descriptions in this
directory that will lead to the services that they
are looking for. The drawback of UDDI is the use
of high-level descriptions. The descriptions that
are supported are very basic and generally are
not machine-readable. Very few organizations are
using the UDDI for service registry as a result.
Even business registries provided by IBM and
Microsoft have discontinued the use of UDDI
(Microsoft, 2006).
Web Service Description
Language (WSDL)
Web Service Description Language (WSDL) is an
XML specification that describes a web service
interface (W3C, 2001). WSDL is designed as an
open content model so that it is extensible and
flexible to be used with various web service bind-
ings. Hence, the web service’s interaction within
an SOA framework is easily machine-readable.
Without such standard description, one must rely
on ad hoc documentation to communicate with
the interfaces for SOA components. WSDLs have
become a widely accepted standard since several
industry-accepted WSDL tools have simplified
the creation and usage of this standard. Lately,
894
WSDLs have been critical to design and execu-
tion of scientific workflows.
ISO 19119
The International Organization for Standardization
(ISO) sponsors geospatial community standards
activities under Technical Committee 211 Geo-
graphic Information / Geoinformatics (ISO/TC
211). While much of the work of this committee
deals with metadata for geographic data and infor-
mation, ISO 19119 is a service standard developed
with an intent to make geospatial resources avail-
able over a distributed environment (Percivall,
2002). A computational aspect of this standard is
the ability to interlink data and services without
prior knowledge about the data or service descrip-
tions. ISO 19119 offers architectures exemplary
of the different kinds of service chaining. Another
feature is the description of a metadata model for
service objects based on service instance, metadata
and type. Models are also provided for interoper-
ability and service distribution.
Service Entry Resource Format (SERF)
National Aeronautics and Space Administration
(NASA)’s Global Change Master Directory
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
(GCMD) includes the Service Entry Resource
Format (SERF), a specification intended to define
information about geospatial data services relevant
to the datasets cataloged in the GCMD. The SERF
consists of both required and optional metadata
fields and is intended to help users locate software
tools and services for use with their data products,
and to decide whether a particular service would
be useful to them (Olsen and Stevens, 2008).
Survey of Web Service
Registry Implementations
Several organizations have implemented service
registries using one or more of these protocols.
Some of these efforts are discussed here.
Global Change Master
Directory (GCMD)
The primary purpose of National Aeronautics and
Space Administration (NASA)’s Global Change
Master Directory (GCMD) is to catalog the wealth
of environmental datasets available from NASA
and other data sources. In recent years, GCMD
has also begun to catalog services relevant to its
metadata holdings, using its SERF service meta-
data specification. Bai et al. (Bai, 2009), describe
how service interoperability requirements are not
fully met by well-known service registries like
GCMD. GCMD is intended as a human search
interface, allowing users to browse descriptions
of services, often related to specific data products.
GCMD allows consumers to search and access to
the geospatial services using keywords that are
too general to be referenced in a service instance
level, preventing machine level interoperability.
Earth Observing System
Clearinghouse (ECHO)
NASA’s Earth Observing System Clearinghouse
(ECHO) also supports the registration of Earth
science data and related services, for use as an
information resource by the science community.
The ECHO service registry is primarily intended
to provide reference to services that enhance the
usability of the Earth science data sets and the
integration of these data and services into service
oriented architectures throughout the Earth science
community. ECHO’s registry is based on web
services standards, as defined by W3C (Burnett,
2007), to include WSDL and UDDI. Registration
of services with ECHO is dependent on multiple
registration steps and multiple WSDL descrip-
tions for each service. The registration process
requires users to be formally registered as not
only ECHO users, but also as ECHO service
providers. Both steps involve interactions with
ECHO administrators. The formal ECHO registry
requires this level of accountability for registra-
tion submissions, but can be overly complex and
intimidating for service providers seeking to share
their functionality (Bai, 2007a). Though the bar-
rier to entry is high, ultimately the ECHO support
for web service standards should be instrumental
in enabling service orchestration, the chaining
of services together for programmatic solutions.
Geospatial One Stop (GOS)
The federal government mandated the manage-
ment, interoperability and public access of all
federally funded spatial datasets through a broad
e-government initiative (OMB Circular A-16).
For end-users, GOS provides a human-searchable
catalog of metadata providing information about
the type and location of datasets. The GOS reg-
istry is comprised of a network of interoperable
catalogs and services that provide an interactive
search capability across all registered resources.
All spatial data collected by federal agencies (or
their agents) are expected to be available through
this registry. GOS uses the National Spatial Data
Infrastructure (NSDI) as the underlying registry
capability, following the Federal Geographic Data
Committee (FGDC) metadata standards. Inter-
nally, GOS maintains a UDDI registry to track
895
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
a list of data producers or sources. Metadata is
subsequently harvested for specific data across the
registered searchable registry services. To register
a data source with GOS, the data producer must
provide access to a registry that is compliant with
GOS supported searches. GOS currently supports
harvesting services for Z39.50, OGC CSW, WAF
(web accessible folders), and OAI-PMH (Open
Archive Initiative Protocol for Metadata Harvest-
ing). The GOS portal provides a map viewer that
can subsequently utilize a number of data access
services to provide data displays to users, to include
OGC WMS, WFS, and ArcIMS. Data producers
may register their data sources through web-based
forms at the GOS website (http://geodata.gov).
GEOSS
Global Earth Observation System of Systems
(GEOSS; IEEE GEOSS, 2009) is an international
infrastructure to connect users, producers and
integrators of environmental information. One
of the GEOSS goals is to make environmental
information publicly available to a broad set of
users. The core components of GEOSS are the
“Components Registry” and the “Standards and In-
teroperability Registry”. The Components Regis-
try’s main purpose is to provide a centralized place
to register and access GEOSS components (e.g.,
web services, software, models). The Standards
and Interoperability Registry’s main purpose is to
provide a centralized place to register and access
standards, protocols and “special agreements”. A
special agreement could be a standard commu-
nity profile or a non standard agreement among
communities. As a “system of systems,” GEOSS
does not support a single registry protocol; rather
multiple data and service registries will exist in the
GEOSS framework, and will communicate each
other via standardized protocols and interfaces.
896
THE OGC CSW STANDARD
Unlike UDDI and other high-level service regis-
tries, domain specific catalogs generally contain
more detailed information. Where these catalogs
are based on standards, this allows for support-
ing solutions based on automatic discovery of
the services registered in the catalog. OGC has
developed a standards-based registry specification
for data and services, namely Catalog Service for
Web (CSW). A CSW-compliant catalog allows
stateless web/HTTP access with standard XML
data being exchanged. The CSW specification
allows for all properties in the metadata to be
searchable, making service and data associations
clearly traversable. CSW supports the ability to
publish and search collections of descriptive infor-
mation (metadata) about geospatial data, services
and related resources (OGC, 2009). Providers of
resources use catalogs to register metadata that
conform to the information model that CSW sup-
ports. The information model includes detailed
descriptions of data and services that are provided.
Client applications can then efficiently search for
geospatial data and services (OGC, 2009). CSW
was recommended for implementation specifica-
tion status in November 2005.
OGC specifies the interfaces, bindings, and
a framework for defining application profiles
required to publish and access a CSW that pro-
vides metadata for geospatial data, services,
and related resource information. Metadata are
represented as generalized properties that can be
queried and returned through catalog services for
resource evaluation and in many cases, invoca-
tion or retrieval of the referenced resource. The
CSW standard and relevant application profiles
are described in more detail in the next section.
CSW Information Structure
and Semantics
OGC specifications allow for flexibility in the
model for information exchange. In case of CSW,
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
application profiles are allowed to be specific about
what information content is to be communicated
through the service. This information follows a
particular syntax from the application profile.
Thus an application profile should address the
following items:
1. Identify information resource types that can
be requested. In the case of a catalog service,
the information resources being described by
the metadata may include geographic data,
imagery, services, controlled vocabularies,
or schemas among a wide variety of possible
types. This sub clause allows the community
to specify or generalize the resource types
being described in the metadata for their
scope of application.
2. Identify a public reference for the informa-
tion being returned by the service (e.g.,
ISO 19115:2003 “Geographic Information
- Metadata”). Include any semantic resources
such as data content model, dictionary,
feature type catalog, code lists, authorities,
taxonomies, etc.
3. Identify named groups of properties (element
sets) that may be requested of the service
(e.g. “brief,” “summary,” or “full”) and the
valid format (syntax) for each element set.
Identify valid schema(s) with respect to a
given format to assist in the validation of
response messages.
A service metadata schema must be defined
for the purpose of interchange of information
within an information community. This schema
should provide a common vocabulary that supports
search, retrieval, display, and association between
the description and the object being described.
Currently, the XML exchanges for CSW are based
on two draft application profiles: ebRIM Registry
Information Model Profile and the 19115/19119
Profile developed by ISO.
ISO 19119
The Catalog Application Profile (Senkler, 2004)
definition, based on the CSW specification, takes
advantage of both 19115 and 19119 for geospatial
data and services, respectively. As ISO 19119
includes extensions that allow for the varying
data/service binding described above, the pro-
file uses the 19115 description with additional
objects designed to include the necessary exten-
sions. For example, 19119 includes a definition
of CSW_CoupledResource, a class that regulates
the interoperation of services and datasets. The
profile excerpt shown in Figure 2 includes the
object SV_ServiceIdentification that would suf-
fice for the metadata functionality of 19115 but
does not satisfy the service requirements of 19119.
The profile described uses an HTTP binding,
but is also capable of using SOAP with XML
binding. ISO 19119:2005 identifies and defines
the architecture patterns for service interfaces
used for geographic information, defines its rela-
tionship to the Open Systems Environment
model, and presents a geographic services tax-
onomy and a list of example geographic services
placed in the services taxonomy. It also prescribes
how to create a platform-neutral service specifica-
tion, how to derive conformant platform-specific
service specifications, and provides guidelines
for the selection and specification of geographic
services from both platform-neutral and platform-
specific perspectives.
ebRIM
The ebXML Registry Information Model (ebRIM)
(Nogueras-Iso, 2005; Chen 2009) has also been
used as a profile for OGC CSW. The Model
provides a format for ordering both the content
and their connections (Martell, 2009) within the
catalog, as shown below in Figure 3.
Extension packages can be used as needed to
support application-specific functionality. Also,
specific components, such as the RegistryObject,
897
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
Figure 2. ISO 19119 profile excerpt
can be substituted where possible by similar ele-
ments. The slots can also be used if no conflicts
are present with either extension packages or
higher-level objects.
George Mason University’s Center for Spatial
Information Science and Systems (CSISS) (Bai,
2007b) uses an ebRIM-style catalog service
designed to provide access to their Landsat im-
ages through a combination of geospatial data
and geospatial services. The data and services
are associated together, letting users tailor the
data via the service interface. This specification
is based on an OGC-compliant model, and thus
is restricted to the same functions as the CSW,
described below.
898
CSW Operations
The operations supported by CSW can be divided
into three classes: service operations which are
operations a client may use to question the service
to determine its capabilities; discovery operations
which a client may use to determine the informa-
tion model of the catalog and to query catalog
records; and management operations which are
used to create or change records in the catalog.
Service Class Operation
GetCapabilities: Generate Capabilities document
Discovery Class Operations
DescribeRecord: Generate Record Schema
GetDomain: Determine value domain and gener-
ate response document
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
Figure 3. ebXML registry information model
GetRecordById: Query repository for record with
specified id and generate response
GetRecords: Execute query and generate re-
sponse document containing result records
Management Class Operations
Transaction: Execute transaction and generate
response document
Harvest: Generate response document that ac-
knowledges receipt of request and execute
request asynchronously.
In general, metadata consumers will invoke
GetCapabilities, DescribeRecords, GetDomain,
GetRecords, and GetRecordById operations
whereas metadata producers will invoke Transac-
tion and Harvest operations.
GetCapabilities Operation
Like all other OGC services, the CSW specifi-
cation mandates the GetCapabilties operation.
This operation allows clients to retrieve service
metadata from a server. CSW servers allow the
HTTP request for GetCapabilities to contain
parameter and value pairs for Request, Service,
Sections, AcceptVersions and AcceptFormats.
The value of the Service parameter should be
“CSW”. The response to a GetCapabilities request
is an XML document containing service metadata
about the server. The document includes sec-
tions on ServiceIdentification, ServiceProvider,
OperationsMetadata, and Filter_Capabilities. Of
these sections, OperationsMetadata is the most
important, as it lists all the operations supported
by the CSW service.
899
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
DescribeRecord Operation
DescribeRecord is a mandatory operation that
allows a client to discover elements of the infor-
mation model supported by the catalog service.
This method requires mandatory keyword-value
pairs for parameters Request, Service, and Ver-
sion. Other supported parameters are Namespace,
TypeName, outputFormat, and schemaLanguage.
The response for the DescribeRecord request con-
tains SchemaComponent elements that describe
each requested schema language.
GetDomain Operation
The GetDomain operation is optional. This op-
eration returns a range of values for a metadata
record element or request parameter.
GetRecords Operation
GetRecords is a mandatory operation that is used
to search for records in the catalog. It uses the
Query element to filter and identify the record.
GetRecords also uses the outputSchema parameter
to indicate which schema shall be used to generate
the response to the GetRecords operation.
GetRecordById Operation
GetRecordById is a mandatory operation that
retrieves a specified record from the catalog based
on the provided identifier. This operation is gener-
ally used after executing another operation to get
back the identifier for the record.
Transaction Operation
Transaction is an optional operation that provides
mechanisms to create, modify, and delete catalog
records. In particular, the Insert action allows one
or more records to be inserted into the catalog. The
Update action uses constraint elements to find the
records to be updated. Similarly, the Delete action
900
allows records to be deleted from the catalog in
conjunction with the constraint element.
Harvest Operation
Harvest is an optional operation that references
data to be inserted or updated in the catalog. The
catalog itself will fetch the data and process into
the catalog.
CSW Use Scenarios
In the context of SOA, Figure 4 illustrates an
example sequence of operations performed by an
SOS consumer. In this figure, the SOS consumer
discovers two instances of SOSs using the GetRe-
cords operation provided by CSW service catalog.
Then, the consumer requests SOS capabilities
documents for each service, using URLs from
the CSW GetRecords response. The consumer
can then make a choice of service by inspecting
the observation offerings from both SOSs, and
invoke the appropriate GetObservation from the
selected SOS is invoked.
Again in the context of SOA, Figure 5 illustrates
an example sequence of operations performed by
two SOS providers (producers). The first pro-
ducer discovers the second SOS from the CSW
catalog. Whenever new sensors are discovered,
the producers can register these sensors with their
SOS. They will subsequently publish sensor ob-
servations that they are able to provide.
THE OGC OCEANS
INTEROPERABILITY EXPERIMENT
Overview of Oceans IE
OOSTethys, the loose collaboration of software
developers and marine scientists who worked with
OGC to define and implement Oceans IE, was
itself the product of several earlier collaborations:
OpenIOOS, the SURA Coastal Ocean Observ-
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web

Figure 4. SOS consumer sequence diagram

Figure 5. SOS sensor data publisher sequence diagram

901
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
ing and Prediction (SCOOP) program, and the
NSF-sponsored Marine Metadata Interoperability
project. OpenIOOS has been demonstrating the
concept that “standards enable innovation” for
several years, thanks to data and service contri-
butions from Ocean Observing Systems (OOSs)
around the country. OpenIOOS.org, which is a
proof of concept portal interface for the distributed
participants, demonstrates a Service-Oriented Ar-
chitecture (SOA) with heavy emphasis on Open
Geospatial Consortium (OGC) web-service speci-
fications. With support from the NOAA Coastal
Services Center and the Office of Naval Research,
the Southeastern Universities Research Associa-
tion (SURA) integrated OpenIOOS.org into the
SCOOP program (http://scoop.sura.org). Through
SCOOP and other programs, SURA has supported
the vision of enabling a distributed laboratory for
research and applications with broad participation
from research institutions working in partnership
with federal agencies and the private sector. The
Marine Metadata Interoperability project, started
in September 2004, now comprises more than 300
members, with a web site consisting of thousands
of pages including guides and references about
marine metadata issues (http://marinemetadata.
org). One of the activities of MMI is an interop-
erability demonstration using an SOA, common
use content standards, and semantic mediation
via ontologies.
In 2005 MMI, OpenIOOS and SCOOP par-
ticipants started working together on a single,
combined interoperability test bed activity of a
distributed, service-oriented architecture to share
real time data. The combined team, which adopted
the name OOSTethys, set up an initial set of meta-
data requirements such as geo-spatial location and
platform type, and agreed on an interface to serve
as a wrapper for each data system. OOSTethys
members decided that working with standards
organizations to select, exercise and advance ap-
propriate standards was the logical path to move
forward toward an integrated ocean observing
system. With the OGC, OOSTethys members
902
started the OGC Ocean Science Interoperability
Experiment in 2007.
Through Oceans IE, a portion of the Ocean
Observing community has reached consensus on
its understanding of various OGC specifications,
solidified demonstrations for Ocean Science appli-
cation areas, hardened software implementations,
and produced a candidate “OGC Best Practices”
document that can be used to inform the broader
ocean science community. To achieve these goals,
the Oceans IE engaged the OGC membership to
assure that any community requirements com-
ing from the Oceans group properly leveraged
the OGC specifications. Change requests on the
OGC specifications were provided to the OGC
Technical Committee to influence the underlying
specifications.
The Oceans IE Phase I, which ended in May
2008, investigated and compared the use of
OGC Web Feature Services (WFS) and Sensor
Observation Services (SOS) for representing
and exchanging point data records from fixed
in-situ marine platforms. This phase of Oceans
IE produced an engineering best practices report
and reference implementations for SOS. Phase I
included a registry to record all the SOS that were
developed. This ad hoc registry was a quick and
temporary solution to the problem of cataloging
and discovering the SOSs. Tools and interfaces
that relied on the registry for SOS information
had to use custom Representational State Trans-
fer (REST) style services to communicate with
the registry. We encountered a growing need
for a standards-based service registry soon after
completion of this initial service registry, as more
participants began utilizing the registry from dif-
ferent applications. Research on available registry
standards thus began during Phase I.
The Oceans IE Phase II ran from March 2009
to December 2009. The following topics were
advanced: Automated instrument metadata/soft-
ware installation via the PUCK protocol (MBARI,
2009), observation offerings of complex systems
(e.g., observations systems containing other sys-
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
tems) such as collection of stations, linking data
from SOS to out-of-band offerings, Semantic
Registry and Services, CSW Registry, and IEEE-
1451/OGC-SWE harmonization. During Phase
II, CSW was investigated in depth along with
other catalog standards. After CSW was selected,
various open source reference implementations
of the specification were also investigated. The
deegree project was chosen as the base CSW
implementation, and the authors developed vari-
ous components to support additional Oceans IE
needs and make the implementation more robust.
CSW Reference Implementations
Multiple organizations have implemented regis-
tries based on CSW. The Oceans IE community
evaluated three such implementations, described
below.
GeoNetwork
GeoNetwork is an open-source software project
developed to provide the geospatial community
a standards-based suite of applications and ser-
vices (GeoNetwork, 2009). Geospatial standards
developed by OGC and ISO are well supported
by GeoNetwork. One of the functions offered
is the cataloging of data and services. Although
powerful in many aspects, the current devel-
opment version of GeoNetwork at the time of
evaluation did not support the latest version of
the CSW specification. A number of ambiguities
in the 2.0.1 specification have been clarified in
2.0.2 and should be reflected in the GeoNetwork
implementation of CSW. In addition, at the time
of our evaluation, the Transaction and Harvest
operations were not supported by GeoNetwork
implementation of CSW. These operations were
critical to our experiment.
GI-cat
GI-cat is open source software developed by
Earth and Space Science Informatics – Labora-
tory (ESSI-Lab) that provides implementations
of the OGC CSW specification, along with a
suite of other OGC standards (Nativi, 2007). GI-
cat uses the ISO application profile information
models for core operations and exposes standard
query interfaces implemented by the profiler
components. The software in developed using the
Java Standard Edition that is capable of running
under Apache Tomcat server. Although GI-cat is a
practical implementation of CSW, the installation
procedures and other compatibility issues were
deemed to be impractical for use in Oceans IE.
Deegree
The deegree project is a Java implementation of
an extensive suite of OGC service specifications.
It also encompasses test clients for the suite of
services. Deegree software is freely available
as open-source software protected by the GNU
Lesser General Public License (GNU LGPL) and
is accessible at http://www.deegree.org/.
The deegree CSW implementation is designed
in such a way that it is able to serve different
metadata formats in parallel even with a single
underlying relational database. This is possible be-
cause the deegree CSW implementation uses Ex-
tensible Stylesheet Language (XSL) (W3Schools,
2010) processing to transform requests as well as
responses into the desired format. The deegree
CSW does not contain a data access module of its
own, rather it uses an OGC Web Feature Service
(WFS), also implemented by deegree as a data
source. It is possible to adjust the deegree CSW to
almost any existing metadata profile without the
need for replication. Because of its flexibility with
respect to configuration and output formats, the
deegree implementation was ideal for our purpose.
It merely required modifying XML configuration
files before installation.
903
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web

Table 1. Summary of features of various CSW

successes were leveraged by developing an open
implementations
source service registry and catalog based on the
Features Geonetwork GI- deegree OGC’s CSW which harvest metadata from the SOS
cat GetCapabilities and DescribeSensor responses.
Open source X X X A real-time use of the CSW-compliant service
Latest version of CSW X X registry is shown in Figure 6. Various SOSs de-
Supports all CSW X X ployed by the IOOS community can be discovered
operations from the OOSTethys/Oceans IE registry. The
Supports many OGC X observations from the registry are displayed in
and ISO standards
real-time at the website openioos.org. All of the
Can support multiple X
application profiles source code developed for the experiment is avail-
Can serve multiple X able for the community to reuse at http://code.
metadata formats in google.com/p/oostethys/. This architecture allows
parallel
for reference implementations to be adapted and
Easy installation and X
configuration
improved easily while keeping the overall usage
No compatibility issues X
as simple as possible.
The OOSTethys/Oceans IE SOS Service Reg-
istry was based on the deegree implementation
of the CSW specification. This implementation
The pros and cons of each of these CSW
allows communication using the ISO 19119 ap-
implementation are summarized in Table 1.
plication profile. Other supporting components
include the TEAM Engine Validator, a web-based
user interface, a GetCapabilities metadata har-
THE OCEANS IE CSW-COMPLIANT
vester, and SOS metadata to ISO metadata con-
SERVICE REGISTRY
verter. Figure 7 illustrates the architecture of the
entire system. Components of the system are
A primary goal of the Oceans Interoperability
described in the following subsections. This ar-
Experiment and the complementary OOSTethys
chitecture is a specific case of SOA that we dis-
project, is to dramatically reduce the time it takes
cussed earlier in the chapter. Service providers
to install, adopt and update standards-compliant
are responsible for storing and serving data col-
web services, focusing on the OGC Sensor
lected from instruments and resulting derived
Observation Service (SOS). Open source Java,
products. In addition, they make their services
PERL, Python, PHP and ASP toolkits serving
available by simplified registration process ex-
ocean observations from NetCDF-formatted files,
posed by CSW interface. End users and tools map
THREDDS, various relational databases and even
to the service consumers in the SOA. The tools
CSV (comma-separated value) text files have been
programmatically discover the services using the
developed. This matches the diverse data needs
CSW interface. Once the services are discovered,
and software expertise of the Integrated Ocean
the tools will invoke the services using the pro-
Observing System community. In many cases
tocol described in the service metadata. OpenIOOS
data providers with limited software experience
is an example of such tools used in OOSTethys/
are able to install an SOS in mere minutes and
Oceans IE.
configure it by editing a simple configuration file.
This easy deployment has led to broad adoption
of SOS by the marine science community. These

904
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web

Figure 6. OpenIOOS: real-time sos observations visualization

Figure 7. OOSTethys/Oceans IE CSW registry system for SOS

905
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
Deegree CSW Registry Package
Deegree’s CSW installation package is a web
archive (WAR) file, which can be dropped easily
into a web server. In our case, the Apache Tom-
cat web server suffices as a container to use for
deployment of the deegree servlet module. It is
also very convenient to use the bundled structured
query language (SQL) scripts to create tables for
the catalog. This implementation supports all of
the operations defined for CSW and is compliant
with the CSW 2.0.2 specifications. During Oceans
IE, degree provided the solutions to any problems
encountered via well supported online forums and
user correspondence.
OGC and ISO geoinformatics standards
played a vital role in development of deegree.
In particular, deegree’s CSW implementation
supports the ISO 19119 standard, and is built
on the deegree implementation of the OGC Web
Feature Service (WFS) – see Figure 8. The WFS
programming interface accepts WFS requests and
replies with WFS responses for a particular data
source. Therefore, the discovery and manager
layers hide the WFS layer from the exposed API.
Figure 8. Dependency between deegree CSW and WFS
906
These components transform incoming catalog
requests into WFS requests and transform WFS
responses into CSW responses.
TEAM Engine Validator
Compliance & Interoperability Testing & Evalu-
ation (CITE) is an ongoing OGC initiative that
develops tests for OGC standards, and makes
those tests available for online access. An integral
part of CITE is a test script interpreter, the Test,
Evaluation, And Measurement (TEAM) Engine.
It executes test scripts written in Compliance
Test Language (CTL) to verify that a designated
implementation complies with the specification.
Being an OGC member, we were able to obtain
the source code for the TEAM Engine for use
in the OOSTethys/Oceans IE Service Registry
interface. The TEAM Engine test for SOS vali-
dation was adapted to integrate into the service
registration process. The incorporation of most
of the TEAM Engine open source code helped
to create a complete and robust CSW registry.
OOSTethys is able to distribute the TEAM Engine
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
as a component of the Service Registry under the
Mozilla 1.1 license.
CSW Registry Web Interface
For Oceans IE, the authors developed a web-based
user interface to the CSW registry for both pro-
ducers and consumers of SOS. This user interface
communicates with the CSW registry by generat-
ing the detailed XML request and response docu-
ments required by CSW operations, thus shielding
users from these complex XML documents. The
web interface for the OOSTethys/Oceans IE SOS
Service Registry is available at: http://scoredev.
itsc.uah.edu/csw/. Two important operations that
are offered through the web interface are service
registration and service discovery.
Service Registration
All of the web service standards used for Oceans IE
are developed and maintained by OGC. The CSW
service registry leverages the fact that all OGC web
services have commonalities in their design and
interface, in particular the GetCapabilities opera-
tion. The response from a GetCapabilities request
is an XML description of the service’s information
content and supported request parameters, and
is therefore both machine- and human-readable.
Since the XML-formatted response conforms to
a predefined schema, clients can easily validate
the response. GetCapabilities responses contain
information on following sections:
• ServiceIdentification: Metadata about
this specific server. The schema of this sec-
tion is the same for all OGC web services.
• ServiceProvider: Metadata about the or-
ganization operating the server. The sche-
ma of this section is also the same for all
OGC web services.
• OperationsMetadata: Metadata about the
operations specified by this service and
implemented by this server, including the
URLs for operation requests. The basic
contents and organization of this section is
the same for all OGC web services, but in-
dividual services can add elements and/or
change the use of optional elements.
Service providers merely need to provide
a valid GetCapabilities URL to register their
service in the registry. Automatic harvesting of
service metadata is achieved by parsing the Get-
Capabilities response from an SOS. The registry
first validates the SOS using the TEAM Engine
Validator. The provider is notified if the SOS is not
in compliance with the OGC SOS specification.
Once the SOS passes the TEAM Engine validation,
the XSL transformation layer translates the SOS
GetCapabilities service metadata to the ISO 19139
service metadata schema required by the deegree
CSW registry package. For the service providers,
all of these complex activities, such as the con-
struction of valid XML insert request to register
their services, are completely hidden. Service
providers can enter an SOS GetCapabilities URL
and press the “Register” button for registration.
Any error encountered during registration will be
displayed to the registrant.
Service Discovery
The users of the registry can also bypass the cre-
ation of XML for discovery or transaction opera-
tions as the easy-to-use web interface builds the
necessary XML from the specification of a few
necessary query components by the user. Pressing
“Get Record By Id” button in the web interface
returns a CSW response for a GetRecordById
request based on the provided identifier, currently
“Service Title”, and element set name. Pressing the
“Get All Records” button returns ISO records for
all the registered services. The underlying CSW
web service can be accessed programmatically
with a valid CSW request.
907
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
SOS Service Metadata to
ISO 19119 Translator
Since the deegree registry interface accepts CSW
requests that conform to the ISO 19119 profile,
the SOS metadata harvested from the GetCa-
pabilities response needs to be converted to an
appropriate ISO record. The authors developed
such a translator using an XSL transformation.
This translator maps almost all available elements
from the GetCapabilities response schema into
the appropriate ISO 19119 schema. The transla-
tor is executed whenever providers register their
service using the CSW interface. Figure 9 shows
a portion of the XSL mapping from SOS service
metadata to an ISO 19119 profile. In this subset
of the XSL mapping, the ObservationOffering’s id
and description are translated into DataIdentifica-
tion’s title and identifier respectively. The CSW
registry thus allows the services to be searched
using these criteria.
FUTURE RESEARCH DIRECTIONS
Since the evolution of web 3.0 technologies, the
future of service registries also seems to moving
towards the semantic-based service registry. Ocean
community efforts related to semantic registries
Figure 9. Portion of XSLT mapping from SOS GetCapabilities to ISO 19119
908
are available at http://oceanobservatories.org/
spaces/display/CIDev/Metadata+Registry and
http://oceanobservatories.org/spaces/display/CI-
Dev/Harvester. The main difficulty with semantic
registry is providing filtering by topics of large
collections of metadata. If there is a thesaurus
(terms with broader and narrower relations), it will
be easier to navigate and discover the metadata
records of interest. Furthermore, initial imple-
mentations of semantic-based service registries
have suffered from performance issues. Much
of the research is now focused on improving the
performance of such registries. Currently, another
research focus area is into ways to describe rela-
tionships between different web services so that
such semantic relationships can be exploited in
service discovery within the service registry.
CONCLUSION
As part of OGC Oceans Interoperability Experi-
ment (Oceans IE), the authors developed a web
service registry compliant with the OGC Catalog
Service for Web (CSW) standard. The registry
evolved with the need for a standards-based catalog
to support a growing number of Sensor Observa-
tion Services (SOS). The resulting registry allows
easy registration of SOSs via harvesting the service
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
metadata from GetCapabilities responses, as well
as easy discovery of these services using the CSW
specifications. A standards-compliant registry for
this endeavor was required to eliminate the need for
discovering the services from ad-hoc registries. In
addition, having a standard CSW registry opened
up avenues for the development of automated
tools. This open source registry uses Extensible
Stylesheet Language (XSL) transformation to
convert SOS metadata into the ISO metadata
that is required by the CSW. Two plug-ins, the
TEAM Engine Validator and CSW registry web
interface, were also developed or enhanced and
integrated to make the CSW registry more robust
and easy to use. The authors believe that the
availability of the OOSTethys/Oceans IE SOS
CSW Registry will promote acceptance of SOS
by the oceanographic community and facilitate
service discovery and reuse. Furthermore, the
OOSTethys/Oceans IE Registry is on schedule
to be registered with GEOSS. This registration
will allow creation of community registry that
includes services from other organizations that
are registered with GEOSS, and interconnected
with OOSTethys/Oceans IE registry.
REFERENCES
Bai, Y., Di, L., Chen, A., & Wei, Y. (2007b). To-
wards a geospatial catalogue federation service.
Photogrammetric Engineering and Remote Sens-
ing, 73, 699–708.
Bai, Y., Di, L., & Keiser, K. (2007a). Review of
NASA ECHO (version 8) Web service registration
process, Technical Whitepaper submitted to the
Earth Science Data Systems’Technology Infusion
Working Group.
Bai, Y., Di, L., & Wei, Y. (2009). A taxonomy of
geospatial services for global service discovery
and interoperability. Computers & Geosciences,
35, 783–790. doi:10.1016/j.cageo.2007.12.018
Burnett, M., Weinstein, B., & Mitchell, A. (2007).
ECHO – enabling interoperability with NASA
earth science data and services. Proceedings from
Geoscience and Remote Sensing Symposium and
IGARSS 2007.
Chen, N., Di, L., Yu, G., & Wei, Y. (2009). Use
of ebRIM-based CSW with sensor observation
services for registry and discovery of remote-
sensing observations. Computers & Geosciences,
35, 360–372. doi:10.1016/j.cageo.2008.08.003
GeoNetwork. (2009). GeoNetwork OpenSource
community website. Retrieved from http://
geonetwork-opensource.org/
IEEE GEOSS. (2009). IEEE Standards Associa-
tion – GEOSS standards registry. Retrieved from
http://seabass.ieee.org/groups/geoss/
Martell, R. (2009). CSW-ebRIM Registry
Service - part 1: ebRIM profile of CSW.
Retrieved from http://portal.opengeospa-
tial.org/modules/admin/license_agreement.
php?suppressHeaders=0&access_license_
id=3&target=http://portal.opengeospatial.org/
files/index.php?artifact_id=31137
MBARI. (2009). Marine plug-and-work consor-
tium. Retrieved from http://www.mbari.org/pw/
puck.htm
Microsoft. (2006). UBR Shutdown FAQ. Retrieved
from http://uddi.microsoft.com/about/FAQshut-
down.htm
Nativi, S., Bigagli, L., Mazzetti, P., Mattia, U.,
& Boldrini, E. (2007). Discovery, query and ac-
cess services for imagery, gridded and coverage
data a clearinghouse solution. Proceedings from
Geoscience and Remote Sensing Symposium and
IGARSS, 2007, 4021–4024.
909
 OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web
Nogueras-Iso, J., Zarazaga-Soria, F., Bejar, R.,
Alvarez, P., & Muro-Medrano, P. (2005). OGC
Catalog Services: a key element for the develop-
ment of Spatial Data Infrastructures. Comput-
ers & Geosciences, 31, 199–209. doi:10.1016/j.
cageo.2004.05.015
OASIS. (2010). Online community for the uni-
versal description, discovery, and integration.
Retrieved from http://uddi.xml.org/.
OGC. (2009) Catalogue service. Retrieved from
http://www.opengeospatial.org/standards/cat
Olsen, L., & Stevens, T. (2008). Service Entry
Resource Format (SERF) standard. (ESDS-RFC-
013v0.1). Retrieved from http://www.esdswg.
org/spg/rfc/esds-rfc-013/ESDS-RFC-013v0.1.pdf
Percivall, G. (2002). ISO 19119 and OGC Service
Architecture. Presented at FIG XXII International
Congress, Washington, D.C.
Percivall, G. (2009). OGC Interoperability Pro-
gram. Retrieved from http://www.opengeospatial.
org/ogc/programs/ip
Senkler, K., Voges, U., & Remke, A. (2004). An
ISO 19115/19119 profile for OGC catalogue ser-
vices CSW 2.0. Presented at the 10th EC GI & GIS
Workshop, ESDI State of the Art, Warsaw, Poland.
W3C. (2001). Web Services Description Lan-
guage (WSDL 1.1). Retrieved from http://www.
w3schools.com/xsl/
W3Schools. (2010). XSLT Tutorial. Retrieved
from http://www.w3schools.com/xsl/
ADDITIONAL READING
Chidambaram, C., Maskey, M., & Ramachandran,
R. (2009). Service Oriented Architecture for Data
Mining at Data Centers and Computing Centers.
Poster presented at the Cloud Computing and Col-
laborative Technologies in Geosciences.
910
Conover, H., Berthiau, G., Botts, M., Goodman,
M., Li, X., & Lu, Y. (2010). Using Sensor Web
Protocols for Environmental Data Acquisition and
Management. Ecological Informatics, 32–41.
doi:10.1016/j.ecoinf.2009.08.009
Conover, H., Drewry, M., Graves, S., Keiser, K.,
Maskey, M., & Smith, M. (in press). SCOOP Data
Management: A Standards-based Distributed In-
formation System for Coastal Data Management.
In Di, L., & Ramapriyan, H. K. (Eds.), Standards-
Based Data and Information Systems for Earth
Observation. Springer-Verlag.
Graves, S., Ramachandran, R., Movva, S.,
Conover, H., Fox, P., & McGuinness, D. (2008).
A distributed knowledge extraction framework
based on semantic Web services. AISR NASA
Investigator Workshop, NASA, College Park,
MD, 2008.
ISO 19119. (2008). Geographic information ser-
vices. Fuger S., Najmi F., & Stojanovic, N. (2005).
ebXML Registry Information Model (Version
3.0). OASIS Standard #OGC regrep-rim-3.0-os,
May 2005.
Movva, S., Ramachandran, R., Maskey, M., &
Li, X. (2009). Building a semantic framework
for e-science. Paper presented at the American
Geophysical Union Fall Meeting.
Shanming, W., & Jianjing, S. (2008). Ontology-
Based Framework for Geospatial Web Services.
Information Science and Engineering., 2, 107–110.
KEY TERMS AND DEFINITIONS
Application Profile: A metadata schema con-
sisting of elements from one or more namespaces
merged and optimized for use in a specific ap-
plication area.
Catalog: A centralized location where informa-
tion regarding the similar resources is stored, and
where they can be searched for easily.
OOSTethys/Oceans IE Service Registry Based on Catalog Service for Web

Catalog Service for Web (CSW): Specifica-
tion that defines the schemas and interfaces to
search, discover, register, and request geospatial
data and services.
Interoperability: Capability of multiple soft-
ware engineering components to communicate
and exchange information.
Reference Implementation: A model or skel-
eton from which other similar implementations
can be derived and customized easily.
Service Oriented Architecture (SOA): An
approach to distributed computing using loosely
coupled web services where communication and
message exchanges are standards based.
Standard: A formal document that specifies a
set of rules to be followed throughout a process.

This work was previously published in Geospatial Web Services: Advances in Information Interoperability, edited by Peisheng
Zhao and Liping Di, pp. 97-117, copyright 2011 by Information Science Reference (an imprint of IGI Global).

911
912

Chapter 56
A Multi-Criteria GIS Site
Selection for Sustainable Cocoa
Development in West Africa:
A Case Study of Nigeria

Tunrayo Alabi
International Institute of Tropical Agriculture, Nigeria

Kai Sonder
International Maize and Wheat Improvement Center, Mexico

Olusoji Oduwole
Cocoa Research Institute of Nigeria, Nigeria

Christopher Okafor
International Institute of Tropical Agriculture, Nigeria

ABSTRACT
Cocoa occupies 6 million hectares in humid coastal West Africa where 70% of the world supply is grown,
90% of which is produced on 2 million family farms of 2 hectares or less. Here, at least 16 million people
depend on cocoa but earn only $100/person/year from the crop. There is need to optimize the farming
system, minimize the environmental impact of technologies, and improve socio-economic dynamics.
This study identifies areas with potential for intensified cocoa farming and where maximum impact to
household income could be achieved without deforestation. The selection involves defining suitability
criteria, preparing an inventory of available data, determining suitability based on identified criteria,
and combining suitability into hierarchical preferences based on weights proposed by local experts.
GIS and Multi-Criteria land Evaluation technique using biophysical, socioeconomic, and demographic
variables were employed in selection. Nineteen administrative units were selected in Nigeria where the
intervention project could be implemented.

DOI: 10.4018/978-1-4666-2038-4.ch056

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
INTRODUCTION
Cocoa (Theobroma cacao) is one of the most
important perennial crops worldwide, with an
estimated world production of 2.8 million tonnes
(t) in 2002 (Food and Agriculture Organization
[FAO], 2003). It is especially a most significant
crop in West Africa where it occupies between 5
and 6 million ha in the coastal humid zone. Nearly
70% of the world cocoa supply is produced in
West Africa (International Cocoa Organization
[ICCO], 2010), 90% of which is grown on nearly
2 million small family farms, the majority with
land holdings of 2 ha or less. It is estimated that
at least 16 million people depend on cocoa for
the majority of their cash income (International
Institute of Tropical Agriculture [IITA], 2005a).
The average cocoa farming family in West
Africa earns US$100 per person per year from
cocoa (IITA, 2002; ICCO, 2007). The long-term
social and economic well-being of these families
and their communities depends upon the viability
and sustainability of the West African cocoa sec-
tor. The sector is also vital to national economies
with cocoa among the top three agricultural export
products in each target country (Bastide & Perret,
2007). To ensure this, cocoa farming needs to
become a profitable, income-generating profes-
sion that allows farmers, both women and men,
to provide adequately for their families.
West African cocoa farmers face numerous
challenges such as production and marketing
inefficiencies as well as stagnant technological
development. For instance, the majority of farm-
ers sell their cocoa individually, a few bags at a
time, to itinerant buyers without access to market
information and this reduces their earnings relative
to the market price.
In Nigeria, like the rest of West Africa, cocoa
cultivation has witnessed significant changes in
the last few decades due to factors such as varia-
tions in international demand, prices, and policy
measures implemented by both the Federal and
State governments (Cocoa Research Institute of
Nigeria [CRIN], 2008). Cocoa was the most im-
portant agricultural export crop during the 1960-
1970 period, earning a significant percentage of
the foreign exchange income (Mustapha, 1999).
The production increased gradually to a maximum
of 308,000 t in 1970-1971 but dropped drastically
to 110,000 t in 1990-1991. This was because of
the diversion of government policy during the
First, Second and Third Development plan periods
(1970-1985) into petroleum and food production,
and the poor price policy of the Marketing Board
which resulted in the abandonment of cocoa farms
(Mustapha, 1999).
In view of the undulating nature of the cocoa
sector as a result of factors such as socio-econom-
ics, government policies, management practices,
and declining natural resources (Bastide & Perret,
2007), it is necessary to seek a viable and sustain-
able cocoa production system in West Africa.
It is therefore important to carefully target
the program’s intervention in the cocoa sector of
the West and Central African countries to ensure
maximum program impact and conformity with
national agricultural sector priorities. This will
include the identification and selection of com-
munities through systematic spatial targeting that
meets multiple criteria. World Cocoa Foundation
(WCF) has asked IITA Sustainable Tree Crop Pro-
gram (STCP) to lead the process of site selection,
building on its recent experience in conducting a
similar multi-criteria and multi-stage systematic
selection of target districts/communities for a
multi-sectoral project in Cote d’Ivoire and Ghana
in collaboration with local expertise.
In view of the complexity involved in the
process of decision making, the method of
multi-criteria land evaluation (MCE) is adopted
in this study (Khoi & Murayama, 2010). MCE is
the process of determining the fitness of a given
parcel of land for a defined use (Stainer, 1991).
An MCE involves the selection of the biophysical
or socio-economic factors, or both, of an area;
the combination of the selected factors with the
decision-makers preferences allows one to create a
913
 A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
composite suitability index (Sui, 1993). The MCE
is an effective tool for multiple criteria decision-
making issues (Malcewski, 2006). The purpose
of the MCE is to investigate a number of choice
possibilities in the light of multiple criteria and
multiple objectives (Cover, 1991). Integration of
the MCE and GIS (GIS-MCE) can help land-use
planners and managers to improve decision-
making processes (Malcewski, 1999). GIS allows
the computation of assessment factors and MCE
aggregates them into a land suitability index.
This paper aims to identify potential land suit-
ability for cocoa, based not only on biophysical
but also socioeconomic and infrastructural vari-
ables in accordance with the framework for land
evaluation developed by the Food Agriculture
Organization of the United Nations (FAO, 1976)
using GIS, a decision support system involving
the integration of spatially referenced data in a
problem solving environment (Cowen,1988), and
to simultaneously examine major socio-economic
Figure 1. The study area showing ecological zones and protected area of the Cocoa Growing Region
of Nigeria
914
factors influencing cocoa farming with respect to
enabling optimum cropping for maximizing pro-
ductivity while reducing environmental impacts.
METHODOLOGY
The Study Area
The study area comprises the cocoa growing
region of Nigeria, consisting of the South West,
South East, and South-South geopolitical zones.
Specifically it consists of the 14 cocoa growing
States (Abia, Adamawa, Akwa Ibom, Cross River,
Delta, Edo, Ekiti, Osun, Ondo, Oyo, Ogun, Kwara,
Kogi, and Taraba).
The cocoa growing area in Nigeria lies mainly
below latitude 8 °N within the lowland rain forest
and forest transition ecological zones (Figure 1).
Total annual rainfall of the cocoa growing areas
of Nigeria ranges from 1200 mm to 3000 mm.
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa

The mean minimum temperature is between 20
°C and 22 °C; the mean maximum during the dry
season varies from 31 °C to 32 °C and in the wet
season from 27 °C to 29 °C (Wood & Lass, 1987).
The elevation of the study area ranges from
50 to 2000 m above sea level. More than 50% of
the area falls within elevation range between 200
and 400 m while about 10% of the areas have
high elevation of above 800 m. The regions of
high elevation occur at the extreme eastern part
of the study area at the border between Cameroon
and Nigeria.
Most of the study area has a high population
density generally above 200 persons/km². Mean
population density of 185 person/km² in the
study area is greater than the national average of
about 153 persons/km². About 46 million of the
national population of 140 million reside within
the 14 cocoa growing States (National Population
Commission [NPC], 2007)
Input Data and Data Sources
The input data used for this study were based
on the selected evaluation factors discussed in
the next part. They include a gridded rainfall,
temperature, relative humidity map, vector soil
map, road network map, and cocoa production
and other relevant socioeconomic data (Table
1). The climate grid surface data were obtained
from Worldclim - Global climate data, database
documented in Hijmans, Cameron, Parra, Jones,
and Jarvis (2005) at resolution of 30 arc seconds
(approximately 1 km) (http://www.worldclim.
org/). Relative Humidity data were obtained from
Climatic Research Unit (CRU) of the University
of East Anglia, UK as described by New, Lister,
Hulme, and Makin (2000). Protected area data
and digital elevation model (DEM) were extracted
for the study area from global databases (SRTM,
2000). Soil depth, soil organic matter, and soil pH
factor maps were extracted from the digital soil
map of the world (FAO, 1974). Road density, an

Table 1. Databases used in this research

Factor Suitable Range Data Source

Rainfall (mm) 1500-3000 http://www.worldclim.org/
Minimum temperature (°C) 18-21 http://www.worldclim.org/
Maximum temperature (°C) 29-32 http://www.worldclim.org/
Relative humidity (%) 70-100 CRU http://www.cru.uea.ac.uk
Elevation (m) 0-300 SRTM, 2000
Organic carbon (%) >2 FAO, 1974
Soil depth (cm) >150 FAO, 1974
Soil pH 5.5-6.5 FAO, 1974
Cocoa production (t) 500 – 30,000 CRIN, 2008
Food crop farm size (ha) 5 - 10 CRIN, 2008
Cocoa farm gate price (Naira/bag) 7500 - 9000 CRIN, 2008
Population density (persons/km²) 50 - 500 NPC, 2007
Road density (km/km²) 0.10 - 0.5 Geospatial Laboratory, IITA
Farmer organization (%) 40-100 CRIN, 2008
Government support 10 - 40 IITA, 2008
(No. Farmer Field Schools in the State)
Protected area World Protected area http://www.wdpa.org/

915
 A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa

indication of the infrastructural development of Cocoa Multi-Criteria Land Evaluation

the region, was extracted from the road network
data by computing the total length of road (km)
divided by area (km²) in each local government
area (LGA). The resolution of all raster factor
maps was set at 1 km × 1 km. These data were
used for delineating areas suitable for sustainable
cocoa production.
The GIS Multi-criteria procedure for the cocoa
cultivation suitability assessment in the cocoa
growing region included several stages that are
framed in Figure 2. The determination of the
relevant factors was the starting step in the as-
sessment and was followed by standardizing the
factors, weighting the factors, combining the fac-
tors with their weights, and finally reclassifying
the suitability into three levels of low, medium,
and high. The procedures and algorithms available

Figure 2. Flow chart of cocoa land suitability assessment

916
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
in ArcGIS 9.3 (Environmental System Research
Institute [ESRI], 2008) were employed to imple-
ment the assessment.
Initially, the factors were selected based on
their relevance to suitability in sustainable cocoa
production and the availability of databases. The
selection of factors is a technical process that is
based on expert knowledge or empirical research.
This was done at a meeting of stakeholders com-
prising cocoa industry partners, government
representatives, cocoa scientists, and other experts.
They participated in selecting the factors, identi-
fying the suitable ranges of the factors, and
evaluating the weights of the factors. Furthermore,
from this meeting, the criteria to be used were
grouped into four thematic areas based on the four
specific objectives of the project. The procedures
for cocoa land suitability assessment based on the
four objectives are described below.
Suitability Assessment Based
on National Cocoa Development
Prioritization and Industry Investments
To identify areas that have been defined by national
policy as priority for cocoa production and where
cocoa industry partners are investing or have in-
vested in production and marketing, the number
of Farmer Field schools (FFS) established in each
cocoa growing State of Nigeria was used as an
indicator for government investment priorities.
The FFS efforts of the States are a reflection of
their response to the cocoa development agenda
of the National Cocoa Development Committee.
The program is a partnership program involving
the State governments and the Federal Ministry
of Agriculture and the various cocoa associations.
Data on the number of FFS for the cocoa grow-
ing region of Nigeria were obtained from IITA,
2008 and a simple attribute query operation was
performed in GIS to select States that had at least
ten FFS. According to the weight assigned to
this factor, all States that met this condition were
included and those that failed to meet this were
excluded from further analysis (Nath et al., 2000;
Son & Shrestha, 2008).
Cocoa Suitability Assessment
Based on Biophysical Factors
Climate plays a fundamental role generally in
agriculture and in cocoa production. The quantity
and quality of yields can be affected by water stress
and heat stress (Kassam, Velthuizen Van, Fischer,
& Shah, 1991). The limitation condition method
of land evaluation described by Sys, Ranst, and
Debaveye (1991a, 1991b) as described by various
researchers, such as Maes, Vereecken, and Darius
(1987) and Rodriguez (1995) was used for rating
the biophysical land suitability classes which
implies that all rating below a given threshold
was considered limiting the overall suitability
of such area.
Binary reclassification method was applied to
each climatic layer. All pixels that met the suit-
able range of rainfall (1500-3000) mm, minimum
temperature (18-21°C), maximum temperature
(29-32°C), and relative humidity (70-100%) were
given codes of 1 and those outside those ranges
were coded 0. The resulting intermediate layers
were combined in a weighted algebraic operation
with higher weights of 55% given to rainfall and
the other climatic factors had weights of 15%
assigned to them (Zuidema et al., 2005). This
resulted in a climatic suitability map (Figure 2).
A similar procedure was applied to soil conditions
as specified in Table 1 and the map of soil suit-
ability was obtained. The combination of climatic
and soil suitability resulted in biophysical land
suitability (Figure 2).
Bowen and Hollinger (2004) reported that
precipitation and temperature follow the “law of
the minimum”. This means if a variable is limiting,
the species cannot be grown, even if all the other
variables are not limiting (Holzkämper, Calancaa,
& Fuhrer, 2010). Hence all areas that did not fall
within the suitable zones of biophysical factors
were excluded from the next step of evaluation.
917
 A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
Cocoa Suitability Assessment Based
on Socio-Economic Environment
The third set objective is to identify an area that
has the potential for impact on cocoa household
income. To evaluate land based on this, we needed
the following socioeconomic parameters; cocoa
production, cocoa farm gate price, and food crop
farm size. Cocoa production data in cocoa grow-
ing LGAs is the quantity (t) of cocoa purchased
in the main cocoa season of the year 2005. This
was chosen to reflect the current status of cocoa
production and also to give an indication of the
gap between the current and potential production
levels. The production data were obtained from
CRIN (2008). Higher production areas are of high
priority in the selection process, hence suitable
LGAs must have an annual production of at least
500 t (Table 1).
Average food crop farm size is another item
of socio-economic data employed in the deci-
sion- making process. It is an indication of high
potential for household food security in the cash
crop economy and a measure of sustainable devel-
opment in cocoa growing areas. LGAs with high
food crop farm size are considered suitable in the
selection process and so a cutoff point of at least
5 ha was set for inclusion in the suitability zones.
In addition, data on farm gate prices measured
in price per bag of cocoa were also included in
socio-economic environmental considerations for
sustainable cocoa production. Areas where prices
are low suggests poor marketing infrastructure
while high prices are good indicators of profit-
ability of cocoa farming in the area. These data
were obtained on LGA levels from CRIN, (2008).
Based on cocoa experts’ opinion, areas with low
prices are the priority for project intervention in
this study; hence the cutoff point for this was
set at a maximum of N9000 per bag in the LGA
(Table 1).
All these socio-economic layers were com-
bined in simple algebraic operations after suitable
918
ranges had been ranked and resulted in the socio-
economic suitability layer (Figure 2).
Cocoa Suitability Assessment
Based on Demographic and
Infrastructural Project Environment
The population density is the number of persons/
km² and this was obtained from the Nigerian
Population census of 2006 (NPC, 2007). At the
stakeholders’ meeting, it was agreed that an area
with a very low population density implied a
sparse population of farmers limiting the number
of farmers that can be reached through the project.
An area with a very high population density, on
the other hand, becomes a constraint on land avail-
ability for cocoa farm land expansion. Hence only
cocoa farming areas where population densities
are between 50 and 500 persons/km² fall within
suitable ranges in the selection process.
Road development and access to infrastructure
was another factor considered in the selection
process. Road density in each LGA was used as
a proxy for access to infrastructure and this was
derived from both paved and unpaved roads.
Road density is defined as the total length of road
network per unit area (km/km2). Digitized road
data were obtained from the GIS database of the
Geospatial Laboratory, IITA. Areas with a road
density from 0.10 to 0.50 were considered useful
in the selection process because places with higher
values correspond with a very high population
density and lower values correspond with water
bodies or forest reserves and are classified as
unsuitable for the project.
The percentage of cocoa farmers registered in
farmers’ organizations was also used in the selec-
tion. Farmers who belong to registered farmers’
organizations have access to input, credit, and
warehousing. One of the project’s objectives is to
identify areas with the potential to enhance pro-
gram efficiency. LGAs that have at least 40% of
their farmers registered in a farmers’ organization
were considered suitable in the study.
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
Spatial overlay of the criteria layers was
performed in iterative multi-stage procedures
(Figure 2). All biophysical criteria layers were
combined in additive raster overlay while similar
procedures were employed for socio-economic
and demographic factors. All the intermediate
outputs were then combined to give the final
priority area. A protected area layer was used to
exclude all of the suitable area that belongs to a
national park, forest or game reserve.
DISCUSSION
Figure 3a presents the result of cocoa suitabil-
ity assessment based on government policy and
investment drives. Government policy has been
one major factor that has influenced the cocoa
sector in Nigeria (Coleman, Akiyama, & Varangis,
1993) and so it is selected as one of the limiting
criteria for the cocoa intervention project. From
the figure, only five States out of the 14 cocoa
producing States were qualified to participate in
the project because of good government drives
to educate the cocoa farmers in order to revamp
the sector. Nigeria as a developing country was
rated the second largest world producer of cocoa
in the 1960s (Adegbola & Abe, 1983), and, for a
long time, the crop has been generating substan-
tial foreign exchange earnings for the country.
However, the production of this important cash
crop for export has suffered a reduction in recent
years in the country owing to a number of factors,
such as government policy, low yield, ageing of
cocoa farms, and farm size (Folayan, Daramola,
& Oguntade, 2006; Oduwole, 2004). Moreover
Oluyole and Sanusi (2009) reported that the
educational level of farmers is one other major
factor that has influenced cocoa production in
Nigeria, implying that strong government sup-
port is necessary if cocoa production is to thrive
again. The FFS is a new approach pioneered by
the Sustainable Tree Crops Program (STCP) for
efficient technology dissemination whereby farm-
ers are trained in integrated crop and postharvest
management practices. The success of FFS has
been found to result in raising farmer incomes,
increasing productivity, and reducing dependency
on costly agrochemicals as well as in improving
cocoa quality (IITA, 2005b). It is therefore impera-
tive that government investment in FFS plays a
crucial role in selecting project implementation
communities. Among the five States that are suit-
able according to government support criteria, the
three top cocoa production States of Ondo, Cross
River, and Osun were selected. The remaining two
States of Edo and Abia, even though they are not
in the high producing States, are both enjoying
good government support (IITA, 2008).
The result of combining all biophysical factors,
soil, elevation, and climate is shown in Figure 3b.
It is clear from Table 2 that about 70% of the study
area is suitable with 19% in the category of mod-
erately suitable and 50% in the high suitability
class. This result also suggests that there are more
areas where cocoa could be grown than where it
is currently being grown in Nigeria. All the areas
that fall within moderate-to-high land suitability
are potential areas where cocoa production could
be explored. The major proportion of Anambra,
Imo, Enugu, Rivers, and Benue States (Figure 3b)
could serve as an expansion zone for cocoa pro-
duction in Nigeria. Even though these States are
not included in the cocoa producing States due to
low production levels, they have good potentials
for the expansion of cocoa cultivation in Nigeria
(CRIN, 2008). Low adoption and production of
cocoa in these States are due not only to bio-
physical constraints but also to some socio-cul-
tural issues, as has been observed by Manyong et
al. (2004).
The result of cocoa land evaluation based on
socio-economic factors, such as cocoa production
level, cocoa farm gate prices, and food crop farm
size, is presented in Figure 3c. Areas that fall
within the low suitability class are about 31% of
the study area while only 4.6% belong to the high
suitability category (Table 2). This result contrasts
919
 A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa

Figure 3. (a) Cocoa growing States showing government support as indicated by number of Farmer
Field Schools; (b) cocoa land suitability; (c) cocoa suitability based on socio-economic factors; (d)
cocoa suitability based on demographic and infrastructural factors

with the suitability obtained for biophysical cri-
teria where a larger percentage of the total area
was found to be highly suitable (Figure 3b).
Moreover it is evident from Figure 3c that high
suitability classes occurred only in Ondo and
Cross River States while moderately suitable
areas were found in Ogun, Ekiti, Edo, and Kogi
States. By socio-economic constraints, Osun
which is the third in the rank of high cocoa pro-
ducing States after Ondo and Cross River had low
suitability and unsuitable categories. This obser-
vation suggests socio-economic factors, such as
farm gate price and food crop farm size, are lim-
iting factors in selecting intervention sites for
cocoa sustainable projects. It therefore follows,
as observed by Son and Shrestha (2008) that,
when interventions are being formulated to im-
prove farmers’ income and preserve land re-

Table 2. Percentage suitability in each category for all factors

Factor High Suitability Moderate Suitability Low Suitability Unsuitable

Land suitability 50.26 18.69 22.85 8.20
Socio-economic 4.62 15.09 30.47 49.82
suitability
Demography & infra- 10.01 31.08 21.09 37.82
structure
All factors 3.43 5.64 3.32 87.62

920
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
sources, an adequate understanding of the socio-
economic context of production is important to
accommodate the real needs of land users.
The result of cocoa land assessment based
on demographic and infrastructural constraints
is presented in Figure 3d and Table 2. Based on
these factors, highly suitable areas (10%, Table 2)
occurred in six cocoa growing States, Oyo, Osun,
Ogun, Kogi, Kwara, and Ekiti. The proportion
of moderately suitable land was much larger at
about 30% (Table 2) and was found in almost all
the 14 cocoa growing States except Taraba and
Adamawa. The higher suitability observed in the
southwestern part of Nigeria is probably influ-
enced by the higher level of farmers’ organization
in this region where well organized cocoa farmers
have been reported (Oduwole, 2004; Oluyole &
Sanusi, 2009). Membership of farmers’ associa-
tions is crucial as it has been found to improve the
adoption of new technologies and to raise farmers’
income (Agbeniyi, Ogunlade & Oluyole, 2010)
Figure 4. The study area showing cocoa suitability based on all biophysical, socio-economic, and de-
mographic factors, including government support
The final selected area based on all factors
is shown in Figure 4 and reveals that five States
have suitable sites. This is due to the limiting
priority condition of government support (Figure
3a). Furthermore areas that were suitable based
on other factors such as land and socio-economic
variables have also been excluded where they fall
within a protected area. Hence it is not surprising
that the highly suitable area was only 3.4% while
the moderately suitable places were only about 6%
(Table 2). More than 87% of the study area was
not suitable for this project’s intervention sites.
Table 3 lists the selected LGAs and it shows
that 19 LGAs were selected, three of which belong
to the high suitability category while seven were
in the moderately suitable area. Nine LGAs were
found in the low suitability category. A total area
of about 20,000 km² was selected. High suitabil-
ity LGAs were distributed evenly among the three
top high-producing States of Ondo, Cross River,
and Osun. Moderately suitable areas can be found
921
 A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa

in all the selected States except in Abia where the
only selected area was in the low suitability cat-
egory.
CONCLUSION AND
RECOMMENDATION
Though the GIS-MCE approach provides an effec-
tive framework for land evaluation, the selection
of assessment factors and the identification of a
suitable range for each factor have a direct influ-
ence on the results. In this study, the factors were
selected based on the experts’ opinion; therefore
they represent a considerable share of the factors
relevant to the suitability of cocoa growing areas
in the region. For instance, the inclusion of the
government priority and investment variable has
a strong influence on the States selected for the
project’s intervention sites.
It can also be concluded that the integration
of spatial databases and expert knowledge sig-
nificantly enhances the decision-making capacity
when land suitability evaluation is undertaken.
Moreover, the approach highlights the partici-
patory decision-making process (Eastman, Jin,
Kyem, & Toledano, 1992). Therefore, it helps to
minimize and solve conflicts among competing
interests in the area of cocoa land-use management.
The GIS-MCE approach has been widely
applied in land suitability analysis (Malcewski,
2006), however, the application of the method in
cocoa sustainable land management is relatively
new in Nigeria. The MCE of soil, topography,
and socioeconomic factors was exemplified to
be useful for delineating areas suitable for cocoa
production in the region. In particular, the in-
volvement of local experts was vital to obtaining
consistent results. The experts played key roles in
the selection of the evaluation factors and in the
determination of the factor weights. The applica-
tion of this paper can be useful for the managers
and planners of cocoa sustainable projects.
The study shows that GIS databases of different
formats and sources can be efficiently integrated
to establish a land suitability assessment for co-
coa production. The methodology is useful for
identifying priority areas for cocoa farming, and

Table 3. Selected Local Government Area (LGA) for cocoa sustainable project

Area Area
State LGA Suitability Class (km²) State LGA Suitability Class (km²)
Cross
River Obudu Moderate 371 Abia Bende Low 591
Cross
Edo Owan East Moderate 959 River Ogoja Low 977
Cross
Ondo Odigbo Moderate 2193 River Obanliku Low 936
Cross
Osun Irewole Moderate 310 River Etung Low 869
Osun Aiyedaade Moderate 1017 Edo Owan West Low 672
Osun Atakumosa West Moderate 868 Edo Etsako West Low 1637
Osun Oriade Moderate 257 Ondo Akure North Low 669
Cross
River Boki High 4868 Osun Atakumosa East Low 385
Ondo Idanre High 1880 Osun Boluwaduro Low 145
Osun Obokun High 474
Total Area 20,079 (km²)

922
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
thus it contributes to improving the efficiency of
conservation and of sustainable land management.
The land suitability information produced
in the recent research is valuable. However, the
socio-economic information was based on a single
survey carried out by the Cocoa Research Insti-
tute of Nigeria in 2005 which has its limitations.
More comprehensive socio-economic factors
and other infrastructural variables could be used
to enhance the accuracy of the results obtained.
Therefore, we recommend that future studies
should consider broader and more detailed socio-
economic information to offer decision-makers a
comprehensive basis for a feasible strategy, and
sound decisions about more sustainable cocoa
production in Nigeria.
REFERENCES
Adegbola, M. O. K., & Abe, J. O. (1983). Cocoa
development programme, Nigeria. Cocoa Re-
search Institute of Nigeria Printing Unit, 9, 3–5.
Agbeniyi, S. O., Ogunlade, M. O., & Oluyole, K.
O. (2010). Fertilizer use and cocoa production in
Cross River State, Nigeria. Journal of Agricultural
and Biological Science, 5(3), 10–13.
Bastide, P., & Perret, C. (2007). Atlas on regional
integration in West Africa: Cocoa. Paris, France:
Organization for Economic Co-operation and
Development.
Bowen, C. R., & Hollinger, S. E. (2004). Geo-
graphic screening of potential alternative crops.
Renewable Agriculture and Food Systems, 19,
141–151.
Cocoa Research Institute of Nigeria. (2008). Ni-
geria Cocoa production survey. Nigeria, Africa:
Cocoa Research Institute of Nigeria (CRIN).
Coleman, J. R., Akiyama, T., & Varangis, P. N.
(1993). How policy changes affected cocoa sec-
tors in sub-Saharan African countries: Vol. 1.
(Research Report No. WPS 1129). Washington,
DC: World Bank.
Cover, S. J. (1991). Integrating multi-criteria
evaluation with geographical information
systems. International Journal of Geo-
graphical Information Systems, 5, 321–339.
doi:10.1080/02693799108927858
Cowen, D. J. (1988). GIS versus CAD versus
DBMS: What are the differences? Photogram-
metric Engineering and Remote Sensing, 54,
1551–1554.
Eastman, J. R., Jin, W., Kyem, P. A. K., & Tole-
dano, F. (1992). Participatory procedures for
multi-criteria evaluation in GIS. In Proceedings
of the Chinese Professionals in GIS (pp. 281-288).
Environmental Systems Research Institute.
(2008). ArcGIS 9.3. Retrieved from http://www.
esri.com/
Folayan, J. A., Daramola, G. A., & Oguntade, A.
E. (2006). Structure and performance evaluation
of cocoa marketing institutions in South-Western
Nigeria: An economic analysis. Journal of Food
Agriculture and Environment, 4(2), 123–128.
Food and Agriculture Organization. (1974). Soil
map of the world (Vol. 1-10). Paris, France: Food
and Agriculture Organization of the United Na-
tions and UNESCO.
Food and Agriculture Organization. (1976). A
framework for land evaluation. FAO Soils Bul-
letin 32.
Food and Agriculture Organization. (2003). Agri-
culture data. Retrieved from http://faostat.fao.org
923
 A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa
Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones,
P. G., & Jarvis, A. (2005). Very high resolution
interpolated climate surfaces for global land
areas. International Journal of Climatology, 25,
1965–1978. doi:10.1002/joc.1276
Holzkämper, A., Calancaa, P., & Fuhrer, J. (2010).
Evaluating climate suitability for agriculture in
Switzerland. Paper presented at the Fifth Biennial
Meeting of the International Congress on Envi-
ronmental Modelling and Software Modelling
for Environment’s Sake, Ottawa, ON, Canada.
International Cocoa Organization. (2007, June 4).
Sustainable cocoa economy: A comprehensive and
participatory approach. Paper presented at the
Thirteenth Meeting ICCO Offices Consultative
Board on the World Cocoa Economy London, UK.
International Cocoa Organization. (2010). Cocoa
year 2009/2010. Quarterly Bulletin of Cocoa
Statistics, 36(2).
International Institute of Tropical Agriculture.
(2002). Sustainable tree crop project (IITA): A
report on the baseline survey on cocoa farming
livelihood of West Africa. Ibadan, Nigeria: Inter-
national Institute of Tropical Agriculture.
International Institute of Tropical Agriculture.
(2005a). STCP impact brief series, issue no 1.
Retrieved from http://www.treecrops.org
International Institute of Tropical Agriculture.
(2005b). STCP impact brief series, issue no 2.
Retrieved from http://www.treecrops.org
International Institute of Tropical Agriculture.
(2008). National cocoa development committee
and sustainable tree crop FFS capacity building
project report. Ibadan, Nigeria: International
Institute of Tropical Agriculture.
924
Kassam, A. H., Velthuizen Van, H. T., Fischer, G.
W., & Shah, M. M. (1991). Agro-egological land
resources assessment for agricultural development
planning - A case study of Kenya - Resources data
base and land productivity - Technical Annex 3.
World Soil Resources Reports, 78.
Khoi, D. D., & Murayama, Y. (2010). Delinea-
tion of suitable cropland areas using a GIS based
multi-criteria evaluation approach in the Tam Dao
National Park Region, Vietnam. Sustainability,
2, 2024–2043. doi:10.3390/su2072024
Maes, J., Vereecken, H., & Darius, P. (1987).
Knowledge processing in land evaluation. In
Beek, K. J., Burrough, P. A., & MacCormack, D.
E. (Eds.), Quantified land evaluation procedures.
Enschede, The Netherlands: ITC.
Malcewski, J. (1999). Spatial multicriteria
decision analysis. In Thill, J.-C. (Ed.), Spatial
multicriteria decision making and analysis: A
geographical information sciences approach (pp.
101–269). Aldershot, UK: Ashgate.
Malcewski, J. (2006). GIS-based multi-
criteria decision analysis: A survey of the
literature. International Journal of Geo-
graphical Information Systems, 20, 703–726.
doi:10.1080/13658810600661508
Manyong, V. M., Okike, I., Legg, C., & Makinde,
K. O. Haan, de N., &. Alabi, T. (2004). NDDC
master plan project: The agriculture and rural
enterprises sector study report. Asokoro, Nigeria:
NDDC.
Mustapha, A. R. (1999). Cocoa farming and
income diversification in South-western Nigeria.
Nigeria: De-Agrarianisation and Rural Employ-
ment Network (DARE).
National Population Commission. (2007). Census
2006 results -36 states and FCT. Retrieved from
http://www.population.gov.ng
A Multi-Criteria GIS Site Selection for Sustainable Cocoa Development in West Africa

New, M., Lister, D., Hulme, M., & Makin, I. SRTM. (2000). US geological survey. Sioux Falls,
(2000). A high-resolution data set of surface cli- SD: EROS Data Center.
mate over global land areas. Climate Research,
Stainer, F. (1991). Landscape planning: A method
21, 1–25. doi:10.3354/cr021001
applied to a growth management example.
Oduwole, O. O. (2004). Adoption of improved ag- Environmental Management, 15, 519–529.
ronomic practices by cocoa farmers in Nigeria: A doi:10.1007/BF02394742
multivariate Tobit analysis. Unpublished doctoral
Sui, D. Z. (1993). Integrating neural networks
dissertation, Federal University of Agriculture,
with GIS for spatial decision making. Operation
Akure, Nigeria.
Geography, 11, 13–20.
Oluyole, K. A., & Sanusi, R. A. (2009). Socio-
Sys, C., Ranst, E., & Debaveye, J. (1991a). Land
economic variables and cocoa production in Cross
evaluation part I: Principles in land evaluation
River State, Nigeria. Journal of Human Ecology
and crop production calculations. Brussels, Bel-
(Delhi, India), 25(1), 5–8.
gium: GADC.
Pereira, J. M. C., & Duckstein, L. A. (1993). Mul-
Sys, C., Ranst, E., & Debaveye, J. (1991b). Land
tiple criteria decision-making approach to GIS-
evaluation part II: Methods in land evaluation.
based land suitability evaluation. International
Brussels, Belgium: GADC.
Journal of Geographical Information Systems,
7, 407–424. doi:10.1080/02693799308901971 Wood, G. A. R., & Lass, R. A. (1987). Cocoa
tropical agriculture series (4th ed.). London,
Rodriguez, O. (1995). Land use conflicts and
UK and New York, NY: Longman Scientific and
planning strategies in urban fringes. A case study
Technical and John Wiley & Sons.
of Western Caracas, Venezuela. Unpublished
doctoral dissertation, University of Ghent, Gh- Zuidema, P. A., Leffelaar, P. A., Gerritsma, W.,
ent, Belgium. Mommer, L., Niels, P. R., & Anten, N. P. R.
(2005). A physiological production model for
Son, N. T., & Shrestha, R. P. (2008). GIS-Assisted
cocoa (Theobroma cacao): Model presentation,
land evaluation for agricultural development in
validation and application. Agricultural Systems,
Mekong Delta, Southern Vietnam. Journal of Sus-
84, 195–225. doi:10.1016/j.agsy.2004.06.015
tainable Development in Africa, 10(2), 875–895.

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 1,
edited by Donald Patrick Albert, pp. 73-87, copyright 2012 by IGI Publishing (an imprint of IGI Global).

925
926

Chapter 57
The Columbia Regional
Geospatial Service
Center System:
A Proven Model for the United States

P. R. Blackwell
Stephen F. Austin State University, USA

Darrel McDonald
Stephen F. Austin State University, USA

ABSTRACT
During the past 20 years, the role of geospatial technology in society has increased dramatically.
However, the impact of these technologies in rural areas remains minimal. In Texas, a federally funded
project called the Columbia Regional Geospatial Service Center System (the System) has emerged as a
model for bringing the benefits of geospatial technology to all portions of society. The model involves
distributed, academically based Centers, each with regional specializations, linked together into a uni-
fied system for addressing critical needs in emergency response, economic development, and natural
resource management. The Centers operate on three focus areas, i.e., data, applications, and training.
The Columbia Center has been in operation for five years and has demonstrated the practical strength
of the System through numerous local and statewide projects, responses to natural disasters, and other
geospatial activities.

DOI: 10.4018/978-1-4666-2038-4.ch057

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
The Columbia Regional Geospatial Service Center System
INTRODUCTION
History
In 2005, the United States Congress established
the Columbia Regional Geospatial Service Center
System through a federal appropriation. Initially,
the System consisted of the Columbia Regional
Geospatial Service Center at Stephen F. Austin
State University (SFASU) in Nacogdoches and
the Regional Geospatial Service Center at the
University of Texas at El Paso. The appropriation,
championed by Senator Kay Bailey Hutchison,
came as recognition of the contributions of SFASU
during the recovery effort after the loss of STS 107,
the Space Shuttle Columbia. But the origins of
the project go back to the 1990s when researchers
at SFASU first recognized the disadvantages that
rural communities suffered in terms of adopting
geospatial technology to address pressing prob-
lems. The work that went on quietly in East Texas
over the next ten years resulted in development
of capabilities that enabled SFA to contribute so
effectively to the recovery effort (Gehman, 2003).
Relevance
Senator Hutchison challenged the Columbia
Center to demonstrate a better way to bring the
advantages of geospatial technology to all parts
of the state, and in so doing create a model for
the nation. During the course of the ensuing
five years, the system has done that. Today the
Columbia Regional Geospatial Service Center
System is comprised of five Centers: the original
two in Nacogdoches and El Paso; the Center for
Geospatial Technology at Texas Tech University
in Lubbock; a facility in the Department for Earth
and Space Sciences at Lamar University; and the
Center for Earth and Environmental Studies at the
Texas A&M International University in Laredo.
The System operates in close coordination with
The Texas Natural Resources Information System
(TNRIS), a state agency mandated to serve the
geospatial needs of Texas. In effect, the Columbia
Center System extends the reach of TNRIS to all
areas of the State.
Mission
The Columbia Regional Geospatial Service Center
System provides regional geospatial support for
emergency preparedness and response, economic
development and resource management. The Sys-
tem serves all interest groups, public and private,
by providing cutting edge technology tempered
with a pragmatic, bottom-up philosophy. This
broad mission allows the System flexibility to
discover new ways to apply geospatial technology
for the common good. At the same time, the three
mission areas give focus to the Systems efforts,
while leveraging the work in each mission area
to the benefit of the other areas. Put another way,
the same geospatial data, applications and training
used for emergency preparedness and response
can be applied to economic development and
natural resource management issues (McDonald
et al., 2007).
THE MODEL
The Columbia Regional Geospatial Service Center
System Model is comprised of eight principles,
each of which defines an important aspect of the
successful program. These principles are a com-
bination of fundamental concepts upon which the
project is based and pragmatic additions that have
developed over the years and refined by experi-
ence. These principles are the foundation of the
Columbia Regional Geospatial Service Center
System (Kroll et al., 2006).
Distributed Centers
The idea for regionally based, distributed centers
dates back to very early thinking about the chal-
lenge of bringing geospatial technology to rural
927
 The Columbia Regional Geospatial Service Center System
areas. The problem persists that state and feder-
ally based programs often do not have sufficient
reach to impact local jurisdictions effectively. An
obvious solution is to place resources where they
are needed, i.e., locally. This idea became a core
concept of the Columbia System model during the
Space Shuttle Columbia recovery effort in 2003.
It was observed that the farther the search ranged
from geospatial resources the more difficult it be-
came to adequately support the effort. As a result,
the concept of “Forward Mapping Centers” was
developed and successfully deployed during the
first critical weeks of the recovery (Blackwell et
al., 2003). Today this idea is reflected not only in
the regional distribution of the Columbia System
Centers, but also in the Geospatial Rapid Response
teams being developed by the System.
Regional Specialization
Further, the needs of different areas vary greatly.
For example, forestry issues that are prevalent
in East Texas hold little interest for the citizens
of El Paso. At the same time international issues
that dominate life in El Paso are less meaningful
in East Texas. Each region has its own concerns
that are best addressed by regional centers. In
addition, each Center is able to concentrate on
local needs and develop a level of expertise that
would be difficult to achieve by a centralized ef-
fort. Other examples of regional specialization are
the research on the Ogallala aquifer by the Center
for Spatial Technology at Texas Tech (Mulligan
et al., 2008) and the coastal process work be-
ing undertaken by the Department of Earth and
Space Sciences at Lamar (Farmer et al., 2010).
These regional specialties are great assets to the
individual region and bring strength and depth
to the System.
Bottom-Up Philosophy
The regional distribution of Centers leads to a dis-
tinctly bottom-up philosophy. Centers concentrate
928
at the local level, working with city and county
governmental agencies, local police and sheriff
offices, fire fighters, appraisal districts and others.
This local focus is one of the defining strengths of
the Columbia Regional Geospatial Service Center
System. The origin of this concept and conviction
in its value comes from hard-earned experience
in dealing with real-world emergencies. Emer-
gencies are local and effective response during
the first critical hours after an event depends on
action by local authorities. Often, however, local
jurisdictions do not have the technical resources
they need to respond effectively. The Columbia
System Model addresses that unmet need with
focused geospatial technology and support at the
local level.
Bridging the Divide
This focus on the local level allows the regional
Centers to bridge the gap between state and fed-
eral programs and local agencies. The Centers’
roles are to “connect the last mile” between these
programs and the individuals and agencies at the
local level that need them. This is essential, par-
ticularly in rural areas where resources are thin
and capacity lacking. Thus, the Columbia Center
System focuses on rural Texas. In most urban
areas, geospatial technology is well entrenched
(Croswell, 2009; DeMers, 2009). In rural parts
of the state, as in most rural areas in America,
geospatial resources often do not exist. Regionally
distributed Centers close the gap between urban
and rural access to geospatial technology, data
and procedures.
Local Data
The System Centers focus on local data. Local
data are often more current and accurate than
data compiled at the state or federal level, and
the most efficient way to build compilations of
large-scale, timely and accurate data across the
states and the nation is with good local data. The
The Columbia Regional Geospatial Service Center System
Columbia Center System is dedicated to this
concept and works with local jurisdictions to
find, vet, update, augment, or create good qual-
ity geospatial data. These data are archived and
distributed by the Centers and passed up the chain
to TNRIS for inclusion in state-level archives
(Figure 1). Coordinating schemas and standards
across domains are challenges that the Columbia
Center System is dedicated to meet. Reliable, ac-
curate and timely local data are essential to any
state or federal level compilation. It is impossible
for state or federal agencies to bring the same
resources to a mapping project as local agencies
can. In terms of local knowledge, timeliness,
detail, workforce, the local effort wins every
time. Often, however, the necessary resources
and expertise are not available at the local level.
This is where the Columbia Regional Geospatial
Service Center System plays an important role in
providing functional geospatial data at all levels.
Figure 1. Data flow model from local to state to federal compilations
System Linkage
While the regional Centers provide appropriate,
customized services to local and regional agencies,
they do not operate in a vacuum. The Centers are
connected to the System by common needs and
a shared mission as well as compatible technol-
ogy. The Centers work together to enhance the
strengths of each Center to address the issues
of each region. Through weekly meetings via
video conferencing, and quarterly face-to-face
coordination meetings, the Centers share their
experiences and capabilities to the benefit of all.
High-performance networking is key to this co-
ordination. In Texas, the Lonestar Education and
Research Network (LEARN) provides high-speed
connectivity between the Centers and TNRIS,
enabling the Centers to operate as one when
necessary. The System structure makes statewide
projects possible as well, allowing the resources
of all Centers to be focused on statewide issues.
In large-scale emergencies, all Centers contribute
to a unified response. Regular training and drills
929
 The Columbia Regional Geospatial Service Center System

keep skills sharp and lead to a strong esprit de
corps across the organization.
Data Replication
The Columbia Center System realizes that single
points of failure are inconsistent with any operation
that is expected to provide critical services. The
System maintains data and services replication to
avoid problems caused by failure of any one part
of the System. The importance of this approach has
been demonstrated many times. In June of 2001,
for example, Tropical Storm Allison devastated
southeast Texas dropping 40 inches of rain over
some areas (LaNore, 2001). Severe flooding in
Houston damaged roads, sewers, hospitals and
many businesses. One of the casualties of this
storm was the equipment room where the Texas
Gigapop resides. The Gigpop connects Texas to
Internet2, a high-speed network for universities,
research centers, and other collaborators (http://
www.internet2.edu). This facility was down for
many days, effectively severing a large part of
the state from the Internet, including SFASU. The
System model prevents any such interruption of
service from a single point of failure by replicating
critical data and services at Centers distributed
across the state. The Columbia Center System’s
data replication encompasses many different
technologies and strategies designed to address
specific needs. Data replication for emergency
response, for instance, is designed not only to
ensure data integrity but to provide robust access
to data through a variety of delivery mechanisms.
each Center’s ability to carry on its academic
mission in addition to meeting the needs of a
regional Center. Secondly, academic institutions
are often progressive in terms of adopting new
technology. This willingness to grow and change
is essential in a field that is as dynamic as geo-
spatial technology. Thirdly, high performance
computing and networking infrastructures avail-
able in universities also are essential in fulfilling
the Columbia Center System mission. Finally,
all three parts of higher education’s mission:
research, education and outreach, are enhanced
by the Columbia Center System activities. This
makes the university environment the ideal place
for Columbia System Centers.
OPERATIONAL MODEL
The Columbia Center System focuses on inte-
grated, regional activity in three broad areas: Data,
Applications, and Training. The three areas and
the technology that overlaps them, form a compre-
hensive program designed to meet the goals and
objectives established for the System (Figure 2).
Figure 2. The Columbia Center System Opera-
tional Model comprises three areas of focus: Data,
Applications and Training that overlap technical
resources

University Based

The Columbia Regional Geospatial Service Center

System Centers are based in academic institu-
tions. This was done for several reasons. Firstly,
participating universities already had established
geospatial expertise and were chosen largely on
the basis of existing capacity. The System built
capacity within the Centers as well, enhancing

930
The Columbia Regional Geospatial Service Center System
Data
As previously mentioned, local data are critical to
successful responses, however, all local data are
not equal. Centers work with local stakeholders
to ensure that the best possible data are identi-
fied and made available to all who need them.
At times, this means revising or creating data
where existing data are inadequate. The System
has projects underway across Texas refining and
extending datasets, from building parcel maps in
the Panhandle to E9-1-1 road-centerline rectifica-
tion in East Texas.
For many applications, geospatial data need
to be integrated at the state level. To meet these
requirements, the System engages in statewide
data projects ranging from electrical power grid
mapping for emergency response to public water
supply critical infrastructure mapping to statewide
E9-1-1 data compilation. With Centers located
across the state, the System is able to bring the same
bottom-up philosophy to statewide projects as it
does to local activities. This capability is recog-
Figure 3. The Columbia Center Systems Common Operational Map provides compressive, map-based
information to all responders during an emergency
nized by state and federal agencies as a significant
advantage (National Research Council, 2007).
Applications
To be useful, geospatial data must be applied to
address specific issues. The Columbia Center Sys-
tem recognizes this fact and places considerable
emphasis on developing customized applications
for specific purposes. Applications range from
entire protocols such as the Wide Area Damage
Assessment (WADA) program developed for the
Texas Military Forces and the System’s vision
of Common Operational Picture, the Common
Operational Map (COM) (Figure 3) to simple
web-based mapping applications such as Road
Closure and Fuel Availability web applications
developed for Nacogdoches County Emergency
Management. In all cases, however, applications
are designed to bring the power of geospatial
technology to those that need it in a way that is
easy to understand, simple to use and affordable.
931
 The Columbia Regional Geospatial Service Center System
Training
The third leg of the System operational model
is training. Even the most elegant geospatial
applications are useless unless people know and
understand them. The Columbia Center System
conducts a comprehensive training program cov-
ering a diverse range of geospatial technologies
and targeting all users from professional GIS
practitioners to the general public.
The System’s training program for the Texas
Military Forces has developed into a new mission
for the Texas State Guard: Wide Area Damage
Assessment (WADA) (McDonald et al., 2008;
Spraggins, 2009). The three-stage program in-
volves land navigation using GPS, field search
protocols and advanced geospatial-based planning
concepts. This program has spawned the concept of
Geospatial Quick Response Teams (QRTs) within
the Texas State Guard and is being expanded to
include Texas National Guard personnel. On-line
training modules under development for this
program will increase the audience and reduce
training costs. In the future, all TXMF personnel
may benefit from this training.
The concept of WADA, as developed by the
Columbia Regional Service Center System and
TXMF, has application beyond Texas (McDonald
et al., 2009). These principles and techniques
complement other methodologies commonly used
for damage assessment such as aerial photography
or satellite image analysis. The recent tragedy in
Haiti illustrates the need for rapid, on-the-ground,
damage assessment. Those assisting with the
relief effort are hampered by the lack of detailed
information about the geography and extent of
the devastation. The WADA initiative provides
such critical information quickly and efficiently.
WADA training is being rolled out for civil
agencies as well as TXMF. WADA training for
police, firefighters, and volunteer first responders
will be made available through each CRGSCS
Center. Distributed Centers allow this training
to be presented to jurisdictions across Texas. The
932
Columbia Center System provides customized
training for other federal agencies such as the
USDA National Forest Services (NFS), National
Weather Service (NWS), and US Army Corp of
Engineers (USACOE). The regional Centers are
able to address the specific needs of agencies
within their areas of interest, providing targeted
training to meet a broad range of needs.
PROOF OF CONCEPT
There were two important developmental pro-
grams in the geospatial sciences program at
Stephen F. Austin State University, which predate
the formation of the Columbia Center on campus.
These programs were instrumental in building geo-
spatial competency that contributed to a successful
response during the Columbia Shuttle incident.
Starting in the mid 1990s, two geospatial labora-
tories were established to introduce GIS into the
curriculum and into research initiatives involving
local entities, agencies and industry. The Forest
Resources Institute (FRI) supported research and
teaching in the College of Forestry and the Hu-
manities Urban Environmental Sciences (HUES)
lab based in Geography concentrated on city/local
government side of GIS applications.
Implementing GIS
FRI built strong relationships with the regional
forest industry while the HUES group worked
closely with the City of Nacogdoches to build
an infrastructure geodatabase. As the programs
matured, entities started to rely upon Stephen F.
Austin State University for leadership in imple-
menting GIS solutions to address service delivery
and public safety issues (Sementelli & McDonald,
2003). Through extended collaborations trust was
built between participating parties that continues
to the present. Recently, Nacogdoches County
expanded their investment in GIS/911 mapping
resources after working with the Columbia Cen-
The Columbia Regional Geospatial Service Center System
ter to improve service delivery and emergency
response capacities.
Similar projects have emerged at partner in-
stitutions of the CRGSCS. For example, at UTEP
the Regional Center for Geospatial Services,
Director Raed Al Douri has become the focal
point for the Mapas del Norte group that links
GIS operations in several city and county entities
along the US/Mexican border extending into New
Mexico. Currently, the Center is working with New
Horizon, a community near El Paso to establish
a GIS solution to managing city infrastructure.
At the Center for Geospatial Technologies at
Texas Tech, Dr. Kevin Mulligan, Director and
staff geospatial scientists are collaborating with
city and county entities to augment parcel data
and build capacity to better serve the public and
create critical infrastructure datasets essential to
successful responses to incidents.
Enhancing Public Safety
Another area that the System partners have been
successful in providing expertise in geospatial
needs is in enhancing public safety. The Columbia
Center was invited to build a base-mapping project
for the East Texas Council of Governments (ET-
COG) to meet requirements of a hazard mitigation
plan directive from FEMA. The Columbia Center,
in conjunction with regional and local entities,
created geodatabases to produce the base maps
needed to meet project objectives.
In another example, at Texas A&M Interna-
tional University (TAMIU), the Center for Earth
and Environmental Studies, under the direction
of Dr. Ken Tobin, coordinated a 14 county hazard
mitigation plan project funded by FEMA. More
recently, the Columbia Center has been tasked
to update the 911 maps for the Deep East Texas
Council of Governments (DETCOG). This project
focused on using current high-resolution photog-
raphy to improve road network representation
on maps used to assign addresses for emergency
response. Moreover, the project included build-
ing a replication solution for local editing by
individual counties with database replication
back to DETCOG offices and offsite storage to
the Columbia Center.
Storm Support
A third area that the Columbia Center model has
been an important aspect in is supporting entities
responding to storms in the Gulf of Mexico region,
particularly hurricanes. Columbia Center support
for local agencies was first initiated just before and
during Hurricane Rita in 2005. As most remember,
in August Hurricane Katrina wreaked havoc on
New Orleans displacing hundreds of thousands
of people. The Columbia Center provided map-
ping support for city and county entities dealing
with the influx of thousands of evacuees. The
maps were primarily informational rather than
analytical, and were well received by the local
entities as a tool to help meet evacuee needs and
questions in a timely fashion. Then, a month later,
Hurricane Rita ripped through East Texas. This
time the Columbia Center operated in conjunction
with local and regional responders. For example,
just hours before Rita made landfall, Columbia
Center personnel delivered county base maps to
Texas Department of Public Safety responders
in the Beaumont area. Throughout the event, the
Columbia Center provided mapping support for a
broad spectrum of responding agencies, ranging
from law enforcement and emergency managers
to electrical companies (McDonald et al., 2009).
In 2008, Hurricane Ike made landfall near
Galveston, Texas. The storm devastated Bolivar
Peninsula and swamped Galveston Island before
moving inland, continuing to damage buildings
and destroy infrastructure as it carved a trail north
through East Texas. Shortly after the storm had
passed Galveston County, officials requested
geospatial assistance to support search and rescue
operations. The Columbia Center deployed a team
to help meet this critical need (Figure 4). Hundreds
of maps later, the Columbia Center was recognized
933
 The Columbia Regional Geospatial Service Center System

Figure 4. Columbia Center South – a self-contained, fully capable GIS facility was imported, set up,
and operated in the Galveston County EOC in Dickerson, Texas within hours of Hurricane Ike’s passing

as an important resource that helped in managing Border Security

a successful response to a tragic incident.
As a result of these deployments, the System
has moved forward with developing county level
base maps (3 counties deep) along the southern
perimeter of Texas from El Paso to Beaumont as
well as extending northward to the Red River. A
final objective of this project is to have base maps
prepared for all 254 Texas counties.
Based on lessons learned during Hurricanes
Katrina, Rita, Gustav and Ike, each Center is de-
veloping two-person, geospatial response teams
capable of responding to incidents anywhere in the
state on short notice. These Rapid Response teams
will be able to operate connected or disconnected
from other resources to meet mapping needs of
any response effort across the state. In addition,
the Lamar Geospatial Lab Associate Director,
Dr. Joe Krueger, is focusing on aligning research
projects that coincide with the Height Moderniza-
tion Project based at Texas A&M University at
Corpus Christi (http://tsrc.cbi.tamucc.edu/THM/
HomePage).
Recently, the System became involved with border
security through work with the Texas Military
Forces. In particular, the Texas State Guard troops
are deployed to the Joint Operations Information
Centers (JOIC) across the Texas borderland to aid
the border security mission. The JOICs combine
the resources of local law enforcement, Texas
Department of Public Safety (TDPS), the Border
Patrol, and other state and federal agencies in a
coordinated effort to increase border security.
The Columbia Center developed a geospatially
based information system called GeoBorderStar
to help JOICs share and use geospatially enabled
information.
This work lead to an invitation for the Columbia
Center System to participate in Project Sidewinder,
a directed action conducted in Brooks County,
Texas in March of 2009. Columbia Center per-
sonnel embedded with the Sidewinder task force
to provide mapping and geospatial analysis sup-
port for the multi-agency task force. The project
demonstrated the importance of mapping in this
type of operation and has led to other initiatives

934
The Columbia Regional Geospatial Service Center System
designed to integrate geospatial technology into
the border security mission in Texas.
Recently, the Texas Military Forces drafted a
requirement directing the Columbia Center to help
integrate TexMap, the TDPS information system,
with TXMF mapping assets. This work will lead
to better integration of information resources for
all stakeholders.
Critical Infrastructure Mapping
The importance of critical infrastructure datasets
has been recognized by research on the role of geo-
spatial sciences in emergency response (National
Research Council, 2007). The Columbia Regional
Geospatial Service Center System has undertaken
several statewide critical infrastructure mapping
projects. Given the size of Texas, the distributed
Centers hold a substantial advantage over any cen-
tralized agency for such projects. These projects
are large scale and require considerable resources.
Originating field activities at each of the Centers
mitigates logistics and travel costs. Project coor-
dination is paramount with this type of endeavor,
and a single project manager is selected to oversee
each project. Three major state wide projects have
been completed or initiated to date.
The Texas Power Grid Mapping Project is
producing a comprehensive geodatabase of the
Texas Power Grid for emergency response. This
project involves compiling existing infrastructure
and spatial data from electrical power companies,
rectifying them against aerial ortho-rectified imag-
ery and integrating the results. Texas Department of
Emergency Management (TDEM) will use the end
product for tracking and mitigating power outages
during storm events.
The Texas Commission on Environmental Qual-
ity (TCEQ) tasked the Columbia Center System
to create a Critical Facilities Infrastructure Map
(CFIM) for the State of Texas. The project involves
mapping public water supply (PWS) infrastructure
for all systems serving fewer than 10,000 customers
in Texas. GPS data collection teams visit each PWS
and record locations for wells, pumps, tanks and
other components, along with information about
capacity, treatment types, storage capacity, and
generator/wiring harness availability. The infor-
mation will be used by TCEQ during emergency
conditions to coordinate relief efforts to restore
clean water as quickly as possible after a natural
or manmade outage.
The Columbia Center System has completed
another project for TCEQ involving updating
boundary data for the various regulatory districts
under TCEQ jurisdictions. This work involved
reviewing plats and legal records, rectifying map
data and constructing a comprehensive geodatabase
of all district boundaries. In many cases, local ap-
praisal districts and courthouses were contacted to
ensure the most accurate and up to date information
were used in the process.
In addition to updating the boundary data, this
project helped TCEQ to update their internal opera-
tions from ArcView 3.x/Shapefiles to a state-of-the-
art ArcGIS/geodatabase operation. The Columbia
Center System continues to provide training and
technical support to TCEQ for the project.
Data Warehousing and Replication
Geospatial data replication is one of the funda-
mental concepts on which the Columbia Regional
Geospatial Service Center System is based. The
System provides replication services for local
and regional entities as well as for critical state
resources. At present, the System is replicating
data for the 12 counties of DETCOG, Panola
County and several cities. In some cases these
replicas serve as live backups and are instantly
available if the primary databases fail. In others,
the Columbia Center hosts the working database
for smaller entities that are not able to maintain
their own geospatial infrastructure. Panola County
is an example of this arrangement. The Columbia
Center is working closely with Panola County,
moving them towards implementing their own
GIS. In the meantime, the Columbia Center hosts
935
 The Columbia Regional Geospatial Service Center System
Panola County data and makes tools available to
the County for editing and using these data from
county offices in Carthage, Texas.
The Columbia Center System partners with
TNRIS on several data replication initiatives. The
Texas Strategic Mapping Program (Stratmap) is
the premier, high-resolution geospatial dataset
for Texas. Stratmap data are mirrored at each of
the Columbia Center System Centers, providing
five-way redundancy and enhanced access to these
important data. Raster data are replicated as well,
including the original 1995 Texas Orthoimagery
Program (TOP) 1-meter CIR data, and USDA
National Agricultural Imagery Program (NAIP)
data for 2004, 2006, 2008 and 2009. These larger
datasets are not fully replicated at each Center.
Instead, each Center maintains their region of
Texas while TNRIS maintains data for all of the
State. Thus, all the data are available from at least
two sources.
The importance of offsite backups of critical
data has been demonstrated repeatedly. Pertinent
case studies can be drawn from the September
11th attack on the World Trade Center and Hur-
ricane Katrina in New Orleans (National Research
Council, 2007). In both cases, both the primary
and backup geospatial data for the area were
lost due to backup data being stored too close
to the main data store. The Columbia Center
System model places replicas of crucial data at
the regional centers, far from the main location.
Where warranted, additional replicas are housed
at other regional centers ensuring uninterrupted
access to critical data.
An innovative project is currently underway
to distribute the L-2 rectified aerial photogra-
phy used to create the 2009 NAIP dataset for
Texas. This is 4 band, 0.7 meter imagery that is
unprocessed other than for L2 geometric correc-
tion (georectification). The 9-terabyte dataset is
split between the Centers, reducing the demand
on any one Center. A web-based discovery tool
provides access to the entire dataset regardless of
its location. A synchronized database keeps track
936
of each image’s location as well as usage metrics.
This system is in beta test at this writing and is
scheduled to go live later this year.
CONCLUSION
The Columbia Regional Geospatial Service Center
System model grew from concepts developed
many years ago and refined by practical experi-
ence with natural and man-made disasters. The
principle components of the model are:
• Distributed Centers
• Regional Specialization
• Bottom-up Philosophy
• Dedicated to Bridging the Divide
• Local Data Emphasis
• Linked Together as a System
• Data Warehousing and Replication
• University Based
These components, combined with a three-
phase operational philosophy based on Data,
Applications and Training comprise a proven
approach to distributive geospatial technology.
Over the past five years, the Columbia Center
system has amassed a track record of successful
projects and deployments; expanding the use of
geospatial technology at the local level, while
spanning Texas with System level activities.
Today, the System stands as a proven model for
delivering the promise of geospatial technology
to everyone.
REFERENCES
Blackwell, P., Garner, J., Henderson, S., & Wil-
liams, J. (2003, April). Space Shuttle Columbia
Recovery Efforts: How Internet2 and Geospatial
Technology Made a Difference. Paper presented
at the Internet2 Spring Members Meeting, Ar-
lington, VA.
The Columbia Regional Geospatial Service Center System
Croswell, P. (2009). The GIS Management Hand-
book. Frankfort, KY: Kessey Dewitt Publications.
DeMers, M. (2009). Fundamentals of Geographic
Information Systems. New York: John Wiley &
Sons, Inc.
Farmer, K., Kruger, J., & Murphy, L. (2010, Feb-
ruary). Hurricane Ike Storm Surge in Chambers
and Jefferson Counties. Poster session presented
at the Annual East Texas Geospatial Conference,
Nacogdoches, TX.
Garey, G. (n.d.). Texas Height Modernization
Project. Retrieved February 11, 2010, from http://
tsrc.cbi.tamucc.edu/THM/HomePage
Gehman, H. (2003). Columbia Accident Investi-
gation Board Report Volume II (CAiB PA 42-43).
Washington, DC: Government Printing Office.
Kroll, J. C., McDonald, D. L., & Blackwell, P.
R. (2006). The Columbia Regional Geospatial
Service Center System: A Model for Rural Geo-
spatial Support for the Nation. In Proceedings of
the 5th Southern Forestry and Natural Resources
GIS Conference (p. 173).
LaNore, S. (n.d.). Tropical Storm Allison: Ex-
treme Texas flood. Retrieved February 11, 2010,
from http://www.examiner.com/x-5182-Dallas-
Weather-Examiner-y2009m6d4-Tropical-Storm-
Allison-Extreme-Texas-Flood-June-2001
McDonald, D. L., Ashton, C., & Blackwell, P. R.
(2009). An Multi-Scalar Open Source Wide Area
Damage Assessment (WADA) Mash-up Applica-
tion Model. In Proceedings of the 2009 Geospatial
Infrastructure Solutions Conference, Tampa, FL.
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 1,
edited by Donald Patrick Albert, pp. 63-74, copyright 2011 by IGI Publishing (an imprint of IGI Global).
McDonald, D. L., Kroll, J. C., & Blackwell, P.
R. (2007). Functionality Issues of Geospatial
Contributions to Emergency Response: The Re-
gional Geospatial Service Center System Model.
In Proceedings of the GITA’s Annual Conference,
San Antonio, TX.
McDonald, D. L., Kroll, J. C., & Blackwell, P. R.
(2008). The Nac-4 Pack: A Proven Set of Essential
Maps for a Successful Emergency Response. In
Proceedings of the GITA Geospatial Infrastructure
Solutions Conference, Seattle, WA.
Mulligan, K., Barbato, L., Warren, A., Van Nice,
C., & Rainwater, L. (2008). A Regional Analysis
of Groundwater depletion for the Ogallala Aquifer
in Texas. Paper presented at the Joint Meeting
of the Geological Society of America, Soil Sci-
ence Society of America, American Society of
Agronomy, Crop Science society of America, Gulf
Coast Association of geological Societies with
the Gulf Coast Section of SEPM, Houston, TX.
National Research Council. (2007). Successful
Response Starts with a Map (pp. 78–85). Wash-
ington, DC: The National Academies Press.
Sementelli, A., & McDonald, D. L. (2003). A
Geographic Information System-Guided Cost-
Effective Analysis for Waterborne Asbestos Reme-
diation. Public Works Management & Policy, 7(3).
Spraggins, M. (2009). TXSG Public Affairs. “This
Is An Exercise... This Is An Exercise” - Disaster
Assessment Exercise. Retrieved February 11, 2010,
from http://www.txsg.state.tx.us/news/article.
aspx?id=20090302
937
938

Chapter 58
Mapping the Distribution of
Tsetse Flies in Eastern Uganda:
A Geoinformatics Approach

Teddy Nakato Ayanlade Ayansina

Ministry of Agriculture, Uganda Obafemi Awolowo University, Nigeria

O. O. Jegede V. F. Olaleye
Obafemi Awolowo University, Nigeria Obafemi Awolowo University, Nigeria
Bolarin Olufemi
University of Ilorin Nigeria

ABSTRACT
This study demonstrates the ability of GIS and Remote Sensing in capturing spatial-temporal data on
land use and land cover classes. The nine land cover classes captured were Built-up area, Secondary
forests, Savannah, Grasslands, and Shrublands containing herbaceous, Rain-fed shrub crops, Fresh water
swamps, Water bodies, and Farmlands. The remote sensed imageries also displayed how the land use
and land cover classes changed between 1986 and 2001, while helping to identify the suitability of the
land cover classes for tsetse fly habitation. In this paper, the authors demonstrate that GIS and remote
sensing coupled with statistical analyses could help immensely in mapping tsetse habitats. Results show
that the tsetse fly habitat area in Eastern Uganda has been decreasing with time due to the increase in
the Savannah and grassland land cover types and urbanization.

1. INTRODUCTION for increased understanding of disease dynamics

Eastern Uganda is one of regions in Africa that have
the highest cases of sleeping sickness recorded.
The continued spread and shifting of the sleeping
sickness distribution within Uganda and the po-
tential for sub-species overlap highlight the need
DOI: 10.4018/978-1-4666-2038-4.ch058
and factors driving transmission. The continued
spread and shifting of the sleeping sickness dis-
tribution within Uganda and the potential for sub-
species overlap highlight the need for increased
understanding of disease dynamics and factors
driving transmission. Previous studies focused
on infection of the disease using parametric and

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Mapping the Distribution of Tsetse Flies in Eastern Uganda
non-parametric technique for their analysis. The
prominent among these are the studies by Ford
(1969), Buyst (1977), Wellde (1989), Khonde
(1995), and Leak (1999). Recently, evidence of dis-
ease spread to districts considered to be previously
uninfected (Fèvre, 2001) has highlighted the need
for increased understanding and implementation
of disease prevention and control. This study aims,
therefore, at using remote sensing and GIS tech-
niques to examine and map the spatio-temporal
landcover dynamics and their relationship with
tsetse fly occurrence in Uganda between 1986
and 2001. Satellite products were used for the
land cover analysis with the aim of assessing the
influence of land cover types on the distribution
of tsetse flies. To achieve the above aim the study
was divided into three phases: (1) satellite data of
1986 and 2001 were used to provide information
on land cover types and thereby identification of
the tsetse habitat; (2) the spatial distribution of a
vector-borne disease were related to the habitat
of the tsetse flies as the vectors; and (3) data on
land cover types, habitat and human population
were analysis to provide information on the spatial
distribution of sleeping sickness and Nagana in
livestock (Curran et al., 2000).
2. STUDY AREA
Uganda is in latitudes 40.0’ North and 10.30’ South
of the equator, and longitudes 300.0’ East and 350.0’
East of Greenwich (Figure 1). The greater part
of Uganda consists of plateaus which are about
800 to 2,000 m (2,600–6,600 ft) in height. Along
the western border, is the Rwenzori Mountains,
Margherita Peak reaches a height of 5,109 m
(16,762 ft), while on the eastern frontier Mount
Elgon rises to 4,321 m (14,178 ft). By contrast,
the Western Rift Valley, which runs from north to
south through the western half of the country, is
below 910 m (3,000 ft). For example, the surface
of Lake Edward, Lake George and Lake Albert (L.
Mobutu Sese Seko) is about 621 m (2,036 ft). The
White Nile has its source in Lake Victoria and as
the Victoria Nile, it runs northward through Lake
Kyoga and then westward to Lake Albert, from
which it emerges as the Albert Nile to resume
its northward course to the Sudan. (Advameg,
2007), Uganda has a typically tropical climate
with little variation in temperature throughout the
year. Distinctive wet and dry seasons characterize
the climate of most of the country, except in the
semi-arid north east. The country’s natural envi-
ronment provided good grazing for cattle, sheep,
and goats, with indigenous breeds dominating
most livestock. Smallholder farmers owned about
95 percent of all cattle, although several hundred
modern commercial ranches were established dur-
ing the 1960s and early 1970s in areas that had
been cleared of tsetse-fly infestation.
3. MATERIALS AND METHODS
Raw data on livestock populations in Uganda was
obtained from the statistical abstracts and back-
ground to the Budget (several years), Ministry of
Finance, Planning and Economic Development;
Small Ruminant Development Study Report,
Ministry of Agriculture, Animal Industry and
Fisheries in Uganda. The data cover the periods
Figure 1. Map of Uganda showing eastern Uganda
939
 Mapping the Distribution of Tsetse Flies in Eastern Uganda
from 1986 to 2001. Satellite products used for
this study include; Land-sat ETM 1986 and 2001.
Image preparation involved the image registration.
To ensure proper image overlay in the GIS envi-
ronment. The land-sat ETM image of 2001 was
formally geo-referenced in projection, Universal
Transverse Mercator (UTM), Zone 36. Arc 1960
and Clarke 1880, were used to geo-reference im-
ages. The standard processes for the analyses of
satellite imagery such as extraction, restoration,
classification, and enhancement were used for the
study. The classification of land-use/Land-cover
were aggregated to 5 classes which include: Sec-
ondary Forests (Agro-forestry), farmlands (shrubs/
fallow), built-up (residential and commercial),
Bare rocks and bare soils (exposed rocks) and
water bodies (ponds, lakes, rivers, streams).
The comparison of the land use/land cover
statistics assisted in identifying the percentage
change, trend and rate of change in vegetation
cover for the period of time (Figure 2). To achieve
Figure 2. Entity relationship model for linking remote sensing, land cover and disease
940
this, the first task was to develop a table showing
the area in hectares and the percentage change for
each year measured against each land use land
cover type. Percentage change to determine the
trend of change can then be calculated by dividing
observed change by sum of changes multiplied
by 100.
(Trend) Percentage Change = Observed change
X 100
Sum of change
4. RESULTS AND DISCUSSION
The land-sat products showed that Eastern Uganda
is covered by built up areas, secondary forests,
savannah, grassland, shrub lands with herbaceous
vegetation, rain-fed shrub crops, freshwater
swamps and water bodies (Table 1 and Table 2).
In 1986, the total area covered by land-sat product
Mapping the Distribution of Tsetse Flies in Eastern Uganda

(the product used for the study) was 9,242,069.841
ha compared to 9,242,069.841 ha in 2001. Irrespec-
tive of the year of data collection, shrub-land with
herbaceous vegetation cover the most extensive
(about 31% in 1986 and 28% in 2001) in Eastern
Uganda. The rain-fed shrub croplands also cover
lager part of the country from the analysis. In 1986
for instance, the area covered by rain-fed shrub
cropland was 1,143,118.44 ha (which is about
13% of the total area) compared to 1,758,760.59
ha (which is about 20% of the total area) of the
land cover in 2001. The area covered by water
bodies was also very extensive both in 1986 and
2001 with 1,846,069.50 ha (which is about 19%
of the total area) and 2,048,790.63 ha (22% of the
total area) covered respectively (Table 1, Table
2, and Table 3).
The freshwater swamps which are dominant
tsetse fly habitats covered 793,783.60 ha (9% of
the total area) and 841,035.49 ha (9% of the total

Table 1. Areas covered by each land cover type - 1986

Eastern Uganda – 1986

Class Names Area (ha) Proportion (%) Suitability for Tsetse flies
Built up Area 675979.0281 7.314151914 *
Cloud cover 155541.8862 1.682976745 *
Secondary Forests 520476.0624 5.631596292 ***
Savannah 791362.7424 8.562613744 *
Grassland 469918.3329 5.084557258 *
Rain-fed Shrubs 2845820.255 30.79202283 **
Herbaceous Crops 1143118.437 12.36864097 **
Fresh Water Swamps 793783.5975 8.588807607 ***
Water Bodies 1846069.499 19.97463264 *
Total 9242069.841 100
*** - Very Suitable
** - Suitable
* - Not Suitable

Table 2. Areas covered by each land cover type - 2001

Eastern Uganda – 2001

Class Names Area (ha) Proportion (%) Suitability for Tsetse flies
Built up Area 383873.1044 4.153540397 *
Secondary Forests 31483.4237 0.340653384 ***
Savannah 561449.3293 6.074930605 *
Grassland 117129.2658 1.26734885 *
Rain-fed Shrubs 2587986.004 28.00223379 **
Herbaceous Crops 1758760.593 19.02994268 **
Fresh Water Swamps 841035.4902 9.100077198 ***
Water Bodies 2048790.628 22.1680929 *
Farm Lands 911562.0025 9.863180199 **
Total 9242069.841 100

941
 Mapping the Distribution of Tsetse Flies in Eastern Uganda

Table 3. Relative change in land cover types

The tsetse suitability for each one of the ag-
Land Cover Types Relative Change gregated land cover classes identified for Eastern
Area (ha) Proportion
Uganda is described in the section. The land
(%) cover suitability for tsetse habitation of Eastern
Built up Area -292105.9237 2.362452841 Uganda is summarized in Table 4. It is worth
Secondary Forests 31483.4237 0.340653384 noting that the degree of suitability was assigned
Savannah 40973.2669 5.132039645 according to the inherent features of the land
Grassland -674233.4766 -5.604624361 cover class only.
Shrublands with Herba- Figure 3 shows the land cover type area distri-
ceous 2118067.671 27.81754874 bution for Eastern Uganda in 1986 and 2001. The
Rain-fed Shrub Crops -1087059.662 17.38416177 areas covered by built-up areas, farmland, savanna,
Fresh Water Swamps -302082.9468 7.717631167 shrub land with herbaceous cover, rain-fed shrub
Water Bodies 1255007.031 21.77402068 land and water bodies were more extensive in
Farm Lands -934507.4965 7.803341979 1986 than in 2001. Conversely, the grassland,
Total 155541.887 84.72722585 secondary forests, and freshwater swamps were
more extensive in 2001 than in 1986.
As shown in Figure 3, there was generally
area) in 1986 and 2001 imageries respectively. slight gradual decrease in the tsetse habitats as
The savannah areas during the period of study of the Shrub lands tend to decrease with an increase
Eastern Uganda covered 791362.74 ha (9% of the in the grasslands land cover type between 1986
total area) and 561449.32 ha (6% of the total area) and 2001. Also there was a decrease in the shrub
in 1986 and 2001 imageries respectively. The built lands with herbaceous cover and rain-fed shrub
up areas which adversely affects tsetse fly habitats crops which are good habitat to tsetse flies. Con-
covered 675,979.03 ha (7% of the total area) and
383,873.10 ha (4% of the total area) of Eastern Table 4. Tsetse suitability for land cover classes
Uganda respectively (Table 1 and Table 2). in eastern Uganda
Analysis (Table 3 and Table 4) showed that
Shrub land with herbaceous cover is moderately Land Cover Types Tsetse fly Species
suitable for the morsitans group and is less suitable fusca palpalis morsitans
for the palpalis group, covers the largest portion Built up Area 0 0 0
of between 28% and 30% of the study area. The Secondary Forests 3 3 2
principal habitat of these forest flies (FAO, 1982) Savannah 0 1 0
is clearly confirmed by the analysis. Similarly, for Grassland 0 0 0
the palpalis group, the single most relevant land Shrub lands with Herbaceous 0 1 2
cover class is the shrub lands with herbaceous Rain-fed Shrub Crops 1 2 1
cover, which accounts for almost 30% percent Fresh Water Swamps 3 3 2
of the distribution. Irrespective of the year of Water Bodies 0 0 0
imagery collection, from the results over 60% of Farm Lands 1 2 1
the area in Eastern Uganda is a potential habitat Tsetse suitability
for tsetse flies. Only 20% of Eastern Uganda
3 High
covered by water bodies is not suitable for tsetse
2 Moderate
fly inhabitation.
1 Low
0 Unsuitable

942
Mapping the Distribution of Tsetse Flies in Eastern Uganda
Figure 3. Land cover type area distribution for eastern Uganda, 1986, 2001
versely, there was an increase in the grassland
which are unfavorable for tsetse habitants between
1986 and 2001. There was an increase in the fresh
water swamps and a decrease in the water bodies
because of the increased growth of fresh water
weeds and plants which are known to be moder-
ately favorable for tsetse habitats.
4.2 Description of the
Land Cover Types
The land cover map of Eastern Uganda in 1986
(Figure 4) and in 2001 (Figure 5) showed the dif-
ferent land cover types the descriptions of which
are given below:
These are continuous closed trees. The main
secondary forest layer consists of closed trees
with crown cover of more than 60 percent. The
vegetation is spread over the area without intervals
or breaks. In Eastern Uganda, most of the areas
in this class are covered by broad-leaved evergreen
trees with a second layer of trees that form a dif-
ferent stratum due to a difference in height and a
third layer of emergent trees higher than the main
stratum. Tsetse suitability: Forests provide favor-
able habitats to several tsetse fly species of the
fusca and palpalis groups and, to a lesser extent,
of morsitans group. With one exception (G. lon-
gipennis), the species of the fusca group are for-
est flies inhabiting either rain forest or isolated
patches of forest, along with riverine forest in the
savannah zones. Gallery forests are the typical
habitat for the flies of the palpalis group. Species
of the morsitans group can be found in forest
edges, forest islands and in riverine forests. (Veg-
etation areas not used by G. morsitans include
very high rainfall areas such as rain forests.) (FAO,
1982). The main shrub land with herbaceous
layer consists of shrubs with crown cover between
15% and 95%. The height is in the ranges from
0.3 to 5 m. In Eastern Uganda, most of the areas
of this class have an open cover (between 20%
and 60%), a second layer of closed to open her-
baceous vegetation and a third layer of emergent
trees and it covers an area of around 35 000 km2
(16 percent of the total surface of the country).
Tsetse suitability: This habitat differs from the
classic savannah only for the presence of the main
shrub layer. Thus, we can assume that it is mod-
erately suitable for the morsitans group and less
so for the palpalis group. This class is deemed
unsuitable for the fusca group (FAO, 1982).
Rain-fed shrub croplands are permanently
cropped area with rain-fed shrub crop(s). The
943
 Mapping the Distribution of Tsetse Flies in Eastern Uganda
Figure 4. Land cover map of eastern Uganda –
1986
crop cover includes orchard(s) and shrub crops
such as plantains, coffee, cotton, bananas, tea,
cocoa and pineapple. In terms of harvested area,
the most widespread crops of this class in Uganda
are plantains, coffee, cotton and bananas (FAO,
2005). The semi-artificial habitats consisting of
various kinds of orchards or other type of planta-
tions (such as bananas, cacao, and coffee) could
be typical habitats for some species of the palpalis
group (e.g., G. tachinoides). On the contrary,
some other crops in this class like cotton fields
are unsuitable for all tsetse species (FAO, 1982).
The main layer consists of grassland (crown
cover is more than 15–100 percent); the height is
in the range of 0.03–3 m, the grassland vegetation
is spread over areas without intervals or breaks. In
Uganda this class covers around 6 000 km2 (2.5
percent of the total surface of the country) in 1986
and 2001 land-sat imageries. This research has
944
Figure 5. Land cover map of eastern Uganda –
2001
however shown that this land cover is unsuitable
for tsetse flies (FAO, 1982).
The main savannah layer consists of grassland
(crown cover is between 15 and 100 percent) with
the height in the range of 0.03 to 1 m. The savan-
nah vegetation is spread over the areas covered
without intervals or breaks. The second layer in
the savannah covered areas consists of sparse
trees. The third layer is mainly sparse shrubs. In
Uganda, this class in the 1986 and 2001 land-sat
imageries of vegetation covers more than 20
000 km2 (8.5 percent of the total surface of the
country). The savannah offers moderately suit-
able habitats for species of the morsitans group
and for some of the palpalis group, and is less
suitable for the fusca group. The limited tree and
shrub cover in the savannah areas however, could
be sufficient for many tsetse species during the
Mapping the Distribution of Tsetse Flies in Eastern Uganda
wet season. Usually however, the area is unable
to support tsetse fly populations during the dry
season (FAO, 1982).
The main layer consists of herbaceous veg-
etation on permanently or temporarily flooded
land (cover is between 15 and 100 percent; the
height is in the range of 0.03–3 m). In Uganda,
this class covers approximately 12 000 km2
(more than 5 percent of the total surface of the
country). In around half of this area, a second
layer of sparse shrubs is present. By definition
Freshwater swamps are associated with humid
environments, and cannot be considered a major
habitat for tsetse flies because of the very limited
presence of woody vegetation. The sparse shrubs
occasionally present in these areas could however
provide atypical habitats to some tsetse species,
particularly these of the palpalis group.
These are perennial natural water bodies. Salin-
ity in these water bodies is less than, <1 000 parts
per million [ppm]. The perennial natural water
bodies included flowing or standing water. In the
Figure 6. Land cover map of eastern Uganda - 1986
Africover map of Uganda for 1986 and 2001, this
class occupies 36 000 km2 (more than 15 percent
of the total surface of the country) and includes
the vast expanses of Lakes Victoria, Albert and
Kyoga (Figure 5, Figure 6, Figure 7, and Figure
9). The land cover here however is not suitable
for tsetse flies (FAO, 2005).
Farmlands are permanently cropped area with
rain-fed tree crops (Figure 6 and Figure 9). The
plantations include the fruit trees (e.g., citrus,
mango, palm, etc.), conifers (e.g., pinus spp.,
Cupressus spp.) and hedging and shade plants.
Among the less typical habitats of tsetse flies,
man-made ones are particularly important from
the point of view of disease transmission, in par-
ticular for Human African Trypanosomiasis
(HAT). Tree plantations are arguably the most
suitable man-created habitat for some tsetse spe-
cies. Plantations of mango are breeding sites for
some species of the palpalis group (e.g., G.
tachinoides and G. palpalis); many mango plan-
tations are grown along riversides, which provide
945
 Mapping the Distribution of Tsetse Flies in Eastern Uganda

Figure 7. Land cover map of eastern Uganda - 2001

Figure 8. Map of eastern Uganda showing Tsetse infested areas – 1986

946
Mapping the Distribution of Tsetse Flies in Eastern Uganda
Figure 9. Map of eastern Uganda showing Tsetse infested areas – 2001
tsetse flies with suitable shelter, particularly so in
the case of old trees with low branches. Examples
of other semi-artificial habitats of this class are
plantations of oil palms and cola nuts and tree
hedges. Untrimmed hedges and tree crops can
also provide a suitable habitat for G. pallidipes
(morsitans group).
Analysis of land cover imageries of Eastern
Uganda (1986) is shown in Figure 6 while the Map
of Eastern Uganda showing the tsetse infested
area is shown in Figure 7. The land cover map
of Eastern Uganda for 2001 is shown in Figure
8 while the Map of Eastern Uganda of the same
area is shown in Figure 9 for 1986. Figure 7 and
Figure 9 showed that there was a decrease in the
tsetse infested areas due to the expansion of the
savannah and grassland land cover types which are
known not to be favorable tsetse habitants. Also,
urbanization that is which increased in the built-
up areas are known not to favor tsetse habitants.
4.3 Analysis of Areas of Different
Vegetative Vigor - NDVI
From Figure 10 and Figure 11, we can clearly see
the different colors for the different vegetative
vigor which are between 1 and -1. Areas of high
vegetative vigor are represented in a pick shade
while areas of low vegetative vigor are indicated
by the dark bluish shade. Areas of low vegetative
vigor are the water bodies in this area showing
that there is no vegetation in these areas and Areas
of high vegetative vigor are area covered with
secondary forests.
From the levels of vegetative vigor, the high-
est areas in 1986 had NDVI of 0.93 and those of
lowest vigor had NDVI of -0.58 while those of
2001 had NDVI of 0.61 as the highest vegetative
vigor and -0.91 as areas of the lowest vegetative
vigor. This shows that there is a very high loss in
vegetative vigor of the vegetation from 1986 to
2001 showing that this area is rapidly tending to
aridity. This difference is not due to the changes
947
 Mapping the Distribution of Tsetse Flies in Eastern Uganda
Figure 10. Map of eastern Uganda showing areas
of different vegetative vigor – NDVI 1986
in climatic seasons because all the images were
taken between December and March which is a
dry season.
4.4 Validation of Land Cover Classes
for Suitability of Tsetse Habitats
1. Rigorous validation of the estimated suit-
ability for tsetse of the land cover classes
defined in Table 4 and 5 and clearly shown
in Figure 6 and Figure 8 was hindered by a
range of practical and conceptual difficulties.
Foremost among the conceptual problems
is the fact that land cover vegetation is only
one component of potential determinant of
the tsetse habitat. Favorable environmental
conditions must include availability of
hosts on which to feed on and convenient
climatic conditions. As a consequence,
suitable vegetation can still represent an
948
Figure 11. Map of eastern Uganda showing areas
of different vegetative vigor – NDVI 2001
unsuitable habitat because of inconclusive
environmental conditions.
2. A second difficulty encountered was re-
lated to the challenge of defining classes of
suitability valid for the study area. Ideally
validation should be based on a number of
sites capable of encompassing the enormous
environmental heterogeneities in Uganda
and Africa at large.
3. Another problem was related to the intra-
group differences in habitat requirements. If
the definitions of forest, riverine and savan-
nah flies for the three groups (fusca, palpalis
and morsitans) are commonly accepted, it is
also true that the ecology and spatial distribu-
tion of the species within each group vary
considerably (e.g., G. longipennis, belonging
to the fusca group, and G. tachinoides, of
the palpalis group, are found in more arid
Mapping the Distribution of Tsetse Flies in Eastern Uganda
environments than the other species in the
same groups).
4. Another complicating factor is the dispersal
of flies, some of which can easily travel hun-
dreds of meters away from their resting and
breeding sites for feeding. It was therefore
difficulty to link trap catches (i.e., apparent
densities) to the vegetation in the immediate
surroundings of the trap; trap catches are
in most cases influenced by the vegetation
mosaic at landscape level.
In other words, it is almost impossible to define
experimental conditions that comply with the
hypothesis of ‘indefinite expanse of one homoge-
neous land cover’ on which the present evaluation
was based. Furthermore, the opportunity to move
in less shaded and less protected environments
is heavily influenced by thermal and humidity
gradients linked to seasonality. The above consid-
erations explained why very broad and qualitative
suitability classes were employed in the project.
5. CONCLUSION
This study demonstrates the ability of Remote
Sensing and GIS in capturing spatial-temporal
data and attempt was made to capture as ac-
curate as possible nine land use and land cover
classes as they change through time and identi-
fying their suitability for tsetse fly habitation.
The study reveals GIS and remote sensing help
immensely in spatial and temporal analyses for
mapping tsetse fly habitats. Eastern Uganda is
one of such regions that have the highest cases
of sleeping sickness recorded from 2000 to 2007.
The concerns over re-current foci and continued
disease spread highlight the absence of adequate
assessment of the differences between current
and historical distributions of sleeping sickness.
The continued spread and shifting of the sleeping
sickness distribution within Uganda and the po-
tential for sub-species overlap highlight the need
for increased understanding of disease dynamics
and factors driving transmission. The land cover
maps produced can be used in several stages of
tsetse and Trypanosoma intervention: mapping
vector habitat, planning baseline entomological
surveys, monitoring the efficacy of tsetse sup-
pression, land-use planning of reclaimed areas
and monitoring the environmental impacts of
intervention strategies.
The results presented here are of a preliminary
nature, but clearly show the capacity of remote
sensing and geographic information systems to
make significant contributions to country, regional
and continental-level analyses of factors affect-
ing tsetse distribution, human population density
and land-use intensity. It is considered that such
analyses will play an increasing role in the future
in the process of resource allocation to improved
food production in the continent through more
effective disease control.
ACKNOWLEDGMENT
The Authors thank the African Regional Centre
for Space Science and Technology Education
in English (ARCSSTE-E), Obafemi Awolowo
University (OAU) Campus, Ile-Ife, Nigeria, for
the scientific contribution to training activities
and fellowship grant support.
REFERENCES
Buyst, H. (1977). The epidemiology of sleeping
sickness in the historical Luangwa valley. An-
nales de la Société Belge de Medecine Tropicale,
57, 349–359.
FAO. (1982). In J. N. Pollock (Ed.).Training
manual for Tsetse control personnel. Vol. 2 Ecol-
ogy and behaviour of tsetse. Retrieved from ftp://
ftp.fao.org/docrep/fao/009/p5444e/p5444e00.pdf
949
 Mapping the Distribution of Tsetse Flies in Eastern Uganda
FAO. (2005). Land Cover Classification Sys-
tem – Classification concepts and user manual
Software version 2. Rome: FAO. Retrieved from
www.glcn-lccs.org/
Feldmann, U., Dyck, V. A., Mattioli, R. C., &
Jannin, J. (2006). Potential impact of tsetse fly
control involving the Sterile Insect Technique
(Vol. 37, pp. 1-723).
Fèvre, E. M., Coleman, P. G., Odiit, M., Magona,
J. W., Welburn, S. C., & Woolhouse, M. E. J.
(2001). The origins of a new Trypanosoma brucei
rhodesiense sleeping sickness outbreak in eastern
Uganda. Lancet, 358, 625–628. doi:10.1016/
S0140-6736(01)05778-6
Ford, J. (1969). The control of the African try-
panosomiases with special reference to land use.
Bulletin of the World Health Organization, 40,
879–982.
Ford, J., & Katondo, K. M. (1975). Revision of
the Glossina distribution map of Africa. In Pro-
ceedings of the 14th of the International Scien-
tific Council for Trypanosomiasis Research and
Control, Dakar, Senegal (No. 109, pp. 23-25).
Ford, J., & Katondo, K. M. (1977a). The distri-
bution of tsetse flies in Africa. Nairobi, Kenya:
Nairobi Organization of African Unity.
Ford, J., & Katondo, K. M. (1977b). Maps of
tsetse flies (Glossina) distribution in Africa, 1973
according to sub-generic groups on scale of 1:5
000 000. Bulletin of Animal Health and Produc-
tion in Africa, 25, 187–193.
Hutchinson, O. C., Fèvre, E. M., Carrington, M.,
& Welburn, S. C. (2003). Lessons learned from
the emergence of a new Trypanosoma brucei rho-
desiense sleeping sickness focus in Uganda. The
Lancet Infectious Diseases, 3, 42–45..doi:10.1016/
S1473-3099(03)00488-2
950
Katondo, K. M. (1984). Revision of second edi-
tion of tsetse distribution maps. Insect Science
and Its Application, 5(5), 381–384.
Khonde, N., Pépin, J., Niyonsenga, T., Milord,
F., & Wals, P. (1995). Epidemiological evidence
for immunity following Trypanosoma brucei
gambiense sleeping sickness. Transactions
of the Royal Society of Tropical Medicine and
Hygiene, 89(6), 607–611. doi:10.1016/0035-
9203(95)90408-5
Leak, S. G. A. (1999). Tsetse Biology and Ecology:
Their Role in the Epidemiology and Control of
Trypanosomosis. Wallingford, UK: CABI.
Ministry of Health. (2004). Uganda sleeping sick-
ness report. Uganda: Ministry of Health.
Paul, J., Curran, P., Atkinson, M., Foody, G. M., &
Milton, E. J. (2000). Linking remote sensing, land
cover and disease. Southampton, UK: University
of Southampton, Department of Geography.
Rodriguez, A. (2005). Trypanosomiasis - Ugan-
da (Kaberamaido). ProMed. Retrieved from
http://www.pubmedcentral.nih.gov/redirect3.
cgi?andandauth=001Z7ISWIFesRlR-eJDdjJ-
hjEG-wI-DVZ_vnK8AJtandreftype=extlinkanda
rtid=1832067andiid=143254andjid=378andFRO
M=Article%7CCitationRefandTO=External%7C
Link%7CURIandarticle-id=1832067andjournal-
id=378andrendering-type=normalandandwww.
promed.org
Rogers, D. J., & Randolph, S. E. (1993). Distribu-
tion of tsetse and ticks in Africa: past, present and
future. Parasitology Today (Personal Ed.), 9(7),
266–271. doi:10.1016/0169-4758(93)90074-P
Rogers, D. J., & Williams, B. G. (1992). Monitor-
ing trypanosomiasis in space and time. Parastio-
log, 106, S77–S92.
Mapping the Distribution of Tsetse Flies in Eastern Uganda

Welburn, S. C., & Odiit, M. (2002). Recent de- Wellde, B. T., Chumo, D. A., Waema, D., Reardon,
velopments in human African trypanosomiasis. M. J., & Smith, D. H. (1989). A history of sleeping
Current Opinion in Infectious Diseases, 15(5), sickness in Kenya. Annals of Tropical Medicine
477–484. and Parasitology, 83(1), 1–11.

This work was previously published in the International Journal of ICT Research and Development in Africa (IJICTRDA),
Volume 1, Issue 2, edited by Blessing Maumbe, pp. 19-34, copyright 2010 by IGI Publishing (an imprint of IGI Global).

951
952

Chapter 59
Open Source Based
Deployment of Environmental
Data into Geospatial
Information Infrastructures
José Gil
Institute of New Imaging Technologies, Universitat Jaume I, Spain

Laura Díaz
Institute of New Imaging Technologies, Universitat Jaume I, Spain

Carlos Granell
Institute for Environment and Sustainability, Italy

Joaquín Huerta
Institute of New Imaging Technologies, Universitat Jaume I, Spain

ABSTRACT
Today, scientists use local and closed geospatial solutions to run their models and store their results. This
may limit their ability to share their models, and results with other interested colleagues. This scenario is
changing with the advent of new factors such as the rapid growth and rise of open source projects, or new
paradigms promoted by government organizations to manage environmental data, such as Infrastructure
for Spatial Information in the European Community (INSPIRE) directive, or the massive use of Web
2.0 techniques where users are looking for applications with a high degree of collaboration, interac-
tiveness, and multimedia effects. Many authors address the versatility of Spatial Data Infrastructures
where resources are shared and accessed via standard service according to complex specifications. In
this context, the authors point out the need to merge the traditional building and maintenance of these
infrastructures, driven by official providers, with these more participative methodologies where users
can participate in creating and integrating information. It seems necessary to develop new geospatial
tools which integrate these new trends. This paper proposes a unified solution offering to the scientific
field an open development framework, based on standards and philosophies focused on new technolo-
gies and scientific needs.

DOI: 10.4018/978-1-4666-2038-4.ch059

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
INTRODUCTION
One of the challenges we are facing today is to
better understand the processes that occur over
or near-surface of our planet (Goodchild, 2008).
The Earth is a multi-dimensional system made of
complex interactions highly interconnected and
continuously evolving at many spatial and tem-
poral scales (GEO Secretariat, 2007). This means
that to understand such interactions, scientists and
environmental experts need to collect and integrate
different data sets referred to physical and biologi-
cal aspects of the Earth (Giuliani et al., 2011). In
addition, these data sets are often georeferenced,
that is, they describe physical phenomena tied to a
concrete geographical location in terms of points,
lines and sets of polygons, and we refer to them
as geospatial data.
Scientists and environmental experts need to
manage vast amounts of these geospatial data to
turn them into meaningful information. The ap-
plicability of complex environmental processes
and models has been the focus of intense research
in this domain (Poch, Comas, Rodríguez-Roda,
Sánchez-Marrè, & Cortés, 2004; Minsker et al.,
2006; Reichert & Schuwirth, 2010). For instance,
models for reduction of energy consumption, com-
putation of atmospheric emissions, and simulation
of forest fire require multidisciplinary teams of
geospatial technologists and environmental ex-
perts. Such models, however, have been tradition-
ally performed using multiple, isolated desktop
geospatial tools: a specialist in Digital Elevation
Models might perform slope computations while
other colleague might apply remote sensing
techniques over satellite imaginary to produce
a snow cover product. In this simple example,
various GIS software packages could be involved
in the activities of a single environmental model.
However, this paradigm of closed solutions, which
operates with local data sets, provides a way of
hiding shared results and collaborative research,
and more often, hindering further analysis tasks.
Better decisions and concrete actions at local
and global scale to underpin sustainable develop-
ment can be taken only based on precise knowledge
of our environment (Lannotta, 2007). However,
geospatial datasets, and other resources are often
isolated, hidden to potential interesting stakehold-
ers that could benefit from them. Therefore, the
ease of access and retrieval of geospatial and
environmental data is a key aspect for information
infrastructures (Yang, Raskin, Goodchild, & Gahe-
gan, 2010). There are consensus-based initiatives
which aim is to promote interoperability, standards
specifications, and seamless integration of services
and information to improve this accessibility to
geospatial resources (see next sections). However,
the participation of users in these infrastructures
has been delayed due to the complexity of cur-
rent methodologies to participate and integrate
new resources; the necessity of skills about how
to incorporate real time data with historical, the
use of standard services specification to deploy
resources and promote interoperable resources,
the integration of data from different domains, or
simply looking for these data sources have become
tedious tasks in current developments (Yang &
Raskin, 2009). This makes scientists continue
using traditional methods like local processing.
The aim of this article is to present an open
source-based framework that enables both:
standard access and processing of data, and also
the deployment of new content into geospatial
information infrastructures. The processing and
access of environmental data sets and the deploy-
ment and sharing of processing results will be
addressed in this paper. All this, in the simplest
way possible, through an immersive and user-
friendly interface, based entirely on geospatial
open source technologies.
The rest of this article is structured as fol-
lows. The following three sections outline the
basic concepts underlying the topics tackled in
this article: architectural styles for information
infrastructures, concrete realizations of geospatial
information infrastructures, and the concept of
953
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
open source applied to geospatial technologies.
Given the pieces of the puzzle, we present a
toolbox that tries to connect all pieces to facili-
tate the sharing, deployment and publication of
new geospatial and environmental data sets into
existing geospatial information infrastructures.
We discuss the applicability of our framework
through a use case in the forest fire domain. The
paper concludes by looking to futures challenges
in automatic deployment of user-generated content
and identifies a number of issues that are in need
of further research.
DESIGNING INFORMATION
INFRASTRUCTURES
Geospatial and environmental research is a field
that involves multidisciplinary teams of scientists
and technologists working together. The applica-
tion of environmental models may span from
local to global scope, what undeniably requires
suitable information infrastructures to share and
do research on the web (Hey & Trefethen, 2005).
This complex scenario demands the use of new
architectural styles which oppose centralized, iso-
lated solutions, and instead support distributed pro-
cessing capabilities and remote communications,
necessary ingredients to successful collaborative
and multidisciplinary research. Service-Oriented
Architectures (SOA) and Resource-Oriented Ar-
chitectures (ROA) are currently the architectural
styles adopted in the development of collaborative,
distributed Web systems and applications.
SOA is an architectural style to design ap-
plications based on a collection of best practices,
principles, interfaces, and patterns related to the
central concept of service (Bell, 2008). In SOA,
services play a key role and become the basic
computing unit to support development and com-
position of larger, more complex services, which
in turn can be used to create flexible, ad hoc and
dynamic applications. The main design principle
behind SOA is that a service is a standards-based,
954
loosely-coupled unit composed of a service inter-
face and a service implementation. The service
interface describes the functional capabilities of a
service. The service implementation codifies what
a service should execute. This principle provides
a clean separation of concerns especially between
service interfaces (what services offer to the public
community) and internal implementations (how
services work). Essentially SOA introduces a new
philosophy for building distributed applications,
where services can be discovered, aggregated,
published, reused, and invoked at the interface
level, independently of the specific technology
used internally to implement each service.
At the time of implementation SOA-based
services must make use of concrete languages
and protocols. Here is where web service technol-
ogy gains importance because it increasingly is
becoming the choice to implement SOA-based
applications. Web services (Papazoglou, 2008)
are, by definition, loosely coupled independent
units and are well described (interface description
contains functional properties), thereby promot-
ing one of the goals of SOA: enabling interoper-
ability or the ability of services to interact with
minimal knowledge of the underlying structure
of other services (Goodchild, Egenhofer, Fegeas,
& Kottman, 1999). Interoperability is achieved
(or optimized) by using standard interfaces. Web
service technology includes various standards such
as Web Service Description Language (WSDL)
for the description of service interfaces, Universal
Description, Discovery and Integration registry
(UDDI) for their advertisement and discovery, and
Simple Object Application Protocol (SOAP) that
enables communication among services (Curbera,
Duftler, Khalaf, Nagy, Mukhi, & Weerawarana,
2002).
As opposed to SOA, focused on distributed
capabilities and services, ROA is an architectural
style devoted to manage distributed, heteroge-
neous domain resources. In SOA client applica-
tions interacts with a distributed Web applications
through delegation, that is, by specifying the
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
desired capability to a service component instead
of directly acting on the resources themselves
(Mazzetti, Nativi, & Caron, 2009). In ROA,
through, client applications interact directly with
the exposed resources.
The main constraints behind ROA-based ap-
plications are the set of architectural principles
known as Representational State Transfer (REST)
(Fielding, 2000), namely:
• Resources should be identified properly
using global identifiers (URI). However,
REST proposes to use URIs not only for
identification but also for physical access
to the actual resource representation. One
important feature is that each resource
must be addressable via a HTTP (Hypertext
Transfer Protocol) URI.
• Uniform interface through the use of
HTTP as the unique application-level pro-
tocol. HTTP has a small, fixes of opera-
tional methods with specific purpose and
meaning. For instance, the GET method is
for retrieving representations of target re-
sources, the POST method for creating new
resources, PUT for updating resources, and
finally, the DELETE method to eliminate a
given resource.
• Resources are manipulated through rep-
resentations since clients and servers ex-
change self-descriptive messages each
another.
• Interaction stateless since servers only re-
cord and manage the state of their resourc-
es they exposes. Client sessions are not
maintained in the server.
• Hypermedia as the engine of application
state, that is the application state is built
following hyperlinks according to the nav-
igation paradigm so that clients can dis-
cover related resources from a given one.
As in the case of SOA-based services, ROA-
based applications can be implemented in many
different ways, such as RESTful web services
(Richardson & Ruby, 2007). Designing REST-
ful web services following the REST constraints
described above has gained popularity within the
Earth science, mobile and location based services
community (Mazzetti, Nativi, & Caron, 2009).
The selection of one or another architectural
style depends often on the context of the target
application. Both approaches should not be dis-
joint but complementary. In this paper, we take
the complementary strategy for the proposed
framework implementation that makes use of
both architectural patterns to communicate with
remote geospatial services. Accessing to common
geospatial services, as we describe in the follow-
ing, are built upon a SOA approach, while recent
open source implementations already expose
RESTful interfaces to manipulate and deploy
remote geospatial resources.
GEOSPATIAL INFORMATION
INFRASTRUCTURES
Several administrations are recognizing the role of
geospatial data in e-commerce, sustainable devel-
opment and government activities, and the ability
to be efficiently coordinated and managed for the
interest of public in general (Rajabifard, Binns,
Masser, & Williamson, 2006). In consequence, the
notion of Spatial Data Infrastructures (SDI) refers
to the specialization of information infrastructures
for the geospatial sciences, allowing discovering,
access, publishing, sharing, maintaining, and
integrating geospatial data (Nebert, 2004).
Many initiatives at different administrative
levels have been recently flourishing to promote
the creation of SDIs (Masser, 2005). These
initiatives coordinate actions and policies that
encourage awareness of institutional agreements,
common standards and effective mechanisms for
data access, harmonization, standardization, and
the development of interoperable geospatial data
and technologies to support decision making for
955
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
multiple purposes. One of these initiatives to elimi-
nate jurisdictional, cultural or domain boundaries
is the Infrastructure for Spatial Information in
the European Community (INSPIRE) Directive
of the European Commission (European Parlia-
ment and Council, 2007). INSPIRE Directive
establishes a legal framework for the creation and
implementation of the European Community SDI,
forcing organizations to publish their geospatial
resources (focusing on the environmental domain)
under standard data services and by promoting the
adaptation of the SOA paradigm and geospatial
web service standards.
In essence, geospatial web services allow users
to access, manage, and process geospatial data
in a service-oriented manner (Granell, Gould, &
Esbri, 2008). The demand for interoperability has
boosted the development of standards and tools
to facilitate data transformation and integration,
mostly in terms of standard service interfaces
specified by Open Geospatial Consortium (OGC1).
The Web Map Service (WMS) (OGC, 2006), the
Web Feature Service (WFS) (OGC, 2005), the
Web Coverage Service (WCS) (OGC, 2008),
the Sensor Observation Services (SOS) (OGC,
2007a) and the Web Processing Services (WPS)
(OGC, 2007b) are some prominent examples of
OGC interfaces for geospatial services with a
clear applicability to environmental applications
(Granell, Díaz, & Gould, 2010). For instance, a
WMS could be used to provide a map of the pro-
tected areas of a given region, a SOS may serve
detailed meteorological observations from near
weather stations, and a WCS may offer satellite
imagery of the area of interests and its surround-
ings. The central building-blocks for data, as
well as service discovery, are provided by the
Catalogue Services for the Web (CSW) (OGC,
2007c). The CSW provides one access point to
users that publish and search environmental data.
In the proposed framework implementation,
we will use a broad list of OGC services (WMS,
WFS, WCS, WPS, and CSW) to let experts to
956
perform not only retrieval but also deployment
of geospatial data sets.
GEOSPATIAL OPEN SOURCE
TECHNOLOGIES
The use of geospatial applications for environ-
mental studies has experienced a significant
boom in the last years (Minsker et al., 2006;
Pezzoli, Marciano, & Robertus, 2006; Morisette
et al., 2009). Its development has led to progress
in making decisions in real time (Ramaprivan,
2008), it has created new relationships between
different areas of knowledge (Ball et al., 2008)
and has developed new processes to understand
the behavior of natural phenomena (Díaz, Granell,
Gould, & Olaya, 2008).
For all these efforts to continue to improve, they
need to be associated with technological advances
that gradually eliminate barriers that scientists
found in their research. This requires advanced
tools (processing, analysis and visualization) and
sophisticated information infrastructures that
provide access to large volumes of data. Given
the wide array of open source technologies for
geospatial data (Steiniger & Bocher, 2009), these
can be a serious solution to address the lack of
mechanisms for democratizing the capacity build-
ing of geospatial information infrastructures.
A key reference in geospatial technologies’
state of the art is the world conference for Free
and Open Source Software for Geospatial (FOS-
S4G2). One can find the latest innovative tools
and developments in the geospatial landscape,
such as OpenGeo Suite or GeoNode, which are
examples of open source technology related to
those proposed in this paper. OpenGeo Suite
(Pumphrey, 2009) is an open source project that
brings together different technologies (PostGIS,
GeoServer, GeoWebCache, GeoExt, OpenLay-
ers) to develop an architecture based on OGC
standards. The main objective is that users can
easily build a platform that provides maps and data
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
via web applications, mobile and desktop clients.
GeoNode provides a new SDI solution by adding
a new application discovery layer so that scientists
can easily discover the data services. The result
is a platform that facilitates the creation, sharing
and collaborative use of geospatial data (Benthall
& Gill, 2010).
Merging our work with other domains not
strictly in touch with geospatial domain, we must
review the take off of many web applications that
following the Web 2.0 philosophy have created
an environment that brings users to participate
very actively and add new content to be accessed
and shared by many users. This way of content
sharing brings new forms of study. Not only with
the incorporation of the social factor but with a
new relationship with the user. Applications are
required to have higher degree of interaction and
attractive interfaces that facilitate complex tasks.
Solutions such as RIA (Rich Internet Applications)
can be a key element in future developments. In
this paper we describe how we propose to merge
these distinct disciplines. Having an underlying
geospatial information infrastructure based on
standard services and components we add on top
a framework composed by mechanisms and Web
2.0 interfaces to assist users in not only accessing
but also deploying and sharing new geospatial
resources in this SDI.
FRAMEWORK FOR DEPLOYMENT
OF ENVIRONMENTAL DATA
In the previous section we have seen numerous
technological trends that can potentially become
the basis of a set of solutions in multiple scenarios.
On one hand, common policies and agreements in
terms of the notion of SDI, understood as geospa-
tial data stores collected by government agencies
and private institutions, where data are published
via open, standards-based services, can be easily
incorporated into scientific applications (Nerbert,
2004). The INSPIRE directive, which creates a
collaborative SDI, accessible, designed to facili-
tate data sharing and standardization of specifica-
tions (European Parliament and Council, 2007),
reinforced by standardization organizations, such
as the Open Geospatial Consortium (OGC) and
International Organization for Standardization
(ISO). On the other hand, the advent of recent Web
2.0 technologies to create interoperable geospatial
web portals (Yang, Li, Xiao, Raskin, & Rambacus,
2007; Wilkins-Diehr, Gannon, Klimeck, Oster, &
Pamidighantam, 2008) and that allows users to
publish and access data in an easy way. The rise
of open-source service-oriented geospatial envi-
ronments and tools, provide a variety of solutions
and knowledge that can be easily extrapolated or
incorporated directly into new scientific applica-
tions (Pierce, Fox, Choi, Guo, Gao, & Ma, 2009).
Why then the diverse geospatial technologies,
open source tools, and geospatial information
infrastructures have not received the same level
interests in environmental applications as in oth-
ers domains? After all, SDI and open source
geospatial technologies presents many benefits
for environmental scientists and experts, since the
intrinsic multidisciplinary aspect in environmental
research can be viewed as a source of expertise
for collaborative applications between different
domains (Westlund, 2010).
From the user’s perspective of environmental
and geospatial sciences, the increasing complexity
of current methodologies, tools and technologies
may lead to underuse such information infrastruc-
tures for efficient analysis and decision-making
tasks. This means scientists who produce interest-
ing results through models and processes often
use information for access, not to publish their
own data sets. This implies in low rates of user
participation in the process of injecting environ-
mental datasets into information infrastructures
for collaborative research.
For this reason, we present an open source
framework based on INSPIRE architecture that has
been extended with some modules for not only ac-
cessing but also deployment of environmental data
957
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
into geospatial information infrastructures (like
SDI). The main characteristics are the following:
• It allows the access and retrieval of envi-
ronmental data by means of interoperable
OGC-based services),
• It allows the integration and processing of
new information by means of deployment
mechanisms,
• It allows the generation and publication of
metadata descriptions in OGC-based cata-
logs for further search.
Figure 1 illustrates the conceptual architecture
of the proposed framework. It basically follows
the INSPIRE technical architecture, that is, three-
layer SOA: the Application layer (top) contains
client side services, application logic and presen-
tation modules; the Service layer (middle) with
server side services that use a standard interface
Figure 1. Multi-layered architecture for the proposed framework
958
for communication; and the Data layer (bottom).
Grey boxes represents the components contained
within the Service Framework. In the following we
outline the main functionality of each layer. Next
section will describe in detail each component in
grey in Figure 1
Application Layer
Our contribution on this layer is to add the Service
Deployer and Publisher that allows user not only
access resources but also deploy new resources in
the SDI and a user interface that following web 2.0
principles offers a entry point to the underlying
SDI that allows users to participate in the building
and maintenance of the SDI.
Following the Model-View-Controller (MVC)
pattern (Freeman, Sierra, & Bates, 2004) to build
the client side, we find the Application layer
is divided into three levels according to their
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
proximity to the user. At the level closest to the
user, we have the visual part of the application,
multimedia effects and interaction with the user.
At the intermediate level, we find the application
logic associated with the controller of the actions
produced by the user. At the lower level we have
client side’s services that allow communication
with the geospatial services deployed at the Ser-
vice layer, through OCG standards and RESTful
technology.
Service Layer
The Service layer is responsible for mediating
between the Application and the Data layer. Its
development focuses on the paradigm INSPIRE/
SDI/SOA to minimize problems of accessibility
and interoperability, so that wherever possible
requests for data on this layer are performed using
OGC standards. However, on certain occasions
such as the insertion of new data, we need to use
a RESTful approach. In this layer reside the in-
stances of the OGC services in which we deploy
the new content that users will provide using
the framework interface. Service Framework is
preconfigured to work with existing instances of
geospatial web services.
Data Layer
In this part we find the final elements for data
storage, such as databases or simply memory disk
regions, where the Service layer can store their
data. It is out of the scope of this paper to describe
further this layer since it will be something man-
aged by the OGC standard service implementa-
tions that we use to deploy new data content and
that reside in the service layer.
OPEN SOURCE BASED
COMPONENTS
As we have seen throughout the paper, there
are many open source projects in the geospatial
domain and in particular many projects that
implement open source components according to
OGC standard specifications. We describe next
the service framework components regarding
the technology used to implement them. For the
Service Framework implementation we have reuse
some of these existing open source components
and also we have self-develop some components
to be offered through an open source license.
The rest of this section describes our solu-
tion from two points of view: interfaces and
technologies used. Figure 2 shows the interfaces
implemented by each component in the Service
Framework. Figure 3 shows the technology that
has been used to implement each of these compo-
nents. Some of these technologies are existing open
source projects and self implemented components
to be delivered as open source projects as well.
Application Layer
The Application layer as we can see in both Figures
2 and 3 is composed of several sub-layers and com-
ponents. At the top of the Application Layer, we
find the components that interact directly with the
user. Next we describe each of these components
regarding the interfaces and the technology used.
User Interface, Map
Viewer and Searcher
Interface. The main purpose of the User Interface
is to facilitate the user interaction with the applica-
tion. The easiness and comfort of use and the level
of interaction are some factors that determine the
success or failure. The User Interface integrates the
other two components that interact directly with
the user: the Map Viewer and the Searcher. The
former component offers a common map-based
959
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures

Figure 2. Interfaces used to expose the components functionality

interface for data visualization and exploration.
The latter, the Searcher component, follows an
OpenSearch3 interface. This simple specification
proposes a standard search interface, used in most
modern browsers to manage custom searches. In
addition, a geospatial extension (OpenSearch-
Geo) has been proposed that adds geographic
filtering capabilities and support for geographic
response formats, obtaining then an optimal ap-
proach to search for geospatial data.
Technology. As we shown in Figure 3, the
interface is implemented using RIA technology.
It proposes a new evolution of the client server
paradigm that allows executing a portion of the
application on the user’s local system. This reduces
the communication with the server and increases
the user’s autonomy (Weaver, Gao, Chin, &
Iverson, 2009).
From among the wide range of valid options
like FLEX, SilverLight or JavaFX, we have se-
lected JavaFX because of its fully integration with
Java which allows us to easily integrate the com-
ponents reused and developed in our framework.
JavaFX is structured according to the classical
concept of scene graph for the implementation
of graphical applications, with a primary scene
node on which iteratively hang the rest of nodes.
These nodes represent mainly behavior anima-

Figure 3. Technology used to implement the components

960
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
tion, transition effects, visual content and user
interaction (Weaver, Gao, Chin, & Iverson, 2009).
In the open source community there exist
several map viewers components available. In
our work OpenLayers has been the chosen one.
OpenLayers is a JavaScript client that offers
interesting features like very good performance,
the ability to consume many different geographic
formats, and client side support of standards such
as OGC Web Map Service (WMS) and Web Fea-
ture Service (WFS). The communication between
the interface and the map viewer component is
done through calls to specific methods (using
the JavaFX AppletStageExtension class) that are
interpreted by the JavaScript runtime engine in
the browser.
Furthermore, OpenLayers has been the con-
tainer extended with a new self developed control
called OpenSearch-geo control which has been
integrated in OpenLayers for searching purposes
(Fonts, Huerta, Díaz, & Granell, 2010). This new
control in the client translates user queries to the
standard OpenSearch protocol, using geospatial
parameters and supported formats, and offers
user the retrieved documents. As in the previous
case, the JavaFX AppletStageExtension class is
the preferred interface.
Controller
Interface. Following the MVC pattern, the logic
of the application is located in the Controller
component. This responds to events produced
by the user, and invokes requests to the model
by delegation to the lower components of the
Application Layer (Service Connector, Service
Deployer and Service Publisher).
Technology. This component is developed
using the Singleton pattern (Freeman, Sierra, &
Bates, 2004) so that any instance requested on it
to access the same content and execution. The
interactions between the view (user interface) and
the controller are performed via a new feature in
JavaFX that synchronizes variables between the
user interface and the controller. Examples are the
list of local resources loaded (localFileResources)
as a view update and the method loadLocalFile()
as an input to the controller (and therefore execu-
tion of actions on the model).
Service Connector
Interface. The Service Connector component in
the application layer contains components that
implement the client side interfaces of the stan-
dard services deployed in the Service layer. For
instance it contains the component Web Process-
ing Service Application Programming Interface
(WPSAPI) that implements the OGC WPS client
interface to be able to access and run distributed
processing services. Another module developed
in the Service connector is the CSWAPI (Catalog
Service for Web API).
Technology. Currently, Service Connector
modules like WPSAPI and CSWAPI are included
in a self–developed Java library. For instance
WPSAPI offers a programmatic way to send
and receive getCapabilities request using Java
objects. It implements objects to send and receive
Process description requests to get the description
of processes in order to see the parameters and the
way a process within a WPS has to be invoked.
Finally, it permits sending an execute request to
invoke one of the available processes and it is
able to retrieve and parse the response in order
to extract the process results. The object model is
scalable to support new versions. Currently ver-
sions 0.4 and 1.0.0 are supported. This component
is programmed also under open source license.
Service Deployer
Until now the described components offered ac-
cessing capabilities to resources deployed in the
infrastructure. The aim of the Service Deployer
component is to assist the user to deploy resources
as services in the infrastructure. These resources
can be either data resources or processes. In
961
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
this paper we focus on the deployment of data
resources.
Interface. The Data Wrapper component al-
lows users to upload geospatial data to be deployed
and delivered as standard OGC data services,
such as WMS, WFS and WCS. The Data Wrapper
interface contains two methods: WrapFile and
WrapData to deploy data available in a File or
included as a parameter in the request.
To deploy data as an OGC Service, this com-
ponent should connect to OGC Service instances
available in the infrastructure. Some of these
OGC standards (such as WFS) include transac-
tional operations in their interface specification.
In this case, the resources can be deployed by
implementing the transactional interface in the
client side. This is the ideal case, because the
component implementation is independent of
service implementations. In other cases like de-
ploying data for visualization purposes as WMS,
the specification does not consider transactional
operations; therefore this component has to deal
with the concrete services implementation and
their own protocol and interface. Currently, in our
prototype implementation this is the case of the
Data Wrapper component when deploying data
in WMS implementations using the Geoserver
RESTful API4.
Technology. The Data Wrapper component
is a self developed Java library in form of API
exposing the described interface. In order to
deploy the data, we have first associated this
component with an instance of OGC Data Service
that implements WMS, WFS and WCS interfaces.
This OGC Services implementation is an open
source project called Geoserver, which offers a
RESTful API to deploy content. Geoserver is then
the responsible for publishing data through OGC
interfaces, transforming the resources processed
in WMS for viewing and WFS and WCS for
download purposes.
962
Service Publisher
Service Publisher component is in charge of de-
ploying data and process descriptions to an open
catalog, so they can be found in the infrastructure.
Interface. Like the Data Wrapper, the Service
Publisher needs to implement an interface to
deploy content in a specific Service type. In this
case the content is not data but metadata and the
service type is the Discovery Service Type, which
recommended interface according to INSPIRE is
the OGC CSW specification. In this case CSW
offers a transactional operation in its interface to
deploy new metadata in the service, thus opposite
to the Data Wrapper the interface implemented
by Service Publisher is a standard interface, the
CSW-T. For these transactional requests, we
build and send a HTTP POST request where the
body is a standard metadata element included in
the standardized XML structure according to the
OGC service.
On the other hand once deployed the metadata,
it will be available for searching purposes through
the CSW interface. Since the CS-W implementa-
tion of choice offers other interfaces, the deployed
metadata will be available not only through the
CSW but also through the Open Search interface.
Technology. The Service Publisher is again a
self-developed Java library delivered with open
source license. The Service Publisher uses the
Service Connector to perform the two main steps
of its workflow: first, uses the Service Connector
to send GetCapabilities request to the services in
which we have deployed the resources and get a
minimum set of metadata elements (available in
the Capabilities document) and second, it uses
the Service Connector CSWAPI to connect to
the catalog(OGC CSW) and deploy this metadata
using the transactional interface, which provides
methods to insert, delete, and update records in
the catalog.
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
Service Layer
As we based our infrastructure on INSPIRE tech-
nical guidelines, the service layer must provide
services for discovery, viewing and downloading
of data available in the infrastructure.
Interface. The INSPIRE technical architecture
shows that the different service types defined
by the Directive are placed in the service layer
resulting in the so-called Service Network. Each
type defines common capabilities offered by a
group of services. Regarding this functionality,
the INSPIRE Service Types we consider are the
Discovery (INSPIRE Network Services Drafting
Team, 2009b), View (INSPIRE Network Services
Drafting Team, 2008) and Download (INSPIRE
Network Services Drafting Team, 2009a) service
types. In addition we also consider the Process-
ing type to add processing capabilities to the SDI
(Díaz, Granell, & Gould, 2008).
Since OGC standards are proven to be mature,
they have been chosen to implement the interfaces
of the services available in the Service layer as
the INSPIRE directive recommends in its imple-
mentation rules (see parenthesis content: such
as OGC CSW (to implement Discovery Service
type), OGC WMS (View Service Types), OGC
WFS, OGC WCS (Download Service types) and
OGC WPS (pointed out to be useful to implement
distributed environmental applications (Granell,
Díaz, & Gould, 2010).
Technology. There are many implementa-
tions of OGC specifications in the open source
community. GeoServer is the project of choice
to deploy data for visualization and download
purposes, that is, through WMS, WFS and WCS
specifications. It offers these interfaces and in its
latest version GeoServer RESTful API allows the
Data Wrapper component to deploy resources and
configure the services.
Geonetwork is another open source project
broadly extended thanks to its implementation of
the OGC CSW specification. This is the project
of choice to deploy the discovery service type in
our framework. Regarding WPS specification,
we choose 52North WPS framework because it
is an open source Java framework that enables
easily design, implementation and deployment
of WPS-based services.
USE CASE
To evaluate our proposal and illustrate with a
practical example the behavior of the described
components we present a particular use case
focused on the evaluation of burned regions on
protected environmental areas in Valencia (Spain).
The analysis of fire impact requires a real time
interaction between different datasets, processes
and users. For these reasons it represents a perfect
case study to look at the overall functionality of
the framework.
A technician must assess the impact of the
some summer fires in the protected areas of Va-
lencia, she should find and access the protected
areas dataset, the burned areas dataset (for a cer-
tain year) and perform the intersection process.
Traditionally the result of the assessment can be
used to write a report or make a decision; here
we only describe how this technician can deploy
it easily in the underlying SDI. Figure 4 shows a
sequence diagram with the operations and actors
involved in each step of this scenario.
The first step is to find the initial data sets, in
our case the protected areas of Valencia and the
burned areas in 2005. For the first data entry
(protected areas), we use a local data provider
using a standard OGC data service. These data
can be found using the OpenSearch-geo client
component.
If the data is not accessible through an OGC
service and it is only available locally like the input
data related to the burned areas in 2005, Service
Framework provides the functionality to upload
local datasets to be deployed as standard OGC
data service. Thus, the user can upload it in the
963
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures

Figure 4. Sequence diagram

Figure 5. Local data deployment

964
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
application. Figure 5 shows a screenshot of how
this step can be performed in the user interface.
The second step shown in the sequence diagram
(Figure 4) is the execution of a distributed process
to perform the intersection and the generation of
the assessment results (Figure 6). The Service
Framework interface allows the user to connect
and execute to existing WPS services, rather than
using heavy desktop applications. In this step we
must select a WPS server and a specific process,
as in our case the Intersection (a process that will
return the protected areas that burned in 2005).
All communication with the service is through
the Service Connector component using the WPS
API. In the diagram of Figure 4, we can see how
internally several WPS interface calls are per-
formed. First a describeProcess request to get the
process description regarding input data and af-
terwards (after the user identifies the correct input
data, in this case references to the WFS where we
Figure 6. Process inputs view and results
have deployed the data), the execute request to
run the process.
The last step consists of deploying the process-
ing results to explore the results (Figure 6). Users
can thus view it and analyze it through the inte-
grated map viewer component based in OpenLay-
ers. Again as we described with the upload of the
burned areas the most interesting feature of the
Service Framework is that the Service Framework
assists the user to deploy the processing results
for visualization and download purposes into
standard services, OGC WMS and OGC WFS
respectively. The Service Framework uses the
GeoServer RESTful API and creates a new data
store accessible through standard services. After
the data deployment, the Service Publisher com-
ponent creates a minimum set of metadata elements
to be published in an instance of CS/W so that
the dataset can be found by other users through
common discovery mechanisms in SDI.
965
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
CONCLUSION
We have presented a framework that follows a
hybrid SDI building methodology. We provide an
INSPIRE-based architecture, whose compo-
nents assist users not only in accessing but also
in deploying resources as INSPIRE-compliant
services. In this way we provide an entry point
for users to participate in SDI building and
maintenance. As a result, users can massively
deploy resources improving their availability in
a standard manner.
To deploy resources automatically as standard
services, the Service Framework uses a concrete
implementation of the OGC WMS, WFS, WCS
specification. The Service Framework uses Ge-
oserver’s own protocol to deploy resources. We
still could migrate the solution to other imple-
mentations, but some programming should be
done. To avoid this, it would be useful to deploy
resources using standard transactional operations.
This would require to have the OGC WMS and
WCS specifications including a transactional
interface like WFS and WPS is considering in its
current and future versions. This would detach
the solution from the vendor implementation.
In this sense OGC standards have shown to be
mature enough to provide specifications to create
interoperable web services to deploy and share
resources and provide the functionality needed
to create distributed applications on SDI.
The use of open source components has the
advantage of giving us full control over the code
to modify everything. A disadvantage, however,
is that they have caused some inconveniences
because most of them use libraries that are not
fully developed in the context of this project. Docu-
mentation scarcity is always a big issue to face.
Therefore, there is a need to invest an important
amount of time and effort in the development to
have satisfying results.
966
ACKNOWLEDGMENT
This work has been partially supported by the
European project EuroGEOSS and by the CENIT
“España Virtual” project through the Instituto
Geográfico Nacional (IGN). We would like to
thank Jorge Suarez from the Conselleria de Medi
Ambient, Aigua, Urbanisme i Habitatge (Environ-
mental, Water, Urban planning and living regional
government of Valencia), who kindly granted the
forest fire data for experimentation purpose.
REFERENCES
Ball, W. P., Brady, D. C., Brooks, M. T., Burns,
R., Cuker, B. E., & Di Toro, D. M. (2008). Pro-
totype system for multidisciplinary shared cyber-
infrastructure: Chesapeake Bay environmental
observatory. Journal of Hydrologic Engineering,
13(10), 960–970. doi:10.1061/(ASCE)1084-
0699(2008)13:10(960)
Bell, M. (2008). Service-Oriented Modeling
(SOA): Service analysis, design, and architecture.
Hoboken, NJ: John Wiley & Sons.
Benthall, S., & Gill, S. (2010). SDI Best Prac-
tices with GeoNode. Retrieved June 1, 2010,
from FOSS4G website: http://2010.foss4g.org/
tutorial10.php
Curbera, F., Duftler, M., Khalaf, R., Nagy, W.,
Mukhi, N., & Weerawarana, S. (2002). Unravelling
the Web Services Web: An Introduction to SOAP,
WSDL, and UDDI. IEEE Internet Computing,
6(2), 86–93. doi:10.1109/4236.991449
Díaz, L., Granell, C., & Gould, M. (2008). Case
study: Geospatial processing services for web
based hydrological applications. In Sample, J. T.,
Shaw, K., Tu, S., & Abdelguerfi, M. (Eds.), Geo-
spatial services and applications for the Internet
(pp. 31–47). New York, NY: Springer.
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
Díaz, L., Granell, C., Gould, M., & Olaya, V.
(2008). An open service network for geospatial
data processing. In Academic Proceedings of the
Free and Open Source Software for Geospatial
Conference, Cape Town, South Africa (pp. 410-
419).
European Parliament and Council. (2007). IN-
SPIRE EU Directive. Retrieved January 28, 2009,
from http://eur-lex.europa.eu/LexUriServ/site/en/
oj/2007/1_108/1_10820070425en00010014.pdf
Fielding, R. (2000). Architectural styles and the
design of network-based software architectures
(Unpublished doctoral dissertation). University
of California, Irvine, CA.
Fonts, O., Huerta, J., Díaz, L., & Granell, C.
(2010). OpenSearch-geo: The simple standard for
geographic web search engines. In Proceedings
of the IV Jornadas de SIG Libre, Girona, Spain.
Retrieved September 1, 2010, from http://www.
sigte.udg.edu/jornadassiglibre/uploads/Articles/
a29.pdf
Freeman, E., Sierra, K., & Bates, B. (2004). Head
first design patterns. Sebastopol, CA: O’Reilly
Media.
Giuliani, G., Ray, N., Schwarzer, S., De Bono,
A., Peduzzi, P., & Dao, H. (2011). Sharing envi-
ronmental data through GEOSS. International
Journal of Applied Geospatial Research, 2(1).
doi:10.4018/jagr.2011010101
Goodchild, M. F. (2008). The use cases of digital
earth. International Journal of Digital Earth,
1(1), 31–42. doi:10.1080/17538940701782528
Goodchild, M. F., Egenhofer, M., Fegeas, R., &
Kottman, C. (1999). Interoperating geographic
information systems. Boston, MA: Kluwer Aca-
demic.
Granell, C., Díaz, L., & Gould, M. (2010).
Service-oriented applications for environmental
models: Reusable geospatial services. Environ-
mental Modelling & Software, 25(2), 182–198.
doi:10.1016/j.envsoft.2009.08.005
Granell, C., Gould, M., & Esbrí, M. A. (2008).
Geospatial web service chaining. In Karimi, H. A.
(Ed.), Handbook of research on geoinformatics
(pp. 189–195). Hershey, PA: Information Science
Reference.
Hey, T., & Trefethen, A. E. (2005). Cyberinfrae-
structure for e-Science. Science, 308(5723),
817–821. doi:10.1126/science.1110410
INSPIRE Network Services Drafting Team.
(2008). Draft technical guidance to implement IN-
SPIRE view services. Retrieved from http://inspire.
jrc.ec.europa.eu/reports/ImplementingRules/
network/Draft_Technical_Guidance_View_Ser-
vices_v1.0.pdf
INSPIRE Network Services Drafting Team.
(2009a). Draft technical guidance for INSPIRE
download services. Retrieved from http://inspire.
jrc.ec.europa.eu/reports/ImplementingRules/
network/Draft_Technical_Guidance_Down-
load_Services_v1.0.pdf
INSPIRE Network Services Drafting Team.
(2009b). Technical guidance for INSPIRE dis-
covery services. Retrieved from http://inspire.
jrc.ec.europa.eu/documents/Network_Services/
Technical%20Guidance%20Discovery%20Ser-
vices%20v2.0.pdf
Lannotta, B. (2007). GEOSS: A global view.
Aerospace America, 45(8), 36–40.
Masser, I. (2005). GIS worlds: Creating spatial
data infrastructures. Redlands, CA: ESRI Press.
967
 Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures

Mazzetti, P., Nativi, S., & Caron, J. (2009). Open Geospatial Consortium Inc. (OGC). (2008).
RESTful implementation of geospatial services Web coverage service implementation specifica-
for Earth and Space science applications. Inter- tion 1.1.2. Redlands, CA: Author.
national Journal of Digital Earth, 2(1), 40–61.
Papazoglou, M. P. (2008). Web services: Principles
doi:10.1080/17538940902866153
& technology. Essex, UK: Person Education.
Minsker, B., Myers, J., Marikos, M., Wentling,
Pezzoli, K., Marciano, R., & Robertus, J. (2006).
T., Downey, S., Liu, Y., et al. (2006). NSCA en-
Regionalizing integrated watershed management:
vironmental cyberinfrastructure demonstration
A strategic vision. In Proceedings of the 7th Annual
project: Creating cyber environments for envi-
International Conference on Digital Government
ronmental engineering and hydrological science
Research, San Diego, CA (pp. 444-445). New
communities. In Proceedings of the ACM/IEEE
York, NY: ACM.
Conference on Supercomputing, Tampa, FL. New
York, NY: ACM. Pierce, M. E., Fox, G. C., Choi, J. Y., Guo, Z.,
Gao, X., & Ma, Y. (2009). Using Web 2.0 for
Morisette, J. T., Richardson, A. D., Knapp, A. K.,
scientific applications and scientific communities.
Fisher, J. I., & Abatzoglou, J. (2009). Tracking
Concurrency and Computation, 21(5), 583–603.
the rhythm of the seasons in the face of global
doi:10.1002/cpe.1365
change: Phenological research in the 21st century.
Frontiers in Ecology and the Environment, 7(5), Poch, M., Comas, J., Rodríguez-Roda, I., Sàn-
253–260. doi:10.1890/070217 chez-Marrè, M., & Cortés, U. (2004). Design-
ing and building real environmental decision
Nerbert, D. (2004). Developing spatial data
support systems. Environmental Modelling &
infrastructure: The SDI cook book, version 2.0.
Software, 19(9), 857–873. doi:10.1016/j.env-
Victoria, Australia: GSDI.
soft.2003.03.007
Open Geospatial Consortium Inc. (OGC). (2005).
Pumphrey, M. (2009). OpenGeo Suite. Retrieved
Web feature service implementation specification
June 1, 2010, from FOSS4G website: http://wiki.
1.1.0. Redlands, CA: Author.
osgeo.org/wiki/FOSS4G_2009_Demonstra-
Open Geospatial Consortium Inc. (OGC). (2006). tion_Theatre
Web map service implementation specification
Rajabifard, A., Binns, A., Masser, I., & William-
1.3.0. Redlands, CA: Author.
son, I. (2006). The role of sub-national govern-
Open Geospatial Consortium Inc. (OGC). (2007a). ment and the private sector in future spatial data
Sensor observation service (Version 1.0.0). Red- infrastructures. International Journal of Geo-
lands, CA: Author. graphical Information Science, 20(7), 727–741.
doi:10.1080/13658810500432224
Open Geospatial Consortium Inc. (OGC). (2007b).
Web processing service (Version 1.0.0). Redlands, Ramaprivan, H. K. (2008). NASA’s Earth science
CA: Author. data systems: A ‘bit of history’ and observations.
In Proceedings of the Cyberinfrastructure for
Open Geospatial Consortium Inc. (OGC). (2007c).
Environmental Observations, Analysis, and Fore-
OpenGIS® catalogue service implementation
casting: A Cyberinformatics Forum.
specification (Version 2.0.2). Redlands, CA:
Author.

968
Open Source Based Deployment of Environmental Data into Geospatial Information Infrastructures
Reichert, P., & Schuwirth, N. (2010). A generic
framework for deriving process stoichiometry in
environmental models. Environmental Modelling
& Software, 25(10), 1241–1251. doi:10.1016/j.
envsoft.2010.03.002
Richardson, L., & Ruby, S. (2007). RESTful Web
Services. Sebastopol, CA: O’Reilly Media.
Secretariat, G. E. O. (2007). The full picture: GEO.
Geneva, Switzerland: Tudor Rose.
Steiniger, S., & Bocher, E. (2009). An overview
on current free and open source desktop GIS de-
velopments. International Journal of Geographi-
cal Information Science, 23(10), 1345–1370.
doi:10.1080/13658810802634956
Weaver, J. L., Gao, W., Chin, S., & Iverson, D.
(2009). Pro JavaFX Platform: Script, Desktop
and Mobile RIA with Java Technology. New York,
NY: APress.
Westlund, S. (2010). The use of geospatial tech-
nologies in disaster management. International
Journal of Applied Geospatial Research, 1(3),
17–30. doi:10.4018/jagr.2010070102
Wilkins-Diehr, N., Gannon, D., Klimeck, G., Oster,
S., & Pamidighantam, S. (2008). TeraGrid science
gateways and their impact on science. Computer,
41(11), 32–42. doi:10.1109/MC.2008.470
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 2,
edited by Donald Patrick Albert, pp. 6-23, copyright 2012 by IGI Publishing (an imprint of IGI Global).
Yang, C., & Raskin, R. (2009). Introduction
to distributed geographic information process-
ing research. International Journal of Geo-
graphical Information Science, 23(5), 553–560.
doi:10.1080/13658810902733682
Yang, C., Raskin, R., Goodchild, M., & Gahegan,
M. (2010). Geospatial Cyberinfrastructure: Past,
present and future. Computers, Environment
and Urban Systems, 34, 264–277. doi:10.1016/j.
compenvurbsys.2010.04.001
Yang, P., Li, W., Xiao, D., Raskin, R., & Rambacus,
M. (2007). Earth information exchange - Sharing
earth science information through interoperable
approach and cyberinfrastructure. In Proceed-
ings of the SPIE Conference on Geoinformatics:
Geospatial Information Science, Nanjing.
ENDNOTES
1
All specifications for service interfaces
defined by OGC are publicly available at
http://www.opengeospatial.org
2
Workshops, tutorials and presentations for
every conference edition are publicly avail-
able at http://foss4g.org
3
The OpenSearch specification and its exten-
sions can be found at http://www.opensearch.
org
4
http://docs.geoserver.org/stable/en/user/
extensions/rest/index.html
969
970

Chapter 60
Application of Topographical
Capture Techniques for
Modelling Virtual Reality:
From the Static Object to
the Human Figure

Mercedes Farjas Abadía

Universidad Politécnica de Madrid, Spain

Manuel Sillero Quintana

Universidad Politécnica de Madrid, Spain

Pedro Ángel Merino Calvo

Universidad Politécnica de Madrid, Spain

ABSTRACT
Since the dawn of time man has attempted to represent the human figure with techniques ranging from
simple drawings to techniques that manage to reflect the movement of body segments. In parallel,
cartographic techniques have developed very advanced capture and 3D representation systems, but
even though they have been applied in recent years to other sciences, they have not been applied yet to
virtual reality. The appearance of the laser acquisition systems has enabled us to acquire data without
discrimination on points and to get quick 3D models. This situation allows us to work directly on the
concept of surface and to analyze it from the uniqueness of the detail, compared to traditional systems
which capture points for, later, imaging surfaces from them. Under this prism, a research group was
formed by graduates in Physical Activity and Sport and in Cartography, in order to bring together both
sciences and to improve techniques of capture and representation of the human body. The road is not
completely gone, but some results have been obtained and are presented in this work.

DOI: 10.4018/978-1-4666-2038-4.ch060

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Application of Topographical Capture Techniques for Modelling Virtual Reality
INTRODUCTION
Nowadays, the term “Virtual Reality” is associ-
ated to multiple disciplines, both in research and
practical applications, but its ultimate goal in all
of them is very similar: to produce an illusion of
reality, which can be subject to a practical, tech-
nical or conceptual application. Since the 50’s,
when they were started to nowadays, the capture,
processing and representation systems have been
varied, evolving in a parallel way to the comput-
ers, essential for the mathematical calculations.
Direct applications of virtual reality range from
medicine to engineering, from computer games to
advertising. In all those fields, simulations provide
an effective and efficient support to its technical
or practical goals.
If we consider the processing hardware and
software, we see that has changed dramatically. In
some cases, it is difficult to distinguish between
real and fictitious when you see some pieces of
work of virtual reality. In a deeper analysis, we can
conclude that the problems are in the collection
of data, since in many cases much of the work is
done through techniques of image designing and
processing.
In this report, we propose a new technique to
capture the information that is going to by virtual-
ized, basing it in real-world data collection using
a 3D scanner. This technique provides us with a
virtual model of reality with metric data, which
we can say that is a perfect representation of real-
ity. The study develops model tests in laboratory
and real environments, trials with static human
models, leaving the door open to the modelling
of the human figure as a moving object.
Before going further in the work, we would
like that you turn the light back with us, so that
when you read, you will understand the steps taken
since the beginning of the topography until the
representation of the human figure by topographic
as a result of biomechanical studies.
At present we are working on virtual reality
techniques, but it has not always been our goal, as
traditionally the development of the topography
has focused on modelling and representation of
terrain in all its aspects. A few years ago, a group
of people considered that topographic techniques
and virtual reality had progressed in two parallel
but not crossed ways, or, in other words, advanc-
ing without looking each other.
Throughout time, topographic and cartographic
engineering technologies have been working on
the acquisition of data using different equipment
and topographic, photogrammetric and geodetic
methods in order to represent 2D and 3D objects,
buildings, statues and archaeological sites, focus-
ing the research on obtaining and modelling those
elements with sufficient metric precision.
It is said that any investigation begins when
we are faced with a question or uncertainty that
makes us question the situation and initiate a
search. And we have the following question:
Can our techniques for capturing and modelling
to contribute something to the virtual reality and
biomechanics?
We believe that one of the fundamentals of
virtual reality is three-dimensional interactive
computer simulation, so that the user is introduced
in an artificial environment that is perceived as real,
allowing you to interact in a virtual world that has
the ability to maintain a two-way correspondence
with the real world. The more closely approximates
the information captured to the reality, the better
the interrelationships between the virtual and the
real. If we apply these concepts to biomechanics,
as a multidisciplinary study of the measurements
of the human body in movement, we can be at
the gates to get almost perfect representations of
the reality with virtual models.
In short, we have to understand both the func-
tional and the physical variability in subject’s
movement, and how these components interact
and how they could be improved. The study of
the dialectical relationship structure-function, is
a basic tool with a strong genetic dependence for
the identification of future athletes.
971
 Application of Topographical Capture Techniques for Modelling Virtual Reality
Under this view, in recent decades, and in
parallel with the evolution of virtual reality,
sports science has tried to analyze and explain
the different aspects of the human figure and its
movement through various sports disciplines, in
order to propose procedures for evaluating and
improving the actions of the athletes by virtual
models. Thus we can observe as the conventional
training methods in elite athletes have been com-
plemented by technologic advances for analyzing
and quantifying the parameters involved in the
performance. At present it is impossible to reach
the maximum goals in sports without the help of
high technology.
From this viewpoint, we considered a quan-
tum leap in the object of study. If we were able
to represent all types of surfaces:
Why not introduce real coordinates and the
variable time in these virtual representations?
If we could represent, controlling the accuracy
in the final model and choosing the proper tech-
nique for that purpose, a surface at a time “A” and
“B” at a later time (e.g. on a glacier to analyze
the displacement of moraines, or on a Cathedral
deterioration over time), the new variable could
also be monitored and would be defined by Δt.
Once we entered into concept of metric rep-
resentation of “dynamic surfaces”, it was an
adventure to go beyond the measured object. It
was an experience to observe the surroundings and
people close to zero velocity, not considering the
scenarios as empty, in an isolated way, but with
his players and considering the human movement
slowed until a “moment” becomes a “taking” in
the data acquisition process.
The results that we obtained, along with the
emergence of new equipment and technologies
in topography, let us accept the hypothesis of the
project, and our ability to look forward again and
again, investing hours and hours (the time vari-
able again) in the design and implementation of
the first tests.
After having the idea, accepting the challenge
and setting the scientific context and the human
972
team, the viability was assured thanks to Leica
Geosystems and the company 3D ICOM, which
always provided us with their last generation
equipments.
We show you now a glimpse of these new
technologies in the field of virtualization and its
application to sports; technologies that should
have assessed by the reader.
VIRTUAL REALITY DEVICES
TO CAPTURE INFORMATION
In a model of virtual reality we have to fulfil
some minimum requirements of simulation or
representative capacity, user-model interaction
and sensory perception of the user. Even as the
last is the most important, it directly depends on
the quality of the simulation we obtain as a result
of the input images we use, because depending
on their proximity to reality we can convince the
users of the accuracy of the parallel reality that
we offer them.
Currently, systems for image capture require
huge processing times and in some cases some-
what tedious, so the incorporation of topographic
techniques (3D scanner) will allow us to capture
the real shape quick and eliminating most of the
post-processing tasks. This situation coupled
with the precision of the shots and the availabil-
ity of space-time metrics, will enable the virtual
reality model, once completed, to interact with
the user in an image very close to reality, where
three-dimensional movements can be controlled
by the metric component that is provided, a way
of solving some of the problems existing today.
This relationship between the use of coordi-
nates of space and time, allows us to overcome
the barriers of space and time, and it could allow
us to set up an environment where information
provided by the model and the perception that the
user receives are interlaced in a direct and two-way;
that is an unknown situation with a huge potential
in the development of virtual reality interfaces.
Application of Topographical Capture Techniques for Modelling Virtual Reality
The application of these results allows us to
move between the topology of three-dimensional
virtual reality or immersive and the non immer-
sive Internet-based, being the last one the more
accepted by users, and where the method can
produce more tangible benefits. We can sustain
this assertion, since in applications where the user
is a component of the virtual world, the nearest is
the image recognition, the better the acceptance of
the system, but if we apply the system to maintain
the heritage of a nation or, in archaeology, as a
method to aid in the exploration and subsequent
modelling, the benefits increase exponentially, al-
lowing us not only to identify the precise site and
its environment, but also knowing their temporal
variations, or enable people located in different
places to make a virtual visit, simultaneously, as
if actually they were in that place, even being in
the other side of the earth.
If we consider the user interface based in
video-mapping, where the person or environment
is filmed in advance and incorporated into the
computer by graphics simulations, the result is,
without any doubt, closer to reality but still not
perfect. The use of 3D scanning systems will al-
low us that the person, image, or environment to
introduce into the model is a clone of the reality, a
young teacher who comes to surprise the viewer,
allowing the illusion is created with realism such
that people can doubt where is the line between
the real and the virtual world.
Today, we are coming into a world where
reality is pure coincidence, and the virtual envi-
ronment with be as normal as the real one, like
a cyberspace where space-time relations are not
just virtual, but will be part of our lives and we
will assume it. In this situation, the better defined
are the models, the better will be the acceptance
by the users. For example, the success of the
clothing shops will depend on the availability of
an implemented system for virtually testing how
fit us some clothing.
In these terms, the virtual world is made up
of geometric shapes produced by a computerized
graphic system; this geometry is interlaced form-
ing meshes, which in turn represent objects, and
the objects virtual scenarios. If we increase the
geometric quality of the original polygons (best
shots and more accuracy), we can offer more
quality in the virtual model. But at present this
quality improvement requires us to have more
processing capacity and consequently increases
the cost. The incorporation of capture systems
based on topographic techniques, allows us to
not only improving the quality of the images,
but also improving and reducing the subsequent
processing times to the minimum, as we shall see
later on specific developments.
At present, the scanner system operating per-
fectly in static 3D objects, but there is a problem
that affects the time delay of the system, so that
the shots with dynamic objects appeared to move;
so that, the capacity to respond to this issue is
today insurmountable, but moving at the pace
that this technology does, this problem will be an
anecdote in a few years. We face a similar situa-
tion to the beginnings of photography when they
were required seconds for taking a good pictures,
and the modern digital cameras are able to take
hundred of high definition pictures per second. It
is only a matter of time, to improve the technique
to obtain the results described in this paper, so that,
we can consider it as a beginning or a window
open for further studies to a fully development of
the system. (See Figure 1)
TOPOGRAPHIC AND
CARTOGRAPHIC TECHNIQUES AS
AN ALTERNATIVE
In parallel with advances in virtual reality, the map-
ping techniques have evolved to very fast allowing
us access to information almost in real time, but
its use has been limited to the representation of
the earth’s surface and in some cases, very recent
ones, to the cataloguing of the artistic heritage.
973
 Application of Topographical Capture Techniques for Modelling Virtual Reality
Figure 1. Human face capture by scanner (2008) Equipment Konica Minolta VI-9i
The increasing use of the advances that oc-
curred in media and technology in recent years
has led to the introduction of graphics systems
as indispensable tools in teaching and learning
processes. The use of the computer as an audio-
visual tool at work, and its computational power,
allows the design of models and processes with
a high potential for development, expression and
operability.
The graphical representation has always been
an essential complement for any type of biome-
chanical analysis, as the kinematic and kinetic
variables obtained from it allow statistical studies
and simulations (digital models) of the movement
of the human body.
At present the results depends on method of
capture, where the variables space and time play
a decisive role in obtaining good results, which
are operational in a short space of time.
In a review of the state of the art, as regards the
representation of human figures, static or dynamic,
we can see that the modern systems apply tech-
niques based on photographic or opto-electronic
systems. Both systems record the spatial position
of a number of spatial markers associated with
the body surface, whose movement supposedly
reflects the movement of body segments. Thus, it
is difficult take decisions regarding the relationship
of markers with the anatomical characteristics of
the structure that lies below.
974
In the course of science, topographic and
cartographic techniques have traditionally been
linked to the study and representation of the ter-
rain, in any of its states or situations, resulting in
digital images, in two or three dimensions, which
were implemented for having maps as a result.
But these monothematic techniques have been
gradually changing its design to become more
generalist and multidisciplinary and, presently,
they have became an essential support for different
studies or scientific projects. They are remarkable
the advances that have occurred in the study and
cataloguing of architectural works or in supporting
the analysis of deformation of materials.
Let’s see two different examples:
1. Gathering data through a continuous digital
picture sensor system. (See Figure 2)
2. Data collection by laser system without
prism.
Currently, the technological development al-
lows the acquisition of information with laser
systems without prism, which allow us the data
acquisition without discrimination on points
(without markers), and therefore they need not
be placed in advance any element facilitate the
calculation process later. As a direct consequence,
we have that the collection of data can be execut-
Application of Topographical Capture Techniques for Modelling Virtual Reality
Figure 2. Capture of the Cibeles fountain by Laser equipment UPM-Leica (2003) CYRAX 2500 Leica-
Geosystem
ed faster and most importantly minimizing the
subsequent process of calculation.
Under these systems, the representation is
done on all of the selected area, obtaining a two-
dimensional and/or three-dimensional continuous
image of the represented figure, in which the
density of points will be accord to the desired
precision. Allowing us to work on the “concept”
surface directly, analyzing it from the uniqueness
of detail, compared to traditional systems which
record points to imagine surfaces from them.
In 2004 a group of people, including the authors
of this work, they posed the following question:
Do topographers’capture systems and their analy-
sis methodology can be used to generate virtual
reality models, and in particular, the representa-
tion of the human body and its biomechanical
analysis?
To find answers, we began research in this
field. We started to work on the doctoral thesis of
Pedro Calvo Merino untitled “Three-dimensional
modelling of the human figure through mapping
techniques and their application the moving ob-
ject”, which is currently in edition process, and
some final works to obtain the Engineering degree
in Topography. We have used some of their results
as part of the contents of this work.
We tried to define a methodology to be able to
capture and generate virtual representation of the
static human figure, and establish some parameters
in order to apply directly the 3D laser scanner on
the sport disciplines in which biomechanics have
a relevant role.
To do so, we have settled the following targets:
1. Analysis of the current methodologies and
their capture parameters applied to static
objects and their virtual representations.
2. Analysis of the mechanisms and parameters
that can influence the acquisition of the hu-
man figure, in order to apply them to sports.
3. Development of methodology of capture
based on a laser system without prism.
4. Representation of the three-dimensional
human figure on a digital format, in real
time and using real coordinates, allowing
its metric analysis.
Based on the above targets, they have been
defined the following specific objectives:
1. Image capture and formation of a 3D model.
2. Analysis of its application to the moving
object.
Since obtaining the first results, we note that
the use of this technology allowed us to work with
975
 Application of Topographical Capture Techniques for Modelling Virtual Reality
both the object and the human figure in real time,
compared to other systems that require a large
computer processing time. At the same time, it
was confirmed that its main advantage would be
the high degree of similarity to the real model,
with a minimum post-processing time, and sub-
sequently allowing modelling the human figure
very close to real time.
Compared to traditional capture systems
through photogrammetric or optoelectronic sys-
tems, the methodology of the 3D laser systems
is reduced in the number of stages.
We believe we are on the way to obtain inertial
parameters suitable for biomechanical modelling
of the human body, involving on it the use of
anatomical points and segments under an intercon-
nected on a local or real reference system, where
there is an interrelationship of those anatomical
points on the surface of the human body with
the space.
This paper will try to define and open a door to
a new system for capturing and representing the
human figure, as a primary step for the study of
human motion in real time. It also try to describe
our findings on trying to develop, validate and
apply a work methodology with direct applica-
tions in sports where the biomechanics and virtual
representation has an important role.
DESCRIPTION OF
CAPTURE SYSTEM
New technologies may allow describing, analyz-
ing, evaluating and modelling statically and cin-
ematically the environment and the human figure
and its movement (cinematic analysis), and then
resolve and optimize the biomechanical analysis,
through representations of the model with details
and temporal patterns which were unthinkable
some years ago.
The capture system using 3D scanner consists
of devices that analyze a real object, inanimate
or animated, to produce a near-real representation
976
of it, as a points cloud with geometry and metric.
This allows us to model not only a rapid model-
ling process, but also, thanks to the availability
of metrics, modifying the position of the model
without changing its real geometry.
3D scanners are classified according to the
measurement system: contact and non-contact
with the object. Focusing on the non-contact
type, are classified into two categories: active
(emit some kind of radiation) and passive (do
not emit any kind of radiation). In this paper we
will focus on the non-contact scanners with active
pulsed laser beam, which are the best suited to
the capture of data for virtual reality applications.
At the moment, they are widely used in many
applications: topographic representations of the
land, industry, maintenance of cultural heritage,
or film and video games edition.
If we focus on the models currently on the
market, there are different systems but they all have
common characteristics in their basic technology
(pulsed laser beam, no markers are required to
objects to capture the environmental conditions
of work, etc.), which we will try to show in the
following paragraphs.
In the pulsed laser beam equipments, it is mea-
sured flight time (going and back) for each point
on the object’s surface. These devices have a timer
with a resolution of 1 picoseconds, allowing us to
produce beam of diameter smaller than 6 mm at
a distance of 50 m from the object. Taking a step
forward, they have been linked to the technique
of laser beam of modulated amplitude, in which
the system scans continuously an object through
a variable beam, by using rotating mirrors. With
one of these systems the team was able to capture
the energy reflected from the object, by comparing
the shape of the reflected wave with modulation
released and, later, calculating the distance of the
object and consequently their spatial coordinates.
As a result, all points were captured from the be-
ginning on three-dimensional coordinates into a
local reference system, allowing its analysis and
conversion to various formats. This is one of the
Application of Topographical Capture Techniques for Modelling Virtual Reality
added values of the 3D laser scanners compared
to traditional capture systems (Barber, Mills, &
Bryan, 2004).
When we make a comparison of traditional
methods based on photogrammetry and optoelec-
tronics compared to laser scanning systems, we
obtain the following advantages of the latter over
the former (Demir et al., 2004):
• Recording of direct points of the object in
3D. Capture in real time.
• Provides a lot of points from the surface,
allowing a more realistic and complete
analysis of the object.
• It is an excellent technique for the de-
scription and representation of irregular
surfaces.
• The final results are available in a very
short period of time.
When “c” is the speed of the pulsed light
and “t” is flight time (going and back), we have
simply that space travel is (t • c) / 2. As speed of
the laser beam is a constant, the calculation of the
distance will depend only on the flight time, which
is currently around 3 picoseconds. Considering
samples of 1mm in the surface, the estimated time
for scanning an object sized 50 x 180 cm (body)
located 50 meters away, with a rate of 200,000
points per second, is approximately 5 seconds.
The distance of a point and its direction (polar
coordinates) is calculated by changing the direc-
tion of the pulse using a rotating mirror method.
The rotating mirror device is commonly used be-
cause they are much lighter than prism and it can
be rotated much faster and with superior accuracy.
To capture technique uses the continuous beam
of laser amplitude modulation, where the scanner
scans an object through the beam of a continuous
variable, using rotating mirrors. With this system
the team is able to capture the energy reflected
from the object, comparing the shape of the re-
flected wave with modulation issued and calculate
the distance of the object and consequently their
spatial coordinates. As a result, all points are
captured from the beginning three-dimensional
coordinates on a local reference system, allowing
its analysis, or export to various formats, this is
one of the added value of the 3D laser scanners
compared to traditional systems of capture.
There are different capture systems based
on laser systems in the market, the most impor-
tant are the ScanStation 2 and the HDS6100 of
Leica-Geosystem (Leica Geosystems, 2005), the
Trimble GX 3D Scanner, and the VI-9i or VI910
of Konica Minolta.
All of them are based on a system of continuous
laser beam, allowing high speed scanning from
100,000 points / sec to 500,000 points / second,
while providing a complete horizontal visual field
(±360º) and ±135º in vertical. They provide high
accuracy (± 0.05 mm in the VI-92), even at great
distances, thanks to the great intensity and the
beam generated by each pulse of short wavelength,
which combined with the optical compensation
schemes and the dynamic expansion of the laser
scanning, allows very accurate captures of surfaces
with different levels of reflection.
Capture methods are often based on laser op-
tical triangulation, providing a points cloud that
automatically generate a polygonal mesh with
all the information connected, thereby geometric
ambiguities are eliminated and the final result is
improved, allowing to apply 24bit textures – real
color and consequently obtaining the continuity
of the elements represented virtually.
The most relevant characteristics of the equip-
ment are:
• Laser scan for high precision and speed cap-
ture using the pulse laser system, with dou-
ble mirror optics and servo-mechanisms.
• Luminosity: Operating with bright light
and complete darkness.
• Allows a scan of 20.000 m3 to 50 meters
and 160.000 m3 to 100 meters.
977
 Application of Topographical Capture Techniques for Modelling Virtual Reality
• The range varies from one computer to an-
other, for example the ScanStation 2 has a
range of 300 m with 90% reflectivity.
• Capture precision < 4 mm to 50 meters
and resolution < 1 mm at distances be-
tween 5 and 50 meters and angles of ± 60
micro-radians.
• Density vertical scanning 1,000 points/col-
umn and horizontal scanning 1000 points/
raw.
• Measurement precision P2P between ±0.05
mm and ±4 mm.
• Capture time less than 5 minutes, de-
pending on the volume: if we consider 1
column/sec for 1000 points/column, 2
columns/sec for 200 points/column but,
considering little volumes, values are 0.3
seconds for 80,000 points and 2.5 seconds
for 300,000 points.
• Visual field complete (±360º) in horizontal
and ±135º in vertical, both under solar light
and complete darkness.
• Direct interface and on real-time connec-
tion with the computer.
• Colour video camera integrated with a res-
olution of 680 x 480 pixel.
How Do They Work?
• Focus the scanner onto the area of interest.
• The video camera integrated in the team,
made a first quick capture, which is sent to
a computer connected to it.
• Implementation of a rapid scan capture in
order to get a 3D preview to provide us in-
formation about the depth of the capture
(blind or dark areas).
• The operator selects the desired area, ac-
curacy, resolution and other required
parameters.
• Starts the capture of the object. To do this
within the system, two mirrors perform a
quick and systematic scan by laser pulses
on the selected area, using the technique
978
of “flying time” to take the measurement
of each pulse of the laser. The system does
not require the use of reflectors (markers).
• For each laser pulse, they are calculated
distances and tiny integrated optical de-
coders perform angular measurements, ob-
taining primary polar coordinates, which
are transformed into Cartesian coordinates
(X, Y, Z).
• The resulting positions are displayed
graphically on the computer screen in real
time (as soon as they are calculated).
CAPTURE AND PROCESSING
SOFTWARE
As an example, we briefly describe the Cyclone
software of Leica Geosystems. Most of them have
a similar nature.
Cyclone software is a modular PC software to
operate the scanner Cyrax. It processes 3D point
clouds in an efficient and manages projects from
3D laser scanner. Makes a thorough record of each
detail, allowing a quickly and safely integration
of different scans, taking only certain points in
common.
The system has a wide range of tools for work-
ing with images and models in three dimensions.
Its most important features are:
• Total control of the scanner and process
from the PC through the communication
of the scanner and the PC by a standard
Ethernet connection.
• Fast capture of the digital video image in
4-bit true colour with 630 x 480 pixels
resolution.
• Adjust parameters in detail: accuracy, den-
sity of points, etc.
• Fast conversion of the point cloud in a tri-
angular mesh surface.
• Generation of digital models using mesh.
Application of Topographical Capture Techniques for Modelling Virtual Reality

• Generation of surfaces and maps of recently, for cataloguing of heritage. On them,

intensity. three-dimensional models were obtained which
• Calculation of volumes on the mesh. allowed carrying out studies, analyzing propos-
• 3D modelling from point clouds, by intel- als and making decisions on geographic areas or
ligent algorithms. archaeological sites. In the following figures we
• Record multiple scans. With overlapping present some of the results. (See Figure 3 and
functions, geometric adjustment, adjust- Figure 4)
ment of spatial coordinates, and so on. The progress occurred in data acquisition
• Instantaneous calculation of distances be- systems with 3D scanning techniques during the
tween two points and from the scanner at last years it has been both in the collection of data
any point of the object. in continuous mode and in bypassing the selection
• Allows displacement, flight, zoom and of points in the field and avoiding the subjectiv-
free-rotation of the 3D models.
• Texturing of the point cloud.
Figure 3. Survey of virtual representation at 1 /
• Viewing selected items with parameterized
500 of the archaeological area of Mleiha in the
intensity.
Emirate of Sharjah with GPS receivers
• Filters to enter data and information based
on distance and intensity ranges.
• Geo-referencing and processing of the
points in space.
• Record multiple scanning processes.
• Measures on the model (ΛX, ΛY, ΛZ), us-
ing P2P techniques.
• Export formats: VRML, DXF, DGN, text
(X, Y, Z, color), PTS, Warefront OBJ, etc.
It is required a PC with the following char-
acteristics:
• Windows 2000 (SP2 or higher), Windows
XP (Professional or Home Edition, SP2).
• Monitor: minimum resolution of 800 x 600 Figure 4. Digital model of the Torreón Guzmanes
pixel. (Avila), produced by a laser three-dimensional
• Pentium 4 2 GHz or higher. 3D Trimble GS200
• 512 MB RAM (1 GB recommended).
• SVGA - OpenGL Graphic Card.
• 10/100 Ethernet Network Card.
• Disk: minimum 10 Gb.

THE CAPTURE OF THE

STATIC OBJECT RESULTS
OBTAINED IN FIELD WORKS

As mentioned above, the first applications of

laser systems were only topographical and,

979
 Application of Topographical Capture Techniques for Modelling Virtual Reality
ity remaining in the results. Data capture is auto-
matically done in the field and, later, in the cabi-
net, it is made the selection of information.
This technique is being used, quite successfully,
in some models as “Tyrannosaurus Rex” by the
Carregie Museum of Natural History (Pittsburgh,
PA) or the “Statue of Liberty” in New York by
The College of Architecture at Texas Tech Uni-
versity in cooperation with the Historic American
Buildings Survery and the National Park Service,
as representative examples.
On the other hand, the use of these systems
has proliferated in the so-called “Reverse engi-
neering” consisting of duplication of pieces or
components without the aid of drawings and/or
documentation, by creating 3D virtual models
from an existing object for obtaining a copy as
accurate as possible.
In terms of the metrics, the modelling process
is concluded with the generation of the mesh of
triangles that represent the surface of the object
by a finite set of surface elements (triangles). But
from the standpoint of representation, the results
can be improved by applying textures to the surface
that resemble those of the constituent material of
the object, simulating different lighting effects
and getting shaded representations of the object
from different viewpoints, including stereoscopic
images and animations.
Figure 5. (A) Modelling La Lonja de Valencia (B) The Central vault of La Lonja de Valencia
980
Our experience with the station Leica TC 305
(direct measure without prism) had allowed us to
work with beams of discrete points, interconnected
by minimum squared adjustments. The equipment
Geosystem Leica Cyrax 2500 made it possible
to continue this line of study using continuous
sequence of points, applying 3D laser technology
to archaeological pieces and other heritage items.
(See Figure 5 and Figure 6)
The methodology consisted of placing the
scanner on a tripod and connected to a laptop.
The scanner was oriented to the selected area to
scan and it began the acquisition of data. The
equipment needed only fifteen minutes to complete
the acquisition providing in real-time the point
cloud and triangle mesh.
The precision was ±4 mm in distances and
±60 microradians in angles. The ability to data
capture reached 1,000 points per second, achieving
a precision in the beam of 1.2 mm. The cloud of
points was managed by the specific software that
has modules for calculation and modelling in two
and three dimensions. As the fourth coordinate,
it was recorded the light intensity.
In the project “La Lonja de Valencia (Spain)”
(See Figure 5 and Figure 6), they were carried out
additional data captures with the laser equipment
in remote areas, in order to represent the vaults
of the main courtyard. It is surprising how easily
is the process of providing textures to the model
Application of Topographical Capture Techniques for Modelling Virtual Reality
Figure 6. Three-dimensional model of the facade of the church of Santa Teresa (Ávila) by the 3D laser
scanner Trimble GS200
using the software provided with the laser scaning
equipment. This task can be performed in situ,
having the final image just after finishing the data
capture of the point cloud.
Within this general framework for data acquisi-
tion, we search for new models of cartographic rep-
resentation and experiment with new technologies,
always maintaining the metric relations between
the data, and creating models or prototypes. In this
way, through the digital models, we extended to
three dimensions the baseline scenario, without
losing the possibility of obtaining traditional
cartographic models.
HUMAN REPRESENTATION
ACHIEVEMENTS IN THE
FIELD WORKS
Since ancient times, men have tried to show their
immediate reality and the environment around
them, either as art or as for technical proposes,
focusing on the representation of the human figure
in motion and trying to exaggerate its traits. Until
today, the representation of the human figure and
its dynamics has been a hotspot for artists, think-
ers and scientists, using variety of materials and
technical supports for that.
As a reality or necessity, it become a science
study: kinanthropometry, which seeks to measure
and represent the human figure, moving or static,
with the greatest realism possible, to apply their
findings to the various sports or in addition to
advances in medicine (Carter, 1982).
The word Kinanthropometry derives from the
Greek “Kinein” which means movement, move-
ment, “Anthropos” which means human, and
finally “Metrein”, which means measurements.
Carter (1982), defined it as “the study of size,
proportion, maturation, body composition and
form, to describe the physical characteristics,
assess and represent the growth, nutrition and
the effects of training”. We can consider it as
an interface between the quantitative morphol-
ogy and physiology, or structure and function to
help interpret the dynamics of growth, exercise,
nutrition and the influence of the physical human
movement.
In the Esparza’s work (1993), we find refer-
ences explaining who the body shape and perfor-
mance are intimately related and the typology of
the athlete plays a decisive role in sports perfor-
mance. This means that the ability to perform work
or exercise will be closely related to the amount
and proportion of different tissues and body seg-
ments that make up the human body and to the
economy in the management of the movements.
This science was born as an interdisciplinary
scientific field (anthropologists, doctors, nutri-
tionists, physiologists, biomechanical, physical
education teachers, among others), to unify
criteria and methods in the assessment study of
981
 Application of Topographical Capture Techniques for Modelling Virtual Reality
man in motion. It aims to analyze the correlation
of morphological and physiological responses,
considered as a bridge between the relationship
structure-function.
Advances in techniques of virtual represen-
tation has evolved dramatically, but always as-
sociated with photogrammetric, optoelectronic
or video systems, to get sensations very close to
reality, but having insufficient density of points
and especially metrics, they have been unrealistic
simulations until now.
In parallel with these developments, topo-
graphic and cartographic techniques have been
progressing in their capture systems, making the
presentation of information almost in real time, but
its implementation has been traditionally limited
to the representation of the earth’s surface and, in
some cases, very recently, to the documentation
of the artistic heritage, but always away from the
systems for virtual representation in applications
which are in the field of topography or cartography.
Today, the graphic representation of the envi-
ronment and humans, as part of it, is a necessary
complement to a number of activities: documenta-
tion of heritage, traffic accidents, reconstruction
of criminal scenarios, civil engineering, anima-
tions and films, etc. The kinematic and kinetic
variables obtained allow statistical studies and
simulations using virtual models (digital models)
to describe the behaviour of human beings and
their environment. However, the techniques used
are limited by the method of capture, in which
the variables space and time play a decisive role
in obtaining operating and accurate results in a
short space of time.
By linking topographic techniques, biome-
chanics, and virtual modelling applications for
the capture and representation of human body
and its movement, we can obtain results both
on the functional variability and on the physical
variability of subject in motion, and understand
how these components interact and how they
could be improved. The study of the dialectical
relationship between structure and function is a
982
basic tool with a strong genetic dependence for
the identification of future athletes.
We can consider that one of the basics of biome-
chanics is the multidisciplinary study of the mea-
surements of the human body, its movement and
derivations thereof. Under this multi-character,
great advances have been developed through the
interest and perseverance of many researchers and
their contributions, to incorporate new methods
and refine the measurement instruments.
In recent decades, and in parallel with the evolu-
tion of other disciplines that work with graphical
representation or use graphic systems as a support,
they have tried to analyze and explain several as-
pects of the human figure and its movement in the
environment that surround it, being a clear example
of this many sporting disciplines, in which it is
necessary to establish procedures for evaluating
and improving the athletes performance. Thus, we
see how conventional training in elite athletes has
been complemented by advances in technology
to analyze and quantify the parameters involved
in the performance. Nowadays, it is impossible to
reach the international level in any sport without
the assistance of high technology.
TECHNIQUES USED TODAY
IN BIOMECHANICS
The use of photography has been for over a century
the most relevant technique applied to biomechani-
cal analysis of human movement. Photogramme-
try is “a set of techniques to obtain information
regarding the size, position and orientation of a
physical object and its environment, through film,
measurement and mathematical calculations on
these measurement”, where the spatial coordinates
of the points of interest (anatomical points) are
calculated from the planar coordinates of images
from at least two cameras (Woltring, 1995).
In these techniques, data acquisition requires
the matching of points to represent by the opera-
tor (passive or active markers in terms of bright-
Application of Topographical Capture Techniques for Modelling Virtual Reality
ness), and the individual capture of each one of
them. After selecting and measuring each point
and with a laborious calculation of the data, it
could be reconstructed the element under study
(in example a double somersault). This technique
of modelling followed a line of work that went
“from the point to the surface” and with a huge
human intervention in the intermediate stages
(Soto, 1995).
In these measurement protocols, entries were
limited to taking the spatial position of a small
number of points (markers), all attached to the
body surface, and whose motion approximates
the motion of the segments that form the union of
these points (Vera, 1989; Zatsiorski & Donskoi,
1988). This capture method represents a partial
aspect of reality, because it transform the informa-
tion obtained from sensors (limited in number and
position) from an recorded image and their further
processing, digitalization and modelling, into a
numerical representation that can be processed
by a computer, forcing all to a high consumption
of resources and time.
Subsequently, they began to be used for opto-
electronic systems, based on the use of TV cam-
eras capable of detecting the position of passive
or active markers in terms of brightness, in real
time. These systems have been employed in the
field of biomechanics of human movement under
laboratory conditions. In the 90s, even today,
There are used systems as ELITE (Elaboratore di
Immagini Televisive, Center of Bioengineering of
Milan, 1983), VICON VX (VInes CONverter for
Biomechanics, University of Strathclycle, 1989),
PREMIUM (Precision Motion Analysis System),
or Expert HirRes Vision 3D (High Resolution 3D
Motion Analysis System). They are classified
according to the sensors they use: PSD (Position
Sensitive Devices), CCD (Charge Coupled De-
vices, online or in area), among others.
All this systems are multi-camera equipments.
The coordinates of the markers are computed in
almost real time, which partly reduces the further
process of data. But we still have a geometric
picture of the human figure, formed by a limited
number of points, so the result is very biased and
poorly defined.
These systems, and other similar with little
variations, always rely on the use of active or
passive markers, being the passive the ones with
a best performance (better space-time resolution
and non-simulated identification of markers), but
it is always necessary to use interpolation proce-
dures to estimate the position of the markers in
the same instant of time, resulting in a distortion
of the final results. All this leads us to have some
systematic errors caused by distortion of the image,
scale errors, accidental errors (incorrect place-
ment of the markers, errors in scanning, error in
the mathematical processing, etc.), which cause
problems in the reconstruction of coordinates in
space-time and, as result, in the representation of
the human figure and its movement.
At present the use of Automatic Motion Cap-
ture 3D, such as VICON, determines the position
of reflective markers attached to the body of the
subject. These are taken up by a set of 6 infrared
cameras that record the movement with frequen-
cies between 60 and 1000 Hz. This capture must
be done indoors, under controlled and calibrated
conditions, as it is necessary to have geometric
reference system. This condition affects the ath-
lete’s actions, because they do not perform on
their natural work conditions and environment.
The main purpose of the system is to collect
the movements of an athlete to apply them to a
later animation work, resulting in a film of a few
seconds of the performed movement. We must
consider that each second of animation requires
approximately one hour of post-processing work
by computer systems and specialized software, as
Mocap and Motion Builder, and after that tedious
process, you get a 3D animation of the athlete
which is feasible for biomechanical analysis
because it include a set of mechanical variables.
This system is widely used in the film industry
(Titanic, Star Wars, The Mask, Harry Potter, etc.).
983
 Application of Topographical Capture Techniques for Modelling Virtual Reality
Within the work done in Spain at present, apart
from the above, you can highlight the following:
• The Institute of Biomechanics of Valencia
(IBV), where the system “Kinescan/IBV”,
similar to VICON, can perform a com-
plete analysis of movements in 3D based
on digital video technology with reflective
markers. This system has a post-process of
the captured images by manual digitizing,
and as a result, the application provides
information about the three-dimensional
movement of people in space: velocity, ac-
celeration, mechanical impulses, moments
angular or energy.
• The Faculty of Physical Activity and Sport
at the University of Granada, particularly
the Biomechanics Research Group. It has
to be highlighted their work on three-di-
mensional photogrammetry, performance
evaluation and quantitative analysis of
sport.
• The group of Biomechanics and
Ergonomics of the Center for Innovation
in the Industrial Enterprise (CINEI),
Department of Technology of the
Universitat Jaume I of Castellon. To mea-
sure the movement they use high speed
B/W digital video cameras (Speedcam
+500), which allow an analysis of 2D mo-
tion. Complementing, they have digital
cameras (used for postural measures) and
specific gloves to measure the position of
the hands (18 sensors Cyberglove).
In conclusion, we note that the common de-
nominator of all these systems is that the meth-
odology for obtaining representative points of
the human figure:
• They use capture system based on video or
digital photography.
984
• It is necessary to use markers attached on
the human body, which limits the number
of metric points to be captured.
• The data collection is done “in door” and
under predefined conditions.
• They have a very laborious post-treatment
process.
• In some cases, between the data collection
and the delivery of the results, there is a
long time and the conditions of the athlete
may have changed.
A NEW PROPOSAL BY THE
USE OF SYSTEMS BASED
ON LASER CAPTURE
Some years ago, our working group was looking
for new capture and modelling systems. We tried to
link capture and mapping cartographic techniques
and the systems used in virtual reality techniques,
which did not allow metric relationships between
the reality and the final model. In this way, we
extended to three dimensions the stage. Without
losing the reference to the traditional mapping
models, we conduct tests that would allow us to
develop a simple working procedure, valid for
obtaining rapid and real representation of reality
with above all, dimensional metrics.
As a subject of study, we focused on the hu-
man body because, as we have seen in previous
examples, the static object was being handled at
that time with very positive results.
The road was not easy because the laser sys-
tems without prism were not very common as
there were only three companies that sold them
and they were very expensive, so that, the num-
ber of equipment in Spain was reduced and its
availability for our tests was very limited. But it
was a fact that its use would allow us to test data
capture without discrimination of markers, and
to capture the whole human figure, not only a
limited number of points. If we add the rapidly in
the point capture, completely and unlimited, and
Application of Topographical Capture Techniques for Modelling Virtual Reality
the small post-processing required to achieve a
three-dimensional digital image with real coordi-
nates (on a grid and capable of performing actual
measurements on the same), the methodology al-
lowed us working with special data on real-time,
where the space-time relationship, it was defined
at the time that the capture was made.
Our first objective was to obtain inertial
parameters suitable for modelling of the human
body, involving on it the use of anatomical points
and segments interconnected under a local or real
reference system where there were an interrelation-
ship between them and the rest of points of the
human body under study. In order to do that, the
following practical experiences were carried out:
• We studied, at the technical level, the work
that had been carried on the heritage, fo-
cusing on statues and irregular patterns. In
this review, we obtained the first informa-
tion to plan further experiences, focusing
on the representation of the human figure.
• Initially, in order to get an idea of the good-
ness of the system and initial capture pa-
rameters, we planned the capture of a static
non-human element; we used a wood mod-
el commonly used by painters. The results
were very exciting because we will have
in a short period of a three-dimensional
model of high quality.
Figure 7. (A) Picture of the laboratory test Figure Madrid -2005 (B) The digital model obtained Madrid
-2005
• In the context of the final work of the
degree “A comparative study of details
from a survey of classical precision and a
lift with 3D laser system: The Warrior of
Xian” (School of Topography, Geodesy
and Cartography), it was carried out a pi-
lot study with a static human figure, show-
ing as result a mesh of points equidistant,
which encouraged us to continue our in-
vestigations in this direction.
• Then, we planned the first study of the
male human figure, using a model (Figure
7). They were made several shots under
different static positions of the body in or-
der to obtain as much information as pos-
sible from the results. The results provided
us with a cloud of point at intervals of 5
mm, so we got a continuous graphical rep-
resentation of the human figure.
• We proceeded to repeat the previous test,
but this time with a female human figure.
(See Figure 8)
In both cases the results were very acceptable
and provided us with a very rich experience for
future experiments.
In a second test, the high-speed capture pro-
tocols of the equipments (1 to 3 seconds) were
used to provide some results on the environment
and movement occurring in it.
985
 Application of Topographical Capture Techniques for Modelling Virtual Reality

Figure 8. (A) Picture of the test with a female model Madrid -2005 (B) Digital pictures obtained by
rotation of the model Madrid -2005 (C) Detail of the obtained digital female model Madrid -2005

The results were more striking than those
obtained previously, because in seconds we had
available on the computer screen a raw 3D image
of the surroundings without any post processing.
However, it appeared the problem of shadows and
blind spots produced by the system itself.
• After analyzing the results obtained for
whole-body models, we proceeded to
perform a test of a body part in more de-
tail.
This was based on the study of human
arm. Implemented within the framework
of a final work of the degree at School of
Topography, Geodesy and Cartography
(Ibáñez & Jiménez, 2006): “An approach
to anthropology in sport by mapping tech-
niques based on laser-scanning systems:
Survey a 3D human arm.”
The results were
acceptable, showing metrics precisions
around 6.3 mm between the mesh points.
An unexpected result was the colour tex-
tures offered by the model after texturing,
once analyzed the colour textures seemed
to correspond with muscle relaxation and
contraction, opening this results another
line of possible future research.
• After carrying out successfully the last
experiences, we were ready to perform

986
Application of Topographical Capture Techniques for Modelling Virtual Reality
the first study of the movement, which
had not been previously performed in the
world. The results were not the most opti-
mal, but they served to correct the capture
parameters and to design the following
experiences.
• Based on the analysis of previous results,
we proceeded to design and conduct a test
in laboratory conditions. We design a ramp
with a known slope for rolling down a
ball which was moving with a controlled
speed. With this test we had two targets:
(See Figure 9)
◦◦ Study of the movement of the ball
rolling down, considering it as a 3D
object.
◦◦ Study of the velocity and acceleration
of the ball in all its way at intervals
of 1 mm.
• With the above results, it was designed an
experiment in real environment conditions:
a race in the running track. We studied the
evolution of the velocities and accelera-
tions alone a 100 meters race. As a result
we can see in the picture below the accel-
eration or deceleration caused by steps of
the athlete, which form the “v shapes” in
the image. (Figure 10)
• At present we keep on working on differ-
ent experiences of motion capture and its
possible applications to sports, without
Figure 9. (A) Picture of the setting of the study of velocity and acceleration of a ball Madrid -2007 (B)
Picture of the results of the study of velocity and acceleration of a ball Madrid -2007
conclusive results but very hopeful. In
parallel, we are working on systems for
real-time 3D representations by virtual
images (holograms), in order to perform
an accurate representation of the human
body. In this vein, we have opened a way
for research by looking at metrics in video
capturing systems and its real 3D represen-
tation, which combined with 3D laser cap-
tures, could allow a process of capture and
analysis almost in real time.
CONCLUSION
In the last five years, the 3D laser capture systems
have had a stunning technological improvement,
and its tendency is to achieve a higher improve-
ment in medium term. The situation is similar to
the development of conventional photography:
100 years ago a man should pose for not less than
5 minutes in order to have a picture, and now in
seconds we got a lot of high quality shots. Thus,
following the example, these systems are evolv-
ing in the same way but in very short period of
time, they are evolving from low resolution and
high capture time to high resolution and shorter
capture times in just one decade.
The members of this working group are in
direct contact with the companies who sell these
987
 Application of Topographical Capture Techniques for Modelling Virtual Reality
Figure 10. Results of the acceleration of a 100 m athleteMadrid - 2008
systems, and based on their experience, they
understand that in 5 years time these equipments
will be affordable to all individuals who need
them (their price today is high) and they will take
shorter to perform a high quality real-time capture
of a moving object. However, the working prin-
ciples and calculations that have been described
in this work will not vary much and, therefore, we
believe that the results we are getting are useful
not only for the development of better equipments
by companies, but also for future researchers in
order to continue this exciting work.
Is worth noting the support and interest aroused
by this study in institutions such as EPOCH-
European Network-European Research Network
on Excellence in Processing Open Cultural
Heritage, IST-2002-507382, framed in the Sixth
Framework Program of the European Union, and
in other sports clubs, research institutions (traffic
accidents, criminal, insurance, etc..) and Technical
Faculties (engineering, aeronautics, seismology,
mining, archaeology, etc.).
As we have commented, the project is cur-
rently underway. The 90% of the program has
been already concluded, so we can offer partial
results of the work done. Some the initial results
will be released in the doctoral thesis of Pedro
Calvo Merino, under the title “Three-dimensional
988
modelling of the human figure using mapping and
its application to the moving object”, which is in
phase of completion.
REFERENCES
Adams, L. (1992). Programación gráfica. Técni-
cas avanzadas de modelado, acabado y animación
3D. Madrid, España: ANAYA Multimedia SA.
Allard, P., Stokes, I. A. F., & Blanchi, J. P. (1995).
Three-dimensional analysis of human movement.
Chicago: Human Kinetics.
Barber, D., Mills, J., & Bryan, P. (2004). Towards
A Standard Specification For Terrestrial Laser
Scaning. In Cultural Heritage. Presented at
International Society for Photogrammetry and
Remote Sensing. Antalya: Istambul.
Borg, C. E., & Margin, M. (2003). Escáneres 3D de
largo alcance: ¿Avanzando hacia una herramienta
híbrida o hacia una metodología híbrida? Datum
XXI, 1(5), 42–46.
Bracci, S., Falletti, F., & Scopigno, M. M. R.
(2004). Explorando David: diagnóstico y estado
de la conservación. Roma, Italia: Giunti Press.
Application of Topographical Capture Techniques for Modelling Virtual Reality
Carter, J. E. L. (1982). Physical structure of
Olympic athletes. Part I: The Montreal Olympic
Games anthropological project. Basel, Switzer-
land: Krager.
Demir, N., Bayram, B., Alkış, Z., Helvaci, C.,
Çetin, I., & Vögtle, T. (2004). Laser scanning
for terrestial photogrammetry, alternativa system
or combined with traditional system? Presented
at International Society for Photogrammetry and
Remote Sensing, Istambul.
Donskoi, D., & Zatsiorski, V. M. (1988). Bio-
mecánica de los ejercicios físicos. La Habana,
Cuba: Pueblo y Educación.
Esparza, F. (1993). Manual de cineantropometría.
Pamplona, España: FEMEDE.
García, P. (n. d.). Kinantropometria o cineantro-
pometria: Definición y alcance social [Versión
electrónica]. Universidad de Caracas. Retrieved
February 8, 2005, from http://www.rendeportin.
com.ve/kinan.htm
Ghosh, S. K. (1979). Analytical photogrammetry
[Forogrametría analítica]. New York: Pergamon
Press.
Gonzalez-Aguilera, D. (2001). Consideraciones
sobre el análisis de la fiabilidad en el patrimonio
edificado. Fiabilidad y fotogrametría arquitec-
tónica [Versión electrónica]. REDCientifica.
Retrieved November 8, 2004, from http://www.
redcientifica.com/imprimir/doc200111070001.
html
Hogarth, B. (1996). Dynamic Anatomic. Köln,
Germany: Benedikt Taschen Verlag GmbH.
Ibáñez, S., & Jiménez, J. (2006). Aproximación a
la antropología en el deporte mediante técnicas
cartográficas basadas en sistemas láser-escáner:
Levantamiento 3D de un brazo humano. Unpub-
lished bachelor dissertation, Universidad Politéc-
nica de Madrid (UPM).
Ikemoto, L., Gelfand, N., & Levoy, M. (2003). A
hierarchical method for aligning warped meshes
[Un método jerárquico para alinear acoplamien-
tos combados]. Paper presented at the Fourth
International Conference on 3D Imaging and
Modeling (3DIM), October 6-10, 2003, Banff,
Alberta, Canada.
Leica Geosystems. (2005). Documentación Téc-
nica y Manuales de Referencia sobre sistemas
láser [Electronic Version]. Retrieved from http://
www.leica-geosystems.com/es/index.htm
Lerma, J. L. (1999). Reconocimiento de materiales
y deterioros en fachadas arquitectónicas. Datum
XXI, 1(0), 25–27.
Leva, P. (1994). Adjustments to Zatsiorsky-Seluy-
anov’s segment inertia parameter. Kinesiology
Department. Bloomington, IN: Indiana University.
Levoy, M., Rusinkiewicz, S., Ginzton, M., &
Ginsberg, J. (2000). The digital Michelangelo
project: 3d scanning of large statues. Department
of Computer Science and Engineering, University
of Washington. Retrieved October 6, 2004, from
http://graphics.stanford.edu/papers/dmich-sig00/
dmich-sig00.html
Moeslund, T. B., Madsen, C. B., & Granum, E.
(2003). Modelling the 3d pose of human arm and
shoulder complex utilisima only two parameters.
Paper presented at the meeting International
Conference of Model-based Imaging, Rendering,
Image Analysis and Graphical Special Effects,
Rocquencourt, Francia (pp. 11-19).
Ojeda, J. C., Martinez, R., Gonzalez, F., &
Sanchez, J. A. (2002). Generación de modelos
tridimensionales de curvas y túneles. Mapping-
Interactivo, Artículo 180. Retrieved October 4,
2004, from http://www.mappinginteractivo.com/
plantilla-egeo.asp?id_articulo=180
989
 Application of Topographical Capture Techniques for Modelling Virtual Reality

Rusinkiewicz, S., & Levoy, M. (2000). QsPlat: Vera, P. (1989). Técnicas de análisis en bio-
A multiresolution point rendering system for mecánica deportiva: estado actual y perspectivas.
large meshes. Paper presented at the meeting Presented at Jornadas Unisport sobre tecnología
SIGGRAPH 2000, Computer Graphics Proceed- del deporte, Málaga.
ings (pp 343-352). Retrieved September 7, 2004,
Vozikus, G. (2004). Lasser Scanning: New method
from http://graphics.stanford.edu/papers/qsplat/
for recording and documentation in Archaeology.
qsplat_paper.pdf
Paper presented at the meeting FIG Working
Salinas, F. S., & Velilla, C. (2003). Estudio mé- Week of Athens. Retrieved October 13, 2004,
trico por fotogrametría terrestre: Documentación from http://www.fig.net/pub/athens/papers/wsa1/
de pequeños monumentos [Versión electrónica]. WSA1_4_Vozikis_et_al.pdf
Mapping-Interactivo, Artículo 175. Retrieved
Warn, H., Marschner, S., Levoy, M., & Hanrohon,
October 4, 2004, from http://www.mappingin-
P. (2000). A practical model for subsurface light
teractivo.com/plantilla-egeo.asp?id_articulo=175
transport. In SIGGRAPH 2000 (Computer Graph-
Shulz, T., & Ingensand, H. (2004). Terrestrial ics Proceedings).
Lasser Scanning: Investigations and Applications
Woltring, H. J. (1995). Smoothong and differentia-
for High Precision Scanning. Paper presented at
tion techineques applied to 3D data. Champaign,
the meeting FIG Working Week of Athens.
IL: Human Kinetic.
Soto, V. M. (1995). Desarrollo de un sistema para
Zatsiorski, V. M., & Donskoi, D. (1988). Bio-
el análisis biomecánico tridimensional del deporte
mecánica de los ejercicios físicos. La Habana,
y la representación gráfica realista del cuerpo
Cuba: Pueblo y Educación.
humano. Tesis Doctoral, Facultad de Ciencias de
la Actividad Física y el Deporte, Universidad de
Granada, España.

This work was previously published in Virtual Technologies for Business and Industrial Applications: Innovative and Syner-
gistic Approaches, edited by N. Raghavendra Rao, pp. 181-200, copyright 2011 by Business Science Reference (an imprint of
IGI Global).

990
 991

Chapter 61
GIS Implementation
in Malaysian Statutory
Development Plan System
Muhammad Faris Abdullah
International Islamic University Malaysia, Malaysia

Alias Abdullah
International Islamic University Malaysia, Malaysia

Rustam Khairi Zahari

International Islamic University Malaysia, Malaysia

ABSTRACT
The chapter presents the current state of GIS implementation in Malaysian development plan system.
It offers an overview of GIS implementation worldwide, touching briefly on the history of GIS, plan-
ners’ early acceptance of the system, factors that promote GIS implementation, level of usage among
planners, and factors that impede successful GIS implementation. At the end, the chapter provides a
comparison between the state of GIS implementation in Malaysian statutory development plan system
with the state of GIS implementation worldwide. The evidence was derived from three main sources:
literature, empirical observation of GIS implementation in Malaysia, and a survey conducted in 2008.

INTRODUCTION
After over two decades since its introduction into
the planning fields, geographical information
systems (GIS) has become one of the important
tools-of-the-trade for planners (Ceccato & Snick-
ars, 2000; Drummond & French, 2008; Gocmen,
2009). Despite planners’ early resentment towards
DOI: 10.4018/978-1-4666-2038-4.ch061
GIS, they have now become one of the most fre-
quent users of the systems (Budic, 2000; Ceccato
& Snickars, 2000; Geertman, 2002; Gilfoyle &
Wong, 1998; Gocmen, 2009). In describing plan-
ners’ early resentment towards GIS, Klosterman
(1997) points out that this was largely due to past
failures of efforts to computerize planning, such
as the failure of large-scale urban modelling.
Early GIS implementation was also expensive,

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 GIS Implementation in Malaysian Statutory Development Plan System
and the software was highly complicated for
planners liking.
Beginning in the middle of 1980s, the adoption
of GIS among local governments in the United
States of America began to increase slowly, and
then sharply in the 1990s (Drummond & French,
2008). Researchers attributed this change in GIS
adoption rate to better and cheaper hardware and
software availability, as well as better GIS data
accessibility (Drummond & French, 2008; Geert-
man, 2002; Gilfoyle & Wong, 1998; Klosterman,
1999).
Key actions by governments also helped to ac-
celerate GIS adoption in the 1990s. For instance,
the publication of Chorley Report in 1987 has
helped to increase GIS awareness and provided
fundamental directions for GIS development in
Britain (Gilfoyle & Wong, 1998). In Wisconsin,
U.S.A., the state-funded Wisconsin Land Informa-
tion Program and the enactment of the Compre-
hensive Planning Law of 1999 have significantly
contributed to increase GIS adoption in the state
(Gocmen, 2009).
Since the 1990s, the use of GIS among planners
has been widespread. Planners began to adopt this
‘new’method in the course of their work, especially
in terms of map-making and land data storing.
GIS-based information systems were developed
and deployed to allow planners and stakeholders
better access to information and data (Craglia &
Signoretta, 2000; Gilfoyle & Wong, 1998; Heeks,
2002). However, widespread implementation
of GIS in the planning field does not translate
into full utilization of GIS application. Research
indicates that planners’ regular use of GIS has
been largely limited to the basic functions of the
systems, such as mapping and accessing land
information, for routine operational and manage-
ment tasks including permit processing, land data
storing and map presentation (Budic, 2000; Gill,
Higgs, & Nevitt, 1999; Klosterman, 1997; Men-
necke & West, 2001). Even at present, prevalent
use of GIS among planners continues to centre
on rudimentary applications while advanced ap-
992
plications, such as spatial analysis and modelling,
remain underutilized (Gocmen, 2009).
While GIS application among planners con-
tinues to be underutilized, the fate of GIS-based
information systems is more difficult to assess,
mainly because such assessment is highly sub-
jective and timing-dependent (Heeks, 2002).
What is considered a success to one person may
be a failure to another, and what is considered a
success today may be a failure tomorrow. The
difficulty in assessing the success and failure of
GIS-based information systems is made worst
by the propensity of system developers to only
report success story. However, several authors
suggest that there are many cases of failed GIS-
based information systems (Abdullah et al, 2002;
Heeks, 2002; Lee & Ahmad, 2000). For instance,
Heeks (2002) says that,
On the basis of…these surveys, one may estimate
that something like one-fifth to one-quarter of
industrialized-country IS (i.e. information system)
projects fall into the category of total failure
category; something like one-third to three-fifths
fall into the partial failure category; and only a
minority fall into success category. (p. 102)
Nevertheless, researchers tend to agree that
organisational factors are more important than
technological ones in ensuring a successful
implementation of an information system (Abdul-
lah et al., 2002; Heeks, 2002; Innes & Simpson,
1993; Ramasubramanian, 1999). Organisational
mission to implement and support information
system, retaining key and trained staff involved
in the development and implementation of the
system, data availability, and system designers’
understanding of users’ needs are among the or-
ganisational factors identified as key ingredients
for a successful information system.
GIS Implementation in Malaysian Statutory Development Plan System
BACKGROUND
Malaysian Statutory
Development Plan System
Town and country planning system in Malaysia is
governed by the Town and Country Planning Act
19761 (Johar, 2003; Abdullah & Ariffin, 2003).
Central to this Act are the requirements for plan-
ning authorities in Malaysia to formulate planning
and development policies through preparation of
statutory development plans and to regulate de-
velopment through development control process.
The statutory development plans required under
the Act are national physical plan, state structure
plan, local plan and special area plan.
The administration of this Act is carried out
at three different governmental levels which are
the Federal, the State and the Local Government
(Ahmad Sarkawi, 2006). Planning authorities are
established at each level of the government to un-
dertake planning of different scale and functions.
The Town and Country Planning Act 1976
was amended in 2001 to make provision for an
improved system of statutory development plans
in Malaysia. The 2001 amendment introduced
additional set of plans into the system, which
Figure 1. Malaysian statutory development plan hierarchy
aims at introducing a more integrated and holistic
approach to planning, and enhancing the control
and regulation of planning in Malaysia (Ahmad
Sarkawi, 2006). Prior to the amendment, the Ma-
laysian statutory development plan system was a
two-tier system consisted of structure plans and
local plans. Following the amendment, the system
now consists of a hierarchy of development plans
ranging from national to local levels. At the top of
the hierarchy is the national physical plan. This
is followed by the state structure plan, the local
plan and the special area plan (Figure 1).
The national physical plan sits at the top of
the hierarchy of statutory development plans in
Malaysia. Prepared by the Federal Department of
Town and Country Planning (FDTCP), the plan
provides strategic development framework for
the whole of West Malaysia. The plan was ap-
proved by the National Physical Planning Coun-
cil in 2005 and shall be in effect until 2020. It is
subject to revision every five years.
The state structure plan is prepared by the
State Department of Town and Country Planning
(State DTCP). The plan translates the development
framework of national physical plan to state level,
and provides planning policies and proposals for
the whole area within the state boundary. The
993
 GIS Implementation in Malaysian Statutory Development Plan System
form and functions of state structure plan remain
similar to the ‘old’ structure plan prior to the 2001
amendment of the Town and Country Planning
Act 1976. The only difference is that the current
state structure plan is a state-wide plan, while the
previous structure plan was prepared for a part of
an area within state boundary. At present, eight
out of eleven states in West Malaysia already had
their state structure plans published in the official
gazette. Once published, the plan becomes a legal
document that guides the planning and develop-
ment within the state.
The 2001 amendment also affects the local
plan. Although the form and functions of local
plan remain the same as before the amendment,
the plan is now a district-wide plan. Previously,
it covered only a part of an area under a local
planning authority’s jurisdiction. Thus, the present
local plan is commonly referred to as the ‘district
local plan’ in order to distinguish it from the ‘old’
local plan.
Local plan is prepared by local planning
authority to set out detail policies and proposals
pertaining to development and use of land in the
district. The plan also translates strategic policies
and proposals of state structure plan into details to
be implemented at the district level. Usually this
involves detailing the exact location, components
and the cost of development. At present, a total of
108 local plans have been published throughout
West Malaysia.
Similar to the local plan, the special area
plan is also prepared by local planning authority.
Nevertheless, special area plan covers a smaller
area in comparison to the local plan and contains
proposals that are more detailed in nature. Special
area plan is prepared for an area that requires
special planning because of its unique character.
At present, special area plans have been prepared
for areas that require special conservation efforts
because of their historical or natural environmental
value, and for areas that are subjected to intense
development pressure.
994
In general, the contents of these statutory
development plans consist of written statements
and maps. The written statements present the
planning policies and development proposals for
the planned area, while the maps act as supporting
documents that clarify the policies and proposals
through graphical representations. The national
physical plan and the state structure plan are subject
to revision every five years. On the other hand,
the revision of a local plan can be performed by
the local planning authority whenever necessary.
Preparations of all statutory development plans
in Malaysia are outsourced to private planning
consultants. A team of planning consultants is
appointed for each statutory development plan
preparation project and is responsible for collect-
ing data, performing analyses and writing reports.
Meanwhile, the planning authorities administer
the preparation process by providing supervision
and funds, as well as handling the publicity and
the gazette process of the plan. The planning
consultants, based on analyses of the collected
data, propose planning policies and development
proposals for the planned area for commentary
by the planning authorities. Once finalised, the
plan will be put on public display before being
published in the government’s gazette. Figure 2
presents the flow chart of a typical statutory de-
velopment plan preparation process in Malaysia.
Although there are other types of development
plans prepared by planning authorities in Malay-
sia, such as regional plans and rural growth cen-
tre plans, their preparations are not bounded by
the requirements in the Town and Country Plan-
ning Act 1976. Thus, these plans are sometimes
termed as non-statutory development plans. For
the purpose of this chapter, any reference to de-
velopment plans in the following discussions will
refer to the statutory development plans.
GIS Implementation in Malaysian Statutory Development Plan System

Figure 2. Typical development plan preparation process and integration with GIS (Adapted from Jabatan
Perancangan Bandar dan Desa Semenanjung Malaysia, 2001 and Jabatan Perancangan Bandar dan
Desa Semenanjung Malaysia, 2007)

995
 GIS Implementation in Malaysian Statutory Development Plan System
History of GIS Implementation
in Malaysian Statutory
Development Plan System
GIS was introduced in Malaysia in the early
1980s. Early record shows that GIS was first
used by Malaysian Department of Agriculture in
1981 (Idrus & Harman Shah, 2006). It was only
over a decade later that GIS was officially used
in planning in Malaysia.
The year 1993 marked the first GIS use in
Malaysian planning when Putrajaya, the new
Malaysian Federal Government Centre, was being
planned (Abu Bakar, 2004). In the process, GIS
(MapInfo version 1) was used, but only for map
presentation. In that same year, the Local Plan for
Parit Buntar and Bagan Serai area was also being
prepared. This presented another opportunity for
FDTCP to try to use GIS in planning. Thus, the
first executive information system for a develop-
ment plan, known as SMaRT, short for Sistem
Maklumat Rancangan Tempatan, or Local Plan
Information System, was developed for that local
plan (Abu Bakar, 2004).
Beginning with these two GIS initiatives, GIS
use in Malaysian planning gradually increased.
Early widespread implementation of GIS centred
on its implementation in the preparation of devel-
opment plans. By 1997, GIS was already embed-
ded into the preparation process of development
plans in Malaysia (Ibrahim et al., 2004). From the
initial focus on development plan preparation, GIS
was then implemented in other activities of plan-
ning including development control and land use
change monitoring, as well as being the platform
on which tailor-made e-planning systems were
developed (Abdullah, 2004).
Today, GIS is a common knowledge among
planners in Malaysia. Formal GIS training is now
offered to all planning graduates at all planning
schools in the country (Ibrahim et al., 2004).
On-demand GIS trainings are also offered by
various parties, including universities and private
GIS companies, to public and private planners.
996
GIS research and development are being pursued
actively with universities playing significant roles
in developing GIS model and exploring new areas
where GIS can add value in the planning process
(Abdullah, Abdullah, & Ibrahim, 2009; Samat,
2006; Tan, 2005).
Information System
Planning Masterplan
In 1996, the Malaysian Government launched
e-government initiatives as part of its efforts to
employ information and communication technol-
ogy (ICT) in order to reinvent the way it operates.
E-government was one of the seven flagship ap-
plications initiated under the Multimedia Super
Corridor programme (Lee & Ahmad, 2000). Under
the programme, government departments were
required to formulate and implement proposals
for integration of ICT in their work in order to
achieve speedier and better service delivery. In line
with the Malaysian Government’s e-government
aspiration, in 1997, FDTCP introduced its Infor-
mation System Planning Masterplan. This master-
plan outlines the department’s approach towards
integration of ICT into planning in Malaysia.
Nevertheless, many of the ICT systems proposed
to be developed under the masterplan have some
GIS elements in them. According to Mohd. Nazri
Abdullah (2004), based on the detail explanation
in the masterplan, it can be concluded that 80%
of the proposed systems would lead to some GIS
applications in planning. Among the ICT systems
proposed by the masterplan are:
• Development and Planning System;
• Planning Monitoring System;
• Planning Approval System;
• Counter Service System;
• Appeal Board System;
• State Planning Council Decision Support
System; and
• Data Warehousing and Information
System.
GIS Implementation in Malaysian Statutory Development Plan System

The masterplan recognizes the benefits of area. The plan would determine the type of de-
GIS to planning, and thus outlines measures to velopment allowable in an area, or land parcel,
integrate GIS into planning in a more coordinated as well as any restrictions (such as development
approach. Among the key thrusts of the masterplan density, plot ratio and building height) associated
pertaining to GIS are to increase GIS awareness to that land. The use of GIS would development
and knowledge among FDTCP officers, and to of an information system and its usage in the
promote GIS use in planning, especially in the development control process. Such information
development plan preparation process. system would allow users, especially council
planners, to quickly search for the land parcel
GIS Implementation in affected by a planning application using com-
Statutory Development Plan puter and to ascertain the planning policies and
Preparation Process the development control requirements related to
that land. Such system would also help speed up
The Information System Planning Masterplan’s the development control process considerably as
proposal to promote GIS use in development compared to manual search and reference tasks,
plan preparation process is perfectly appropriate especially when a plan may consist of more than
for several reasons. Firstly, development plan one hundred policies (refer Table 1). The less time
preparation involves the production of a signifi- taken to process a planning application would fit
cant number of maps. For instance, Terengganu very well into the e-government initiatives, which
State Structure Plan contains 62 maps and Perak seek for speedier service delivery.
State Structure Plan contains 52 maps (Table 1).
Nevertheless, it has to be noted that these are just Problems Faced During Early
the number of maps included in the published GIS Implementation Period
plan. The number of maps prepared throughout
the entire plan preparation process is notably Following the introduction of the Information
higher since there are also other reports like the System Planning Masterplan, FDTCP began its
inception report and the report of survey (Figure promotion of GIS. Officers from the department
2), which have to be produced in the process and
these reports contain a sizeable number of maps Table 1. Number of maps in published state
as well. In fact, report of survey usually contains structure plans
more maps than the published plan itself. There-
fore, even though the number of maps included No. State Structure Plan No. of No. of
maps policies
in Pahang State Structure Plan is only 15 (Figure
1. Terengganu State Structure 69 158
2), the total number of maps produced during the
Plan
Plan’s entire preparation process is considerably
2. Perak State Structure Plan 52 49
higher. Due to the high number of maps produced
3. Melaka State Structure Plan 29 165
during development plan preparation process,
4. Pulau Pinang State Struc- 29 79
many opportunities to use GIS arise in that pro- ture Plan
cess. Maps, which were usually prepared using 5. Selangor State Structure 26 100
computer-aided design (CAD) software (and in Plan
some cases using Microsoft PowerPoint) previ- 6. Negeri Sembilan State 25 111
Structure Plan
ously, can now be prepared using GIS software.
7. Pahang State Structure Plan 15 29
Secondly, statutory development plan forms
the basis for development control for the planned 8. Johor State Structure Plan n/a n/a

997
 GIS Implementation in Malaysian Statutory Development Plan System
were sent to GIS trainings and road shows were
conducted at various town planning departments
in the country. FDTCP also conducted several
workshops to discuss in detail the areas where
GIS can be used in the development plan prepa-
ration process and, later on, introduced GIS to
private planning consultants who were involved
in development plan preparation projects (Ibrahim
et al., 2004).
Early attempts to integrate GIS into Malaysian
statutory development plan preparation process
were not without problems. When initially in-
troduced, GIS received lukewarm reaction from
FDTCP officers, created confusion among private
planning consultants and failed to achieve what it
was intended to do. Mohd Ali Abu Bakar (2004)
reports that despite the promotion and the road
shows, GIS was not initially well received by
many officers from the professional category,
such as planners. Many of them perceived GIS
merely as a mapping tool and thus only useful
to sub-professional officers, such as technical
assistants and technicians.
When GIS was introduced to private planning
consultants, the reception was not any better as
well. To many of them, GIS was completely a
new thing and most were unclear about how GIS
was to be integrated into the development plan
preparation process. When further clarification
was sought, FDTCP’s response suggested that
what the department wanted was for digital maps
to be used in the preparation of the plans. How-
ever, as stated by Mansor, et al. (2004), initially
many consultants still decided not to use GIS. As
digital plans and maps could be produced easily
using CAD software, not all consultants decided
to venture into using GIS software in their work.
Instead, many of them opted to continue using
CAD software, which they were already using
prior to the GIS introduction.
Early implementation of GIS in the develop-
ment plan preparation process also faced the
problem of standardisation, or the lack of it (Ba-
hagian Teknologi Maklumat, 2006). At that time,
998
FDTCP did not provide any detail guidelines on
land use colour classification, database structure,
metadata structure and the like. Lack of detailed
guidelines led to non-standardised GIS outputs
prepared by the few consultants who decided to
use GIS in the plan preparation.
The absence of detailed guidelines from
FDTCP and lack of GIS knowledge among con-
sultants also resulted in limited use of GIS in the
development plan preparation process. GIS was
viewed as a map drawing tool and thus, used as
one. Maps were drawn using GIS software but
without data attributes. Spatial analyses were
still conducted manually despite the availability
of GIS software and digital maps. However, the
results of these analyses were then drawn onto
the maps using GIS.
Quality Control of GIS Outputs
Having learnt lessons from the problems faced
during the early period of GIS implementation
in the development plan preparation process, in
1999, FDTCP introduced measures to control
the quality of GIS outputs from the process.
A guideline on GIS application for local plan
preparation process was produced (Abu Bakar,
2004). Although brief in nature, the guideline was
rather useful as foundation towards standardisa-
tion of GIS outputs. The guideline was tested on
several local plan preparation projects before it
was finally improved. In 2001, it was officially
included in the revised FDTCP’s Manual for
Local Plan Preparation (Jabatan Perancangan
Bandar dan Desa Semenanjung Malaysia, 2001;
Abu Bakar, 2004).
Similarly, the FDTCP’s Manual for State
Structure Plan Preparation was also revised in
2003 to include guidelines for GIS application in
the plan preparation process. Other guidelines con-
cerning map preparation, metadata and database
structure were also produced. These guidelines
are constantly revised and improved. Today, the
level of details provided in these guidelines is
GIS Implementation in Malaysian Statutory Development Plan System
quite remarkable. For instance, the latest version
of FDTCP’s Manual Penyediaan Pelan GIS (or
GIS Plan Preparation Manual) goes to the extent
of providing step-by-step instructions on how to
prepare land use data layers (Bahagian Teknologi
Maklumat, 2007b). The guideline for Format
Metadata dan Struktur Pengkalan Data Sistem
Maklumat Geografi (GIS) Kajian Rancangan Tem-
patan (or Format for Metadata and GIS Database
Systems for Local Plan) is similarly detailed in
contents. The 78-page guideline includes samples
of metadata format for each land use layer and
pre-determines the name of each of the layers
(Bahagian Teknologi Maklumat, 2007a).
Another measure taken by FDTCP to ensure
standardisation of GIS outputs is the appointment
of GIS consultant into the team of consultants
that are involved in preparing a development
plan. When GIS was initially introduced into
the development plan preparation process, the
responsibility to prepare GIS maps was on each
of the planning consultant in the team. Such ar-
rangement lacked the necessary coordination to
ensure that all the maps produced were uniformly
formatted. With GIS consultant in the team, the
responsibility to coordinate maps preparation is
on the GIS consultant. Because of their advanced
knowledge in GIS technologies, GIS consultants
are also tasked to conduct GIS analyses, develop
GIS database and develop executive information
system for the plan (Jabatan Perancangan Bandar
dan Desa Semenanjung Malaysia, 2007). In terms
of map preparation per se, the responsibility
remains with individual consultants in the team.
However, increasingly, in many development plan
preparation projects, GIS consultants are also
given the task to prepare maps on behalf of the
planning consultants. In such cases, the planning
consultants still provide the required information
for GIS maps preparation to GIS consultants..
Level of GIS Application in Statutory
Development Plan Preparation
Despite having GIS consultants’ expertise at dis-
posal, present GIS use in most statutory develop-
ment plan preparation projects is still limited to
map preparation and presentation with minimum
GIS analyses being performed (Ibrahim et al.,
2004; Nor Salehi Kassim, personal communica-
tion, December 11, 2008). In most instances, GIS
analyses performed are restricted to basic analyses
like overlays and catchment (buffer) analyses (G.
L. Chua, personal communication, December
12, 2008). Nevertheless, although rare, there
were cases of advanced GIS analyses performed
in development plan preparation, such as in the
preparation of the national physical plan and the
Kulim District Local Plan (Abdullah, Abdullah, &
Shahbuddin et al., 2009). In both cases, GIS was
coupled with decision support systems to conduct
analysis and generate maps for land suitability
(Amandus Jr, 2006; Kassim & Islam, 2009).
In explaining the reason behind the limited use
of GIS analyses in the development plan prepara-
tion process, Ibrahim, et al. (2004) states that:
…the current limitations in terms of GIS use in
development plan preparation arisen due to in-
adequate comprehension among private planning
consultants on the analyses capability of GIS and
also on how GIS can help to ease their work. In
addition, the preparation of development plan
does not solely involve analyses of spatial data,
in many instances, analyses of aspatial data, or
a combination of both, are equally important.
Marrying the spatial analyses with aspatial ones
might be a little on the complicated side to many
planning consultants. (p. 3)
Although planning consultants who are in-
volved in the development plan preparation are
aware of GIS functions, they are not aware of GIS
full capability for analysis. In particular, they do
not know whether any of the GIS analyses can be
999
 GIS Implementation in Malaysian Statutory Development Plan System
used to replace what they have been doing manu-
ally. In the same time, GIS consultants also tend
to overlook their responsibility to advice planning
consultants on GIS analyses because they are
too preoccupied with developing GIS database
and the executive information system, which are
deemed as the core of their work. Additionally,
they receive no request for GIS analyses to be
performed from planning consultants (G.L. Chua,
personal communication, December 12, 2008).
The superficial knowledge of GIS among
planning consultants, as described above, can be
attributed to the lack of GIS training received. It
should be noted that many of the planning consul-
tants involved in preparing development plans are
seniors in their fields. Many never received any
formal GIS training during their university days.
Although commercial GIS training is available at
present, senior consultants prefer to not attend the
training sessions. Instead, younger staff are sent
for the training. A common idea shared by many
senior planning consultants is that GIS (and ICT)
suits younger people better.
Executive Information System for
Statutory Development Plan
Executive information system (EIS) is another
aspect of GIS application in the statutory develop-
ment plan preparation process. The requirement
to produce EIS for development plan is now
mandatory in all development plan preparation
projects (Jabatan Perancangan Bandar dan Desa
Semenanjung Malaysia, 2007). EIS for develop-
ment plan is basically a computer system that stores
all GIS maps prepared for a development plan.
Users can then access these maps through the EIS
and perform some GIS functions using the toolbar
on the system’s graphical user interface. Usually,
the toolbar includes only basic GIS functions such
as calculating distance, accessing information
of land parcels and displaying selected layers of
information. There are also EIS that stores the
digital version of the published development plan
1000
as well. As such, the system is frequently used by
the planning authorities’ officers as a substitute
for the hardcopy version of the plan.
The development of EIS for development plan
is advancing. Earlier, EIS was developed mainly as
a stand-alone system to store GIS maps and reports
produced during the development plan preparation
process. The database of the system was fixed and
only updated when revision was done to the plan.
However, recently, EIS began to be developed as
a multi-user system and deployed over local area
network. The system also now includes features
of a charting2 system, where the database can be
easily updated through the system’s interface.
With these features, planners are able to update the
land information in the database without having
to actually use GIS software.
EIS for development plan targets three types
of users, which are:
• Executives, including council planners,
who make decisions about planning
applications;
• System administrators who are usually
planners and technical staff of planning
authorities;
• The general public, including developers
and land owners.
Access to information stored in EIS data-
base is restricted according to the type of users.
Executive users have access to all information
including the digital version of both the report
of survey and the published development plan.
However, they do not have access to update the
GIS database. Public users neither have access to
the reports nor to update GIS database. Instead,
they can only access information regarding the
proposed land use of a particular land parcel and
the development restrictions attached to that land.
System administrators, obviously, have access to
all information and to update the GIS database.
Besides being used as a medium for quick ac-
cess to planning information, EIS is also used in
GIS Implementation in Malaysian Statutory Development Plan System
development control, counter services and public
exhibition of development plan. FDTCP has also
begun to encourage the development of internet-
based EIS (Jabatan Perancangan Bandar dan Desa
Semenanjung Malaysia, 2007). It is hoped that by
being internet-based, future EIS can be accessed
more easily by users, especially the public. With
the present system, public users can only access
the system at planning authorities’ offices, either
at the information kiosk or service counter.
EIS Implementation Problems
While EIS development is mandatory for all
statutory development plans in Malaysia, the
success of its implementation is uncertain due
to lack of research on this issue. Nonetheless, it
seems that some planning authorities are more
successful in implementing EIS than others. EIS
implementation at planning authorities such as at
Melaka State DTCP, Petaling Jaya City Council
and Seberang Perai Municipal Council are often
referred to, by those in the planning fraternity,
as examples of successful EIS (G. L. Chua, per-
sonal communication, December 12, 2008; Mohd
Ramzi Mohd Hussain, personal communication,
December 10, 2008).
As mentioned previously, EIS for development
plan is developed during the plan preparation
process. The cost to prepare the plan, including
EIS, is jointly funded by FDTCP and the planning
authority whose area is being planned for. In most
cases, the former contributes towards a major
portion of the funding. Once EIS development
is completed, the GIS consultant will create an
installation disc of the system and hand it over to
the planning authority. A manual on how to use
the system is also prepared and printed out by the
GIS consultant for the planning authority’s use.
It is then up to the planning authority to install,
maintain and upgrade the system. Yet, it is com-
mon for planning authorities to fail to maintain
the system. Often times, the reasons cited for
such failure are lack of personnel and financial
capacities.
In hindsight, the EIS for development plan is ac-
tually forced upon planning authorities by FDTCP.
When EIS became a mandatory requirement in
development plan preparation, planning authori-
ties had to accept the system when it is handed
over to them although they may not have the
resources to maintain the system. Unfortunately,
many of the planning authorities do not have the
resources to maintain the system. The problem
of not having sufficient personnel, especially
trained personnel, and funds to maintain EIS is
quite significant among many planning authori-
ties, especially those operating in the rural area.
In terms of personnel, some planning authorities
in the rural area do not even have a planner on
their payroll. In cases like this, it is usually the
technical assistant who will act as the planner for
the area. There are also cases where a planning
authority has only one staff, either a planner or
a technical assistant, to handle planning matters
of the area. To worsen the matter, some of these
planning authorities are not even equipped with
a computer in which the system can be installed
(Aishah Abdullah, personal communication,
November 18, 2008; Muhammad Hakimi Mohd.
Hussain, personal communication, November 18,
2008). Their existing computers are obsolete that
they do not have the right specifications to run
EIS applications.
GIS-Based Statutory Development
Plan Monitoring System
The monitoring of statutory development plan
implementation has not been given sufficient
attention by planning authorities in Malaysia,
even at present. In fact, many planning authori-
ties do not monitor the implementation of their
development plans. In the event where monitor-
ing of development plan is undertaken, the focus
is solely on land use change. Meanwhile, other
aspects of the plans, such as protection of envi-
1001
 GIS Implementation in Malaysian Statutory Development Plan System
ronmental resources, management of traffic and
provision of public facilities, are not included in
the monitoring exercise.
Although the monitoring of development plan
implementation is not widely practiced, the need
for it is clearly identified, especially for the state
structure plan. The Manual for Preparation of
Structure Plan proposes a procedure that can be
employed by States DTCP to monitor their state
structure plans (Jabatan Perancangan Bandar dan
Desa Semenanjung Malaysia, 2007). Additionally,
all published state structure plans include a section
that outlines the monitoring programme that the
States DTCP intend to undertake. Interestingly,
while the manual does not specifically propose the
use of any ICT application in state structure plan
monitoring, the plans describe the possibility of
GIS being used for monitoring purposes.
The only GIS-based monitoring system for
state structure plan in Malaysia is the one devel-
oped by Negeri Sembilan State DTCP. Known as
GIS9, the system was developed for the purpose
of monitoring land use change in the state and
evaluating the change in relation to the land use
allocation proposed in Negeri Sembilan State
Structure Plan. The system is placed on a server
located at the Negeri Sembilan State DTCP office.
The system is also connected to all local planning
authorities in the state in order to enable them
to update the system’s database when change of
land use occurs in their area. The updated data
will then be used to determine the actual land
use allocation in the state. This information will
then be compared with the land use allocation
proposed by the state structure plan. Thereafter,
the difference between actual land use allocation
and the proposed one will be calculated and the
results will be presented in terms of percentage
(UGisP, 2006).
1002
GIS-Based Monitoring System
Implementation Problems
Despite being potentially useful, the implementa-
tion of GIS9 is rather unsuccessful. This is due to
the failure of local planning authorities in updating
the data for their area and poor system’s infra-
structure. As aforementioned, the system is very
dependent on local planning authorities updating
the land use data for their area whenever there is
a change of land use. However, the local planning
authorities failed to do this. As in the case of EIS
previously, lack of personnel was often cited as
one of the reasons for their failure to update the
data (Anuar Maidin, personal communication,
November 11, 2008). As a result, the land use data
in the system database is outdated and does not
reflect the actual land use allocation in the state.
Thus, the data cannot be used to provide accurate
monitoring results.
The system is also plagued by poor infra-
structure set up. The server experiences too
many problems that lead to numerous down time,
which at times extending well over two months
(Haibernarisal Baijuri, personal communication,
November 16, 2008). The long down time leads
to data updating exercise being put on hold and
discourages users from using the system.
Impacts of GIS Implementation in
Statutory Development Plan System
The implementation of GIS in statutory devel-
opment plan system has impacted Malaysian
planning scene in a number of ways. Perhaps, the
biggest impact is how it has contributed towards
the increase of GIS critical mass in the country.
Although Rogers’ diffusion of innovation theory
(as cited in Ramasubramanian, 1999) suggests that
critical mass precedes adoption of technology, in
the case of GIS implementation in Malaysia, the
reverse may have happened. When FDTCP intro-
duced GIS in the development plan preparation
process, few planners were aware of what GIS
GIS Implementation in Malaysian Statutory Development Plan System
was all about. Nevertheless, FDTCP had already
decided to make GIS implementation in the plan
preparation process mandatory. This created the
demand for planners with knowledge on GIS.
Warnecke, Beatie & Lyday (as cited in Drum-
mond & French, 2008) traced the beginning of
large-scale introduction of GIS into planning back
to 1984. So, while it was already 13 years that
planners elsewhere in the world have been using
GIS in their work, many Malaysian planners still
found GIS a completely new invention when it
was officially introduced into Malaysian develop-
ment plan preparation process in 1997 (Mansor
Ibrahim et al., 2004). However, the result of a
survey conducted among planners in Malaysia
revealed that by 2004, already more than 50% of
the planners have used GIS software. A similar
survey in 2007 showed that the figure had risen
to almost 70% (Abdullah, Abdullah, & Ibrahim et
al., 2005). Looking from this angle, one can see
the profound impact of GIS implementation in
development plan preparation process in creating
GIS critical mass in Malaysian planning. It took
only 10 years since GIS was officially introduced
into the development plan preparation process for it
to become widespread among Malaysian planners.
Based on our observation, we would like to
offer two main reasons for the popularity of GIS
among Malaysian planners after its use was made
mandatory in the development plan preparation
process. Firstly, in the late 1990s, Malaysia and
several other countries in the Asian region were
faced with economic recession due to the cur-
rency crisis. Prior to the recession, Malaysia was
developing rapidly and many private planners
were involved as consultants in the construc-
tion industry, especially in terms of formulating
masterplans for new townships and submitting
planning applications for new development.
However, during the recession, the construction
industry in Malaysia was badly hit that it almost
halted. Suddenly, many of the private planners
found that there was no job in the construction
industry. Around the same time, FDTCP began
commissioning a new batch of development plan
preparation projects and was looking for private
planners to be appointed as consultants in order
to prepare these plans. Since planning jobs were
scarce, many private planners had no choice but to
get involved in those development plan prepara-
tion projects. It was through their involvement in
those projects that they began to obtain exposure
to GIS and GIS software.
Planners from the public sector also involved
in development plan preparation projects, al-
though not as consultants, but usually as a project
manager, as a technical committee member or as
owner of the plan. Thus, similar to private plan-
ners, their involvement in the projects helped to
expose them to GIS. Additionally, many of them
had the benefit of attending formal GIS training.
FDTCP, for instance, had established a programme
under which GIS trainings are offered to planners
from the public sector. The programme, known as
Geotechnical Spatial Analysis Research and De-
velopment, was established in 2000 as a platform
for FDTCP to undertake research and training
on spatial planning and decision-making tools
relevant to planning such as GIS, multi-criteria
evaluation and spatial decision support systems
(Jabatan Perancangan Bandar dan Desa Semenan-
jung Malaysia, 2008).
Secondly, the implementation of GIS in the
development plan preparation process has also
led to planning schools in the country to begin
to offer, or place greater emphasis on, GIS sub-
jects with the aim of producing more marketable
graduates. Graduates are not only trained through
GIS-specific subjects, but also through heavy GIS
integration, both theoretically and practically,
within other subjects such as design studios, urban
design, planning theory and practice, environmen-
tal planning and so on (Ibrahim et al., 2004). In fact,
during the late 1990s and early 2000s, having the
knowledge to operate GIS software became one of
the, if not the most, important criteria for planning
graduates to secure employment at planning firms
and authorities. Today, although the importance
1003
 GIS Implementation in Malaysian Statutory Development Plan System
of GIS knowledge in securing employment has
reduced slightly (but still important nonetheless),
planning schools continue to offer GIS-specific
subjects and the integration of GIS in other related
subjects remains high. These ensure that planning
graduates in the country continue to get exposure
to GIS, at least into the foreseeable future. In turn,
it helps to retain the critical mass needed to sup-
port continued use of GIS in Malaysian planning.
FUTURE RESEARCH DIRECTIONS
IN GIS APPLICATION IN
MALAYSIAN STATUTORY
DEVELOPMENT PLAN SYSTEM
Recent developments indicate that the use of GIS
in Malaysian statutory development plan system
will continue to flourish. Two areas of great op-
portunity lie in the integration of GIS with decision
support systems in the process of development
plan preparation and the development of GIS-
based systems for monitoring development plan
implementation.
Within the past few years, efforts to institu-
tionalise decision support systems into Malaysian
planning have been on-going (Abdullah, Abdullah,
& Ibrahim, 2009). Experimental projects in which
GIS was coupled with decision support systems
in order to provide better decision making during
the development plan preparation process were
undertaken. Whether those projects have added
value to the plan preparation process is, however,
unclear at the moment. Presently, it seems that
there is lack of research on this issue. Nevertheless,
those projects had contributed towards creating
among Malaysian planners on the integration of
GIS with decision support systems and had laid
the foundation for further application of GIS and
decision support systems. FDTCP is also work-
ing on a guideline on decision support systems
implementation in the development plan prepara-
tion process at the moment. How this guideline
will turn out remains to be seen.
1004
Recently, the development plan monitoring
is getting increasing attention from planning au-
thorities. FDTCP is currently working to develop
a GIS-based development plan monitoring model
(Nor Salehi Kassim, personal communication,
November 4, 2008). There are also efforts to
develop GIS-based systems for development
plan monitoring at other planning authorities. For
instance, Selangor DTCP will begin developing
their own GIS-based systems soon and Kuala
Lumpur City Hall is also looking into the idea of
having GIS-based systems to help them monitor
their local plan.
CONCLUSION
Looking at GIS implementation in Malaysian
statutory development plan system, one can iden-
tify several common problems in relation to GIS
implementation in planning worldwide.
These are:
• The use of GIS is largely limited to basic
functions.
• The implementation failure is due to or-
ganisational factors rather than technical.
• Data availability limits the success of GIS-
based information systems.
The implementation of GIS in Malaysian de-
velopment plan preparation process faces several
persistent problems. For instance, advanced GIS
applications in the preparation of development
plans are still underutilized. Despite increased GIS
awareness among planners, GIS application in the
development plan preparation process continues
to be rudimentary in nature. This is most likely
caused by insufficient GIS knowledge among
senior planning consultants who are involved in
the plan preparation. To mitigate the problem, in
the short term, it may be necessary for FDTCP
to produce a guideline that details out the types
of GIS analyses required in the plan preparation
GIS Implementation in Malaysian Statutory Development Plan System
process. At present, the same measure is already
undertaken by FDTCP in order to promote the
use of planning analyses in development plan
preparation process. In its Manual for Preparation
of Structure Plan, FDTCP has identified a list of
required planning analyses that need to be under-
taken during the state structure plan preparation
process. These analyses include, for example,
consistency analysis between plans’ objectives
and policies, policy impact analysis, land require-
ment analysis and cost benefit analysis (Jabatan
Perancangan Bandar dan Desa Semenanjung
Malaysia, 2007). Similar approach may be taken
by FDTCP for GIS analyses.
In the long run, the number of GIS analyses
used in the development plan preparation may
increase as younger planners with GIS technical
know-how become planning consultants in the
future. Since GIS is relatively new in Malaysia,
so are GIS courses at planning schools in the
country. While the senior planners may not have
the opportunity to learn GIS during their univer-
sity days, younger planners have had formal GIS
training at university.
Many researchers point out that strong organi-
sational support is important to ensure successful
GIS implementation (Abdullah, A. et al., 2002;
Innes & Simpson, 1993; Ramasubramanian,
1999). Thus, relevant parties must continue the
drive towards advanced application of GIS in
statutory development plan system. More GIS re-
search should be encouraged to discover the areas
in which advanced GIS application can add value
to the development plan preparation process, and
funds must be made available for such research.
The need for strong organisational support is
more visible in the case of GIS-based information
systems implementation. EIS for development
plan could prove useful for planning authori-
ties in implementing the policies and proposals
of their development plans, and as a medium
to convey planning information to the general
public. However, its implementation, especially
at local authority level, is plagued by the problem
of inadequate financial and personnel resources.
Although FDTCP has been funding the develop-
ment cost of EIS, it is one-off in nature and this
proved to be insufficient. As Alias Abdullah,
Muhammad Faris Abdullah & Fauzan Nordin
(2003) point out, information system like EIS is
a ‘living’ system and thus:
...require maintenance and system capacity build-
ing from time to time. Accordingly, sufficient
funding must be made available. Funding is re-
quired for manpower to operate and maintain the
system, and for capacity building, which include
hardware and software upgrades as well as the
overall improvement of the system. (p. 79)
For planning authorities that face financial
restriction in maintaining their EIS, regular fund-
ing from FDTCP may be necessary to subsidise
part of the maintenance cost. Additionally, the
planning authorities must be committed to EIS
and not use the subsidy for other purposes, except
for maintaining the EIS.
The effort to utilize GIS in monitoring the
implementation of development plan has been
severely hampered by the problem of data un-
availability. This is hardly surprising because
researchers have found that data unavailability
continues to be one of the most prominent fac-
tors that limit successful GIS implementation in
planning (Gilfoyle & Wong, 1998; Mennecke &
West, 2001). Therefore, to overcome this problem,
solutions have to be focused towards ensuring
data availability. Very often data available are
generally inaccurate and incomplete, but these
are also the best available data (Klosterman &
Abdullah, A., 2008). Thus, it is important that any
GIS-based systems for monitoring development
plan implementation do not require extensive data
sets that are difficult to obtain in order for it to
work. The system must also be flexible enough
to accommodate the best available data.
Another interesting development pertaining
to GIS implementation in Malaysian statutory
1005
 GIS Implementation in Malaysian Statutory Development Plan System
development plan preparation process is the ef-
fort to institutionalise decision support systems
in the process. These efforts are welcomed but it
has to be implemented with care. For planners
who are unfamiliar with, or have little knowl-
edge of, decision support systems, it could be
a daunting task to apply it in their work. If its
implementation is rushed, it would only lead to
planners finding it too difficult and consequently
losing interest to continue to use decision support
systems. Researchers have suggested that one of
the important characteristics of successful tech-
nological implementation is that it must allow for
incremental and small trials (Innes & Simpson,
1993; Ramasubramanian, 1999). Planners must be
exposed slowly to decision support systems. The
decision support technology and models adopted
must be able to sustain planners’ interest in the
system and relevant to the planning tasks. We
must not adopt complex models which are only
of pedagogic value but not useful for plan making
purposes (Batty, 2004).
REFERENCES
Abdullah, A., Abdullah, M. F., & Nordin, M. F.
(2002). Spatial information system as a tool in
managing urban development process. Paper
presented at the 18th EAROPH World Planning
Congress.
Abdullah, A., Abdullah, M. F., & Nordin, M. F.
(2003). Managing urban development process
by using spatial information system. Planning
Malaysia, 1, 71–92.
Abdullah, A., Abdullah, M. F., Shahbuddin, M.
N. A., & Klosterman, R. E. (2009). Modelling
the reality in planning: SPDSS experience in
Malaysia. Malaysian Journal of Environmental
Management, 10(1), 119–132.
1006
Abdullah, M. F., Abdullah, A., & Ibrahim, M.
(2009). Institutionalisation of spatial planning
and decision support systems for planning and
governance in Malaysia. In Davide, G., & Abdul-
lah, A. (Eds.), Spatial decision support for urban
and environmental planning: a collection of case
studies. Shah Alam, Malaysia: Arah Publications.
Abdullah, M. F., Abdullah, A., Ibrahim, M., &
Abdul Samad, D. (2005). A study on the demand
of spatial planning and decision support system in
Malaysia. Planning Malaysia, 3, 47–60.
Abdullah, M. F., & Ariffin, I. (2003). Incorporat-
ing sustainable development objectives into de-
velopment plans through strategic environmental
assessment. Planning Malaysia, 1, 1–18.
Abdullah, M. N. (2004). Pelaksanaan sistem
maklumat geografi di Jabatan Perancangan Bandar
dan Desa Negeri: dilema dan potensi. Malaysian
Townplan, 2(1), 29–37.
Abu Bakar, M. A. (2004). Sistem maklumat geo-
grafi di Jabatan Perancangan Bandar dan Desa
Semenanjung Malaysia. Malaysian Townplan,
2(1), 19–23.
Ahmad Sarkawi, A. (2006). The legal consid-
erations of neighbouring lands in development
planning: the Malaysian context. Newcastle-
upon-Tyne, United Kingdom.
Amandus, P. Jr. (2006). MCDM and land suit-
ability analysis for spatial development: case
study Kulim local plan. In Shamsuddin, K. (Ed.),
Implementing Multi Criteria Decision Making in
Malaysian Town Planning. Kuala Lumpur: Fed-
eral Department of Town and Country Planning
Peninsular Malaysia.
Bahagian Teknologi Maklumat. (2007a). Format
Metadata dan Struktur Pengkalan Data Sistem
Maklumat Geografi Kajian Rancangan Tempatan.
Kuala Lumpur: Jabatan Perancangan Bandar dan
Desa Semenanjung Malaysia.
GIS Implementation in Malaysian Statutory Development Plan System
Bahagian Teknologi Maklumat. (2007b). Manual
Penyediaan Pelan GIS. Kuala Lumpur: Jabatan
Perancangan Bandar dan Desa Semenanjung
Malaysia.
Batty, M. (2004). Dissecting the streams of plan-
ning history: technology versus policy through
models. Environment and Planning. B, Planning
& Design, 31, 326–330. doi:10.1068/b3103ed
Budic, D. Z. (2000). Geographic information sci-
ence implications for urban and regional planning.
URISA, 12(2), 81–93.
Ceccato, V. A., & Snickars, F. (2000). Adapting
GIS technology to the needs of local planning.
Environment and Planning. B, Planning & Design,
27, 923–937. doi:10.1068/b26103
Craglia, M., & Signoretta, P. (2000). From global
to local: the development of local geographic
information strategies in the United Kingdom.
Environment and Planning. B, Planning & Design,
27, 777–788. doi:10.1068/b2651
Drummond, W. J., & French, S. P. (2008).
The Future of GIS in Planning: Converging
Technologies and Diverging Interests. Journal
of the American Planning Association. Ameri-
can Planning Association, 74(2), 161–174.
doi:10.1080/01944360801982146
Geertman, S. (2002). Participatory planning and
GIS: a PSS to bridge the gap. Environment and
Planning. B, Planning & Design, 29, 21–35.
doi:10.1068/b2760
Gilfoyle, I., & Wong, C. (1998). Computer ap-
plications in planning: Twenty years’. Plan-
ning Practice and Research, 13(2), 191–197.
doi:10.1080/02697459816193
Gill, S., Higgs, G., & Nevitt, P. (1999). GIS in
planning departments: Preliminary results from
a survey of local planning authorities. Plan-
ning Practice and Research, 14(3), 341–361.
doi:10.1080/02697459915643
Gocmen, Z. A. (2009). GIS use for planning in
Wisconsin’s public agencies. Paper presented at
the 11th International Conference on Computers
in Urban Planning and Urban Management.
Heeks, R. (2002). Information Systems and De-
veloping Countries: Failure, Success, and Local
Improvisations. The Information Society, 18(2),
101–112. doi:10.1080/01972240290075039
Ibrahim, M., Abdullah, M. F., & Abdul Samad,
D. (2004). Geographical information system
(GIS) in Malaysian planning education. Paper
presented at the 7th International Seminar on GIS
for Developing Countries (GISDECO).
Idrus, S., & Harman Shah, A. H. (2006). ‘Me-
tabolisme’ GIS di Malaysia. Buletin Geospatial
Sektor Awam, 1, 13–19.
Innes, J. E., & Simpson, D. M. (1993). Implement-
ing GIS for planning. Journal of the American Plan-
ning Association. American Planning Association,
59(2), 230. doi:10.1080/01944369308975872
Jabatan Perancangan Bandar dan Desa Semenan-
jung Malaysia. (2001). Manual Penyediaan
Rancangan Tempatan (Pindaan 2001). Kuala
Lumpur: Jabatan Perancangan Bandar dan Desa
Semenanjung Malaysia.
Jabatan Perancangan Bandar dan Desa Semenan-
jung Malaysia. (2007). Manual Penyediaan
Rancangan Struktur. Kuala Lumpur: Jabatan
Perancangan Bandar dan Desa Semenanjung
Malaysia.
Jabatan Perancangan Bandar dan Desa Semenan-
jung Malaysia. (2008). Geotechnical spatial analy-
sis research and development: fokus dan agenda.
Retrieved December 12, 2008, from http://www.
townplan.gov.my/penyelidikan_gsard.php
Johar, F. (2003). Environmental concern in local
planning. Planning Malaysia, 1, 19–34.
1007
 GIS Implementation in Malaysian Statutory Development Plan System
Kassim, N. S., & Islam, R. (2009). Spatial AHP
for the National Physical Plan of Malaysia. In
Davide, G., & Abdullah, A. (Eds.), Spatial decision
support for urban and environmental planning: a
collection of case studies. Shah Alam, Malaysia:
Arah Publications.
Klosterman, R. E. (1997). Planning support sys-
tems: a new perspective on computer-aided plan-
ning. Journal of Planning Education and Research,
17(1), 45–54. doi:10.1177/0739456X9701700105
Klosterman, R. E. (1999). New perspectives on
planning support systems. Environment and
Planning. B, Planning & Design, 26, 317–320.
doi:10.1068/b260393
Klosterman, R. E., & Abdullah, A. (2008). Com-
puter technologies in urban planning. Paper
presented at the Medina Forum Seminar on Devel-
oping and Deploying Planning Support Systems.
Lee, L. M., & Ahmad, M. J. (2000). Local authority
networked development approval system. Paper
presented at the Planning Digital Conference.
Maklumat, B. T. (2006). Sistem pengurusan data
JPBD Semenanjung Malaysia. Buletin Geospatial
Sektor Awam, 1/2006, 26–30.
Mennecke, B. E., & West, L. A. J. (2001). Geo-
graphic Information Systems in Developing Coun-
tries: Issues in Data Collection, Implementation
and Management. (Technology Information).
Journal of Global Information Management,
9(4), 44–54.
Ramasubramanian, L. (1999). GIS Imple-
mentation in Developing Countries: Learning
from Organisational Theory and Reflective
Practice. Transactions in GIS, 3(4), 359–380.
doi:10.1111/1467-9671.00028
Samat, N. (2006). Applications of geographic
information systems in urban land use planning
in Malaysia. Paper presented at the 4th Taipei
International Conference on Digital Earth.
1008
Tan, T. S. (2005). Electronic local authority man-
agement system. Planning Malaysia, 3, 27–46.
UGisP. (2006). GIS9: Pembangunan Pengkalan
Data Sistem Maklumat Geografi Negeri Sembilan
Darul Khusus - Laporan Akhir Fasa III.
ADDITIONAL READING
Abdullah, A. Kamalruddin Shamsuddin, & Abdul-
lah, M. F. (Eds.). (2004). Applications of planning
and decision support systems. Kuala Lumpur:
Bureau of Consultancy & Entrepreneurship, In-
ternational Islamic University Malaysia.
Abdullah, M. F., Abdullah, A., & Rustam Khairi
Zahari. (2009). GIS application in monitoring
development plan implementation in Malaysia:
weaknesses and potential. Paper presented at the
11th International Conference on Computers in
Urban Planning and Urban Management.
Brail, R. K., & Klosterman, R. E. (Eds.). (2001).
Planning support systems: integrating geographic
information systems, models, and visualization
tools. Redlands, CA: ESRI Press.
Bruton, M. J. (2007). Malaysia: the planning of
a nation. Kuala Lumpur: PERSADA.
Budic, D. Z. (1993). GIS use among southeast-
ern local governments. Journal of Urban and
Regional Information Systems Association, 6(6),
529–546.
Geertman, S., & Stillwell, J. (Eds.). (2003). Plan-
ning support systems in practice. Heidelberg,
Germany: Springer.
Goh, B. L. (1990). Urban planning in Malaysia.
Petaling Jaya, Malaysia: Tempo Publishing.
Huxhold, W. (1991). An introduction to urban
geographic information systems. New York:
Oxford University Press.
GIS Implementation in Malaysian Statutory Development Plan System
Johar, F. (2006). Environmental sustainability in
selected local plans in Malaysia. In Kadouf, H.
A., & Al-Junid, S. Z. (Eds.), Land use planning
and environmental sustainability in Malaysia:
policies and trends (pp. 255–274). Kuala Lumpur:
Research Centre, International Islamic University
Malaysia.
Kadouf, H. A., & Al-Junid, S. Z. (Eds.). (2006).
Land use planning and environmental sustain-
ability in Malaysia: policies and trends. Kuala
Lumpur: Research Centre, International Islamic
University Malaysia.
Kadouf, H. A., & Maidin, A. J. (2006). Theory
and practice in land use planning in Malaysia:
an overview. In Kadouf, H. A., & Al-Junid, S.
Z. (Eds.), Land use planning and environmental
sustainability in Malaysia: policies and trends
(pp. 77–113). Kuala Lumpur: Research Centre,
International Islamic University Malaysia.
Kaniclides, A., & Kimble, C. (1995). A develop-
ment framework for executive information system.
Paper presented at the GRONICS ‘95, Groningen,
The Netherlands.
Kasmin, A. M. (1999). Kerangka pelaksanaan
dasar rancangan struktur. Shah Alam, Malaysia:
Pusat R&D, Pejabat Setiausaha Kerajaan Negeri
Selangor.
Kassim, N. S. (2003). Konsep dan proses e-
submission dalam permohonan perancangan.
Malaysian Townplan, 1(1), 19–25.
Matori, A. N. (2007). Development of a user
interface application for information delivery - a
case study of Ipoh City Council’s Department
of Town Planning. Buletin Geospatial Sektor
Awam, 1, 24–29.
Muhammad, Z. (2000). Integrated rural planning
and development approach - the Malaysian experi-
ence. Habitat Malaysia, 3, 1–34.
Munir, A. B., & Zulhuda, S. (2007). Becoming
e-cities: legal issues and challenges. Paper pre-
sented at the 2nd International Symposium on
Knowledge Cities, Shah Alam.
Nedovic-Budic, Z. (1998). The impact of GIS tech-
nology. Environment and Planning. B, Planning
& Design, 25, 681–692. doi:10.1068/b250681
Rogers, E. (1983). Diffusion of Innovations (3rd
ed.). New York: Free Press.
Shamsuddin, K. (Ed.). (2006). Implementing
multi-criteria decision making in Malaysian town
planning. Kuala Lumpur: Federal Department of
Town and Country Planning Peninsular Malaysia.
Stillwell, J., Geertman, S., & Openshaw, S. (Eds.).
(1999). Geographical information and planning.
Heidelberg, Germany: Springer.
Vonk, G., Geertman, S., & Schot, P. (2005).
Bottlenecks blocking widespread usage of plan-
ning support systems. Environment & Planning
A, 37, 909–924. doi:10.1068/a3712
Wan Ismail, W. H. (2007). Pematuhan peran-
cangan (teori). Paper presented at the Mesyuarat
Pegawai Kanan Perancang Bandar dan Desa
Malaysia ke 21.
Wegener, M. (1998). GIS and spatial planning.
Environment and Planning. B, Planning & De-
sign, 25, 48–52.
Yaakup, A. (2004). GIS as a tool for development
planning and monitoring. Paper presented at the
NGIS 2004.
Yaakup, A., Abu Bakar, S. Z., & Sulaiman, S.
(2009). Decision support systems for urban
sustainability in Malaysia. Malaysian Journal
of Environmental Management, 10(1), 101–118.
Yaakup, A., Johar, F., Maidin, M. A., & Ahmad,
E. F. (2004). GIS and decision support systems
for Malaysian development plan studies. Paper
presented at the International Seminar on SPDSS
2004.
1009
 GIS Implementation in Malaysian Statutory Development Plan System
KEY TERMS AND DEFINITIONS
Decision Support Systems: Computerized
systems that can be used to help users in making
decision over complex and ill-structured problems.
Development Plan: A plan which is usually
prepared by planners to guide or promote de-
velopment in an area. It usually contains maps
of the area, and planning policies and proposals
for the area.
Executive Information Systems: Comput-
erized systems that are developed to provide
important information in a concise manner to
senior executives. This information can be used
by the executives in making decision over a busi-
ness problem.
Federal Department of Town and Country
Planning: Is the planning authority at the Federal
level that is responsible in charting the general
direction of planning and development for the
nation. Its functions also include promoting town
and country planning in Malaysia and providing
advisory services to the Federal Government and
planning authorities at State and Local levels. It
also provides technical and monetary supports to
State and Local level planning authorities in the
preparation of their development plans.
Geographical Information Systems (GIS):
Computerized systems that can be used to store,
analyze and display spatially-referenced data.
GIS-Based Systems: Computerized systems
which are developed using GIS software as the
platform.
Local Planning Authority: Local planning
authority of an area is usually the local authority
of the area, such as city hall, the municipal council
or the district council. It is responsible in regulat-
This work was previously published in Handbook of Research on E-Planning: ICTs for Urban Development and Monitoring,
edited by Carlos Nunes Silva, pp. 435-454, copyright 2010 by Information Science Reference (an imprint of IGI Global).
1010
ing, controlling and determining the planning and
development direction of its area.
State Department of Town and Country
Planning: Each state in the Peninsular Malaysia
has its own Town and Country Planning Depart-
ment which oversees the planning and develop-
ment matters within the State boundary and pro-
vides advisory services to the State government.
State Town and Country Planning Department
also acts as local planning authority, handling
development control affairs for areas within the
state, which are not under the jurisdiction of any
local authority.
Statutory Development Plan: Development
plan which preparation is required by law.
ENDNOTES
1
States in East Malaysia (e.g., Sabah and
Sarawak), and the three Federal Territories
(e.g., Kuala Lumpur, Putrajaya and Labuan)
in Malaysia are not bounded by the Town
and Country Planning Act 1976. They, how-
ever, have their own planning acts which are
almost similar to the Act.
2
Charting is a process where information
about planning applications is ‘drawn’ onto
standard cadastral map. Traditionally, this
was done by hand where draughtsman would
draw on the map the boundary of land parcel
where planning applications were submitted,
as well as other basic information about the
applications. Many planning authorities have
now replaced their manual charting with
GIS.
 1011

Chapter 62
3D City Modeling and
Visualization for Smart
Phone Applications
Juha Hyyppä Zhengjun Liu
Finnish Geodetic Institute, Finland Chinese Academy of Surveying and Mapping,
China
Lingli Zhu
Finnish Geodetic Institute, Finland Harri Kaartinen
Finnish Geodetic Institute, Finland
Anttoni Jaakkola
Finnish Geodetic Institute, Finland

ABSTRACT
Smartphones with larger screens, powerful processors, abundant memory, and an open operation system
provide many possibilities for 3D city or photorealistic model applications. 3D city or photorealistic
models can be used by the users to locate themselves in the 3D world, or they can be used as methods
for visualizing the surrounding environment once a smartphone has already located the phone by other
means, e.g. by using GNSS, and then to provide an interface in the form of a 3D model for the location-
based services. In principle, 3D models can be also used for positioning purposes. For example, match-
ing of images exported from the smartphone and then registering them in the existing 3D photorealistic
world provides the position of the image capture. In that process, the central computer can do a similar
image matching task when the users locate themselves interactively into the 3D world. As the benefits
of 3D city models are obvious, this chapter demonstrates the technology used to provide photorealistic
3D city models and focus on 3D data acquisition and the methods available in 3D city modeling, and
the development of 3D display technology for smartphone applications. Currently, global geoinformatic
data providers, such as Google, Nokia (NAVTEQ), and TomTom (Tele Atlas), are expanding their prod-
ucts from 2D to 3D. This chapter is a presentation of a case study of 3D data acquisition, modeling and
mapping, and visualization for a smartphone, including an example based on data collected by mobile
laser scanning data from the Tapiola (Espoo, Finland) test field.

DOI: 10.4018/978-1-4666-2038-4.ch062

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 3D City Modeling and Visualization for Smart Phone Applications
INTRODUCTION
In the past, visualization of 3D information was
limited to personal computers. Today, the increase
in the capabilities of mobile devices (powerful
processors, abundant memory, and open opera-
tion systems) provides possibilities for users to
utilize 3D information for purposes such as
navigation and location-based services (LBS).
“Future developments in navigation and other
location-enabled solutions will rely heavily on 3D
mapping capabilities,” said Cliff Fox, executive
vice president, NAVTEQ Maps, in November,
2010 (NAVTEQ). Many location-based service
agencies (e.g. Google, Nokia (NAVTEQ), and
TomTom (Tele Atlas)) are currently expanding
their products from 2D to 3D. In 2007, Google was
the first to introduced panoramic images (360°)
of street views (2D) in the company’s Google
Earth (GE) service. In December of 2010, their
products were expanded to 3D street views and 3D
buildings. After several months, in April of 2011,
collaborating with Microsoft, Nokia launched Ovi
Maps 3D where currently twenty cities are avail-
able in 3D views. Intense competition is usually
an indication of heavy user demand. This has
created a practical need for rapid 3D geographi-
cal data acquisition and modeling. The 3D laser
scanning market is also expected to double by
2015 (Geospatial World, 2011).
Currently, there are two primary means for
3D geographic data acquisition: photogrammetry
and laser scanning. Photogrammetry is the tech-
nology of deriving 3D data from 2D images by
mono-plotting (using the existing height model),
stereo-imagery interpretation or multi-imagery
block adjustment. According to different vehicle
platforms for optical sensors, 2D images can
be separated into several categories: satellite
images, aerial images, and terrestrial images as
well as image video sequences. Methodologies
for 3D modeling from satellite images can be
found in numerous publications, and the follow-
ing is a representative sample of such documents
1012
(Baltsavias, Pateraki, & Zhang, 2001; Baltsavias,
Zhang, & Eisenbeiss, 2005; Fraser, Baltsavias &
Güen, 2002; Grün, Zhang & Eisenbeiss, 2005;
Jacobsen, 2004; Lee, Shan & Bethel, 2002; Poli,
2005; Sörgel, Szhulz, Toennessen & Stilla, 2005;
Sörgel, Thoennessen, Brenner & Stilla 2006;
Sörgel, Michaelsen, Thiele, Cadario & Thoennes-
sen, 2009; Toutin, 2004; Zhang & Grün, 2006).
Content on 3D modeling using aerial images
has been presented by the following (Ahmadi,
Zoej, Ebedi, Moghaddam & Mohammadzadeh,
2010; Baillard, Schmid, Zisserman & Fitzgibbo,
1999; Baltsavias et al., 2005; Bescoby, 2006; Le
Besnerais, Sanfourche & Champagnat, 2008; Col-
lins et al., 1998; Cord, Jordan & Cocquerez, 2001;
Fischer et al, 1998, Fradkin, Maitre & Roux, 2001;
Grün & Wang, 1998a; Jaynes, Riseman & Hanson,
2003; Jung, 2004; Kim, Park, Kim, Jung & Kim,
2004; Mayer, 1999; Paparoditis, Cord, Jordan &
Cocquerez, 1998; Peng & Liu, 2005; Suveg &
Vosselman, 2004; Tournaire, Bredif, Boldo &
Durupt, 2010). Literatures on 3D modeling using
terrestrial images have been contributed by the
following (Fraser, 1996; Fraser & Cronk, 2009;
Grün, Remondino & Zhang, 2004; Luhmann,
2010; Remondino, 2006). Papers related to 3D
modeling using image sequences can be obtained
from the following sources (Bethmann, Herd, Luh-
mann & Ohm, 2009; Cornelis, Leibe, Cornelis &
Van Gool, 2008; D’Apuzzo, 2003; Fitzgibbon &
Zisserman, 1998; Forlani, Roncella & Remondino,
2005; Pollefeys et al., 2008; Remondino, 2002;
Remondino, 2004; Tian, Gerke, Vosselman &
Zhu, 2010). However, despite the research efforts
focused on photogrammetry over many decades,
the level of automation is still relatively low.
The current technology in laser scanning (LS)
offers an alternative solution for the acquisition
of 3D geographic data. LS is also referred to as
Light Detection And Ranging (LiDAR) because
it uses a laser to illuminate the earth’s surface and
a photodiode to register the backscatter radiation.
Typically, the time it takes for the laser beam to
reach the target and come back to the source (de-
3D City Modeling and Visualization for Smart Phone Applications
lay) is used to measure the distance to the target,
i.e. with the speed of the light. After GPS, IMU,
and scanning mechanism were attached to laser
ranging measurements, first for military purposes
in 1980s, and later on for surveying purposes,
the field of science currently known as Airborne
Laser Scanning (ALS) was born. The typical
operating wavelength range used in LiDAR is
between 250 nm -1600 nm, and in LS the most
common wavelength range is close to 1000 nm
e.g. 905nm or1064nm. However, they are easy
to be absorbed by the eye. An alternative laser
with 1550nm wavelength is eye-safe at much
higher power levels. Commercial ALS systems
have been available since 1994. Currently, ALS
systems are extremely feasible solutions for map-
ping large areas with a high degree of accuracy,
e.g. to determine elevations and forest models.
Elevation accuracies of 5-10 cm are common,
whereas planimetric accuracies range between
20-80 cm, depending on the flying height and
IMU characteristics. Nation-wide LS data acqui-
sition projects for multipurpose uses have been
conducted in or are envisaged in several countries.
Elevation models, stand-wise forest inventories,
and corridor mapping are good examples of
commercial activities. In the Nordic Countries,
stand-wise forest inventory is being increasingly
assisted by ALS, which has the effect of reduc-
ing the costs of inventory data. Literature on 3D
modeling using ALS data can be found in sources
such as the following (Ackermann, 1999; Baltsav-
ias, 2004; Dorninger & Pfeifer, 2008; Hyyppä &
Engdahl, 2000; Hyyppä, Matikainen, Kaartinen,
Yu, Hyyppä & Rönnholm, 2007; Hyyppä et al.,
2008; Hyyppä et al., 2009; Kada & Mckinley,
2009; Melzer, 2007; Rottensteiner, Trinder, Clode
& Kubik, 2005a; Sampath & Shan, 2010; Sohn,
Huang & Tao, 2008; Verma, Kumar & Hsu, 2006;
Vosselman, Gorte, Sithole & Rabbani, 2004).
Terrestrial Laser Scanning (TLS) differs from
ALS in that the scanner is mounted in a fixed
position, i.e. on a tripod, and ranging takes place
across a large spherical angle-of-view. The scans
are primarily used for the reconstruction of man-
made targets, e.g. in the fields of architecture,
civil engineering, archaeology, cultural heritage,
and robotics (Vosselman & Maas, 2010). TLS is
a feasible alternative when carrying out detailed
surveys over small areas, with typically radii for
high-density scans being less than 20-30 meters.
Approaches applied in 3D modeling using TLS
data are presented in the following examples of
sources (Arayici, 2007; Buckley, Howell, Enge
& Kurz, 2008; Pu, 2008; Pu & Vosselman, 2009).
Mobile Laser Scanning (MLS), also called
Mobile Terrestrial Laser Scanning, is currently
a rapidly developing area in LS with laser scan-
ners, GNSS and IMU, being mounted onboard a
moving vehicle.
MLS can be considered to fill the gap between
ALS and TLS. In MLS, data collection can be
performed either in the so-called stop-and-go mode
or in a continuous mode. The stop-and-go mode
corresponds to conventional TLS measurements,
and therefore MLS, hereafter, is used to refer to
the continuous model, i.e. the use of continuous
scanning measurements along the drive track. In
addition to laser scanners, MLS data acquisition
sensors can include accessories such as digital
cameras or video cameras. Research results about
3D modeling from MLS have been addressed
by researchers such as the following (Brenner,
2009; Früh & Zakhor, 2004; Haala, Peter, Kremer
& Hunter, 2008; Jaakkola, Hyyppä, Hyyppä &
Kukko, 2008; Jaakkola et al., 2010; Zhao & Shi-
basaki, 2003; Zhao & Shibasaki, 2005; Zhu et al.,
2009; Zhu, Hyyppä, Kukko, Kaartinen & Chen,
2011). However, the point density of laser data has
great impact on the methodological development
and quality of the produced 3D models.
The integration of photogrammetry and laser
scanning provide new technology solutions for
3D modeling. By combining the good height-
assessment accuracy of a laser scanner and the
good planimetric accuracy of aerial images, both
high accuracy and higher automation can, in
theory, be obtained (Kaartinen & Hyyppä, 2006).
1013
 3D City Modeling and Visualization for Smart Phone Applications
The use of data integration greatly improves the
level of automation (Brenner, 2005). Method-
ologies for data fusion from photogrammetric
images and laser scanning have been contributed
by researchers such as the following (Awrangjeb,
Ravanbakhsh & Fraser, 2010; Becker & Haala,
2007; Brenner, 2005; Früh & Zakhor, 2004; Haala
& Kada, 2010; Habib et al., 2010; Kaartinen &
Hyyppä, 2006; Ma, 2004; Pu & Vosselman, 2009;
Sohn & Dowman, 2007; Vosselman, 2002; Vu,
Yamazaki & Matsouka, 2009).
The development of desktop mapping applica-
tions with 3D geo-visualization capabilities has
been explored over its long history (Koller et al.,
1994). Over the past 10 years, various 3D visual-
ization systems for geographical data have been
developed. Many commercial GIS and remote-
sensing software vendors such as ESRI (ArcGIS
ArcScene and ArcGlobe) and ERDAS (IMAGINE
VirtualGIS) have integrated 3D mapping facilities
to facilitatew visualization and spatial analyses.
Other companies and agencies, e.g., Google
(Google Earth) and NASA (WorldWind), provide
online 3D virtual globe visualization software for
free viewing of the 3D Earth environment and use
of location-based services. Techniques supporting
very large scenes, very high-resolution satellite
images and aerial photos with high visualization
performance, include the use of multi-resolution
terrain models and high level of detail (LOD).
Examples of studies looking into these are (Bösch,
Goswamiz & Pajarolax, 2009; Garland, 1999;
Hoppe, 1998; Lindstrom & Pascucci, 2001; Livny,
Sokolovsky, Grinshpoun & El-Sana, 2008; Sch-
neider & Westermann, 2006; Ulrich, 2000). As
regards quad-tree-based or oc-tree-based scene
management and tile based terrain rendering,
the following researchers have looked into them
(Amara & Marsault, 2009; Blow, 2000; Livny,
Kogan & El-Sasa, 2009).
Since a mobile device has limited hardware
resources and far less graphical computation
power built into it, the 3D visualization techniques
developed for desktop computers cannot be ap-
1014
plied without some adaptation measures. Although
NAVTEQ and some other navigation companies
have supported 3D landmark visualization in
their navigation devices, the development of a
high performance, photo-realistic 3D visualiza-
tion engine to support smartphone navigation and
LBS is still a work-in-progress (Bodgan & Coors,
2009; Coors, 2008; Goetz & Zipf, 2010).
This chapter endeavors to contribute to the
procedure of 3D city model reconstruction, and
also to the state-of-the-art of 3D modeling and the
visualization aspects, e.g. smartphone 3D visual-
ization for navigation and LBS approaches and
techniques. The complete workflow of 3D model
reconstruction and visualization from data acquisi-
tion to photorealistic 3D model reconstruction and
also to the implementation of 3D engine as a case
study is also presented. In the last section, future
research aspects and challenges are discussed.
ACQUISITION OF GEOMETRY
FOR 3D CITY MODELS
The complete process of 3D city model recon-
struction covers two steps: acquisition of geom-
etry and texture mapping. The approaches for
acquisition of geometry heavily rely on the data
sources. Currently two primary means are widely
used for geo-data acquisition: laser scanning and
photogrammetry. In the following sections, the
basics of laser scanning and photogrammetry
and the methodologies for 3D building geometry
extraction and reconstruction will be introduced.
In addition, some useful commercial software and
systems and also some free open sources will be
addressed.
Basics of Laser Scanning
In order to understand the basics of laser scanning,
in the following we will summarize the operation
principle of laser scanners, highlight major laser
scanning parameters and talk about basics of point
3D City Modeling and Visualization for Smart Phone Applications
cloud processing (Hyyppä, Wagner, Hollaus, &
Hyyppä, 2009).
Basic Operation Principle
The laser scanning is a LiDAR-based method for
acquiring range measurements typically from a
moving platform and the precise orientation of
these measurements. Short laser pulses (4-10 ns)
are emitted with a high frequency (e.g. 100-400
kHz) and are continuously deflected in across-
flight direction using various scanning mecha-
nisms. The position and rotation of the sensor are
continuously recorded along the flight path using
a Global Navigation Satellite System (GNSS)
and an Inertial Measurement Unit (IMU). The
recorded measurements of sensor position and
orientation, beam deflection and range can be
converted to a georeferenced three-dimensional
(3D) point cloud representing the measured tar-
gets. Since it is a-priori not known which targets
generated the echoes, i.e., the points are dummy
in that sense, typically surface models or points
forming the models are required for converting
the 3D point cloud into geospatial data products
such as Digital Terrain Models (DTM), Digital
Surface Models (DSM), Canopy Height Model
CHM or normalized Digital Surface Model
(nDSM) (CHM=nDSM=DSM-DTM). These
basic geospatial products are increasingly used in
a broad range of applications, e.g. in hydrology,
construction engineering, infrastructure monitor-
ing, and forestry. When using points, filtering
techniques for point classification has to be ap-
plied. Overviews of laser scanning can be found
in (Baltsavias, 1999; Fowler, Samberg, Flood &
Greaves, 2006; Wehr & Lohr, 1999). As ALS is
a rapidly growing technology, several technical
improvements have occurred since these reviews
have been published. Specifically, the sampling
frequency, positioning accuracy, and number
of recorded echoes have been improved during
the last years. Also, small-footprint waveform-
digitizing ALS systems have become available
(Wagner, Ullrich, Ducic, Melzer & Studnicka,
2006) compared to previously developed large-
footprint systems (Blair, Rabine & Hofton, 1999;
Garvin, 1996; Rabine & Martin, 1996) mainly used
to test space-borne LiDAR techniques.
Laser Scanning Parameters
Important technical parameters to characterize the
physical properties of laser scanner systems are
the laser wavelength (μm), pulse duration (ns),
pulse energy (mJ), pulse repetition frequency/
rate (PRF, PRR) (kHz), beamwidth (mrad), scan
angle (deg), scan rate (Hz), flying height (m), and
size (m) of laser footprint on the ground. Some of
these parameters related to the data acquisition,
which can be affected by the user are as follows:
The flight altitude and the scan angle - for
a selected laser scanner, the main user-selected
parameters are flight altitude and maximum scan
angle (FOV/2, Field of View). Flight altitudes
can vary between 10 to 6,000 m, depending on
the systems and flying platform (UAV, helicopter,
aircraft). The flight altitude also determines the
PRF of systems, which previously allowed one
pulse in the air and presently increasingly allow-
ing multiple pulses in the air. The type of scanner
used (scanning mechanism) and, for a fixed PRF
and scanner rate, the maximum scan angle and
flying height determines the area coverage and the
distribution and density of points for a given pulse
ALS system. The density of the pulses on the target
is the main parameter affecting the quality of the
surveyed data. Increase in flight height increases
planimetric errors (errors in plane direction (X-Y)
are mainly due to IMU).
Pulse repetition frequency - In 1993, the pulse
rate (pulse repetition frequency, PRF) was 2 kHz,
while the best systems in 2007 provided samples
at a rate of 150-200 kHz. Today, some ALS sys-
tems allow multiple pulses in the air or are using
several laser guns to send multiple pulses. Those
systems having only one pulse in the air, they are
1015
 3D City Modeling and Visualization for Smart Phone Applications
limited by maximum PRF from higher altitudes
(the speed of light acts as a limitation).
Beam size – the effect of beam size on the
quality of the model obtained has not been well
studied and some findings are contradictory. The
ability to detect either the ground or tree crowns
depends not only on the beam size but also on
the signal strength and receiver sensitivity, the
pulse detection algorithm and the proper sampling
(footprint size versus sampling density).
Basics of Point Cloud Processing
Laser points have coordinate, echo, reflectance,
and time information to be used in the classifica-
tion. Statistical classification methods are com-
monly used. Points can be divided into various
classes using intensity values, echoes (e.g. only,
the first of many, intermediate or last pulse),
height information, aerial/hyperspectral image
features of the same point or time. Height can be
relative to the ground or absolute values. Various
filtering techniques have been developed for low
point detection, power line, building, and ground
classification purposes.
The classification of low points - Removal of
low points is an important pre-processing part
of data analysis and is usually done before the
ground classification. Low points are below the
ground surface and their origin may be manholes
on the street, multiple reflections from trees or
buildings or spreading of the too-strong returns
in the receiver. A single or group of points may
cause an anomaly in the correct ground surface if
not removed before ground point classification.
The classification of ground points – photo-
grammetrists have developed various methods
for obtaining DTMs from laser scanning point
clouds. For example, Kraus and Pfeifer (1998)
developed a DTM algorithm for which laser points
between terrain points and non-terrain points were
distinguished using an iterative prediction of the
DTM and weights attached to each laser point,
depending on the vertical distance between the
1016
expected DTM level and the corresponding laser
point. The method is implemented in SCOP++
(Kraus & Otepka, 2005). Axelsson (2000) de-
veloped a progressive TIN densification method
which is implemented into TerraScan software.
A comparison of the filtering techniques used
for DTM extraction can be found in a report on
ISPRS comparison of filters (Sithole & Vossel-
man, 2004). Selection of the filtering strategy is
not a simple process. In practice, the amount of
interactive work determines the final quality of
the product.
Classification of building points – Previous
research has shown that automatic detection of
buildings from laser scanner data is possible with
relatively good accuracy (e.g. Forlani, Nardinoc-
chi, Scaioni & Zingaretti, 2006; Hug, 1997; Mati-
kainen, Hyyppä & Hyyppä, 2003; Rottensteiner
et al., 2005a, Rottensteiner, Trinder, Clode &
Kubik, 2005b; Vögtle & Steinle, 2003; Zhang,
Yan & Chen, 2006). Different types of features
have been used to separate buildings and vegeta-
tion, including local co-planarity, height texture or
surface roughness, reflectance information from
the images or laser scanning, height differences
between the first pulse and last pulse laser scanner
data, and shape and size of objects.
Basics of Photogrammetry
The fundamental task of photogrammetry is that
3D information is recovered from 2D images of a
scene. When an image is taken from a 3D scene,
depth information is lost. Reconstructing 3D ob-
jects from 2D images by the means of photogram-
metry usually needs georeferencing image space
points to object space points. When images are
georeferenced to object space coordinate system,
interior orientation (IO), and exterior orientation
(EO) of images should be determined. Generally
speaking, IO can be derived from camera calibra-
tion of laboratory. Therefore the determination of
EO is a crucial step for georeferencing in any type
of imagery from terrestrial, airborne or satellite
3D City Modeling and Visualization for Smart Phone Applications

platforms. In photogrammetry, the georeferencing a 3D transformation, for instance, in the process

methods have direct georeferencing and indirect of absolute orientation, 3D model coordinates are
georeferencing according to different ways in transformed to object space coordinates.
deriving EO of images. When information is at- The 2D Projective transformation using eight
tained directly by means of measurements from parameters is typically used for the transforma-
sensors on-board vehicle, this approach is called tion between images or between an image and a
direct georeferencing (Skaloud, 1999). As regard planar object.
to indirect georeferencing method, EO of the image
is determined by performing block adjustment or a1 X + b1 Y + c1
stereo model set-up. The technological develop- X* =
a 0 X + b0 Y + 1
ment in the navigation, positioning, and orientation
systems, results that the direct georeferencing is
widely used. However the fundamental 2D or 3D a 2 X + b2 Y + c 2
Y* = (2)
transformation methods and collinearity equations a 0 X + b0 Y + 1
are still frequently utilized, especially when differ-
ent data sources for consistent 3D modeling and
From a single image geometry, it shows the
mapping e.g. automated texture mapping. In the
fundamental characteristic of the frame image,
following part, some widely used transformation
which is that the perspective centre, the image
methods are introduced.
point and the corresponding object point are on a
Generally, coordinate transformation in 3D
straight line in 3D space. It is also the elementary
can be decomposed to a single effect, such as
condition of collinearity equations.
translation, uniform scale, rotation, reflection,
Collinearity equations (from Mikhail, 2001):
non-uniform scale etc. In photogrammetry, 3D
linear transformation usually is limited to seven
x –xo =
parameters transformation composed of a uniform
scale, three translations, and three rotations. a11 (X − Xl ) +a 21 (Y −Yl ) +a 31 (Z − Zl )
−f 
a13 (X − Xl ) +a 23 (Y −Yl ) + a 33 (Z − Zl )
X *  X  t 
     x
Y *  = kM Y  + t  (1)
     y
 *     y −yo =
Z 
   Z  tz 
a12 (X − Xl ) +a 22 (Y −Yl ) +a 32 (Z − Zl )
−f 
a13 (X − Xl ) +a 23 (Y −Yl ) + a 33 (Z − Zl )
Where X, Y, Z --- original coordinates; X * , Y * ,
Z * --- the transformed coordinates; k --- scale (3)
factor; tx , ty ,tz --- translation factors; M --- rota-
Where
tion matrix; aij --- coefficient of rotation matrix,
f --- calibrated focal length;
i =1, 2, 3;
aij --- coefficient of rotation matrix, i =1, 2, 3; j
j = 1, 2, 3. = 1, 2, 3;
Xl , Yl , Zl --- perspective centre coordinates in
Typical seven parameters transformation is also object space coordinate system;
called Helmert transformation. It is often used in

1017
 3D City Modeling and Visualization for Smart Phone Applications

X, Y, Z --- point coordinates in object space; x, y
--- point coordinates in image plane;
x o , yo --- image principal point coordinates.
The object space coordinates can be calculated
from collinearity equations when initial Z value
is assumed known:
X − Xl =
a11 (x − xo ) + a12 (y − yo ) + a13 (−f )
(Z − Z )
a 31 (x − xo ) + a 32 (y − yo ) + a 33 (−f )
l
Y −Yl =
a21 (x − xo ) + a22 (y − yo ) + a23 (−f )
(Z − Zl )
a 31 (x − xo ) + a 32 (y − yo ) + a 33 (−f )
(4)
Collinearity equations are very famous in pho-
togrammetry. It can be applied to various tasks,
such as space intersection, space resection, and
the relative orientation and so on.
Methodologies for 3D
Building Geometry Extraction
and Reconstruction
We have divided the methods for 3D building
geometry extraction and reconstruction to those
consisting of using aerial images and those con-
sisting of laser scanning and hybrid methods.
Building Extraction Using Aerial Images
Digital aerial photogrammetry supports accurate
measurement of points and structures, which are
usually defined by a human operator (Brenner,
2005). Manual work is time consuming and
expensive even when using today’s state-of-the-
art digital workstations, so a lot of emphasis has
been placed on the development of automation
in data acquisition. Despite significant research
efforts in the past, the low degree of automation
achieved has remained the major problem. Thus,
the majority of development work has focused on
semi-automatic systems, in which, for example,
recognition and interpretation tasks are performed
by the human operator, whereas modeling and pre-
cise measurements are supported by automation.
In the following some approaches of how to apply
automation in building extraction using aerial
images are briefly described (see also Table 1).
On automatic approaches the 2D line segments
can be extracted automatically from two (stereo
pair) or several aerial images covering the same
area. These 2D segments can be searched in the
whole image area or regions of interest (ROI),
which can be given by the operator or found by
using existing ground plans or by using image

Table 1. Features of presented approaches using aerial images

Approach No of Feature search area Extracted features Constructed 3D Additional image

images primitives information used
Haala, 1996 2 ROIs determined by 2D lines 3D rectangles Grey value variance be-
DSM tween images
Henricsson & N Given by operator 2D lines Planes Color attributes for differ-
Baltsavias, 1997 ent sides of 2D lines
Fischer et al., 1998 N Segmentation of whole Points, lines, and 3D corners
images areas
Baillard & Zisser- 6 Whole images 2D lines Half planes to both Radiometry used for line
man, 1999 sides of 3D lines grouping, half-plane com-
puting, and verification
Süveg & Vossel- 2 2D ground plans 2D corners 3D corners Gradients and texture used
man, 2004 in solution evaluation

1018
3D City Modeling and Visualization for Smart Phone Applications
matching to generate DSM, which is then used
to find ROIs. After the 2D features are extracted
from the images, they are constructed to 3D fea-
tures. The approaches described here use either
3D rectangles, matched planes or 3D corners as
building primitives, which are then refined and
assembled to create extracted buildings.
In Haala (1996), as shown in Figure 1, a stereo
pair is used to create a DSM using image match-
ing. This DSM is used to find ROIs, areas where
buildings may exist; and to create a disparity map
between the left and right image of a stereo pair.
2D line segments are extracted from image ROIs.
The disparity map is used to group the 2D line
segments to 3D segments taking into account the
orientation of the 2D segments, e.g. by grouping
them to longer lines or to parallel lines. These
3D segments are grouped in 3D rectangles. 3D
rectangles are then used to obtain approximate
parameters for a saddleback building, and these
parameters are then improved by using an esti-
mation, which minimizes the distance between
the 3D line segments and the line segments of
a saddleback wire-frame model. The estimation
error and the grey value variance of the images
Figure 1. Building reconstruction according to Haala (1996) (Figure based on Brenner, 2005)
are used to select one model from the possible
reconstruction results (Brenner, 2005).
Henricsson and Baltsavias (1997) approach
extracts generic planar patches, which are then
assembled to complete roofs. A generic 3D patch
is non-vertical, roughly planar and encloses a
compact 2D polygonal enclosure with similar
photometric and chromatic attributes along its
boundary. The approach is depicted in Figure 2.
The operator gives a ROI in the master image
where 2D line segments are extracted. Extracted
segments from this one image are then matched
to other images by maximizing an “edginess
measure” along the epipolar line. The edginess
measure is a function of the gradient (magnitude
and direction) in the other images. Geometric and
photometric constraints are also used to reduce
the number of mismatches. This creates a set of
3D segments, which are used to hypothetical
planes. A similarity grouping based on the similar-
ity in proximity, orientation, and photometric and
chromatic region attributes (Henricsson, 1996) is
used to find the most evident and consistent planes.
Vertical walls are added afterwards by projecting
the eaves of the roof to a DTM. Along the ex-
1019
 3D City Modeling and Visualization for Smart Phone Applications
Figure 2. Building extraction according to Henricsson and Baltsavias (1997) (Figure based on Brenner,
2005)
tracted contours, color attributes are computed
for the left and right flanking regions. This infor-
mation is used during the stereo matching, planar
grouping, and similarity grouping.
Fischer et al. (1998) propose a model-based
hierarchical concept for modeling. A generic
modeling approach is employed which relies on
the well-defined combination of building part
models. Starting from low-level image features,
data-driven and model-driven processes are com-
bined within a multilevel aggregation hierarchy,
thereby using a tight coupling of the 2D image
and 3D object modeling and processing. This
formalizes the idea of an early transition to 3D,
as realized in the previously presented approach-
es of Haala (1996) and Henricsson and Baltsavias
(1997). The coupling on each layer does not only
involve a projection of the geometry, but also
propagates constraints and quality measures. The
following layers were given as hierarchy levels
(Braun et al., 1995): scene, object, object part,
aggregated feature, feature, voxel. Accordingly,
for image space: image, aspect, aspect part, ag-
gregated feature, feature, and pixel. This hierarchy
was partly realized for building corners. In this
1020
case, basic features are points, lines, and regions.
They can be extracted by a polymorphous seg-
mentation from the images (Fuchs, 1998). Corners
are aggregated features (one level up in the hier-
archy), consisting of at least three intersecting
planes, a corresponding number of line segments
and a single point. Corners can be obtained on
the basis of a polymorphous segmentation (Lang,
1999). Still one level up, an aggregation of corners
into object parts takes place, in this case into
simple building primitives called ‘connectors’ and
`terminals’. The aggregation of those object parts
finally leads to objects (which themselves are part
of the scene, which is the highest level (Brenner,
2005). The building extraction strategy consists
of three main tasks: 1) reconstruction of 3D cor-
ners (2D to 3D); 2) Generation of building hy-
potheses (3D to 3D); and 3) Verification of build-
ing hypotheses (3D to 2D). In step 1 the
extracted 2D corner hypotheses of different im-
ages are used to derive 3D corner reconstructions
by multi-image parameter estimation. In step 2
the 3D corners are used for indexing into a library
of parameterized building parts, which explain
these corners. The indexed building parts are ag-
3D City Modeling and Visualization for Smart Phone Applications
gregated to complete building hypotheses by
successive steps of merging and connecting build-
ing parts. In step 3 the 3D building hypotheses
are projected back into the original aerial images
and compared to originally extracted features.
The degree of similarity is computed and hypoth-
eses either confirmed or rejected. Furthermore,
geometric reasoning of the images may derive
previously undetected image features. These three
steps are repeated until no further hypotheses can
be generated. Verified building hypotheses predict
earlier unobserved 2D primitives giving addi-
tional information for reconstructing previously
undetected corners. This initiates a further itera-
tion of the three tasks.
The key idea in the approach by Baillard
and Zisserman (1999), as shown in Figure 3,
is to determine 3D planar facets by using both
linear features and their image neighborhoods
over multiple views. The advantage compared to
other approaches is that by using the dihedral line
and left/ right image patches in its vicinity, only
relatively local information is exploited (Brenner,
2005). Six overlapping images are used to extract
2D line segments, which are generated to 3D line
segments by using a line grouping (Baillard et al.,
1999) based on the geometry and radiometry of
Figure 3. Building extraction according to Baillard and Zisserman (1999) (Figure based on Brenner, 2005)
the images. Since the half planes are constrained
by the 3D line segment through which they pass,
only their slope has to be determined. This is done
by a search over all possible angles (the geometric
transformation is given by the plane equation and
the exterior orientation of images), computing
a heuristic correspondence measure of image
similarity; if the plane is correct the intensities at
corresponding pixels will be highly correlated. The
correspondence measure is based on a weighted
correlation between all of the images. The correla-
tion is computed in the neighborhood of points of
interest, which are selected by applying an edge
detector to one image using a very low threshold
of the gradient. After half planes are detected, line
segments and half planes are grouped on the basis
of collinearity and coplanarity, which reduce the
total number of planar patches. Additional lines
are introduced by intersecting existing planar
patches and the extent of each patch is obtained
from a heuristic grouping. Each surface is then
verified using a similarity measure computed
over all images.
Süveg and Vosselman (2004) use existing 2D
ground plans to constrain possible solutions. First,
to be able to project the ground plan to the correct
location in the images, an approximate height is
1021
 3D City Modeling and Visualization for Smart Phone Applications
obtained by projecting the given ground plan to
the images and shifting it along the epipolar line
and evaluating the score function based on the
length of the extracted image lines inside the
ground plan area. In a second step, the ground
plans are partitioned into rectangles, which are
then considered separately. As in Fischer et al.
(1998), 3D corners are used as primitives. They
are obtained from 2D corners, extracted from the
images using Föstner’s operator. Extracted 2D
corners are labeled as wall or roof corners; if a
corner may correspond to a wall primitive it is
labeled as a wall corner, otherwise it is labeled as
a roof corner. This labeling together with geomet-
ric constraints is used when matching the 2D
corners from different images to obtain 3D corner
primitives. Using 3D corners and the 2D ground
plan, an initial estimation for 3D volumetric
primitives is set up. This estimation is then refined
using a gradient-weighted least square fitting in
the images. The fit is evaluated using a metric
based on two measures: the contour measure,
which checks for large image gradients along the
wire frame model projected to the image; and the
texture measure, which evaluates intensity differ-
ences over two images. Finally, all 3D primitives
of a building are estimated again simultaneously,
taking into account the relations between the
primitives, which are introduced as weighted
constraints into the estimation. (Brenner, 2005)
On semi-automatic systems the operator
measures 3D points in a predefined way. In Cy-
berCity Modeler (Grün & Wang, 1998a; Grün
& Wang, 1998b; Grün & Wang, 1999a; Grün &
Wang, 1999b) manual stereo measurement is used
to collect the roof points in a certain order and
point coding, and this information is used in an
automatic roof topology generation. In inJECT
(Gülch & Müller, 2001) the operator selects from
predefined building models and adapts this model
to match the building being measured by placing
a few points to the roof of the building in the other
image of a stereo pair. Automation is applied to
minimize the amount of manually measured points
1022
by operator guidance and by image matching,
which is used to find corresponding points on the
other image automatically.
The above methods of building extraction
serve for nadir aerial images. Currently one of
the highlighted 3D modeling technologies is so
called C3 technology. It utilizes oblique aerial /
UAV images for automatic 3D model generation.
It makes a great progress in the level of automa-
tion. As it is said, ‘C3’s models are generated with
little human intervention. First, a plane equipped
with a custom-designed package of professional-
grade digital single-lens reflex cameras takes
aerial photos. Four cameras look out along the
main compass points, at oblique angles to the
ground, to image buildings from the side as well
as above. Additional cameras capture overlap-
ping images from their own carefully determined
angles, producing a final set that contains all
the information needed for a full 3-D rendering
of a city’s buildings. Machine-vision software
developed by C3 compares pairs of overlapping
images to gauge depth, just as our brains use stereo
vision, to produce a richly detailed 3-D model’
(Technology Review of C3 technologies, 2011).
Building Extraction Using Laser
Scanner Data and Hybrid Methods
Due to the development of scanning systems and
improvements in the accuracy of direct georef-
erencing, airborne laser scanning (ALS) became
a feasible technology to provide range data in
the early 1990s. At that time, ALS was already
considered as a mature technology (Baltsavias,
1999). ALS provides dense point clouds (more
than 10 returns per m2 are possible today) with 3D
coordinates. This makes range data segmentation
relatively easy (Brenner, 2005).
The integration of laser point clouds and
photogrammetric processes with aerial photos
also provides new technological solutions. By
combining the good height assessment accuracy
of a laser scanner and good planimetric accuracy
3D City Modeling and Visualization for Smart Phone Applications

of aerial images, both high accuracy and higher
automation can in theory be obtained. However,
despite the progress that has been made with the
integrating laser scanning systems and digital
images, automated processing of the resulting
datasets is at a very early research stage (Brenner,
2005).
Approaches for building extraction using laser
scanner data either use laser scanner data as it
is (point clouds) or regularized data in raster or
grid format. Whereas automatic building extrac-
tion methods based on aerial imagery typically
concentrate in finding edges, methods based on
laser scanner data typically concentrate on finding
planes, and the edges are detected either as plane
intersections, height jump edges in DSM or given
as additional data (ground plans). Therefore it is
well grounded that hybrid systems integrating the
power of edge detection of aerial imagery and plane
determination accuracy of laser scanner data are
developed. This is coming more evident as digital
cameras are integrated with modern laser scan-
ning systems. Aerial images can also be utilized
in creating a realistic surface texture to model
buildings and giving additional information to
laser scanner data interpretation. In the following
some approaches of how to apply automation in
building extraction using laser scanner data with
or without aerial images are briefly described
(see also Table 2).
Brenner and Haala (1998) approach uses in-
terpolated raster DSMs (usually from laser scan-
ning) and 2D ground plans as data sources for an
automatic and/or semiautomatic reconstruction
process. First, ground plans are divided into rect-
angular primitives using a heuristic algorithm.
For each of the 2D primitives, a number of dif-
ferent 3D parametric primitives from a fixed set
of standard types are instantiated and their optimal
parameters are estimated. The best instantiation
is selected based on area and slope thresholds and
the final fit error. The 3D primitive selection and
parameters can be changed later on using a semi-
automatic extension. A human operator can use
aerial images for the refinement of the automatic
reconstruction when semi-automatic post-pro-
cessing is performed. Rectangles of ground plan
decomposition can be modified or added interac-
tively, and these rectangles are again used for 3D
primitive matching. This interactive mode can
also be used if no ground plan is available. The

Table 2. Features of presented approaches using laser scanner data with or without aerial images

Approach Used data Ground plan Aerial images used? Principle

Brenner, 1998 Raster DSM Yes Interpretation, ground Ground plan division and 3D primitive
plan manipulation matching
Vosselman, 1999 Laser points No No Plane extraction using Hough transform
Vosselman & Dijkman, Laser points Yes No Hough transform, ground plan division,
2001 and 3D primitive matching
Brenner, 2000 Raster DSM Yes No Planes extracted using RANSAC and
rule based acceptance, roof structure by
searching correct plane intersections
Rottensteiner & Briese, Raster DSM No Suggested to be used for Plane extraction using seed regions and
2003 interpretation and wire region growing, plane intersections, and
frame refinement jump edges detected
Hofmann, 2003 Laser points Yes No Detection of planes using clustering of
=> TIN 3D triangle parameter space
TerraSolid, 2003 Laser points No Possible to use for inter- Point classification and plane matching
pretation and measure-
ment

1023
 3D City Modeling and Visualization for Smart Phone Applications
final object representation is obtained by merging
all 3D primitives.
Vosselman (1999) approach uses non-regular-
ized laser scanner data. Roof faces are extracted
using a Hough transform which is extended to
a 3D transformation, where points in planes are
clustered in cluster space. The cluster space is
defined by the slopes in x- and y-direction and a
scalar in the direction of the plane normal. Each
point in cluster space defines a plane in object
space and vice versa. The bin of the cluster space
with the largest number of intersecting planes, i.e.
object points, is selected and this plane estimation
is improved using least squares fitting of points
that are within a given threshold to the observed
plane. All points in the data set are triangulated in
2D, and if all three points of a triangle are within
a given distance to the plane, the triangle is clas-
sified as belonging to this plane. If the size of the
connected triangles exceeds a given threshold, a
planar face has been found. Another least squares
fitting is performed using only the points of the
determined planar face. After removing the points
of the found component from the cluster space,
the bin with the next highest count is selected and
analyzed, and so forth. The edges are found by the
intersection of faces and analysis of height dis-
continuities. No ground plans are used. The main
building orientation is determined and used as a
constraint for the edge orientation. Roof topology
is built by bridging the gaps in the detected edges.
The use of geometric constraints is proposed to
enforce building regularities.
In a later approach (Vosselman & Dijkman,
2001; Vosselman & Süveg, 2001) ground plans
are used to prevent the false roof faces. If build-
ing outlines are not available, they are manually
drawn in a display of the laser points with color-
coded heights. The concave ground plan corners
are extended to cut the building area into smaller
regions and Hough-based plane extraction is con-
strained to these regions. Split-and merge is used
to obtain the final face. This model may lead to
under-segmentation of the ground plan thus gen-
1024
eralizing the building model. In order to preserve
more detail in the model, another reconstruction
strategy has been explored. If the point cloud
within a building polygon can be represented by a
simple roof shape (flat, shed, gable, hip, gambrel,
spherical, or cylindrical roof) the model of this roof
is fitted to the points with a robust least squares
estimation. Building models can be decomposed
interactively so that all parts correspond to the
above mentioned shape primitives. If the point
cloud is such that all roof planes can be detected
automatically, an automatic reconstruction is
attempted based on the intersection of detected
neighboring roof faces and the detection of height
jump edge between the roof faces. If the two above
situations do not apply, the building polygon is
split into two or more parts until each part fulfils
one of the two above conditions. Optionally, point
clouds are edited to remove outlier points that
would disable an automatic roof reconstruction.
Brenner (2000b) uses regularized DSM where
planar faces are extracted using a random sampling
consensus (Fischler & Bolles 1981; RANSAC).
The RANSAC procedure uses as small an initial
data set as feasible and enlarges this set with
consistent data when possible. For example, given
the task of fitting an plane to a set of 3D points,
the RANSAC approach would be to select a set
of three points, compute the parameters of the
implied plane, and count the number of points
that are close enough to that plane to suggest
their compatibility with it (i.e., their deviations
are small enough to be measurement errors). If
there are enough compatible points, RANSAC
would employ a smoothing technique such as least
squares, to compute an improved estimate for the
parameters of the plane now that a set of mutually
consistent points has been identified. The normal
vectors of the extracted faces and ground plan
edges are computed, and faces are labeled based
on the orientation between these normal vectors. A
set of rules expressing possible labeling sequences,
i.e. possible relationships between faces and the
ground plan edges, is used to either accept or reject
3D City Modeling and Visualization for Smart Phone Applications
the extracted face. A plane is generated for every
accepted face, and final roof topology is obtained
by a global search procedure, which selects the
correct plane intersections. The search procedure
uses a discrete relaxation to reduce the number
of possible plane intersections, junctions, and
constrained tree search to cut down search space.
Usually several solutions remain after relaxation
and a subsequent backtracking search with forward
checking can be used to find them. Regularity is
enforced by using additional constraints and least
squares adjustment (Brenner, 2000a).
Also Rottensteiner and Briese (2003) use regu-
larized DSM. Planar roof segments are detected
using a segmentation based on the analysis of
the variations of the DSM normal vectors and/or
a segmentation of aerial image grey levels, and
expanded by region growing algorithms. Plane
intersections and step edges are detected and a
polyhedral model is derived. The parameters of
the building models are improved by a consis-
tent estimation procedure using all the available
information, i.e. laser points are assigned to roof
planes and direct observations for the planimetric
coordinates of building vertices at step edges are
introduced. Hypotheses about geometric con-
straints between planes are introduced to improve
the models further, and parameter estimation is
repeated. Aerial images are suggested to be used
to detect small planar segments and to improve the
geometric accuracy of laser scanner data derived
wire frame models.
Hofmann, Maas and Streilein (2003) approach
uses point clouds obtained by a pre-segmentation
of airborne laser scanner data. It is a plane-based
approach that presumes that buildings are charac-
terized by planes. It utilizes a TIN-structure that is
calculated into the point cloud. The method only
uses point clouds of the laser scanner data that
contain one building. In order to get such point
clouds, polygons coarsely framing the building
can be used to extract the points (e.g. in ArcGIS).
The polygons can be created manually or map
or ground plan information can be used. The
parameters of every TIN-mesh, which define its
position in space uniquely, are mapped into a 3D
triangle-mesh parameter space. As all triangles
of a roof face have similar parameters, they form
clusters in the parameter space. Those clusters that
represent roof faces are detected with a cluster
analysis technique. By analyzing the clusters,
significant roof planes are derived from the 3D
triangle parameter space while taking common
knowledge of roofs into account. However, no
prior knowledge of the roof as e.g. the number of
roof faces is required. The obtained roof planes
are intersected in accordance to their position in
space. By analyzing the intersected roof faces,
the roof outlines are determined and the ground
plan is derived.
TerraSolid (2003) approach extracts buildings
using the classification of the laser scanner de-
rived points. Laser scanner points are classified as
buildings, ground, and high, medium or low veg-
etation. Parameters controlling the classification
are given based on the terrain type and building
characteristics on the target area. Building roofs
are automatically found based on laser hits on
planar surfaces, resulting in vectorized planes of
each roof. Roof boundaries can also be created or
modified manually, and aerial images with known
exterior orientation can be applied (hybrid system)
to measure building boundaries.
Kaartinen and Hyyppä (2006) collected build-
ing extraction methods from eleven research agen-
cies with four testing areas. The data were provided
by the Finnish Geodetic Institute (FGI), namely in
Senaatti, Hermanni and Espoonlahti and one site,
Amiens, provided by the Institut Geographique
National (IGN). Data sources contain ALS data,
aerial images and the ground plan.
Building extraction methods were analyzed and
evaluated from the aspects of the time consumed,
the level of automation, the level of detail, the
geometric accuracy, the total relative building
area and shape dissimilarity. Details of the results
can be found from Kaartinen and Hyyppä (2006).
1025
 3D City Modeling and Visualization for Smart Phone Applications
Commercial Systems, Software,
and Open Sources
In the above we have addressed the means of
geo-data acquisition and methodologies of 3D
model reconstruction. Current status shows that
data processing is still far from fully automation.
Data are processed mainly by the use of existing
systems and software. Table 3 gives a list of current
available systems, software, and open sources.
ACQUISITION OF PHOTOREALISTIC
TEXTURES FOR 3D CITY MODELS
In the section 2, 3D geometry reconstruction of
buildings from Laser scanning and Photogram-
metry have been introduced. However, geometric
models of buildings exclude societal information,
e.g., landmarks, related to the real scene. Models
with textures not only offer rich content to guide
people’s lives, e.g., personal navigation, but they
also provide good visualization.
Photorealistic textures make use of images
taken from the real scene, by perspective correc-
tion and image mosaicing to project it onto the
surfaces of geometric models. There are two ways
for texture mapping: automatic texture mapping
and software-assisted texture mapping.
Table 3. Commercial systems and software
The use of commercial system or software
Photogrammetry, Geoinformation and Remote sensing
system
3D modeling from images
3D modeling from point cloud data (TLS software survey)
Professional 3D software
Free 3D modeling sources (3D modeling applications)
1026
Automatic Texture Mapping
Currently, most systems like ALS and MLS contain
both laser scanner and camera, integrated with
GNSS and IMU. The corresponding relationship
between laser point cloud and images can be
calculated by collinearity equations or projec-
tive transformation (see section 2.2) since the
interior orientation (IO) and exterior orientation
(EO) information of each image are known (IO
parameters and image distortion parameters can
be obtained from camera calibration informa-
tion and EO can be acquired from documented
GNSS and IMU information). The RGB values
from the images can be easily projected onto the
mesh surfaces of geometric models according to
the corresponding relations.
The advantage of automatic texture mapping is
high efficiency. However, object occlusions on the
images would lead to an unpleasant visualization.
When images are collected by mobile systems, the
positions of the cameras on the platform are fixed.
It is fairly often that the case of object occlusion
happens, for example, a car or a tree in front of
a building facade. Some publications related to
object occlusion issues have been addressed by
e.g. (Curless & Levoy, 1996; Früh, Jain & Zakhor,
2005; Stamos & Allen, 2002; Stulp & Fisher,
2001). The method from Früh et al. (2005) was
Name of the system or software
BAE (SOCET SET); ERDAS; EspaCity; ESRI; INPHO; Intergraph Z/I
imaging.
Apollo Photo3D; Photometrix; PhotoModeler; Renoir.
Rhino 3D; 3D Reshaper; Cloudworx; Cyclone; Geomagic; Kubit; Innov-
mentric Polyworks Software; Multigen Creator; Multigen Vega Prime;
Paraform; Pointools; Rapidform; Realworks; Solidworks; Terrasolid.
AutoCAD 2000; Alias Wavefront (Maya); 3D Studio Max
Blender; K-3D; Art of Illusion; SOFTIMAGE|XSI Mod Tool; Zmod-
eler; TopMod3d; Google SketchUp; AutoQ3D Community – 3D Editor;
Anim8or; Seamless3d; BRL-CAD; 3DPlus; 3D Canvas; eDrawings;
Blink 3D; Minos; freeCAD; Bishop3D; K3DSurf; DesignWorkshop Lite;
GDesign; Sweet Home 3D; trueSpace; Alibre Design Xpress; 3DVIA
Shape; DAZ 3D; Wings3d.
3D City Modeling and Visualization for Smart Phone Applications
based on the MLS point cloud. Four steps were
taken: i) Separate the trajectory of scanning into
approximate straight-line pieces; ii) Transform
laser points of each piece into a depth image;
iii) Split the scan points into a foreground layer
(e.g. a car or a tree) and a background layer (e.g.
a building facade) according to the most frequent
depth value; iv) Remove the foreground layer and
fill the holes on the background using the linear
interpolation between the closest right and left
structure boundary point horizontally. And then a
similar interpolation is done vertically, using the
already filled in points as valid boundary points.
When a large area occlusion happened, additional
geometric information was obtained from stereo-
imagery. Therefore, full automation to solve the
object occlusions is a challenge work.
For photorealistic 3D models, we always
make efforts towards two goals: automation and
good visual appearance. It is a challenge work to
develop a fully automatic method for 3D models
with good photorealistic visual appearances.
Software-Assisted Texture Mapping
The images are taken manually without needed
orientation information. In this case, it is common
to utilize existing 3D software for texture map-
ping. The main work for software-assisted texture
mapping is texture preparation. It includes image
perspective correction, image mosaicing, and
object occlusion removal from building facades.
Perspective correction is indispensability for
texture preparation. When objects obstruct the line
of sight in front of a building or if buildings are
high, the images have to be taken at an oblique
angle. In addition, even if images were taken at
approximately right angles to the building facade,
they still need some perspective correction.
Usually a building facade is corresponding to
an individual texture. Due to the limited field of
view of the camera when covering a large build-
ing facade, one image only covered part of the
building facade. Thus, several images needed to be
combined to create an image mosaic of the facade.
The advantage of images taken manually is that
the position and pose of the camera can be chosen.
In some cases, it enables to avoid object occlu-
sions. However, in the majority of cases, object
occlusions still exist. There is a large work load
for manual removal of object occlusions. Usually
it yields a good and realistic visual appearance. In
generally speaking, manual work load and good
realistic visual appearance are proportional.
After the preparation of textures, texture map-
ping can be performed by 3D software. The map-
ping process is to apply mathematical functions
to assign each point on the texture to each point
on the object surface. Different software usually
employs different texture mapping methods such
as UV mapping, UVW mapping and polynomial
texture mapping.
SMARTPHONE 3D VISUALIZATION
FOR NAVIGATION AND LBS:
APPROACHES AND TECHNIQUES
The provision of 3D graphical navigation ap-
plications and location-based services for mo-
bile devices has been envisaged for a long time
(Zlatanova & Verbree, 2003). One of the benefits
of 3D navigation, according to Coors, Kray, Laakso
and Elting (2005), is that people can recognize
landmarks and find routes in cities easier by using
3D models than using a symbolic 2D map because
of the highly visual correspondence between map
objects and real world objects.
One of the major challenges for a 3D navi-
gation system is the development of a highly
efficient 3D navigation engine. Although some
desktop 3D navigation applications have been
introduced in the past decades (Sharkawi, Ujang
& Abdul-Rahman, 2007), the 3D capabilities of
mobile devices are restricted in several aspects:
dedicated 3D hardware chip, floating point units,
hardware division circuits, memory bandwidth,
1027
 3D City Modeling and Visualization for Smart Phone Applications
and CPU speed (Zlatanova & Verbree, 2003), and
the lack of an accurate integrated positioning sys-
tem. Besides, the visibility of scene management
on the low-resolution screens of mobile devices
needs adapting.
Following recent improvements implemented
in smartphones in regard to increased computa-
tional and graphic capabilities, improved power
management and the reduced price of the GPS
module, the navigation application is now be-
coming a standard feature in smartphones, tablet
computers, and other portable devices. This makes
feasible the development of a 3D personal navi-
gation visualization engine. Although there have
been already a few industrial Portable Navigation
Devices (PND), e.g. Garmin Nuvi 765T GPS,
support visualization of 3D models on 2D maps
from different view angles, the development of
a real 3D navigation engine, which seamlessly
supports indoor/outdoor navigation, is still a
challenge with smartphones.
This section mainly introduces different ap-
proaches and architectures of the 3D personal
navigation visualization engine to support 3D
navigation and LBS application. We firstly pro-
ceed to introduce a prototype smartphone-based
real-time 3D engine. This prototype engine has
so far been able to achieve high performance in
real-time 3D rendering of very complex urban
models; this has been demonstrated in indoor/
outdoor navigation within the Shanghai EXPO
area and within the Tapiola/FGI study area in
Finland. We also discuss another 3D visualization
engine implementation approach based on client/
server architectures, which use the server’s 3D-
rendering capabilities and provides a 3D- render-
ing service to the client portable devices through a
communication network. Finally, the advantages
and disadvantages of these 2D vs. 3D navigation
methods are investigated and some future trends
are discussed.
1028
Approaches to 3D Personal
Navigation Visualization
A smartphone-based 3D visualization engine
generates 3D graphical representations of the
navigation context environment, organizes them
into scene graphs, and allows mobile users real-
time navigation and interaction within the 3D
scene. To date, however, all this is still a real chal-
lenge when considering the limited computation
and graphical capabilities of mobile hardware
systems and the low bandwidth of the wireless
communication network when compared to the
vast volume of massive geometrical and texture
data of the surrounding 3D city models and terrain
data that must be transferred.
Fortunately, given the new generations of
iPhones or Android Smartphone systems and
iPad/Xoom tablet computers, the computation
and graphical capabilities of this genre of hard-
ware have increased dramatically due to the use
of multi-core systems and new graphical chips,
which make it affordable to run high-performance
3D graphical applications. Alongside this de-
velopment, the emerging 4G communication
technology, e.g., TD-LTE, means that wireless
communication network speeds will ultimately
satisfy the need for a fast approach to 3D model
data transfer.
The development of a smartphone-based 3D
navigation visualization engine with powerful
rendering capability is normally based on low-
level high-performance 3D graphics API, e.g.,
OpenGL ES API, which is now available on almost
all modern smartphone OSs, e.g., iOS, Android,
Symbian 3, and MeeGo.
Here we proceed to introduce a prototype
smartphone-based real-time 3D engine, which
was initially developed for the Symbian environ-
ment using standard C++ and OpenGL ES, and
it can be ported to other mobile complex urban
models as was demonstrated in the indoor/outdoor
navigation trial conducted within the Shanghai
EXPO area and within the Tapiola/FGI study area
3D City Modeling and Visualization for Smart Phone Applications
Figure 4. The architecture of the 3D navigation visualization engine
in Finland. The engine is basically structured as
having three layers: an API interface layer, a core
engine layer, and a data management layer, and
these are illustrated in Figure 4. The top layer is the
interface layer, which is used for cross-component
functional operation, request handling, and data
exchange (Chen et al., 2010).
The core engine layer takes care of the 3D
scene visualization, e.g., scene management, scene
rendering, mathematical computation, coordinate
transformation, and model object management.
To enable the 3D scene visualization for a large
area, the objects in the scene are classified into
two categories within this layer. One is for the 3D
models (such as buildings, tree and poles), while
the other is for the texture of the land surface,
which consists of ortho-rectified digital aerial
photos. All of the objects are subdivided into
squared tiles with geographic references derived
from the original data. Using this data processing
technique, the models and images are tailored to
satisfy the navigation application, balancing be-
tween performance and visualization effects.
The data management layer accesses the 3D
models and ground-texture images persistent on
the flash disk of the mobile phone via an inde-
pendent thread (or active objects for a Symbian-
based system).
As regards the general object relationships in
the 3D Visualization Engine (3DVE), there is a
key object called NaviSceneControl, which func-
tions as the hub of the 3DVE. It is implemented
as a visual user interface (UI) and controls itself
with a window for displaying a 3D scene in the
application. It is also responsible for the following:
• Connecting with Interface Objects in the
API Interface Layer to exchange informa-
tion with the main application and other
components(Chen et al, 2010; Liu et al.
2010);
• Creating and displaying the scene’s
Renderable Objects such as terrain surface,
3D models, landmarks, and planned route;
• Importing incoming Point of Interest
Information (POI) from the Interface
1029
 3D City Modeling and Visualization for Smart Phone Applications
Layer and re-organizing this information
as Geometry Objects;
• Importing keyboard/command information
from Control Info in Interface Layer and
processing operations according to the dif-
ferent event types using Platform Layers’
(S60) event-processing mechanism;
• Updating view information and exporting
to the ViewInfo and LmkInViewList in the
API Interface Layer.
The Renderable Objects encapsulate the
drawing operation of the Geometry Objects. The
persistence of the geometry objects is managed
by the Cache Access Objects, which use the plat-
form’s (S60) multi-thread techniques, i.e., active
objects, to access local caches from flash memory.
In the SDK Layer, Symbian C++ provides a
common language tool for developing the engine.
All objects in 3DVE are implemented using this
Symbian C++ and its open C++ extension. In
addition, the OpenGL ES library is also used to
provide APIs to realize the rendering facilities,
specifically for Renderable Objects in the Core
Engine Layer.
The engine has been tested within the Shanghai
EXPO area for outdoor navigation demonstration
as well as within the Tapiola/FGI study area for
outdoor/indoor navigation demonstration. The
Shanghai EXPO area has over 200 pavilion models
Figure 5. Demonstrations within the Shanghai EXPO area and in Finland using a Nokia 2610 smart-
phone: a. China Pavilion; b. Virtual navigation within the Shanghai EXPO area; c. A 3D view of the
city center of Tapiola, Finland; d. FGI’s building
1030
and high-resolution aerial photos. The ordinary
pavilion mesh consists of thousands to tens of
thousands of vertexes and dozens of texture images
for each pavilion. The rendering performance has
been successfully improved to less than 1 second
of the update interval on a Nokia 6710 navigator
mobile phone. Figure 5a, 5b show the screenshot
of the China Pavilion and the virtual navigation
in the EXPO area respectively.
Two test areas in Finland were selected for
indoor/outdoor tests, as shown in Figure 5c, 5d.
Figure 5c demonstrates the outdoor navigation
within the Tapiola area, while Figure 5d demon-
strates the indoor navigation along the corridors
of FGI’s 3rd floor premises. The positioning ap-
proach for indoor environment was arranged
through a combination of the smartphone’s blue-
tooth (BT) and WLAN connected to the access
points (AP) installed within the building (Pei et
al. 2010), and this was achieved with a position-
ing accuracy of about 5 meters.
Image-Based 3D Visualization Engine
Compared to a smartphone-based 3D navigation
engine, the image-based 3D navigation engine
retrieves 2D perspective images of the 3D geo-
virtual environments from remote servers via a
high-speed network.
3D City Modeling and Visualization for Smart Phone Applications
A typical example is the Web Perspective View
Service (WPVS) and its extension WPVS++ as
proposed by (Hildebrandt et al. 2010; Hagedorn
et al. 2010). This approach generates a set of
2D images of the perspective views of 3D geo-
virtual environments (e.g., virtual 3D city models)
using the server-based 3D rendering service.
Retrieval was continuous from the client side on
the light-weight devices. To provide users with a
high degree of interactivity, the WPVS includes
additional thematic information encoded into
non-color image layers for local reconstruction
of the visual representations of the remote 3D
geo-virtual environments and corresponding
functionalities for the exploration and analysis of
the presented 3DGeoVEs by simple clients, e.g.,
by requesting 3D positions, retrieving object IDs,
and measuring distances. A procedure called View
Process (VP) transforms commands for updating
the visualization from the controller, which pro-
cesses user-interaction events into the calling of
service operations for requesting images or other
functionalities. In order to reduce the latency
between user input and display updates, which
can be introduced using the narrow bandwidth of
the network and increased frame rates, a client-
side 3D reconstruction and rendering techniques
based on image-based modeling and rendering
(IBMR), point-based modeling and rendering
(PBMR), and the nested, partial visualization
pipeline (NPVP) pattern, are employed. Since
the WPVS is provided as an open, interoperable
HTTP-based interface, it can be easily integrated
into web-based applications or mobile systems to
provide smartphone 3D navigation.
WPVS supports interaction with the generated
image by proposing a series of APIs. This includes
techniques for maneuvering the virtual camera
through the 3D world and for retrieving informa-
tion about the visualized objects. For example, a
GetView with style specifications and parameters
should be used to request specific content (e.g.,
feature type or building usage) for the requested
image layers. A GetPosition is proposed to retrieve
3D geo-coordinates for a specific 2D pixel posi-
tion in the generated image. A GetFeatureInfo
operation should provide extra information about
the geo-objects at a specific pixel position. Us-
ing these APIs, the client provides an extendable
interactive method between the mobile device
and the remote server.
A key advantage of this image-based approach
is that the complexity exposed to a client for
displaying a visual representation is reduced to a
constant factor, primarily depending on the image
resolution, and thus it provides better scalability
with the increasing complexity of 3D city models,
concurrent users, and the computation required for
real-time photorealistic 3D visualization. Finally,
since the 2D representative images of the 3D geo-
virtual environments in this approach depend on
the server’s 3D rendering service, the limitation
of this method is that it can only be accessed via
a high-speed network environment, and high-
performance 3D rendering servers are preferred to
provide multi-user service. An example is shown
in the Figure 6.
Techniques of 3D Personal
Navigation Visualization
Tilling is a technique that is used to subdivide
original full-resolution terrain texture images to
generate squared tiles of predefined sizes with
reduced resolution. It is widely applied in large-
area terrain visualization on desktop applications,
e.g., Google Earth. The tile structure used in the
3DVE engine is efficient for 3D models and ter-
rain textures management. An XY grid starts in
the lower left hand corner of the Earth, defined as
(-90, -180) in latitude and longitude (see Figure
7). The coordinates of each box are from the lower
left hand corner of the box. The initial point is
considered to be X = 0, Y = 0. With a given tile
size, it is easy to find the box’s X-Y coordinates for
a given latitude and/or longitude, and vice versa.
The tiling strategy is very important for a
mobile system because we just need to load the
1031
 3D City Modeling and Visualization for Smart Phone Applications
Figure 6. An example of WPVS service showing the feasibility of the image-based 3D visualization
engine on mobile devices (courtesy of Hagedorn et al., 2010)
Figure 7. Illustration of the tiling technique
implemented in 3DVE
small images and 3D models according to the
extent of the view from the current viewing angle,
rather than the entire images with full resolution
and all 3D models.
Using a static level of detail technique could
also significantly improve the rendering per-
formance by reducing the complexity of 3D
object representation as it moves away from the
viewer according to some key factors such as
object importance, eye-space speed or position.
This significantly decreases the workload on the
1032
graphics rendering pipeline. It is especially im-
portant for 3D model data streaming between the
memory and flash disk, vertex transformations,
and texture mapping (Wikipedia, 2011). To do so,
the 3D models should be classified into different
layers and importance according to the service
purpose of the LBS applications. Meanwhile, the
maximum viewing height and distance between
the scene camera and the visible 3D objects must
also be defined accordingly.
Since building models, e.g., within the
Shanghai EXPO area for our demonstration, are
normally very complex, usually with thousands
of vertexes, tens to hundreds of wall textures
for a single building model, it is not possible to
acquire acceptable performance by rendering a
3D scene using OpenGL’s instant mode given the
current mainstream smartphone’s graphical and
computational capabilities. Using the Vertex Buf-
fer Object (VBO), the rendering performance can
be enhanced by a factor of 10-30 In order to use
VBO, the vertex coordinates of the models need
to be linearly stored in the disk file. Meanwhile, to
support easy inclusion of 3rd party 3D models, the
local model-space coordinate system and WGS84
coordinate system are used simultaneously in 3D
models. To import a 3D model with a local model-
space coordinate system into the engine, the user
3D City Modeling and Visualization for Smart Phone Applications

needs to know only the latitude, longitude, and
the azimuth of the local origin of the 3D model.
An automatic geometry transformation for the
3D model into the universal WGS84 coordinate
system is performed before visualization using the
orientation information. Thanks to this, the model
author does not need to know which coordinate
system will be used in the model’s visualization.
To reduce the data size of the 3D models and
to improve the visualization performance, the
original *.3ds file created with 3D Max Studio
software from the aerial photo and CAD draw-
ing is compressed and reformatted to fulfill the
requirements of the mobile device. In order to
support model simplification and tiling, a utility
software is developed to create a tile of models
and put each model into its corresponding location
grid, packing all small textures of a model into
one or two images of size 512x512 or 256x256
using the bin packing algorithm (Lodi, Martello &
Monaci, 2002). This is done in conjunction with
file formatting from *.3ds file to support VBO
rendering, which results in significant time sav-
ings in reducing graphical rendering and texture
loading times.
2D vs. 3D Navigation Application:
A Comparison from the
User’s Point of View
In theory, compared to traditional 2D maps, 3D
mobile maps involve volumetric instead of flat rep-
resentation of space, realistic instead of symbolic
representation of objects, more variable views
that are directional and bound to a first-person
perspective, more degrees of freedom in move-
ment, and dynamically changing object details.
In order to facilitate better understanding of
the differences between 2D and 3D navigation,
two experiments and surveys were conducted by
the 3D-NAVI-EXPO project team. Thirty-two
volunteers as testers were brought to the Shanghai
EXPO area (where they had not been before) and
to a street unknown to the testers in Finland. They
were then asked to use the developed 3D personal
navigation prototype software on a GPS-enabled
smartphone respectively to find and navigate to
a destination. Each tester was required to fill in a
feedback form at the end of the test. A summary
of the feedback comparisons from the users is
illustrated in Table 4.
It is an indisputable fact that a 3D map has the
volumetric concept, which is an important dimen-
sion of the cityscape. Rich and realistic visual
details of 3D models help the user to identify an
object easier. Further, 3DPN supports the virtual
reality capability, which ensures that the user
obtains an impression of a place without being
present there in person. However, there are also
some disadvantages of the 3DPN, and the follow-
ing are some examples:
• The ambiguous textual or missing details
of an object can mislead the user when try-
ing to rapidly understand the physical en-
vironment based on the 3D models.

Table 4. Comparison of user experiences with 2D vs. 3D navigation applications

Navigation Activity
Identifying the position
Finding the direction
Navigating to the destination
2D
Known landmarks, street names, relative directions,
street crossings, cardinal directions.
Known landmarks, street names, relative directions,
street crossings, cardinal directions.
Consumes fewer resources to maintain, but takes
more time to identify the target while approaching
the destination. Easy to get a general picture.
3D
Building shapes, facades in general, facade
details, and relative directions.
Finds the direction while identifying the posi-
tion. Even one building is enough.
Costs more for resources, but easy to identify
the target while approaching the destination.
Difficult to get a general picture.

1033
 3D City Modeling and Visualization for Smart Phone Applications
• Inaccurate position and heading without
any processing can introduce a disaster
into the 3D visualization. For example, if a
user gets an incorrect position or heading,
the user will obtain a problematic 3D scene
with the wrong orientation and position.
• Environmental elements, such as light
and weather, also influence the user’s
understanding.
• The field of view in 3DPN cannot simulate
the field of view of a human being exactly
due to issues such as the limitation of the
screen size of a smartphone.
Some efforts have been made to tackle these
disadvantages, e.g., Pei et al. (2010) developed a
method of using built-in sensors in a smartphone to
assist 3D personal navigation in a GPS-degraded
area. Recognition of motion is firstly applied
to detect the context by using the data from a
built-in tri-axis accelerometer. Afterwards, a map
matching algorithm is implemented to snap the
GPS position onto the middle of the nearest road.
Finally, in order to avoid the heading drift during
the map matching procedure, a Bayesian Network
-based heading change detection method is applied
by combining the orientation readings from both
the GPS and the digital compass integrated in
the smartphone. With this improvement, the user
experience has been greatly improved in the 3D-
NAVI-EXPO project (Pei et al. 2010). We believe
that further study of the other disadvantages of the
3DPN is needed and that such study will improve
the user experience.
A CASE STUDY: 3D MODELS FOR
TAPIOLA AREA (ESPOO, FINLAND)
Main contents of the case study were originally
from Zhu et al. (2011). It presents a full proce-
dure of the generalization of photorealistic 3D
models from MLS data. The test site is located
in the Tapiola area, Espoo, Finland. It is an area
1034
of commercial buildings, including shopping
centers, banks, government agencies, bookstores,
and high-rise residential buildings, with the tall-
est building being 45 m in height. The data were
collected by the Finnish Geodetic Institute (FGI)
ROAMER system, which was a laser scanning
system mounted on a mobile platform. The 3D
models were constructed from the collected data
in two steps: geometry modeling and texture
mapping. The resulting models are applied to a
smartphone for personal navigation.
Data Acquisition
The FGI ROAMER system is a mobile laser
scanning system which can be mounted on vari-
ous vehicles. The hardware sub-systems of the
ROAMER include the following: (1) a FARO laser
scanner; (2) a GPS-INS navigation equipment; (3)
a camera system; (4) synchronization electronics;
and (5) a mechanical support structure. The mirror
rotation frequency, or profile measuring frequency
of the FARO LS, is typically set to 24 Hz, 49 Hz
or 61 Hz in mobile applications, and the verti-
cal angular resolution can be set to 0.009–0.288
degrees. For platform speeds of 50–60 km/h, the
profile spacing is about 30 cm when using the
profile measuring frequency of 49 Hz. With a
frequency of 49 Hz, the profile interval is less than
20 cm when the speed of the mapping unit is kept
below 40 km/h, and the point resolution along the
profile is still 2.5–5 cm, which is sufficient for
the practical ranges of 20–40 m, respectively, in
the urban environment (Kukko, 2009).
The ROAMER configuration as used in Tapi-
ola can be seen from Table 5. The data included
altogether about 160,000 profiles, each profile
having 2,150 points with 3D coordinates and
return intensity. Data collection lasted about one
hour, and covered an area 180 m by 280 m. The
laser data were transformed into a map coordinate
system (ETRS-TM35FIN with GRS80 ellipsoidal
height). Due to narrow streets and high buildings,
images from the ROAMER system did not meet the
3D City Modeling and Visualization for Smart Phone Applications
Table 5. The ROAMER’s acquisition parameters
Date May 12th 2010
Laser scanner Faro Photon™ 120
Navigation system NovAtel SPAN™
Laser point measuring 244 kHz
frequency
IMU frequency 100 kHz
GPS frequency 1 Hz
Data synchronization Synchronizer by FGI, scanner
as master
Cameras Two AVT Pike
Profile measuring fre- 49 Hz
quency
requirement for high-quality textures. Therefore,
the images were taken separately using a Canon
EOS 400D digital camera.
Geometry Reconstruction
Geometry reconstruction from raw MLS data
includes the following steps: noise point filtering,
object classification (ground, buildings, and the
other objects), detection of planar surfaces, key
point extraction, and surface meshing. Our goal in
geometry reconstruction was to utilize some key
points in building model construction. Detailed
algorithms for noise points filtering, ground point
classification, building point extraction, detection
of planar surfaces, and derivation of the buildings’
key points are presented in each section.
• Noise points filtering: The acquired data
were composed of 340 million points with
georeferenced XYZ coordinates. A noise
point here refers to a point with markedly
deviation from the other points. We used
the data of 2D projections (x-y, x-z, y-z)
respectively to filter out these noise points.
The data in each 2D projection were dis-
tributed into 10-by-1 bins. A three-dimen-
sional histogram of bivariate data was ap-
plied to calculate the number of elements
falling into each bin, and the positions
of each bin center were calculated. The
threshold (T) for the number of the points
in each bin (e.g. 800) can be defined by the
user according to the different density of
the point cloud and the size of a dataset.
• Object classification: After noise remov-
al, the number of points is greatly reduced.
These points contained ground points,
building points, and other object points,
which needed to separate.
Ground Point Classification
The algorithm was developed for a relatively flat
area. In order to facilitate the description, Table
6 depicts abbreviations used.
This algorithm was developed by considering
the most frequently occurring height value in the
data. For the MLS data in the relatively flat area,
ground points usually hold the most frequently
occurring height value, especially in an urban area
without a large area of low vegetation. In this
algorithm, we used 0.25m as a tolerance taking
into account the influence of measure accuracy
of a few decimeters in the ROAMER system and
also the unevenness of the terrain. In addition, a
3m threshold for Zd was used mainly considering
that it is possible that data in some bins contain
only the points from the reflection area due to the
incomplete removal of the noise points. After the
algorithm was performed, ground points are ex-
tracted.
Building Point Classification
After ground point extraction, the remaining data
contain buildings and other objects. In 3D city
model creation, building point classification is a
key step for building model reconstruction. Un-
like ALS data scanned from the several hundreds
or even thousands of meters above the top of the
object, MLS data are collected close to the ground
passing along a street. Detailed 3D information of
1035
 3D City Modeling and Visualization for Smart Phone Applications

Table 6. A list of abbreviations

Abbreviation Description
Zf_data The most frequently occurring height value for the dataset;
Zf_grid The most frequently occurring height value for data in each grid;
Z_min The minimum height value for the data in one file;
Zmin_grid The minimum height value for the data in the grid;
Zd The difference between Zf_data and Zmin_grid
Algorithm 1. Ground point classification
1: Calculate Zf_data: mode (height value of data)
2: Compare the difference between Zf_data and Z_min
3: if the difference <=1m then
points with the height value <= Zf_data +0.25m accepted as ground points
4: else
grid points in XY plane into 10*10 bins
5: for each bin
calculate Zf_grid and absolute value of Zd: abs(Zd)
6: if abs(Zd) <= 3m then
compare the difference between Zf_grid and Zmin_grid
8: if the difference <= 0.25m then
points with the height value <= Zf_grid +0.25m as ground points
9: else
points with the height value <= Zmin_grid +0.25m as ground points
10: end if
11: end if
12: end for
13: merge all ground points from each bin
14: end if

the vertical planes of buildings can be obtained.
The algorithms were developed based on the as-
sumption that the building facades are vertical
with respect to the ground. Figure 8 shows the
principle of the algorithm. We first utilize two
cutoff boxes (cutoff box 1 and cutoff box 2) to
find the overlap between them. Overlapping areas
reveal high buildings such as B1, B2, and B5. After
the overlap parts are detected, the whole buildings
are extracted. Then these building points were
removed from the data. The lower buildings were
then extracted from the remaining data by setting
two other cutoff boxes (cutoff box 3 and cutoff
box 4). Then the overlaps between them can be
found, which are B3, B4, B6, and B7. However, it
is possible that the above detected buildings still
contain other points (e.g. tree points). Therefore,
these non-building points need to be removed.
We employ image processing technology for
non-building point removal. The x-y coordinates
of the rough building points (e.g. B1, B2, and B5)
are gridded into predefined pixel size and trans-
formed them to a binary image. By image param-
eter control (e.g. image region properties: area
size, eccentricity value, major axis length and
minor axis length, and morphological structure
element), non-building points can be removed.
After image processing, the binary image is trans-
formed back to 3D points. Similarly, this image
processing method is also applied to the detected
lower buildings (e.g. B3, B4, B6, and B7) for
non-building point removal. The resulting build-
ing points are obtained by merging these two sets
of cleaned points. The benefit of this algorithm
is that trees can then be effectively isolated from
buildings. Figure 9 shows the procedure of the

1036
3D City Modeling and Visualization for Smart Phone Applications

Figure 8. The basic principle of the algorithm of building point classification

transformation from 3D points to a 2D binary
image, then applying image processing technol-
ogy for non-building point removal, and trans-
forming the cleaned image to 3D points.
Key Point Extraction and Surface Meshes -
One of our goals was to generate building models
that require low amounts of memory and can thus
be applied for applications such as mobile de-
vices with limited memory, and rapid rendering
for visualization. The buildings extracted by ap-
plying the above steps still have a huge amount
of points. Therefore, key point extraction was
essential. We took two steps for key point extrac-
tion: identify the planar building facade and extract
the key points. There are two popularly used
planar detection methods including surface normal
consistency and region growth. Here the algorithm
for coplanar point detection is realized as follows:
1. Two neighbor points (P1, P2) were selected
randomly to form a vector (V1).
2. A third point (P3) was added to form the
vector V2 with P1. If V2 was not parallel
to V1, then it was considered to be the basic
plane.
3. If V2 and V1 were collinear, then this point
(P3) was taken as a coplanar point. If not,
then the next step was taken.
4. P4 is any point in the dataset. P4 and P2
formed vector V3. If it satisfies the coplanar
condition V3· (V1×V2) = 0, it is regarded
as a coplanar point.
5. When a plane was identified, the points of
this plane were removed from the dataset.
6. Steps (1) to (5) were repeated until all points
were allocated to a certain plane.

Figure 9. An example of the procedure of transformation of 3D point clouds into 2D binary images,
non-building points filtering on images, then into 3D point clouds

1037
 3D City Modeling and Visualization for Smart Phone Applications
Key points here refer to the corners of the
building facades. The corners could be deduced
from the minimum and maximum values of the
X or Y coordinates and the Z values. In order to
achieve high quality (e.g., completeness and cor-
rectness) of building geometry, manual checking
and correction are needed. Manual checking and
correction are performed by visually comparing
the identified building planes and corners with
terrestrial images and also making correction
by resetting parameters in the algorithm for bet-
ter results. Usually completed models can be
achieved by using building roof information from
airborne-based data and the building facade from
MLS data. However in our data the advantage is
that the buildings are scanned from all sides. This
means that complete models can be created with
manual assistance.
Following key point extraction, the meshes
were constructed for the surfaces. When dealing
with complex buildings, they were divided into
cubes according to different heights. The meshing
surfaces were imported to 3ds Max for further op-
erations, e.g., surface merging and small geometry
adjustment for good visualization.
Texture Mapping
A detailed procedure of geometry reconstruction
from MLS data has been addressed in the above
Figure 10. 3D models of Tapiola
1038
steps. However building geometric models do
not contain landmark information in the scene,
therefore it is important to provide photorealistic
visualization for the models.
Due to narrow streets and high buildings, im-
ages from the ROAMER system did not meet the
requirement for high-quality textures. Therefore,
the images were taken separately using a Canon
EOS 400D digital camera. Thus manual texture
preparation, e.g., the image perspective correc-
tion, noise point removal, and image mosaicing,
is needed. These images were used to provide the
textures of the building facades. For complete 3D
models, the textures of the building roofs were
also needed. We made use of aerial images from
website to obtain the textures of the building roofs.
Each roof plane was made to correspond to one
texture. Image processing software (Corel Photo
Paint) was utilized in texture preparation. The
3ds Max software was used in texture mapping.
Figure 10 shows the results of 3D models in the
Tapiola area.
Implementation into a 3D Engine
Recently, the development of 3D game engines
has been rapid and, currently, there are a number
of free and commercial options available, such as
Blender, realXtend, and CryEngine 3. As mobile
devices have improved in processing power and
3D City Modeling and Visualization for Smart Phone Applications
graphics performance, also 3D game engines ca-
pable of running on mobile devices have started
to emerge and some examples that run on mobile
devices as well as desktop computers include
ShiVa3D, SIO2, Unity, and Unreal Engine 3. This
is a significant progress, as previously the devel-
opment of a 3D engine for a mobile device may
have taken multiple years and has been limited
to a single platform and application.
With the modern game engines and develop-
ment tools, it is now extremely easy to deploy
the produced 3D model into a game engine with
a graphical user interface. The game engine also
allows for further application development that
may include networked, interactive, real-time
and multiplayer aspects. Also, many of the game
engines feature simultaneous multi-platform
development, which means that the same model
and the program code can be utilized on many
different platforms with minimal effort.
The Tapiola 3D model is implemented with
Unity 3D in the Android environment. It is ac-
cessible from Android Market (http://market.
android.com/details?id=com.FGI.Tapiola3D).
Figure 11 shows an example of the scene on an
Android device.
Figure 11. A screenshot of the Tapiola 3D model running on an Android device
FUTURE ASPECTS AND
CHALLENGES
The development of sensors for 3D data acquisition
provides more and more solutions for 3D model
generation. Current laser scanning has been able
to offer a dense point cloud, e.g. ALS enables up
to 40 points /m2, MLS up to 1.3Mpoints/s prov-
ing tens of thousands of points /m2. Due to the
increase in the density of point clouds, the mea-
surement accuracy and automation level have also
greatly improved. The poor capability to detect
breaklines will be improved by systems proving
higher density of point clouds. For the extraction
of building geometry for the models of fly-through
views, some successful examples (more automa-
tion and better results) show that good results can
be achieved by the use of both laser and ground
plan data. Thus, high resolution 3D models for
walk-through views require both airborne-based
and terrestrially-collected data. Airborne-based
data provide building roof information whereas
terrestrial data offer detailed building facade
information. Their combination increases the
completeness and the level of detail of the 3D
models, and in the near future large areas can be
created in 3D for personal navigation.
1039
 3D City Modeling and Visualization for Smart Phone Applications
Despite much effort in the development of
automated methods, fully-automatic methods are
not yet available. As regards building geometry
reconstruction, although the use of both laser
scanning and images improves the level of automa-
tion, difficulties still arise when complex models
need to be reconstructed. Many of the developed
methods are scene-based. As regards photorealistic
textures, object occlusions have caused the main
problems. Manual work can result in good visual
appearances, but it leads to the lower efficiency
of the reconstruction process. Thus, a high degree
of automation with good visual appearance and
high extraction accuracy provides challenges also
for future work.
Due to the limited hardware resources and
graphical computation power of mobile devices,
3D visualization techniques for the specified
mobile platform need to be developed. The chal-
lenge related to real-time fast 3D rendering and
display of large datasets and of complex urban
environments still remains. We introduced two
main approaches to 3D navigation’s visualization
engines. With each engine, personal 3D naviga-
tion in an urban environment is feasible. The 3D
visualization engine developed in conjunction
with the 3D-NAVI-EXPO project realized a
high-performance real-time 3D rendering of very
complex urban models as we have demonstrated
in the experimental areas. In order to support
highly efficient visualization, techniques for 3D
model simplification and texture processing play
an important role alongside tiling, level of detail,
and vertex buffer objects.
Our example in the Android Market (http://
market.android.com/details?id=com.FGI.Tapi-
ola3D) shows the current state of the art in creating
visually good appearance textures, high-quality
geometry, and 3D visualization in a single demo
for near-future personal geospatial infrastructures.
1040
REFERENCES
Ackermann, F. (1999). Airborne laser scanning
-present status and future expectations. ISPRS
Journal of Photogrammetry and Remote Sensing,
54, 64–67. doi:10.1016/S0924-2716(99)00009-X
Ahmadi, S., Zoej, M. J. V., Ebadi, H., Moghad-
dam, H. A., & Mohammadzadeh, A. (2010). Au-
tomatic urban building boundary extraction from
high resolution aerial images using an innovative
model of active contours. International Journal of
Applied Earth Observation and Geoinformation,
12(3), 150–157. doi:10.1016/j.jag.2010.02.001
Amara, Y., & Marsault, X. (2009). A GPU tile-
load-map architecture for terrain rendering:
Theory and applications. The Visual Computer,
25(8), 805–824. doi:10.1007/s00371-008-0305-1
Arayici, Y. (2007). An approach for real world
data modeling with the 3D terrestrial laser scan-
ner for built environment. Journal of Automation
in Construction, 16(6), 816–829. doi:10.1016/j.
autcon.2007.02.008
Awrangjeb, M., Ravanbakhsh, M., & Fraser, C.
S. (2010). Automatic detection of residential
buildings using LIDAR data and multispectral
imagery. ISPRS Journal of Photogrammetry and
Remote Sensing, 65(5), 457–467. doi:10.1016/j.
isprsjprs.2010.06.001
Axelsson, P. (2000). DEM generation from laser
scanner data using TIN models. In D. Fritsch, &
M. Molenaar (Eds.), The International Archives
of the Photogrammetry and Remote Sensing, Vol.
XXXIII, Part B4/1 (pp.110-117).
Baillard, C., Schmid, C., Zisserman, A., & Fitzgib-
bon, A. (1999). Automatic line matching and 3D
reconstruction of buildings from multiple views.
The International Archives of the Photogramme-
try and Remote Sensing, Vol. XXXII, Part 3-2W5
(pp. 69-80).
3D City Modeling and Visualization for Smart Phone Applications
Baillard, C., & Zisserman, A. (1999). Automatic
reconstruction of piecewise planar models from
multiple views. Proceedings of the IEEE Confer-
ence on Computer Vision and Pattern Recognition,
(pp. 559-565).
Baltsavias, E., Zhang, L., & Eisenbeiss, H. (2005,
May). DSM generation and interior orientation
determination of IKONOS images using a testfield
in Switzerland. Paper presented at the ISPRS Han-
nover Workshop 2005 on “High-Resolution Earth
Imaging for Geospatial Information,” Germany.
Baltsavias, E. P. (1999). A comparison between
photogrammetry and laser scanning. ISPRS
Journal of Photogrammetry and Remote Sensing,
54, 83–94. doi:10.1016/S0924-2716(99)00014-3
Baltsavias, E. P. (2004). Object extraction and
revision by image analysis using existing geo-
data and knowledge: current status and steps
towards operational systems. ISPRS Journal of
Photogrammetry and Remote Sensing, 58(3-4),
129–151. doi:10.1016/j.isprsjprs.2003.09.002
Baltsavias, E. P., Pateraki, M., & Zhang, L. (2001,
September). Radiometric and geometric evalua-
tion of IKONOS geo images and their use for 3D
building modeling. Paper presented at the Joint
ISPRS Workshop on “High Resolution Mapping
from Space 2001”. Retrieved from http://e-
collection.library.ethz.ch/eserv/eth:25224/eth-
25224-01.pdf
Becker, S., & Haala, N. (2007). Combined feature
extraction for facade reconstruction. Proceedings
of the ISPRS Workshop Laser Scanning 2007 and
SilviLaser 2007, (pp. 241-247).
Bescoby, D. J. (2006). Detecting Roman land
boundaries in aerial photographs using Radon
transforms. Journal of Archaeological Science,
33(5), 735–743. doi:10.1016/j.jas.2005.10.012
Bethmann, F., Herd, B., Luhmann, T., & Ohm,
J. (2009). Free- Form surface measurement with
image sequences under consideration of disturbing
objects. Optical 3D. Measurement Techniques,
2, 51–61.
Blair, J. B., Rabine, D. L., & Hofton, M. A.
(1999). The laser vegetation imaging sensor:
A medium-altitude, digitization only, airborne
laser altimeter for mapping vegetation and to-
pography. ISPRS Journal of Photogrammetry
and Remote Sensing, 54, 115–122. doi:10.1016/
S0924-2716(99)00002-7
Blow, J. (2000). Terrain rendering at high levels
of detail. In J. Blow (Ed.), Game Developer’s
Conference Proceedings. San Jose, CA, CMP
Media LLC. Retrieved from http://number-none.
com/blow/papers/terrain_rendering.pdf
BlueFAQs. (2011). 30 Free 3D modeling applica-
tions. Retrieved June 8, 2011, from http://bluefaqs.
com/2009/05/30-free-3d-modeling-applications/
Bodgan, J., & Coors, V. (2009). Using 3D urban
models for pedestrian navigation support. Pro-
ceedings of the ISPRS working group III/4, IV/8,
IV/5: GeoWeb 2009 Academic Track – Cityscapes,
(pp. 9-15).
Bösch, J., Goswamiz, P., & Pajarolax, R. (2009).
RASTeR: Simple and efficient terrain rendering
on the GPU. In Ebert, D., & Krüger, J. (Eds.),
Eurographics 2009 (pp. 35–42).
Braun, C., Kolbe, T. H., Lang, F., Schickler,
W., Steinhage, V., & Cremers, A. B. (1995).
Models for photogrammetric building recon-
struction. Computer Graphics, 19(1), 109–118.
doi:10.1016/0097-8493(94)00126-J
Brenner, C. (2000a). Dreidimensionale Gebäuder-
ekunstruktion aus digitalen Oberflächenmodellen
und Grundrissen. Doctoral thesis, Universität
Stuttgart, Stuttgart, Germany. Retrieved June
10, 2011, from http://elib.uni-stuttgart.de/opus/
volltexte/2001/806/
1041
 3D City Modeling and Visualization for Smart Phone Applications
Brenner, C. (2000b). Towards fully automatic
generation of city models. In I. Dowman, & L.
Janssen (Eds.), The International Archives of
the Photogrammetry and Remote Sensing: Vol.
XXXIII, Part B3 (pp. 85-92).
Brenner, C. (2005). Building reconstruction
from images and laser scanning. International
Journal of Applied Earth Observation and Geo-
information, 6(3-4), 187–198. doi:10.1016/j.
jag.2004.10.006
Brenner, C. (2009). Extraction of features from
mobile laser scanning data for future driver as-
sistance systems. In Sester, Z., Bernard, L., &
Paelke, V. (Eds.), Advances in GIScience, Lecture
Notes in Geoinformation and Cartography (pp.
25–42). Springer Verlag. doi:10.1007/978-3-642-
00318-9_2
Brenner, C., & Haala, N. (1998). Rapid acquisition
of virtual reality city models from multiple data
sources. In Chikatsu, H., & Shimizu, E. (Eds.),
The International Archives of the Photogrammetry
and Remote Sensing (Vol. XXXII, pp. 323–330).
Buckley, S. J., Howell, J. A., Enge, H. D., & Kurz,
T. (2008). Terrestrial laser scanning in geology:
data acquisition, processing and accuracy con-
siderations. Journal of the Geological Society,
165(3), 625–638. doi:10.1144/0016-76492007-
100
Chen, R., Kuusniemi, H., Hyyppä, J., Zhang, J.,
Takala, J., & Kuittinen, R. (2010). Going 3D:
Personal nav and LBS. GPSWorld, 21(2), 14–18.
Collins, R. T., Jaynes, C. O., Cheng, Y. Q., Wang,
X., Stolle, F., Riseman, E. M., & Hanson, A. R.
(1998). The ascender system: Automated site
modeling from multiple aerial images. Computer
Vision and Image Understanding, 72(2), 143–162.
doi:10.1006/cviu.1998.0729
Coors, V. (2008). Synthetic textures for 3d urban
models in pedestrian navigation. Proceedings of
TAMoCo 2008: Techniques and Applications for
Mobile Commerce, (pp. 46–53).
1042
Coors, V., Kray, C., Laakso, K., & Elting, C.
(2005). Presenting route instructions on mobile
devices. In Dykes, J., MacEachren, A. M., &
Kraak, M.-J. (Eds.), Exploring geo-visualization
(pp. 529–550). Elsevier Ltd.
Cord, M., Jordan, M., & Cocquerez, J.-P. (2001).
Accurate building structure recovery from high
resolution aerial images. Computer Vision and Im-
age Understanding, 82(2), 138–173. doi:10.1006/
cviu.2001.0905
Cornelis, N., Leibe, B., Cornelis, K., & Van Gool,
L. (2008). 3D urban scene modeling integrating
recognition and reconstruction. International
Journal of Computer Vision, 78(2-3), 121–141.
doi:10.1007/s11263-007-0081-9
Curless, B., & Levoy, M. (1996, August). A volu-
metric method for building complex models from
range images. Paper presented at the 23rd Annual
Conference on Computer Graphics and Interac-
tive Techniques. Retrieved from http://graphics.
stanford.edu/papers/volrange/volrange.pdf
D’Apuzzo, N. (2003). Surface measurement and
tracking of human body parts from multi station
video sequences. PhD Thesis, Institute of Geodesy
and Photogrammetry, ETH Zurich, Switzerland.
Retrieved from http://webfarm102b.ethz.ch/
berichte/Blaue_Berichte_PDF/81.pdf
Dorninger, P., & Pfeifer, N. (2008). A compre-
hensive automated 3D approach for building
extraction, reconstruction, and regularization
from airborne laser scanning point clouds. Sen-
sors (Basel, Switzerland), 8(11), 7323–7343.
doi:10.3390/s8117323
Fischer, A., Kolbe, T. H., Lang, F., Cremers, A.
B., Förstner, W., Pluemer, L., & Steinhage, V.
(1998). Extracting buildings from aerial images
using hierarchical aggregation in 2D and 3D.
Computer Vision and Image Understanding, 72(2),
185–203. doi:10.1006/cviu.1998.0721
3D City Modeling and Visualization for Smart Phone Applications
Fischler, M. A., & Bolles, R. C. (1981). Random
sample consensus: a paradigm for model fitting
with applications to image analysis and automatec
cartography. Communications of the ACM, 24(6),
381–395. doi:10.1145/358669.358692
Fitzgibbon, A., & Zisserman, A. (1998). Automatic
3D model acquisition and generation of new im-
ages from video sequence. Proceedings of Euro-
pean Signal Processing Conference (EUSIPCO
‘98), (pp. 1261-1269).
Forlani, G., Nardinocchi, C., Scaioni, M., & Zinga-
retti, P. (2006). Complete classification of raw
LIDAR data and 3D reconstruction of buildings.
Pattern Analysis & Applications, 8(4), 357–374.
doi:10.1007/s10044-005-0018-2
Forlani, G., Roncella, R., & Remondino, F.
(2005). Structure and motion reconstruction of
short mobile mapping image sequences. Optical
3-D Measurement Techniques VII: Vol. 1, (pp.
265-274).
Fowler, R., Samberg, A., Flood, M., & Greaves,
T. (2006). Topographic and terrestrial lidar. In
Maune, D. (Ed.), Digital elevation model tech-
nologies & applications (2nd ed., pp. 199–252).
Bethesda, MD: American Society for Photogram-
metry and Remote Sensing.
Fradkin, M., Maitre, H., & Roux, M. (2001).
Building detection from multiple aerial images in
dense urban areas. Computer Vision and Image
Understanding, 82(3), 181–207. doi:10.1006/
cviu.2001.0917
Fraser, C. S. (1996). Network design. In Atkin-
son, K. B. (Ed.), Close-range photogrammetry
and machine vision (pp. 256–282). UK: Whittles
Publishing.
Fraser, C. S., Baltsavias, E., & Gruen, A. (2002).
Processing of IKONOS imagery for submetre
3D positioning and building extraction. Inter-
national Journal of Photogrammetry and Remote
Sensing, 56(3), 177–194. doi:10.1016/S0924-
2716(02)00045-X
Fraser, C. S., & Cronk, S. (2009). A hybrid mea-
surement approach for close-range photogram-
metry. ISPRS Journal of Photogrammetry and
Remote Sensing, 64(3), 328–333. doi:10.1016/j.
isprsjprs.2008.09.009
Früh, C., Jain, S., & Zakhor, A. (2005). Data
processing algorithms for generating textured
3D building facade meshes from laser scans
and camera images. International Journal of
Computer Vision, 61(2), 159–184. doi:10.1023/
B:VISI.0000043756.03810.dd
Früh, C., & Zakhor, A. (2004). An automated
method for large-scale, ground-based city
model acquisition. International Journal of
Computer Vision, 60(1), 5–24. doi:10.1023/
B:VISI.0000027787.82851.b6
Fuchs, C. (1998). Extraktion polymorpher Bild-
strukturen und ihre topologische und geometrische
Gruppierung. PhD thesis, Institut für Photogram-
metrie, Rheinische Friedrich-Wilhelms-Univer-
sität zu Bonn, Germany.
Garland, M. (1999). Multiresolution modeling:
Survey and future opportunities. Proccedings of
Eurographics, 99, 111–131.
Garvin, J. B. (1996). Geological Society of Lon-
don Special Publication: Vol. 110. Topographic
characterization and monitoring of volcanoes via
aircraft laser altimetry (pp. 137–153).
Geospatial World. (2011). 3D laser scanning
market to double by 2015. Retrieved September
3, 2011, from http://geospatialworld.net/index.
php?option=com_content&view=article&id=2
2771%3A3d-laser-scanning-market-to-double-
by-2015&catid=72%3Abusiness-market-survey-
research&Itemid=1
1043
 3D City Modeling and Visualization for Smart Phone Applications
Goetz, M., & Zipf, A. (2010). Open issues in
bringing 3D to location based services (LBS) - A
review focusing on 3D data streaming and 3D
indoor navigation. In T. H. Kolbe, G. König, &
C. Nagel (Eds.), The International Archives of
the Photogrammetry, Remote Sensing and Spatial
Information Sciences, Vol. XXXVIII-4/W15 (pp.
121-124).
Grün, A., Remondino, F., & Zhang, L. (2004).
3D modeling and visualization of large cultural
heritage sites at very high resolution: The Bamiyan
valley and its standing Buddhas. International
Archives of Photogrammetry and Remote Sensing,
XXXV(Part B5), 603–608.
Grün, A., & Wang, X. (1998a). CC-Modeler:
A topology generator for 3-D city models. IS-
PRS Journal of Photogrammetry and Remote
Sensing, 53(5), 286–295. doi:10.1016/S0924-
2716(98)00011-2
Grün, A., & Wang, X. (1998b). Acquisition and
management of urban 3D data using CyberCity
modeler. Retrieved June 10, 2011, from www.
cybercity.tv/pub/ccm_3D_data.pdf
Grün, A., & Wang, X. (1999a, September). Cyber-
City modeller, a tool for interactive 3-D city model
generation. Paper presented at the Photogram-
metric Week 99, Wichmann Veriag, Heideiberg.
Grün, A., & Wang, X. (1999b, November). Cyber-
City spatial information system (CC-SIS): A new
concept for the management of 3-D city models in
a hybrid GIS. Paper presented in The 20th Asian
Conference on Remote Sensing. Retrieved from
http://www.geod.ethz.ch/p02/general/persons/
AG_pub/cc-sis_acrs.pdf
Grün, A., Zhang, L., & Eisenbeiss, H. (2005,
March). 3D Precision processing of high resolution
satellite imagery. Paper presented at the ASPRS
2005 Annual Conference. Retrieved from ftp://
ftp.ecn.purdue.edu/jshan/proceedings/asprs2005/
Files/0048.pdf
1044
Gülch, E., & Müller, H. (2001, June). Semi-
automatic building acquisition. Paper presented
at the Third International Workshop on Automatic
Extraction of Man-Made Objects from Aerial and
Space Images. Retrieved from http://www.ipb.
uni-bonn.de/projects/semi-automatic.html
Haala, N. (1996). Gebäuderekonstruktion durch
Kombination von Bild- und Höhendaten. Ph.D
thesis, Universität Stuttgart, Stuttgart, Germany.
Retrieved from http://elib.uni-stuttgart.de/opus/
volltexte/2006/2714/pdf/Haala_diss.pdf
Haala, N., & Kada, M. (2010). An update on auto-
matic 3D building reconstruction. ISPRS Journal
of Photogrammetry and Remote Sensing, 65(6),
570–580. doi:10.1016/j.isprsjprs.2010.09.006
Haala, N., Peter, M., Kremer, J., & Hunter, G.
(2008). Mobile lidar mapping for 3D point cloud
collection in urban areas – A performance test.
The International Archives of the Photogram-
metry and Remote Sensing, Vol. XXXVII, Part
B5, V, 1119–1124.
Hagedorn, B., Hildebrandt, D., & Dollner, J.
(2010). Towards advanced and interactive web
perspective view services. In Neutens, T., & De
Mauer, P. (Eds.), Developments in 3D Geo-Infor-
mation Sciences, Lecture Notes on Geoinformation
and Cartography (pp. 33–51). Springer Verlag.
doi:10.1007/978-3-642-04791-6_3
Henricsson, O. (1996). Analysis of image
structures using color attributes and similarity
relations. Ph.D thesis, Institut für Geodäsie und
Photogrammetrie, ETH Zürich, Zürich. Retrieved
from http://e-collection.library.ethz.ch/eserv/
eth:40304/eth-40304-01.pdf
Henricsson, O., & Baltsavias, E. (1997). 3-D build-
ing reconstruction with ARUBA: A qualitative and
quantitative evaluation. In Grün, A., Baltsavias, E.,
& Henricsson, O. (Eds.), Automatic extraction of
man-made objects from aerial and space images
(II) (pp. 65–76). Basel, Switzerland: Birkhäuser.
doi:10.1007/978-3-0348-8906-3_7
3D City Modeling and Visualization for Smart Phone Applications
Hildebrandt, D., Hagedorn, B., & Dollner, J.
(2010). Image-based, interactive visualization
of complex 3D geovirtual environments on light-
weight devices. Proceedings of the 7th Interna-
tional Symposium on LBS and TeleCartography,
(pp. 52-71).
Hofmann, A., Maas, H.-G., & Streilein, A. (2003).
Derivation of roof types by cluster analysis in
parameter spaces of airborne laser scanner point
clouds. Proceedings of the ISPRS Working Group
III/3 Workshop, 3-D Reconstruction from Airborne
Lasers Canner and InSAR Data.
Hoppe, H. (1998). Smooth view-dependent level-
of-detail control and its application to terrain ren-
dering. Proceedings of Ninth IEEE Visualization
(VIS ‘98), (pp. 35-42).
Hug, C. (1997). Extracting artificial surface ob-
jects from airborne laser scanner data. In Grün,
A., Baltsavias, E. P., & Henricson, O. (Eds.),
Automatic extraction of man-made objects from
aerial and space images (II) (pp. 203–212). Basel,
Switzerland: Birkhauser Verlag. doi:10.1007/978-
3-0348-8906-3_20
Hyyppä, J., & Engdahl, M. (2000). Verification
of the capability of repeat-pass ERS-1/2 SAR in-
terferometry, to provide digital elevations models
and the impact of tree height and canopy closure
on SAR-derived terrain height in boreal forests.
The Photogrammetric Journal of Finland, 16(2),
16–26.
Hyyppä, J., Hyyppä, H., Leckie, D., Gougeon, F.,
Yu, X., & Maltamo, M. (2008). Review of meth-
ods of small-footprint airborne laser scanning for
extracting forest inventory data in boreal forests.
International Journal of Remote Sensing, 29(5),
1339–1366. doi:10.1080/01431160701736489
Hyyppä, J., Hyyppä, H., Yu, X., Kaartinen, H.,
Kukko, A., & Holopainen, M. (2009). Forest
inventory using small-footprint airborne lidar.
In Shan, J., & Toth, C. (Eds.), Topographic laser
ranging and scanning: Principles and processing
(pp. 335–370). London, UK: CRC Press, Taylor
& Fracis.
Hyyppä, J., Matikainen, L., & Kaartinen, H. Yu.
X., Hyyppä, H., & Ronnholm, P. (2007). Improv-
ing automation in map updating based on national
laser scanning, classification trees, object-based
change detection and 3D object reconstruction.
Proceedings 2007 Urban Remote Sensing Confer-
ence, (pp. 1-10).
Hyyppä, J., Wagner, W., Hollaus, M., & Hyyppä,
H. (2009). Airborne laser scanning. In Warner,
T. A., Nellis, M. D., & Foody, G. M. (Eds.), The
SAGE handbook of remote sensing (pp. 199–212).
London, UK: SAGE Publications.
Jaakkola, A., Hyyppä, J., Hyyppä, H., & Kukko,
A. (2008). Retrieval algorithms for road surface
modeling using laser-based mobile mapping.
Sensors (Basel, Switzerland), 8(9), 5238–5249.
doi:10.3390/s8095238
Jaakkola, A., Hyyppä, J., Kukko, A., Yu, X.,
Kaartinen, H., Lehtomäki, M., & Lin, Y. (2010).
A low-cost multi-sensoral mobile mapping
system and its feasibility for tree measure-
ments. ISPRS Journal of Photogrammetry and
Remote Sensing, 65(6), 514–522. doi:10.1016/j.
isprsjprs.2010.08.002
Jacobsen, K. (2004). DTM generation by SPOT5
HRS. International Archives of Photogrammetry
and Remote Sensing, XXXV(Part B1), 439–444.
Jaynes, C., Riseman, E., & Hanson, A. (2003).
Recognition and reconstruction of buildings from
multiple aerial images. Computer Vision and Im-
age Understanding, 90(1), 68–98. doi:10.1016/
S1077-3142(03)00027-4
1045
 3D City Modeling and Visualization for Smart Phone Applications
Jung, F. (2004). Detecting building changes from
multitemporal aerial stereopairs. ISPRS Journal
of Photogrammetry and Remote Sensing, 58(3-
4), 187–201. doi:10.1016/j.isprsjprs.2003.09.005
Kaartinen, H., & Hyyppä, J. (2006). EuroSDR-
project commission 3 “evaluation of building
extraction”, final report. (EuroSDR, Official
Publication No. 50). MAA: European Spatial
Data Research.
Kada, M., & McKinley, L. (2009). 3D build-
ing reconstruction from LiDAR based on a cell
decomposition approach. In The International
Archives of the Photogrammetry, Remote Sensing
and Spatial Information Sciences, Vol. XXXVIII,
Part 3/W4 (pp. 47-52).
Kager, H. (2004). Discrepancies between overlap-
ping laser scanning strips- Simultaneous fitting
of aerial laser scanner strips. The International
Archives of the Photogrammetry, Remote Sensing
and Spatial Information Science, Vol. XXXV, Part
B/1 (pp. 555 – 560).
Kim, T., Park, S.-R., Kim, M.-G., Jung, S., &
Kim, K.-O. (2004). Tracking road centerlines
from high resolution remote sensing images by
least squares correlation matching. Photogram-
metric Engineering and Remote Sensing, 70(12),
1417–1422.
Koller, D., Lindstrom, P., Ribarsky, W., Hodges,
L., Faust, N., & Turner, G. (1994).Virtual GIS:
A real-time 3D geographic information system.
Proceedings of ARPA Image Understanding
Workshop 94, i(pp. 94-100).
Kraus, K., & Otepka, J. (2005, September). DTM
modeling and visualization-The SCOP approach.
Paper presented at the Photogrammetric Week 05,
Stuttgart, Germany. Retrieved from http://www.
ipf.tuwien.ac.at/publications/kk_jo_photogram-
metric_week_2005.pdf
1046
Kraus, K., & Pfeifer, N. (1998). Determination of
terrain models in wooded areas with airborne laser
scanner data. ISPRS Journal of Photogrammetry
and Remote Sensing, 53, 193–203. doi:10.1016/
S0924-2716(98)00009-4
Kukko, A. (2009). Road environment mapper—3D
data capturing with mobile mapping. Licentiate’s
Thesis, Helsinki University of Technology, Espoo
Finland.
Lang, F. (1999). Geometrische und semantische
Rekonstruktion von Geb¨auden durch Ableitung
von 3D-Geb¨audeecken. Ph.D thesis, Institut
für Photogrammetrie, Rheinische Friedrich-
Wilhelms-Universität zu Bonn, Bonn, Germany.
Retrieved from http://elib.uni-stuttgart.de/opus/
volltexte/2006/2714/pdf/Haala_diss.pdf
Le Besnerais, G., Sanfourche, G. M., & Champag-
nat, F. (2008). Dense height map estimation from
oblique aerial image sequences. Computer Vision
and Image Understanding, 109(2), 204–225.
doi:10.1016/j.cviu.2007.07.003
Lee, D. S., Shan, J., & Bethel, J. S. (2002). Class-
guided building extraction from IKONOS imag-
ery. Photogrammetric Engineering and Remote
Sensing, 69(2), 143–150.
Lindstrom, P., & Pascucci, V. (2001). Visualiza-
tion of large terrains made easy. Proceedings of
IEEE Visualization, 01, 363–370.
Liu, Z., Zhang, J., & Yan, Q. (2010, September).
Implementation of a 3D personal navigation
visualization engine and its demonstration in
Shanghai World Exposition 2010. Paper presented
at the 7th International Symposium on LBS and
TeleCartography, Guangzhou, China.
Livny, Y., Kogan, Z., & El-Sana, J. (2009). Seam-
less patches for GPU-based terrain rendering. The
Visual Computer, 25(3), 197–208. doi:10.1007/
s00371-008-0214-3
3D City Modeling and Visualization for Smart Phone Applications
Livny, Y., Sokolovsky, N., Grinshpoun, T., & El-
Sana, J. (2008). A GPU persistent grid mapping
for terrain rendering. The Visual Computer, 24(2),
139–153. doi:10.1007/s00371-007-0180-1
Lodi, A., Martello, S., & Monaci, M. (2002). Two-
dimensional packing problems: A survey. Euro-
pean Journal of Operational Research, 141(2),
241–252. doi:10.1016/S0377-2217(02)00123-6
Luhmann, T. (2010). Close range photogram-
metry for industrial applications. ISPRS Journal
of Photogrammetry and Remote Sensing, 64(3),
558–569. doi:10.1016/j.isprsjprs.2010.06.003
Ma, R. (2004). Building model reconstruction
from LiDAR data and aerial photographs. Doc-
toral thesis, The Ohio State University, USA.
Retrieved from http://www.asprs.org/a/publica-
tions/proceedings/reno2006/0088.pdf
Matikainen, L., Hyyppä, J., & Hyyppä, H. (2003).
Automatic detection of buildings from laser scan-
ner data for map updating. In H.-G. Maas, G. Vos-
selman, & A. Streilein (Eds.), The International
Archives of Photogrammetry, Remote Sensing and
Spatial Information Sciences, Vol. XXXIV, Part 3/
W13 (pp. 218-224).
Mayer, H. (1999). Automatic object extraction
from aerial imagery - A survey focusing on build-
ings. Computer Vision and Image Understanding,
74(2), 138–149. doi:10.1006/cviu.1999.0750
McGlone, J. C., Mikhail, E. M., Bethel, J., & Mul-
len, R. (Eds.). (2004). Manual of photogrammetry
(5th ed.). USA: American Society of Photogram-
metry and Remote Sensing.
Melzer, T. (2007). Non-parametric segmentation
of ALS point clouds using mean shift. Journal
of Applied Geodesy, 1(3), 159–170. doi:10.1515/
jag.2007.018
Mikhail, E. M., Bethel, J., & McGlone, J. C.
(2001). Introduction to modern photogrammetry.
New York, NY: John Wiley & Sons, Inc.
NAVTEQ. (2010). NAVTEQ website. Retrieved
December 11, 2010, from http://corporate.navteq.
com/webapps/NewsUserServlet?action=NewsDe
tail&newsId=946&lang=en&englishonly=false
Paparoditis, N., Cord, M., Jordan, M., & Coc-
querez, J.-P. (1998). Building detection and
reconstruction from mid- and high-resolution
aerial imagery. Computer Vision and Image
Understanding, 72(2), 122–142. doi:10.1006/
cviu.1998.0722
Pei, L., Chen, R., Liu, J., Tenhunen, T., Kuus-
niemi, H., & Chen, Y. (2010). An inquiry-based
Bluetooth indoor positioning approach for the
Finnish Pavilion at Shanghai World Expo 2010.
Proceedings of IEEE/ION PLANS 2010, (pp.
1002-1009).
Peng, J., & Liu, Y. (2005). Model and context-
driven building extraction in dense urban aerial
images. International Journal of Remote Sensing,
26(7), 1289–1307. doi:10.1080/0143116051233
1326675
Poli, D. (2005). Modeling of spaceborne linear ar-
ray sensors. Doctoral thesis, Institute of Geodesy
and Photogrammetry, ETH Zurich, Switzerland.
Retrieved from http://webfarm102b.ethz.ch/
berichte/Blaue_Berichte_PDF/85.pdf
Pollefeys, M., Nister, D., Frahm, J.-M., Akbar-
zadeh, A., Mordohai, P., & Clipp, B. (2008).
Detailed real-time urban 3D reconstruction from
video. International Journal of Computer Vision,
78, 143–167. doi:10.1007/s11263-007-0086-4
Pu, S. (2008). Automatic building modeling from
terrestrial laser scanning. In Oosterom, P., Zlatano-
va, S., Penninga, F., & Fendel, E. (Eds.), Advances
in 3D geoinformation systems (pp. 147–160).
Berlin, Germany: Springer. doi:10.1007/978-3-
540-72135-2_9
1047
 3D City Modeling and Visualization for Smart Phone Applications
Pu, S., & Vosselman, G. (2009). Knowledge
based reconstruction of building models from
terrestrial laser scanning data. ISPRS Journal
of Photogrammetry and Remote Sensing, 64(6),
575–584. doi:10.1016/j.isprsjprs.2009.04.001
Rabin, R. M., & Martin, D. W. (1996). Satel-
lite observations of shallow cumulus coverage
over the central United States: An exploration
of land use impact on cloud cover. Journal of
Geophysical Research, 101(D3), 7149–7155.
doi:10.1029/95JD02891
Remondino, F. (2002). Human body reconstruction
from image sequences. In Van Gool, L. (Ed.), Pat-
tern Recognition (Vol. 2449, pp. 50–57). Lecture
Notes in Computer Science Zurich, Switzerland:
Springer Verlag. doi:10.1007/3-540-45783-6_7
Remondino, F. (2004). 3-D reconstruction of static
human body shape from image sequence. Journal
of Computer Vision and Image Understanding,
93(1), 65–85. doi:10.1016/j.cviu.2003.08.006
Rottensteiner, F., & Briese, Ch. (2003). Automatic
generation of building models from LIDAR data
and the integration of aerial images. In: H.-G.
Maas, G. Vosselman, & A. Streilein (Eds.), Pro-
ceedings of the ISPRS WG III/3 Workshop on ´3-D
Reconstruction from Airborne Laserscanner and
InSAR data´, vol. XXXIV, Part 3/W13 (pp. 174-
180). Dresden: IAPRS.
Rottensteiner, F., Trinder, J., Clode, S., & Ku-
bik, K. (2005a). Automated delineation of roof
planes from LiDAR data. In G. Vosselman, &
C. Brenner (Eds.), The International Archives of
the Photogrammetry, Remote Sensing and Spatial
Information Sciences, Vol. XXXVI, Part 3/W4
(pp. 221-226).
Rottensteiner, F., Trinder, J., Clode, S., & Kubik,
K. (2005b). Using the Dempster–Shafer method
for the fusion of LIDAR data and multispectral
images for building detection. Information Fusion,
6(4), 283–300. doi:10.1016/j.inffus.2004.06.004
1048
Sampath, A., & Shan, J. (2010). Segmentation
and reconstruction of polyhedral building roofs
from aerial LiDAR point clouds. IEEE Transac-
tions on Geoscience and Remote Sensing, 48(3),
1554–1567. doi:10.1109/TGRS.2009.2030180
Schneider, J., & Westermann, R. (2006). GPU-
friendly high-quality terrain rendering. Journal
of WSCG 2006, 14(1213-6972), 49-56.
Sharkawi, H., Ujang, M. U., & Abdul-Rahman, A.
(2007, November). Indoor and outdoor navigation
based on 3D game engine. Paper presented at the
Joint International Symposium and Exhibition
on Geoinformation and International Sympo-
sium and Exhibition on GPS/GNSS 2007, Johor
Bahru, Johor.
Simple 3D. (2011). 3D Scanners, digitizers, and
software for making 3D models and 3D measure-
ments. Retrieved June 8, 2011, from http://www.
simple3d.com/#software
Sithole, G., & Vosselman, G. (2004). Experimen-
tal comparison of filter algorithms for bare Earth
extraction from airborne laser scanning point
clouds. ISPRS Journal of Photogrammetry and
Remote Sensing, 59(1-2), 85–101. doi:10.1016/j.
isprsjprs.2004.05.004
Skaloud, J. (1999). Optimizing georeferencing of
airborne survey systems by INS/DGPS. Doctoral
thesis, University of Calgary, Canada. Retrieved
from https://dspace.ucalgary.ca/bitstream/1880/
25307/2/38506Skaloud.pdf
Sohn, G., & Dowman, I. (2007). Data fusion of
high-resolution satellite imagery and LiDAR data
for automatic building extraction. ISPRS Journal
of Photogrammetry and Remote Sensing, 62(1),
43–63. doi:10.1016/j.isprsjprs.2007.01.001
Sohn, G., Huang, X., & Tao, V. (2008). Using a
binary space partitioning tree for reconstructing
polyhedral building models from airborne LiDAR
data. Photogrammetric Engineering and Remote
Sensing, 74(11), 1425–1438.
3D City Modeling and Visualization for Smart Phone Applications
Sörgel, U., Michaelsen, E., Thiele, A., Cadario,
E., & Thoennessen, U. (2009). Stereo analysis of
high-resolution SAR images for building height
estimation in cases of orthogonal aspect direc-
tions. ISPRS Journal of Photogrammetry and
Remote Sensing, 64(5), 490–500. doi:10.1016/j.
isprsjprs.2008.10.007
Sörgel, U., Schulz, K., Thoennessen, U., & Stilla,
U. (2005). Integration of 3D data in SAR mission
planning and image interpretation in urban areas.
An International Journal on Information Fusion,
6(4), 301–310. doi:10.1016/j.inffus.2004.06.007
Sörgel, U., Thoennessen, U., Brenner, A., &
Stilla, U. (2006). High-resolution SAR data. New
opportunities and challenges for the analysis of
urban areas. IEE Proceedings. Radar, Sonar
and Navigation, 153(3), 294–300. doi:10.1049/
ip-rsn:20045088
Stamos, I., & Allen, P. K. (2002). Geometry and
texture recovery of scenes of large scale. Computer
Vision and Image Understanding, 88(2), 94–118.
doi:10.1006/cviu.2002.0963
Stulp, F., & Fisher, R. (2001). Reconstruction of
surfaces behind occlusions in range images. Pro-
ceedings of the 3rd International Conference on
3-D Digital Images and Modeling, (pp. 232-239).
Süveg, I., & Vosselman, G. (2004). Reconstruction
of 3D building models from aerial images and
maps. ISPRS Journal of Photogrammetry and
Remote Sensing, 58(3-4), 202–224. doi:10.1016/j.
isprsjprs.2003.09.006
Technology Review Website. (2011). Technology
review of C3 technologies. Retrieved November
20, 2011, from http://www.technologyreview.
com/computing/37466/
TerraSolid. (2003). TerraScan user’s guide.
Tian, Y., Gerke, M., Vosselman, G., & Zhu, Q.
(2010). Knowledge-based building reconstruction
from terrestrial video sequences. ISPRS Journal
of Photogrammetry and Remote Sensing, 65(4),
395–408. doi:10.1016/j.isprsjprs.2010.05.001
TLS software survey. (2011). ISPRS website.
Retrieved June 8, 2011 from http://www.cage.
curtin.edu.au/~gordonsj/isprs_wgv3/data/ispr-
swgv3_tls_software_survey.pdf
Tournaire, O., Bredif, M., Boldo, D., & Durupt,
M. (2010). An efficient stochastic approach for
building footprint extraction from digital elevation
models. ISPRS Journal of Photogrammetry and
Remote Sensing, 65(4), 317–327. doi:10.1016/j.
isprsjprs.2010.02.002
Toutin, T. (2004). Comparison of stereo-extracted
DTM from different high-resolution sensors:
SPOT-5, EROS-A, IKONOS-II, and QuickBird.
IEEE Transactions on Geoscience and Remote
Sensing, 42(10), 2121–2129. doi:10.1109/
TGRS.2004.834641
Ulrich, T. (2000, July). Rendering massive terrains
using chunked level of detail. Paper presented
at the SIGGRAPH 2000 Conference Abstracts
and Applications, Scaling up to Massive Virtual
Worlds. Retrieved from http://tulrich.com/geek-
stuff/sig-notes.pdf
Verma, V., Kumar, R., & Hsu, S. (2006). 3D build-
ing detection and modeling from aerial LiDAR
data. Proceedings of the 2006 IEEE Computer
Society Conference on Computer Vision and
Pattern Recognition (CVPR’06, (pp. 2213-2220).
Vögtle, T., & Steinle, E. (2003). On the quality
of object classification and automated building
modeling based on laserscanning data. In H.-G.
Maas, G. Vosselman, & A. Streilein (Eds.), The
International Archives of Photogrammetry, Re-
mote Sensing and Spatial Information Sciences,
Vol. XXXIV, Part 3/W13 (7p.).
1049
 3D City Modeling and Visualization for Smart Phone Applications
Vosselman, G. (1999). Building reconstruction
using planar faces in very high density height
data. In The International Archives of the Photo-
grammetry and Remote Sensing, Vol. XXXII, Part
3-2W5 (pp. 87-92).
Vosselman, G. (2002). Fusion of laser scanning
data, maps, and aerial photographs for building
reconstruction. In IEEE International Geoscience
and Remote Sensing Symposium (IGARSS’02),
Vol. 1 (pp.85-88).
Vosselman, G., & Dijkman, S. (2001). 3D build-
ing model reconstruction from point clouds and
ground plans. In The International Archives of
the Photogrammetry and Remote Sensing, Vol.
XXXIV, Part 3W4 (pp. 37-43).
Vosselman, G., Gorte, B. G. H., Sithole, G., &
Rabbani, T. (2004). Recognising structure in la-
ser scanner point clouds. In M. Thies, B. Koch,
H. Spiecker, & H. Weinacker (Eds.), The Inter-
national Archives of Photogrammetry, Remote
Sensing and Spatial Information Sciences: Vol.
XXXVI, Part 8/W2 (pp.33-38).
Vosselman, G., & Maas, H.-G. (Eds.). (2010).
Airborne and terrestrial laser scanning. Scotland:
Whittles Publishing.
Vosselman, G., & Süveg, I. (2001). Map based
building reconstruction from laser data and images.
In Automatic Extraction of Man-Made Objects
from Aerial and Space Images (III) (pp. 231–239).
Ascona, Switzerland: Balkema Publishers.
Vu, T., Yamazaki, F., & Matsuoka, M. (2009).
Multi-scale solution for building extraction from
LiDAR and image data. International Journal of
Applied Earth Observation and Geoinformation,
11(4), 281–289. doi:10.1016/j.jag.2009.03.005
1050
Wagner, W., Ullrich, A., Ducic, V., Melzer, T.,
& Studnicka, N. (2006). Gaussian decomposi-
tion and calibration of a novelsmall-footprint
full-waveform digitising airborne laser scan-
ner. ISPRS Journal of Photogrammetry and
Remote Sensing, 60(2), 100–112. doi:10.1016/j.
isprsjprs.2005.12.001
Wehr, A., & Lohr, U. (1999). Airborne laser
scanning – An introduction and overview. IS-
PRS Journal of Photogrammetry and Remote
Sensing, 54(2-3), 68–82. doi:10.1016/S0924-
2716(99)00011-8
Wikipedia. (2011). Level of detail. Retrieved
June 8, 2011, from http://en.wikipedia.org/wiki/
Level_of_detail
Zhang, K., Yan, J., & Chen, S. (2006). Automatic
construction of building footprints from airborne
LIDAR data. IEEE Transactions on Geosci-
ence and Remote Sensing, 44(9), 2523–2533.
doi:10.1109/TGRS.2006.874137
Zhang, L., & Gruen, A. (2006). Multi-image
matching for DSM generation from IKONOS
imagery. ISPRS Journal of Photogrammetry and
Remote Sensing, 60(3), 195–211. doi:10.1016/j.
isprsjprs.2006.01.001
Zhao, H., & Shibasaki, R. (2003). Reconstructing
a textured CAD model of an urban environment
using vehicle-borne laser range scanners and line
cameras. Machine Vision and Applications, 14(1),
35–41. doi:10.1007/s00138-002-0099-5
Zhao, H., & Shibasaki, R. (2005). Updating digital
geographic database using vehicle-borne laser
scanners and line cameras. Photogrammetric
Engineering and Remote Sensing, 71(4), 415–424.
3D City Modeling and Visualization for Smart Phone Applications
Zhu, L., Hyyppä, J., Kukko, A., Jaakkola, A.,
Lehtomäki, M., & Kaartinen, H. … Chen, Y. (2009,
May). 3D city model for mobile phone by using
MMS data. Paper presented at the IEEE 2009
Urban Remote Sensing Joint Events. Retrieved
from http://ieeexplore.ieee.org/xpl/freeabs_all.
jsp?arnumber=5137691
Zhu, L., Hyyppä, J., Kukko, A., Kaartinen, H., &
Chen, R. (2011). Photorealistic 3D city modeling
from mobile laser scanning data. Remote Sensing,
3(7), 1406–1426. doi:10.3390/rs3071406
Zlatanova, S., & Verbree, E. (2003). Technological
developments within 3D location-based services.
Proceedings of the International Symposium
and Exhibition on Geoinformation 2003, (pp.
153-160).
ADDITIONAL READING
Alho, P., Kukko, A., Hyyppä, H., Kaartinen, H.,
& Hyyppä, J. (2009). Application of boat-based
laser scanning for river survey. [Letters to ESEX].
Earth Surface Processes and Landforms, 34(13),
1831–1838. doi:10.1002/esp.1879
Brenner, C. (2005). Building reconstruction
from images and laser scanning. International
Journal of Applied Earth Observation and Geo-
information, 6(3-4), 187–198. doi:10.1016/j.
jag.2004.10.006
Chen, R., Kuusniemi, H., Hyyppä, J., Zhang, J.,
Takala, J., & Kuittinen, R. (2010). Going 3D:
Personal nav and LBS. GPSWorld, 21(2), 14–18.
Gülch, E., Kaartinen, H., & Hyyppä, J. (2009).
Quality of buildings extracted from airborne laser
scanning data: Results of an empirical investiga-
tion on 3D building reconstruction, 535-574. In J.
Shan, & C. Toth (Eds.), Topographic laser ranging
and scanning: Principles and processing. London,
UK: CRC Press, Taylor & Francis.
Hyyppä, J. (2011). State of the art in laser scanning.
In Fritsch, D. (Ed.), Photogrammetric Week’11 (pp.
203–216). Berlin, Germany: Wichmann Verlag.
Hyyppä, J., Hyyppä, H., Yu, X., Kaartinen, H.,
Kukko, A., & Holopainen, M. (2009). Forest
inventory using small-footprint airborne lidar.
In Shan, J., & Toth, C. (Eds.), Topographic laser
ranging and scanning: Principles and processing
(pp. 335–370). London, UK: CRC Press, Taylor
& Francis.
Hyyppä, J., Jaakkola, A., Hyyppä, H., Kaartinen,
H., Kukko, A., Holopainen, M., et al. (2009). Map
updating and change detection using vehicle-based
laser scanning. Proceedings of the 2009 Urban
Remote Sensing Joint Event, (pp. 1-6).
Jaakkola, A., Hyyppä, J., Kukko, A., Yu, X.,
Kaartinen, H., Lehtomäki, M., & Lin, Y. (2010).
A low-cost multi-sensoral mobile mapping
system and its feasibility for tree measure-
ments. ISPRS Journal of Photogrammetry and
Remote Sensing, 65(6), 514–522. doi:10.1016/j.
isprsjprs.2010.08.002
Kukko, A. (2009). Road environment mapper—3D
data capturing with mobile mapping. Licentiate’s
Thesis, Helsinki University of Technology, Espoo
Finland.
Kukko, A., & Hyyppä, J. (2009). Small-footprint
laser scanning simulator for system validation,
error assessment, and algorithm development.
Photogrammetric Engineering and Remote Sens-
ing, 75(10), 1177–1189.
Lehtomäki, M., Jaakkola, A., Hyyppä, J., Kukko,
A., & Kaartinen, H. (2010). Detection of vertical
pole-like objects in a road environment using
vehicle-based laser scanning data. Remote Sens-
ing, 2(3), 641–664. doi:10.3390/rs2030641
Lin, Y., & Hyyppä, J. (2011). K-segments-based
geometric modeling of VLS scan lines. IEEE
Geoscience and Remote Sensing Letters, 8(1),
93–97. doi:10.1109/LGRS.2010.2051940
1051
 3D City Modeling and Visualization for Smart Phone Applications
Lin, Y., Jaakkola, A., Hyyppä, J., & Kaartinen, H.
(2010). From TLS to VLS: Biomass estimation
at individual tree level. Remote Sensing, 2(8),
1864–1879. doi:10.3390/rs2081864
Matikainen, L., Hyyppä, J., Ahokas, E., Markelin,
L., & Kaartinen, H. (2010). Automatic detection
of buildings and changes in buildings for updat-
ing of maps. Remote Sensing, 2(5), 1217–1248.
doi:10.3390/rs2051217
Mikhail, E. M., Bethel, J., & McGlone, J. C.
(2001). Introduction to modern photogrammetry.
New York, NY: John Wiley & Sons, Inc.
Puttonen, E., Jaakkola, A., Litkey, P., & Hyyppä, J.
(2011). Tree classification with fused mobile laser
scanning and hyperspectral data. Sensors (Basel,
Switzerland), 11(5), 5158–5182. doi:10.3390/
s110505158
Rönnholm, P., Hyyppä, H., Hyyppä, J., & Hag-
grén, H. (2009). Orientation of airborne laser
scanning point clouds with multi-view, multi-scale
image blocks. Sensors (Basel, Switzerland), 9(8),
6008–6027. doi:10.3390/s90806008
Vaaja, M., Hyyppä, J., Kukko, A., Kaartinen, H.,
Hyyppä, H., & Alho, P. (2011). Mapping topog-
raphy changes and elevation accuracies using
a mobile laser scanner. Remote Sensing, 3(3),
587–600. doi:10.3390/rs3030587
Vosselman, G., & Maas, H.-G. (Eds.). (2010).
Airborne and terrestrial laser scanning. Scotland:
Whittles Publishing.
Wagner, W., Hyyppä, J., Ullrich, A., Lehner, H.,
Briese, C., & Kaasalainen, S. (2008). Radiometric
calibration of full-waveform small-footprint air-
borne laser scanners. The International Archives of
the Photogrammetry. Remote Sensing and Spatial
Information Sciences, XXXVII(Part B1), 163–168.
This work was previously published in Ubiquitous Positioning and Mobile Location-Based Services in Smart Phones, edited
by Ruizhi Chen, pp. 254-296, copyright 2012 by Information Science Reference (an imprint of IGI Global).
1052
Zhu, L., Hyyppä, J., Kukko, A., Kaartinen, H., &
Chen, R. (2011). Photorealistic 3D city modeling
from mobile laser scanning data. Remote Sensing,
3(7), 1406–1426. doi:10.3390/rs3071406
KEY TERMS AND DEFINITIONS
3D city Models: Scenes are represented by
geometric elements such as points, lines and
surfaces in 3D space, and rendering the textures
onto the surfaces of geometry.
3D Modeling: It is the process to develop the
mathematics methods to represent the objects in
3D space.
Airborne Laser Scanning: a method based
on laser (LiDAR) range measurements from an
aircraft consisting of a laser scanner, GPS/IMU
and a computer for collecting point clouds typi-
cally for topographic surveys.
Laser Scanning: a method based on laser
(LiDAR) range measurements providing a point
cloud of objects scanned.
Mobile Laser Scanning: a method based
on laser (LiDAR) range measurements from a
system which consists of a laser scanner, GPS/
IMU and also a computer, mounted on a moving
platform e.g. a van, a car or a boat for collecting
point clouds from the surrounding environment.
Photogrammetry: It is the process of acquir-
ing information from images by establishing the
geometric relationship between images and scene
and interpreting the images.
Terrestrial Laser Scanning: a method based
on laser (LiDAR) range measurements from a
laser scanner mounted on a tripod for collecting
point clouds from the surrounding environment.
 1053

Chapter 63
Assessing the Success of
an Industrial Cluster
Neil Reid
The University of Toledo, USA

Bruce W. Smith
Bowling Green State University, USA

ABSTRACT
Industrial clusters have received considerable attention as a regional development strategy. While their
efficacy has been debated by academics, clusters have become popular among practitioners. Despite
clusters’ acceptance, there have been few attempts to measure their success or their impact on constituent
firms. This paper outlines and discusses the metrics developed to evaluate the success of the northwest
Ohio greenhouse cluster. The cluster was launched in 2004 to help the industry become more competi-
tive though collaborative problem solving. In identifying success metrics, the authors were cognizant
of the fact that they had to reflect the cluster’s objectives and goals. Thus metrics that measured the
impact of branding and marketing efforts, reducing energy costs, and increasing collaboration among
cluster stakeholders were developed. The work reported in this paper is only the beginning phases of a
longer-term, on-going effort to track the progress and success of the northwest Ohio greenhouse cluster.

INTRODUCTION have widely adopted the cluster strategy. In 2002,

Cluster-based economic development strategies
have generated substantial attention in recent
years. Even though the benefits of clusters have
been questioned in the academic literature (Mar-
tin & Sunley, 2003; Taylor, 2006), practitioners
DOI: 10.4018/978-1-4666-2038-4.ch063
van der Linde (2002) identified over 800 cluster
projects globally. More recently, Sölvell (2008)
reported that there were over one thousand cluster
initiatives in Europe alone. Clusters have also been
embraced in the United States. For example, the
National Governors Association (NGA, 2007)
published Cluster-Based Strategies for Growing
State Economies and subsequently announced

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

that it was launching a policy academy for states
using cluster analysis and innovation-based
economic development strategies. Also, Muro
and Fikri (2011, p. 1) argued that cluster-based
strategies are particularly attractive for governors
concerned with revitalizing state economies since
they “provide a direct route to economic renewal
because they build on existing assets to promote
growth in regions by enhancing the interactions
by which firms complete transactions, share ideas,
start new enterprises, and create jobs.”
Despite the acceptance of clusters by many
academics and policy makers, there have been
few attempts to objectively measure the success of
clusters or the impact of a cluster on the constitu-
ent firms (Feser et al., 2008; Fromhold-Eisebith
& Eisebith, 2005; Schmiedeberg, 2010). Sölvell
(2008) reported that only five cluster organiza-
tions out of 50 he surveyed had completed cluster
evaluations. A few academic researchers have
investigated the impact of clusters on specific
economic conditions. For example, Feser et al.
(2008) analyzed the affect of clusters on job growth
and new business formation in Appalachia. Also
Delgado et al. (2010) investigated the relationship
between clusters and entrepreneurship. As a final
example, Fowler and Kleit (2010) assessed the
impact of cluster activity on areas’ poverty rates.
While such studies contribute to our general un-
derstanding of regional development, they do not
necessarily inform about measuring the success
of a specific cluster project.
Fromhold-Eisebith and Eisebith (2005) sug-
gested that the heterogeneity of clusters leads
to intricate methodological issues that inhibit
evaluation. Schmiedeberg (2010) and Sölvell
(2008) discuss a variety of such problems in their
overviews of cluster policy evaluation. Moreover,
Fromhold-Eisebith and Eisebith (2005) noted that
there may be a vested interest on the part of cluster
management and public officials not to have an
evaluation because it may yield undesirable re-
sults. Nonetheless, for those persons managing a
cluster, an evaluation of the progress of a cluster
1054
Assessing the Success of an Industrial Cluster
is necessary in order to attract funding and partici-
pants (Sölvell, 2008). Those were the motivations
behind the greenhouse cluster management team
developing measures to assess the success of the
northwest Ohio greenhouse cluster.
PURPOSE
The purpose of this paper is to discuss the initial
measures that have been developed to evaluate the
success of a greenhouse cluster which has been
operating in northwest Ohio since 2004. Success
for this paper is defined as the benefits that the
cluster has generated for the participating firms and
the local industry. An assessment of the impact of
the cluster on the regional economy is outside the
domain of this paper. The work reported in this
paper is only the beginning phases of a longer-
term, on-going effort to track the progress and
success of the cluster.
In the remainder of this paper, we begin by
providing background information on the clus-
ter, which is necessary to understand the metrics
that have been developed. Also data collection
procedures are outlined. This is followed by a
discussion of the results.
THE NORTHWEST OHIO
GREENHOUSE CLUSTER
While northwest Ohio’s greenhouse industry can
trace its existence back to the late 19th century, the
organization of the industry as a formal industrial
cluster began in 2004. The cluster was initiated by
university researchers at the University of Toledo
and Bowling Green State University, with strong
support from U.S. Congresswoman Marcy Kaptur.
Funding for the cluster project has been provided
by the United States Department of Agriculture.
The members of the cluster, including green-
house owners, university researchers, industry
suppliers, and other community stakeholders, have
Assessing the Success of an Industrial Cluster
worked collaboratively to identify and implement
solutions to common problems that the local in-
dustry has been experiencing. Those challenges
include a weak market presence, international
competition, dated infrastructure and production
technology, and high and rising energy costs
(LaFary et al., 2006). As a result of these issues,
many greenhouse operators have been pessimistic
about their industry’s future. In a 2004 survey,
only 27% of the growers in northwest Ohio were
optimistic about the future of the industry (Gatrell
et al., 2009).
The geographic footprint of the cluster com-
prises six counties in northwest Ohio (Figure 1).
The region covers an area of 2,189 square miles
and has a population of 789,452. The region’s main
urban area is Toledo (located in Lucas County),
with a population of 287,208 (U.S. Bureau of the
Census, 2010). The six counties are home to 81
greenhouse businesses that have over 5,125,717
square feet under glass or other protection (U.S.
Census of Agriculture, 2007). The industry com-
prises a mix of retail and wholesale greenhouses
whose main product line comprises floriculture
products (bedding plants, hanging baskets, etc.)
Figure 1. Northwest Ohio greenhouse cluster region
When, in 2004, Congresswoman Kaptur ap-
proached the University of Toledo to develop and
implement a strategy to assist the industry, uni-
versity researchers, in partnership with colleagues
from Bowling Green State University (and later
Ohio State University and Indiana State Univer-
sity) decided to adopt a cluster-based economic
development approach (Reid & Carroll, 2006). A
cluster approach seemed appropriate for the
greenhouse industry because initial analyses found
a fragmented local industry composed of small
and medium-sized enterprises that were family-
owned and multi-generational in nature. There is
a critical mass of greenhouses (80+) in the region
that has led to the creation of some external
economies of scale. For example, a small number
of suppliers to the local industry exists in the
region. From the perspective of the university
researchers, however, the key to helping the in-
dustry become more competitive was in helping
the growers to see the benefits of and engaging
in collaborative projects and initiatives.
We define cluster-based economic develop-
ment as a network driven economic strategy
built on collaboration and joint action among the
1055

participants which, when combined with external
economies of scale, generate what Schmitz (1995)
refers to as collective efficiencies. The goal is to
create a culture of collaboration that can occur
when growers, suppliers, researchers, economic
development agencies, and other community
stakeholders strategically join forces to confront
the competitive challenges facing the industry
– competitive challenges that due to the lack of
resources individual growers cannot successfully
address by themselves.
Meaningful collaboration can be problematic
since firms within a particular industry often
view each other as competitors. In the case of the
northwest Ohio greenhouse industry, enhancing
collaboration has been significantly impacted
by the grower culture. Most of the greenhouses
have been owned by the same family for multiple
generations, resulting in a complex network of
embedded relationships. While one might expect
this to be a positive characteristic from the per-
spective of the relationships within the industry,
this is not the case. Not all of these embedded
relationships are positive. Moreover, many of
the growers are independent and working col-
laboratively does not come easy, even in the face
of common threats. Consequently, an ongoing
challenge for the cluster management team has
been the fostering of collaboration.
Unlike many cluster projects which are top-
down and expert-driven, this particular one has
attempted to be grass roots and grower-driven The
grower-led nature of the cluster makes grower
collaboration and the existence of a strong social
network among the participants essential.
The cluster has undertaken a variety of projects
to enhance the industry’s competitiveness. The first
cluster project was the creation of a brand identity,
Maumee Valley Growers (MVG), in 2005. Starting
in 2006, the MVG brand was used in a number
of collaborative marketing initiatives that were
designed to enhance awareness of the industry
among residents of northwest Ohio and to con-
vince people to purchase more locally-produced
1056
Assessing the Success of an Industrial Cluster
products. The implementation of the branding
project and the subsequent marketing initiative was
slow because the growers had never previously
engaged in joint marketing efforts. For example,
a 2004 survey found that nearly two-thirds of
the northwest Ohio growers reported a lack of
marketing expertise and market information. Thus
they were unfamiliar, and initially uncomfortable,
with larger scale marketing projects. However, the
very process of creating the brand and the then
designing and implementing a marketing initiative
utilizing the brand brought the growers together,
cemented existing relationships, and created new
relationships (Reid & Carroll, 2008).
Following the launch of the brand and associ-
ated marketing campaign, the next challenge that
the growers tackled was high natural gas costs.
To manage natural gas costs, a number of the
growers came together in 2007 and combined
their purchases of natural gas into a single pur-
chase, thereby generating significant economies
of scale. In making these purchases the growers
worked with an energy consultant who made the
purchase on their behalf.
Natural gas is purchased on the spot and
futures markets and reflects current and future
demand conditions (Augustine et al., 2006). As
a result many growers had been spending inordi-
nate amounts of time researching gas prices and
agonizing over buying decisions, often second-
guessing themselves. The introduction of an
expert consultant and the economies of scale that
resulted from bulk purchases removed many of
the uncertainties of purchasing and resulted in
lower natural gas costs.
Despite the demonstrated benefits of the pro-
gram, however, many growers were slow to agree
to participate. As with the branding and market-
ing initiative, there was an inherent resistance to
collaboration. While the process of going-it-alone
in purchasing natural gas was stressful and chal-
lenging, it still meant that individual growers
were in charge of their own decision-making.
Participating in the joint purchasing program
Assessing the Success of an Industrial Cluster
required relinquishing critical decision-making
to the group. Once the group got large, it meant
individual growers relinquishing decision-making
to a decision-making body of which they were
not necessarily a participant. By 2009, buying
decisions were made by a nine-grower advisory
panel. In addition to the expertise and knowledge
of the energy consultant, the growers on the panel
also brought their particular wisdom and expertise
to the process. Having an energy consultant also
reduced the influence of gas marketers, whose
advice to the growers often varied dramatically
in its quality and was rarely in the best interests
of the growers.
DATA
As Schmiedeberg (2010) discussed, data avail-
ability has a major impact on the type of cluster
evaluation one utilizes. Ideally one would ac-
quire economic data, such as sales, employment,
and productivity growth, for the businesses in
the greenhouse cluster in order to evaluate the
cluster’s success (Sölvell, 2008). However, the
greenhouse owners are very reticent to share such
information. In addition, we had little success in
acquiring information about those owners’ social
characteristics that studies ((Madill et al., 2004;
Watts et al., 2006) have reported impact business
owners participation in clusters. Moreover, the
greenhouses are small firms and therefore little
data about them exist in published databases.
Consequently it was necessary to create indirect
indicators by which we could begin to triangulate
the success of the cluster. Sölvell (2008) provides
an excellent discussion of the role of perceptual
and analytical data, as well as hard data and soft
indicators in assessing clusters. In this project, we
employ a variety of types of information, includ-
ing hard and soft data and perceptual indicators.
A 2009 survey of greenhouse operators was one
important source of information. The survey was
sent to 60 greenhouses, of which 57 responded,
for a response rate of 95%. This survey provided
insights into growers’ perceptions of the value
of the cluster to them. Supporting these data are
personal observations by the researchers obtained
while working with the cluster since its inception.
A second source of data was a telephone survey
of northwest Ohio residents that was conducted
during 2008 (Reid et al., 2009). One objective of
this telephone survey was to provide greenhouse
owners with information on the characteristics
and motivations of persons buying greenhouse
products. Another objective was to determine
consumers’ awareness of the MVG brand that was
being used by growers in collaborative market-
ing initiatives. A total of 2,388 adults completed
the survey, resulting in a sampling error of 2.7%.
Respondents were selected by randomly selecting
telephone numbers from an electronic directory
of active residential telephone lines. The response
rate was 56%.
Sölvell (2008) suggests that measures such as
membership and network meetings are indicators
of strengthening cluster dynamics. In the case
of greenhouse cluster, one of the authors was
instrumental in the establishment of the cluster.
He also served as a member of the Board of Direc-
tors of the Maumee Valley Grower Association
and attends monthly stakeholder meetings. This
active participation provides rich insights into the
growers’ culture and attitudes toward all aspects
of the cluster. Due to the key role of the cluster
manager in a cluster project (Coletti, 2010) he/
she is a good source of insights into the operation
of all phases of the cluster.
To measure the development and growth of
social relations within the cluster, social network
analysis was used. Social network analysis de-
scribes and analyzes the relationships among a
group of people and/or organizations (de Nooy et
al., 2005). In social network analysis, the focus is
on the relationships among individuals, with the
underlying premise that the behavior of people and
organizations is impacted by, and in turn, shapes
their social networks. We focused specifically on
1057

the collaboration and advice/support networks
within the cluster. Moreover, we examine the
changes over time in those networks. This ap-
proach was chosen since standardized values have
not been set for social network metrics making it
difficult to state definitively that any given metric
value is either high or low. However, one can
monitor trends in the metrics to gauge evolution
of the network.
In 2007 and 2009, social network surveys of
the participants in the greenhouse cluster were
undertaken. In 2007, there were 111 people were
identified as being involved in the cluster, includ-
ing growers and representatives from suppliers,
academia, government agencies, and the economic
development field, whereas that number was 154
in 2009. The data were collected using the roster-
recall method (Steiner & Ploder, 2008; Ter Wal
& Boschma, 2009). In this approach, each of the
survey participants was given a list of all members
of the cluster project. They were asked to indicate
those individuals with whom, during the past 12
months, they had collaborated on a project and
who had given them advice or support over the
previous 12 months. The return rate was 66.7%
in 2007 and 43.0% in 2009. More specifically, the
analysis is based on the 46 people who returned
the questionnaire in both 2007 and 2009. These 46
people have been involved in the cluster since its
inception. The group is composed of 33 growers
and 13 non-growers, including academics, com-
munity stakeholders, and economic development
practitioners.
RESULTS
The results for collaboration and advice/support
networks will be reviewed. This is followed by
a discussion of the findings for the branding/
marketing and natural gas projects respectively.
1058
Assessing the Success of an Industrial Cluster
Networks
Various networking opportunities have been in-
corporated into cluster activities to build the social
network. These include monthly stakeholder meet-
ings, an annual dinner, and other periodic events.
One positive indicator of relationship building is
that average attendance at the monthly stakeholder
meetings increased from 15.0 in 2004 to 23.3 in
2010. Moreover, as a result of their participation
in the cluster, some growers have noted that they
no longer feel as isolated as they did previously.
In the 2009 grower survey, 80% of the respon-
dents said that they have more interaction with
their peers as a result of the cluster project. In
addition, nearly 60% of the respondents reported
greater access to university researchers as a result
of their participation in the cluster. The fact that
growers continue to attend stakeholder meetings
on a regular basis indicates that the meetings and
the project more broadly, provide value-added for
the growers. As small business owners surviving
on razor-thin margins, they do not have the time
to attend meetings that they perceive to have
little value.
An analysis of the social network data is another
avenue for examining collaboration and advice/
support networks. Cluster-based economic devel-
opment is an active and participatory collaborative
strategy that relies on a well-connected network.
One indicator of the connectedness within a
network is the number of ties. The number of
ties in both the collaboration and advice/support
networks decreased slightly over the two-year
period. In the case of the collaboration network,
the number of ties declined by 3.3% from 512 to
495. Similarly, the number of ties in the advice/
support network declined by 4.4% from 454 ties
to 434 ties. Explaining the decline in the number
of ties is challenging because most indicators
of cluster activity, such as meeting attendance,
participation in cluster projects, and the like,
increased over the study period. One hypothesis
is that people over time reduce the number of
Assessing the Success of an Industrial Cluster
people with whom they interact by focusing
on those individuals with whom they discover
interaction to be the most beneficial. The desire
to reduce the number of ties may be driven by
the high cost of maintaining ties in any network
(Kastelle & Steen, 2010). It is also possible that
people’s sets of links with other cluster participants
change over time as their need for different types
of information changes.
Another network measure is density. Density
is the ratio of actual node connections to the to-
tal potential node connections within a network
(de Nooy et al., 2005). The greater the number
of links (ties) between nodes (growers and non-
grower members of the cluster), the higher will be
the density of the network. Density is important
because ideas and information can travel more
efficiently through a higher density network. High
density networks also tend to me more robust and
can more easily withstand the loss of members.
Also, Krätke (2002, p. 38) argued that network
density impacts “the productive capacity of the
cluster as a regional production system.” Ucinet
6 (Borgatti et al., 2002.) was used to compute
densities and the other social network metrics.
The density of the collaboration network
was 0.248 in 2007 and 0.242 in 2009, whereas
the advice and support network had a density
of 0.219 in 2007 and 0.209 in 2009. Decreasing
densities are to be expected since the number of
nodes remained the same, but the number of ties
decreased over the study period.
Both networks had relatively low densities
in both years. As noted previously, standardized
thresholds have not been established for social
network metrics so it is difficult to make definitive
statements that any given value is either high or
low. Moreover, finding directly comparable stud-
ies is problematic due to variations in questions
asked, data collection methodologies, etc. Also
density values will vary depending on the size
of the network. Nonetheless Krätke (2002, pp.
38-39) stated that: “the ‘realistic’ value range of
the network density in production clusters tends
to be between 0 and 0.4.” It is important to note
that there is no agreement on the level of density
that is beneficial as opposed to detrimental to a
given network (Provan et al., 2007).
The generally low density in the greenhouse
cluster in part reflects the history of grower rela-
tions mentioned previously. These results are
consistent with other research that has shown
that actor characteristics and motivations impact
their networking activities (Madill et al., 2004).
As Bathelt (2005) suggested, formation of a
cluster does not cause firms to immediately start
working for the common good. Historical rela-
tions and power asymmetries among the actors
initially are major barriers that must be overcome
(Bathelt et al., 2004).
If we could have acquired more greenhouse
production data, as well as data describing owners’
social characteristics, then we could have used
GIS analyses. The spatial distribution of cluster
greenhouses and the total industry could have been
examined. Also the differences between cluster
member and non-members in terms of production
attributes and owner characteristics could have
been analyzed. Such analyses would not only
provide another perspective on the success of
the cluster, but also they might produce insights
as to the type of greenhouses and owners that are
attracted to the cluster, which could be useful in
recruiting non-member greenhouses to the cluster.
In addition to density, we examined the in-
degree of centralization in the network. Degree of
centralization measures the degree of inequality in
a network as a percentage of that of a perfect star
network of the same size, where a star network is
the most centralized network possible (Hanneman,
2005). In a star network, all the links in the network
center on one node (person) so that that one node
(person) has the most influence or power in the
network, as exemplified by node A in Figure 2.
Cross et al. (2006, p. 39) argued that many new
networks are characterized by high degrees of
centralization, with perhaps 15% of the members
accounting for 50% of the ties.
1059

Figure 2. Hypothetical star network
In both the collaboration and the advice/sup-
port networks, the degree of centralization dimin-
ished somewhat between the two years. Central-
ization decreased from 21.4% in 2007 to 13.8%
in 2009 in the collaboration network and the re-
spective percentages for the advice/support net-
work were 51.7% and 15.4%. In the case of the
collaboration network in 2007, the 15% of the
participants with the most ties in the network ac-
counted for less than 34.7% of all ties and by 2009
that groups’ percentage of all ties had decreased
to 31.6%. In the advice/support network, the re-
spective percentages were 32.9% in 2007 and
15.8% in 2009. It is notable that the advice/sup-
port network showed more decrease than the
collaboration network. The decrease in centraliza-
tion is important to the goals of the cluster. As
noted earlier, the cluster is a “grass roots” initia-
tive in which the strategy has been to build col-
laboration and support networks among the
cluster participants. The decline in centralization
suggests that there is a broadening of influence
among the various people in the network, which
is in keeping with the goals of the project.
The change in centralization within the advice
and support networks is also evident when one
1060
Assessing the Success of an Industrial Cluster
examines the network graphs. Figure 3 shows the
participants’ degree-in centralities in the networks;
size of the node is proportional to the node’s cen-
trality. Centrality is the number of connections (ties
or links) which a person has with other people in the
network. According to Steiner and Plodder (2008),
centrality provides insights into the interaction
among people in the network. Moreover, a high
centrality is indicative of more opportunities to
receive information. Also, centrality is a measure
of a person’s influence in the network (Lee & Su,
2010). Centrality can be measured as degree-out
which is a respondent reporting their ties to other
people. Conversely, degree-in is the number of
times a person is mentioned by other people in
the network. Degree-in is preferred because it
eliminates chances of self-reporting bias.
An examination of the two graphs reveals the
centrality of a number of growers had increased
between 2007 and 2009. Those growers enhanced
their interaction with other people in the network
and improved their opportunities for receiving
advice and support from network members. In
addition, their influence expanded over the study
period. In general, some growers were playing a
greater role in the advice and support network in
2007 than they had in 2009. This result is rein-
forced in the 2009 grower survey which found
that nearly 80% of growers indicated that they
had more interactions with other growers as a
result of their participation in the cluster.
We also examined growth in the network. The
collaboration network grew by 183 ties whereas
the advice/support network produced 181 new
ties. Preferential attachment is one explanation
of how a network expands. According to Ter
Wahl and Boshma (2009, p. 743), “the process of
preferential attachment describes the growth of a
network in which the probability that a new node
will link to a certain other node is proportional
to the number of links that node already has. An
outcome of this probabilistic process is that central
firms tend to become more central, whereas pe-
ripherally positioned firms tend to stay peripheral.”
Assessing the Success of an Industrial Cluster

Figure 3. Network graphs showing in-degree centrality

1061

According to the preferential attachment notion,
we would expect a positive correlation between
the number of ties which a node had in 2007 and
the number of ties added between 2007 and 2009.
The data do not support that expectation. In the
case of the collaboration network, the correlation
was 0.042 and it was 0.054 for the advice/support
network, and neither correlation coefficient was
significant at the 0.05 level. This result suggests
that the development of new ties is not simply
benefiting central people in the networks, which
is a positive finding given that one of the goals
of the cluster project is to grow the network by
getting individuals who previously had little or no
participation to become more active participants.
One phenomenon impacting the development
of new ties in the cluster is the presence of a
management team working at fostering mutu-
ally beneficial ties for the long term benefit of
the cluster. In fact, one member of the cluster
functioned as a network weaver whose job was
to build ties among participants which would
benefit the cluster.
In general, the social network metrics reveal
that the network among the 46 people had not
substantially changed over the two year period.
While the number of ties had declined somewhat,
the density of the networks had remained approxi-
mately the same. These results may have occurred
because two years is too short a time-frame for
substantial change to occur. Also they reflect the
fact that the greenhouse industry is generally a
conservative, slow changing industry. In the case
of the advice/support network, there is evidence
of a broadening of influence and increasing role
of growers within the network.
GIS can be applied to investigate the spatial
structure of social relationships within a cluster.
Even though the SNA nodes and their associated
metrics are aspatial, one can create a spatial com-
ponent by geocoding the node (peoples’) locations.
We were unable to acquire locations for all nodes
and therefore could not pursue this line of analysis.
By mapping the links among cluster members,
1062
Assessing the Success of an Industrial Cluster
the geography of the social relationships will be
revealed. One can assess the impact of distance
on the collaboration and advice/support networks.
Also spatial biases in the social network can be
identified. As one example, one might find that
most linkages are highly localized causing growers
to miss potentially productive linkages with more
distant people. The management team can then
work to bridge such relationship gaps (structural
holes). In general, GIS can be applied in various
ways to assess critical social relationships within
a cluster. Other examples of the use of GIS with
SNA can be found in Reid et al. (2009).
Branding and Marketing
The thrust of the marketing campaigns between
2005 and 2010 have been fairly traditional in
nature. For example, in the spring/early summer
of 2010 a number of northwest Ohio greenhouses
participated in a collaborative marketing cam-
paign. The campaign comprised commercials on
local cable television, advertisements in three local
newspapers, a direct mailing initiative, and adver-
tising on the grower website (maumeevalleygrow-
ers.com). The marketing also had a number of less
traditional elements, including “wrapping” a local
transit authority bus in the Maumee Valley Grower
logo. Coffee cups, refrigerator magnets, t-shirts,
and an annual calendar are also used to promote
the local greenhouse industry. The growers have
also participated in and contributed to a number
of local community efforts including Habitat for
Humanity, Extreme Makeover: Home Edition,
America in Bloom, and a variety of beautification
plantings throughout Lucas County.
To assess the outcomes of the branding and
marketing campaigns, data obtained from both
the consumer and producer surveys and par-
ticipating growers were used. In the consumer
survey, residents of northwest Ohio were asked
if they recognized the MVG brand and 41% of
the respondents indicated that they did. However,
there was considerable inter-county variation in
Assessing the Success of an Industrial Cluster
brand recognition. The highest level of brand
recognition occurred in Lucas County where 47%
or those surveyed recognized the brand. This is
perhaps not surprising given that Lucas County
is the most heavily populated county in northwest
Ohio and has been the focus of MVG marketing
dollars. For example, there have been some mar-
keting initiatives, e.g., the TARTA bus wrap, that
have been almost exclusively focused on Lucas
County. The lowest levels of brand recognition
(18.7%) are found in the most geographically
distant county, Erie County. Unfortunately, we
do not have data for years other than 2008 so we
cannot assess the extent to which recognition of
the brand has changed over time. However, given
that the brand was only introduced to the public as
recently as 2005 we feel that a recognition level of
47% within the region’s core county is reasonable.
We also learned that television was the most effec-
tive media for promoting the brand with 34.9% of
survey respondents having heard of the brand via
that medium, followed by newspaper (23%), and
word-of-mouth (22.7%). Through the survey, we
also learned about the motivations of consumer
purchasing preferences. The survey showed that
consumers who shopped at greenhouses valued
product quality and the opportunity to purchase
from locally-owned businesses versus “Big-Box”
store customers who valued price and convenience
(Reid et al., 2009).
Growers participated in a number of collabora-
tive marketing initiatives utilizing the Maumee
Valley Grower brand. The major benefit of this
collaborative marketing was that growers were
able to get better exposure for their businesses
and their products than they did when marketing
individually. In the case of print advertising, for
example, growers would historically place a small
advertisement in one of the local newspapers.
This advertisement would often appear along with
other advertisements for unrelated businesses and
was not particularly prominent. By collaborating,
growers were able to place full-page advertise-
ments in local newspapers, thereby giving their
individual business advertising space that was
more prominent. For example, one full page
advertisement that ran for ten days in the Toledo
Blade in the spring of 2010 cost a total of $15,423.
With 14 growers participating in this initiative,
the per-grower cost was $1,108.81. Only by col-
laborating was such a prominent advertisement
possible. The advertisement included coupons
for $5 off purchases of $25 or more. A total of
3,353 coupons were redeemed. An indicator of the
growers’ perception of the success of this initiative
is the fact that the growers have taken increasing
responsibility for the financing, creation, and
execution of collaborative marketing initiatives.
Between 2005 and 2007 all of the marketing was
paid for by a USDA grant. Beginning in 2008, the
growers started to contribute their own dollars to
defray a portion of the collaborative marketing
costs so the growers paid 67.7% of marketing
costs in 2009. By 2010, their contribution had
increased to 81% and they will pay 100% of col-
laborative marketing costs in 2011. Growers have
also taken greater ownership of the actual creation
of marketing initiatives. Up until 2009 the grow-
ers utilized the services of a professional brand-
ing and marketing firm. The growers have since
created a grower committee that is responsible
for creating and executing the growers’ annual
marketing campaign.
Another approach to assessing the impact of the
brand and marketing initiative is asking growers
about the impact that it has had on their business.
As part of the 2009 grower survey, growers were
asked if participation in the cluster marketing
initiative had helped them attract new customers.
Given the nature of the industry, we expected that
the branding/marketing program would provide
greater benefit to the retail rather than the whole-
sale growers. This is because the retail growers
sell directly to the consumer while the wholesale
growers sell to another retailer (mostly “Box
Box” stores like Walmart and Home Depot) who
in turn sells their product to the customer. A chi-
square analysis of a contingency table (Table 1)
1063

Table 1. Growers attracting new customers*
Attract New Customers? Retailers Wholesalers
Yes 26 5
No 11 12
*Significantly different at the 0.01 level of significance
indicated a significant difference between retailers
and wholesalers responses. Whereas 29.4% of
the wholesalers indicated that participation in the
project had helped them attract new customers,
70.3% of retailers indicated that participation in
the project had helped them attract new customers.
Given that wholesalers have no direct relationship
with the final consumer, it is interesting that five
of them indicated that participation in the project
had helped them attract new customers. Perhaps
this is the result of higher retail sales causing
higher wholesale sales.
GIS has long been applied in retail location
studies; see for example Clarke, 1998. A future
endeavor, which will be useful from an assessment
viewpoint, is to use appropriate GIS procedures
to analyze the retail greenhouses. As one example,
a map of retail greenhouse market potential in the
region will be generated using census block group
data. Building of an index of sales potential that
can be mapped with conventional data sources
will be a challenge since the demographics of
greenhouse consumers are only reported for large
regions of the country. If the potential can be
reasonably estimated, then market penetration by
cluster firms can be assessed and opportunities
for expansion of sales can be gauged. If indi-
vidual retailers will provide their sales data, then
more detailed analysis will be possible. These
efforts will provide a different assessment of the
market initiatives of the cluster.
Natural Gas Purchasing
The natural gas purchasing program has been suc-
cessful in various respects. From the perspective of
managing costs, it has been beneficial. In the first
1064
Assessing the Success of an Industrial Cluster
year of the program (2007), participating growers
collectively saved $150,000, with individual firms
reducing natural gas costs by between 8 and 15%.
The gas program also has been successful in
terms of attracting new firms. Growers elsewhere
in Ohio and in neighboring southeastern Michi-
gan began participating in the purchasing pool in
2009. By March 2011, the program had a total of
102 participating growers (90 in Ohio and 12 in
Michigan) whose collective savings for the year
are forecast to be between $400,000 and $500,000.
The program has become particularly attractive to
small growers who benefit from the economies of
scale resulting from the purchasing pool.
Another benefit of the gas program is that it
has provided money for the cluster organization
(the Maumee Valley Growers Association). The
organization receives a commission on natural gas
purchased through the program, which has gener-
ated over $60,000 for the organization since 2007.
Because of its success, the gas program has
led to other collaborative endeavors to manage
growers’ costs. For example, the model is being
extended to electric power purchasing. While still
in the early phases, this project will generate a
5.0% savings for firms over a two-year period.
Also this approach has been extended to purchas-
ing other greenhouse inputs. In the case of one
supplier, the growers were able to negotiate a 20%
discount with no minimum order, instead of the
customary 5% discount that also had a minimum
order requirement. Moreover, 5% of the orders’
dollar value is returned to the cluster organization
to reinvest into the program.
From the perspective of evaluation, the gas
program is the easiest to assess. For one thing,
the nature of the project results in detailed records
being kept, and those records are kept by the
program manager so they are available in con-
trast to growers’ records which are inaccessible.
Moreover, it focuses directly on the firms’ bot-
tom lines, not indirect actions such as the amount
of collaboration among participants. As Sölvell
(2008, p. 60) argued evaluations should partially
Assessing the Success of an Industrial Cluster
focus on the performance of firms in the cluster,
not measures, such as “... number of meetings,
extension of personal networks, and reports that
have been carried out and presented.” Because one
can easily define and measure performance in the
gas project (dollar savings to growers), growers
more readily embrace it than other initiatives.
DISCUSSION AND CONCLUSION
Evaluation of cluster outcomes is important, but the
topic has rarely been addressed in the literature. No
doubt people involved in building and/or managing
a cluster may lack the time or skill sets to execute
sophisticated evaluation schemes. Such is the case
with the northwest Ohio greenhouse cluster. For
these reasons, it is important to develop metrics
and procedures that can be compiled easily, while
still providing meaningful information. Metrics of
success must also be appropriate for the cluster in
question and must reflect the goals of the cluster.
For example, one of the advantages of clusters that
are often cited in the literature is that it enhances the
innovative capacity of the industry. In the case of
the northwest Ohio greenhouse industry, enhancing
innovative capacity is not a primary goal at this time,
and therefore it is not included in our evaluation.
There is evidence that the greenhouse cluster
has had a positive impact on its participants. The
most notable success is the natural gas program. It
has saved participants substantial money and gained
acceptance by growers located outside the local
region. It is easier to quantify the impact of the gas
program because the collection of monetary data
is a necessary part of operating the program. Other
hard data, such as sales of greenhouse products, are
not available to the research team.
The evidence regarding the impact of the brand-
ing and marketing program is less solid than in the
case of the gas project. On one hand, a notable
percentage of surveyed northwest Ohio residents
had heard of the MVG brand, but that percentage
was very uneven across the cluster region. Also,
we do not know if brand recognition has led to
greater sales. We need to collect point-of-sale data
to validate such a relationship. The perception of
many growers is that cluster activities have enabled
them to attract more customers. However, we can-
not measure the increase in sales. Moreover we
do not know that the increased sales, if they exist,
have been caused by the cluster. In fact, a general
problem in cluster evaluation is to determine if
observed effects have been caused by the cluster
or if there are other “explanatory drivers” (Sölvell,
2008, p. 62).
Measurement of collaboration is problematic.
While some indirect indicators, such as grower
interviews and participation in cluster activities,
are positive, we cannot gauge the extent to which
collaboration has generated positive economic im-
pacts. Social network analysis provides a promising
venue for assessing collaboration, but that technique
has inherent problems in both data collection and
interpretation (Ter Wal & Boschma, 2009; Brandt
et al., 2009). Future research plans include making
greater efforts to create more useful indicators of
collaboration and the impact of collaboration and
joint actions.
A major challenge will be the acquisition of
economic data on the greenhouses in the cluster
region. At various places in this paper, we have
identified the need for sales, employment, and
size data, as well as other information. To date,
greenhouse owners have been most resistant to
providing specifics on their operations. With the
acquisition of more data, it will be possible to
improve our existing measures of success and use
GIS to provide valuable insights into social rela-
tions in the industry, market potential, and the like.
Ideally the measures that are developed should
not only help gauge the success of a cluster, but also
they should help guide future decisions (Schmie-
deberg, 2010). An example of this is using social
network analysis to build desired connections within
the cluster. As suggested by Brandt et al. (2009),
one can identify isolated people and fragments of
the network that perhaps can be better integrated
1065

with the main cluster core. Another example is the
finding from the consumer survey that individuals
who shop at retail greenhouses value product qual-
ity and purchasing from locally-owned businesses.
This information has helped shape advertising
campaigns.
The metrics developed to date provide a baseline
for future comparison. One piece of perceptual data
we believe is significant is the fact that 60% of the
growers stated in 2009 that they were optimistic
about the future of their business as a result of the
cluster. This statistic is in marked contrast to the
27% who were optimistic about the future of the
industry in 2004. Due to the intergenerational nature
of this industry, such optimism is critical. If adult
children, many of who are college educated and have
viable career opportunities outside of the industry,
are going to take over the family business when
their parents retire then a high degree of optimism
is highly desirable. At the same time, this growth
in optimism occurred while northwest Ohio expe-
rienced extreme economic distress. Therefore, this
cluster may very well have had some real impact
on the economic health of this industry.
ACKNOWLEDGMENT
This paper was invited for publication in the Inter-
national Journal of Applied Geospatial Research
with expansion and revision from its earlier form
published in the Papers of the Applied Geography
Conferences, Volume 33, 2010, and is adapted
herein with permission.
REFERENCES
Augustine, C., Broxson, B., & Peterson, S. (2006).
Understanding natural gas markets: A policy
analysis study by Lexecon, an FTI Company.
Retrieved March 28, 2011, from http://www.
api.org/aboutoilgas/upload/UNDERSTAND-
ING_NATURAL_GAS_MARKETS.pdf
1066
Assessing the Success of an Industrial Cluster
Bathelt, H. (2005). Geographies of produc-
tion: Growth regimes in spatial perspective (ii)
– knowledge creation and growth in clusters.
Progress in Human Geography, 29(2), 204–216.
doi:10.1191/0309132505ph539pr
Bathelt, H., Malmberg, A., & Maskell, P. (2004).
Clusters and knowledge: Local buzz, global
pipelines and the process of knowledge creation.
Progress in Human Geography, 28(1), 31–56.
doi:10.1191/0309132504ph469oa
Borgatti, S., Everett, M., & Freeman, L. (2002).
Ucinet for Windows: Software for social network
analysis. Harvard, MA: Analytic Technologies.
Retrieved March 12, 2011, from http://www.
analytictech.com
Brandt, A., Hahn, C., Krätke, S., & Kiese, M.
(2009). Metropolitan regions in the knowledge
economy: Network analysis as a strategic informa-
tion tool. Tijdschrift voor Economische en Sociale
Geografie, 100(2), 236–249. doi:10.1111/j.1467-
9663.2009.00532.x
Clarke, C. (1998). Changing methods of location
planning for retail companies. GeoJournal, 45(4),
289–298. doi:10.1023/A:1006995106736
Coletti, M. (2010). Technology and industrial
clusters: How different are they to manage? Sci-
ence & Public Policy, 37(9), 679–688. doi:10.31
52/030234210X12778118264413
Cross, R., Laseter, T., Parker, A., & Velasquez, G.
(2006). Using social network analysis to improve
communities of practice. California Management
Review, 49(1), 32–60.
de Nooy, W., Mrvar, A., & Batagelj, V. (2005).
Exploratory network analysis with pajek. Cam-
bridge, UK: Cambridge University Press.
Delgado, M., Porter, M., & Stern, S. (2010).
Clusters and entrepreneurship. Journal of Eco-
nomic Geography, 10(4), 495–518. doi:10.1093/
jeg/lbq010
Assessing the Success of an Industrial Cluster
Feser, E., Renski, R., & Goldstein, H. (2008).
Clusters and economic development outcomes:
An analysis of the link between clustering and in-
dustry growth. Economic Development Quarterly,
22(4), 324–344. doi:10.1177/0891242408325419
Fowler, C., & Kleit, R. (2010). Linking economic
development and equity: An analysis of the effects
of industrial clusters on the poverty rate. Seattle,
WA: Community Vitality Project, Evans School
of Public Affairs, University of Washington. Re-
trieved March 11, 2011, from http://cvp.evans.
washington.edu/wp-content/uploads/2010/07/
fowler_and_kleit_working_paper_2010.pdf
Fromhold-Eisebith,M.,&Eisebith,G.(2005).Look-
ing behind facades: Evaluating effects of (automo-
tive) cluster promotion. Regional Studies, 42(10),
1343–1356. doi:10.1080/00343400701654202
Gatrell, J., Reid, N., Steiger, T., Carroll, M.,
& Smith, B. (2009). “Value”-Chains: Identity,
Tradition, and Ohio’s Flori (culture) Industry.
Applied Geography (Sevenoaks, England), 29(3),
346–357. doi:10.1016/j.apgeog.2008.12.001
Hanneman, R. (2005). Introduction to social net-
work networks. Retrieved March 14, 2011, from
http://analytictech.com/networks.pdf
Kastelle, T., & Steen, J. (2010). Are small world
networks always best for innovation? Innovation:
Management. Policy & Practice, 12(1), 75–87.
doi:10.5172/impp.12.1.75
Krätke, S. (2002). Network analysis of production
clusters: The Potsdam/Babelsberg film industry as
an example. European Planning Studies, 10(1),
27–54. doi:10.1080/09654310120099254
LaFary, E., Gatrell, J., Reid, N., & Lindquist, P.
(2006). Specialized agriculture: Local markets &
global competitors in Ohio’s greenhouse industry.
In Gatrell, J., & Reid, N. (Eds.), Enterprising
worlds: A geographic perspective on economics,
environments, and ethics (pp. 57–70). Dordrecht,
The Netherlands: Springer-Verlag.
Lee, P., & Su, H. (2010). Investigating the structure
of regional innovation system research through
keyword co-occurrence and social network analy-
sis. Innovation: Management. Policy & Practice,
12(1), 26–40. doi:10.5172/impp.12.1.26
Madill, J., Haines, G. Jr, & Riding, A. (2004).
Networks and linkages among firms and organi-
zations in the Ottawa-region technology cluster.
Entrepreneurship & Regional Development, 16(5),
351–368. doi:10.1080/0898562042000188414
Martin, R., & Sunley, P. (2003). Deconstructing
clusters: Chaotic concept or policy panacea?
Journal of Economic Geography, 3(1), 5–35.
doi:10.1093/jeg/3.1.5
Muro, M., & Fikri, K. (2011). Job creation on a
budget: How regional industry clusters can add
jobs, bolster entrepreneurship, and spark inno-
vation. Washington, DC: Brookings-Rockefeller
Project on State and Metropolitan Innovation,
Brookings Institution. Retrieved March 10, 2011,
from http://www.brookings.edu/~/media/Files/
rc/papers/2011/0119_clusters_muro/0119_clus-
ters_muro.pdf
National Governors Association (NGA).
(2007). Cluster-based strategies for grow-
ing state economies. Retrieved March
11, 2011, from http://www.nga.org/Files/
pdf/0702INNOVATIONCLUSTERS.PDF
Provan, K., Fish, A., & Sydow, J. (2007). Inter-
organizational networks at the network level: A
review of the empirical literature on whole net-
works. Journal of Management, 33(3), 479–516.
doi:10.1177/0149206307302554
Reid, N., & Carroll, M. (2006). Collaborating to
compete: The case of the northwest Ohio green-
house cluster. In Gatrell, J., & Reid, N. (Eds.),
Enterprising worlds: A geographic perspective on
economics, environments, and ethics (pp. 41–56).
Dordrecht, The Netherlands: Springer-Verlag.
1067

Reid, N., & Carroll, M. (2008). Creating trust
through branding: The case of northwest Ohio’s
greenhouse cluster. In Stringer, C., & LeHeron,
R. (Eds.), Agri-food commodity chains and glo-
balizing networks (pp. 174–188). London, UK:
Ashgate.
Reid, N., Carroll, M., Smith, B., & Frizado,
J. (2009). GIS and economic development. In
Gatrell, J., & Jensen, R. (Eds.), Geotechnolo-
gies and the environment: Socioeconomic and
planning applications (pp. 5–28). Heidelberg,
Germany: Springer-Verlag. doi:10.1007/978-1-
4020-9642-6_2
Reid, N., Smith, B., Haase, D., Ross, P., Mirzoy-
ants, A., & Gatrell, J. (2009). Marketing & growing
placed-based clusters: The case of the northwest
Ohio greenhouse industry. Papers of the Applied
Geography Conferences, 32, 40–46.
Schmiedeberg, C. (2010). Evaluation of cluster
policy: A methodological overview. Evaluation,
16(4), 389–412. doi:10.1177/1356389010381184
Schmitz, H. (1995). Collective efficiency:
Growth path for small scale industry. The Jour-
nal of Development Studies, 31(4), 529–566.
doi:10.1080/00220389508422377
Sölvell, Ö. (2008). Clusters: Balancing evolution-
ary and constructive forces. Stockholm, Sweden:
Ivory Tower.
Steiner, M., & Ploder, M. (2008). Structure and
strategy within heterogeneity: Multiple dimensions
of regional networking. Regional Studies, 42(6),
793–815. doi:10.1080/00343400701861310
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 3,
edited by Donald Patrick Albert, pp. 21-36, copyright 2012 by IGI Publishing (an imprint of IGI Global).
1068
Assessing the Success of an Industrial Cluster
Taylor, M. (2006). Clusters and local economic
growth: Unpacking the cluster model. In Gatrell,
J., & Reid, N. (Eds.), Enterprising worlds: A
geographic perspective on economics, environ-
ments & ethics (pp. 99–117). Dordrecht, The
Netherlands: Springer-Verlag. doi:10.1007/1-
4020-5226-X_9
Ter Wal, A., & Boschma, R. (2009). Applying
social network analysis in economic geography:
Framing some key analytic issues. The Annals of
Regional Science, 43(3), 739–756. doi:10.1007/
s00168-008-0258-3
U.S. Census Bureau. (2010). Census of population
2010. Retrieved November 5, 2010, from http://
factfinder2.census.gov
U.S. Department of Agriculture. (2007). U.S.
Census of agriculture 2007. Retrieved March 28,
2011, from http://www.agcensus.usda.gov/Pub-
lications/2007/Full_Report/Volume_1,_Chap-
ter_2_County_Level/Ohio/st39_2_035_035.pdf
van der Linde, C. (2002). The cluster meta-study;
list of clusters and bibliography. Boston, MA: In-
stitute for Strategy and Competitiveness, Harvard
Business School. Retrieved March 12, 2011, from
http://www.isc.hbs.edu/MetaStudy2002Bib.pdf
Watts, D., Wood, A., & Wardle, P. (2006).
Owner-managers, clusters and local embed-
dedness: Small firms in the Sheffield (UK)
metal-working cluster. Entrepreneurship
and Regional Development, 18(3), 185–205.
doi:10.1080/08985620600680141
 1069

Chapter 64
A Geospatial Analysis
of Convective Rainfall
Regions within Tropical
Cyclones after Landfall
Corene J. Matyas
University of Florida, USA

ABSTRACT
In this article, the author utilizes a GIS to spatially analyze radar reflectivity returns during the 24 hours
following 43 tropical cyclone (TC) landfalls. The positions of convective rainfall regions and their areal
extent are then examined according to storm intensity, motion, vertical wind shear, time until extra-
tropical transition, time after landfall, and distance from the coastline. As forward velocity increases in
conjunction with an extratropical transition, these regions move outward, shift from the right side to the
front of the TC, and grow in size. A similar radial shift, but with a decrease in areal extent, occurs as
TCs weaken. Further quantification of the shapes of these regions could yield a more spatially accurate
assessment of where TCs may produce high rainfall totals.

INTRODUCTION example, more than 150 lives were lost from

Tropical cyclones (TCs) can produce high rainfall
totals that lead to flooding both near the coastline
where fast winds and storm surges are also of
concern, and hundreds of kilometers away from
the location of landfall (Rappaport, 2000). For
DOI: 10.4018/978-1-4666-2038-4.ch064
fresh water flooding as Hurricane Camille (1969)
produced record rainfall totals nearing 700 mm
in Virginia. Several days earlier, most deaths that
occurred during the landfall of Hurricane Camille
resulted from the high storm surge (Schwarz,
1970). Recently, the original rainfall climatology
and persistence (R-CLIPER) statistical model
developed to verify rainfall predictions for TCs

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
(Marchok, Rogers, & Tuleya, 2007) has been
upgraded to include the effects of vertical wind
shear and topography as these characteristics
have been shown to cause asymmetries in TC
rain fields (Lonfat, Rogers, Marchok, & Marks,
2007). However, as recently as 2007, the Ameri-
can Meteorological Society stated that rainfall
predictions for landfalling TCs are still in need
of improvement (AMS, 2007).
Although stratiform clouds can cover ten
times more area within a TC than convective
clouds (Jorgensen, 1984), the higher rainfall
rates produced by convective clouds can produce
flooding rainfall in a relatively short amount of
time (Geerts et al., 2000; Elsberry, 2002). Even
though Hurricane Floyd (1999) was moving at a
relatively fast forward velocity of 9 m s-1 during
landfall, the high rain rates produced by convec-
tive processes caused more than 500 mm of rain
to accumulate (Lawrence et al., 2001; Atallah &
Bosart, 2003). Thus, identifying the conditions
under which convective clouds cover a large area
within a TC’s rain fields, and where these areas
of convection will exist relative to the circulation
center of a TC is an important task.
This study employs geospatial research tech-
niques to examine convective rainfall regions that
exist during the 24-hour period after TCs make
landfall. Specifically, three research goals are
pursued: 1) determine where regions of heavy
rainfall exist in relation to the circulation center
of the TC under different conditions, 2) determine
the areal extent, or size, of these rainfall regions,
and 3) determine the conditions under which the
largest areas of heavy rainfall tend to form. To
achieve these research goals, radar reflectivity
returns are analyzed within a GIS to locate the
heavy rainfall regions and to calculate the spatial
properties of these regions as they relate to six
factors. These factors are a) storm intensity, b)
vertical wind shear, c) storm motion, d) whether
or not a TC becomes extratropical, e) the time
elapsed after landfall, and f) the distance that the
storm and its convection are located relative to the
1070
coastline. In addition to achieving these research
goals, this study analyzes a large number of TCs
in a GIS framework.
BACKGROUND
Willoughby, Marks, and Feinberg (1984) describe
the regions of convection that develop in both
the core of a hurricane (within 100 km of the
circulation center), and in the outer rainbands
that can extend more than 360 km from the
circulation center of the storm. Convection is
enhanced by increased surface friction along the
coastline during landfall (Tuleya, 1994; Frank
& Ritchie, 1999), and fast tangential winds can
advect moisture counterclockwise into the left
front quadrant in the core of a hurricane (Parrish,
Burpee, Marks, & Grebe, 1982). Konrad, Meaux,
and Meaux (2002) found that intense TCs have
increased rainfall near the core of the storm. On
the other hand, tropical depressions have very
little organized convection in their core (Frank &
Ritchie, 1999). These findings suggest that storm
intensity could exhibit a strong relationship with
both the location of convection relative to the
circulation center of the storm and the size of the
area covered by convective rainfall.
A recent modeling study by Kimball (2008)
found that rainfall spreads outwards from the
circulation center as a TC tracks inland, and shifts
from a maximum on the right side of the storm
to the front of the storm in a counterclockwise
direction. Observational studies of individual TCs
have demonstrated that convection within TCs
may increase or decrease in spatial extent after
landfall (e.g., Bluestein & Hazen, 1989; Atallah
& Bosart, 2003; Medlin, Kimball, & Blackwell,
2007; Matyas, 2008). Time elapsed after landfall
may demonstrate an association with these trends
in the location and areal extent of TC convec-
tion. However, the influence of the atmosphere,
internal storm dynamics, and the earth’s surface
may exhibit an even stronger association with the
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
position and size of convective regions during
the post-landfall period. TCs that remain near to
the coastline after landfall may be able to advect
moisture into their circulation so that rainfall does
not diminish as quickly as it would if the storm
had moved directly inland (Medlin et al., 2007).
Thus, the location of the TC or its convection
relative to the coastline could be related to the
position and/or the areal extent of convection.
Several observational studies have shown
that vertical wind shear causes an asymmetrical
distribution of rainfall about the center of a TC so
that the majority of rainfall occurs in the downs-
hear direction (Molinari, Moore, & Idone, 1999;
Corbosiero & Molinari, 2002; Cecil, 2007). Other
studies have focused on the asymmetries in the
rain fields of TCs produced by storm motion (e.g.,
Burpee & Black, 1989; Rodgers, Baik, & Pierce,
1994; Lonfat, Marks, & Chen, 2004), finding that
rainfall shifts to a location ahead of the storm as
forward velocity increases. When Corbosiero
and Molinari (2003) and Chen, Knaff, and Marks
(2006) examined the contributions of both storm
motion and vertical wind shear to the distribu-
tion of TC rainfall, they determined that vertical
wind shear exerts the dominant influence, while
storm motion controls rainfall distribution only
when vertical wind shear is weak. However, these
authors only examine lightning flashes and areas
of rainfall within 300 and 330 km of the storm
centers, respectively, whereas research by Lonfat
et al. (2004) show that convective rainfall can oc-
cur 500 km from the circulation center of a TC.
When TCs become extratropical, they ex-
perience both strong vertical wind shear and an
increase in forward velocity (Atallah & Bosart,
2003; Jones et al., 2003), so it stands to reason
that the time until a TC becomes extratropical
should exhibit a relationship to the distribution of
rainfall about the circulation center of the storm
in a pattern similar to that imposed by shear and
storm motion. TCs that transition into extratropical
systems tend to have well-developed rain fields
on the left side of their storm track (Atallah &
Bosart, 2003; Matyas, 2008) as rainfall expands
ahead of the circulation center. For TCs that do not
experience the rainfall-enhancing dynamics that
occur during an extratropical transition, convective
rainfall may decrease as the storm moves inland,
and is removed from its source of latent energy
(Tuleya, 1994; Kimball, 2008), and is surrounded
by drier continental air masses (Cubukcu, Pfeffer,
& Dietrich, 2000; Hill & Lackmann, 2009).
DATA
Radar reflectivity returns from multiple Weather
Surveillance Radar-1988 Doppler (WSR-88D)
sites are utilized to determine where heavy rainfall
is located relative to the TC’s center of circulation.
These data are obtained from the National Climatic
Data Center’s NEXRAD data archive (http://www.
ncdc.noaa.gov/nexradinv/). Base reflectivity data
from the Level III WSR-88D products are utilized
whenever available. In cases where only the raw
Level II data are available, data from the lowest
scan elevation are utilized for the spatial analy-
sis. Additional information pertaining to Level
II and III radar data are available in the Federal
Meteorological Handbook 11-Part C (Office of
the Federal Coordinator for Meteorology, 2006).
Archives for many of the 159 WSR-88D sites
extend from 1995 through the present, allowing
landfalling TCs during 1995-2008 to be examined
in the current study. Due to problems with the
archiving system at the National Climatic Data
Center, data for several TCs are not available at
the time of this study, including all TCs from the
1998 season. For a TC to be analyzed in the current
study, its rain fields must be within range of the
WSR-88D network for a 24-hour period following
the time of landfall. Thus, TCs that either dissipate
within 24 hours of landfall (e.g., Matthew (2004)),
or that did not remain over land for 24 consecutive
hours after landfall (e.g., Charley (2004), Wilma
(2005)) are not included in the study. As a second
U.S. landfall occurs for some TCs more than 24
1071
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
hours after the first landfall, these subsequent
landfalls are also analyzed. Overall, this study
examines 43 U.S. landfalls from 38 TCs.
Data pertaining to the latitude and longitude
of the circulation centers of the TCs are obtained
from two sources. The position of the circulation
center at the time of landfall is taken from the
National Hurricane Center (NHC)’s Hurricane
Season Tropical Cyclone Report for each storm
(http://www.nhc.noaa.gov/pastall.shtml). These
reports also provide the date and time of landfall.
For this study, the time of landfall is rounded to
the nearest half-hour. The position of the storm
at all other analysis times, as well as its intensity
as measured by the maximum sustained wind
speed, forward velocity and heading and stage
of the storm, are obtained from the Hurricane
Database (HURDAT) maintained by the NHC
(2006). This database contains observations for
all TCs in the North Atlantic Ocean basin every
six hours at 0000, 0600, 1200, and 1800 UTC.
A linear interpolation is performed to obtain
data in three-hourly time steps beginning at the
time of landfall and ending 24 hours later. This
process yields nine observation times for each
TC landfall, or 387 total observation times for
the 43 landfalls examined. The number of hours
until a TC becomes extratropical is also calculated
at each observation time for TCs that became
extratropical either during the study period or up
to 72 hours after landfall.
The velocity and direction of the vertical wind
shear are acquired from the Statistical Hurricane
Intensity Scheme (SHIPS) database (DeMaria &
Kaplan, 1994; DeMaria, Mainelli, Shay, Knaff, &
Kaplan, 2005). SHIPS data are available for TCs in
both the Atlantic and eastern Pacific Ocean basins
at 0000, 0600, 1200, and 1800. The vertical wind
shear is the difference betweeen the 200 and 850
hPa winds calculated for an annular region located
200-800 km from the circulation center of each
TC. The SHIPS data are also interpolated linearly
to obtain data in three-hourly time steps beginning
at the time of landfall and ending 24 hours later.
1072
ANALYSIS
To identify regions of heavy rainfall within the
rain fields of a TC and to calculate their spatial
attributes, the radar reflectivity data are converted
into a georeferenced format and imported into a
GIS. Working with each three-hourly observation
separately, the data from adjoining WSR-88D sta-
tions are first combined into a single raster layer
with a grid size of one square kilometer. A small
amount of error is introduced into the analysis due
to differences in calibration among the different
radars and the range-dependent beam geometry
and elevation (Anagnostou, 2004). At locations
where reflectivity values are available from mul-
tiple WSR-88D sites, the highest reflectivity value
is retained. Figure 1 demonstrates the combining
of the radar reflectivity from multiple WSR-88D
sites within the GIS. Next, an interpolation is per-
formed using inverse distance weighting to create
contour lines that connect areas having the same
reflectivity value. As the level III reflectivity data
are available in 5 dBZ increments, the contour
lines are also placed in 5 dBZ increments. The
polylines created by the interpolation are then
converted into polygons, and spatial attributes
are calculated for each polygon, including area
and the latitude and longitude of the centroid, or
center of mass.
As this study is concerned with identifying
regions of heavy rainfall that may have the po-
tential to produce flooding, polygons larger than
500 km2 in area that enclose reflectivity values
of 40 dBZ or greater are retained for further
analysis. As shown in Figure 2, multiple convec-
tive regions of 40 dBZ reflectivity values over
500 km2 in area may exist for a TC at a given
observation time, and all such areas are analyzed
in the current study. TC researchers such as Jor-
gensen (1984) have used 40 dBZ reflectivity
values to denote heavy rainfall within convective
rainbands. Reflectivity values of 40 dBZ may
produce rainfall rates of approximately 21.6 mm
hr-1 (0.85 in hr-1) according to the reflectivity to
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Figure 1. Radar reflectivity values for Hurricane Ivan (2004) at the time of landfall; The highlighted
polygons represent regions having reflectivity values greater than or equal to 40 dBZ that are greater
than 500 km2 in a real extent

Figure 2. Position and intensity of Ivan (2004) and the 40 dBZ polygons that are analyzed at each
3-hourly observation time

1073
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
rain-rate conversion developed by Rosenfeld,
Wolff, and Atlas (1993) for tropical systems.
This study seeks to determine which of six
factors are most closely associated with the posi-
tion of the convective regions relative to the
circulation center of the storm, and the areal extent
of these regions. These factors are the intensity
of the storm as measured by the maximum sus-
tained wind speed, the speed of the vertical wind
shear and storm forward motion, whether the TC
becomes extratropical, time that has elapsed since
landfall, and the position of the TC and its rainfall
regions relative to the coastline. Therefore, the
attribute table of each convective polygon is
expanded to include additional information for
its observation time. Columns added to the at-
tribute table include the storm name, year, month,
day, observation time, circulation center latitude
and longitude, maximum sustained wind speed,
direction and speed of vertical wind shear and
storm motion, the number of hours since landfall,
and the number of hours until the storm becomes
extratropical, if applicable. The distance and bear-
ing of each polygon centroid relative to the cir-
culation center of the TC at each observation time
is then calculated using spherical trigonometry.
The next step in the analysis is to recalculate
the positions of the polygon centroids so that
their bearings are in coordinates that are relative
to the heading of the storm. First, the heading of
the storm is subtracted from the bearing of the
centroid. Next, 360° is added to negative values.
Thus, polygon centroids located at bearings 0-90°
are in the right-front quadrant, and centroids with
bearings of 90-180°, 180-270°, and 270-360° are
located in the right rear, left rear, and left front
quadrants respectively. A similar process is per-
formed to determine the shear-relative quadrants.
For this calculation, the heading of the vertical
wind shear is subtracted from the bearing of the
centroid, and 360° is added to negative values.
Polygon centroids with bearings of 0-90° are
located in the downshear right quadrant, while 90-
180°, 180-270°, and 270-360° are located upshear
1074
right, upshear left, and downshear left respectively.
The GIS is also utilized to determine the distance
from each polygon centroid to the nearest point
on the U.S. coastline. In all, 21 columns of data
exist for each convective region.
Two strategies are employed to determine
which of the six factors has the strongest asso-
ciation with the radial position of the convective
regions relative to the circulation center of the
TC. First, correlation coefficients are calculated
between the distance of the convective region
relative to the circulation center of the TC and the
variables representing the six factors. As the data
are not normally distributed, they are ranked from
highest to lowest values and Spearman rank cor-
relation coefficients (Wilks, 1995) are calculated.
The second strategy to explore patterns in
the location of convection under the six factors
involves grouping the convective regions ac-
cording to sub-classifications of each condition.
Observations are grouped by the intensity of the
storm as determined by the velocity of the maxi-
mum sustained winds: 1) greater than 33 m s-1
for hurricanes, 2) 17-33 m s-1 for tropical storms,
or 3) less than 17 m s-1 for tropical depressions.
The convective regions are placed into one of
three groups based on the speed of the vertical
wind shear: 1) less than 5 m s-1, 2) 5-10 m s-1,
or 3) greater than 10 m s-1. This breakdown of
slow, medium, and fast vertical wind shear is
utilized to render the results comparable to those
of Corbosiero and Molinari (2002). These same
divisions are used to group the convective regions
according to the forward velocity of the TC. Four
categories are developed to examine convection
relative to the time that a TC becomes extratropi-
cal: 1) less than 24 hours, 2) 24-48 hours, 3) 48-72
hours, or 4) observations from TCs that did not
become extratropical within 72 hours of landfall.
The number of hours after landfall (0, 3, 6, 9,
12, 15, 18, 21, and 24) is also used to group the
observations. Distance relative to the coastline is
the final factor by which the convective regions
are divided into three categories: 1) more than
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
25 km offshore, 2) within 25 km of the coastline
either on or offshore, or 3) more than 25 km in-
land. These distances represent natural breaks in
the data. To examine the radial positions of the
convective regions, the percent of observations
in 100 km-wide annular rings extending outward
from the circulation center is calculated for each
condition-group. The distances of convective re-
gions that are greater than 500 km are combined.
To characterize the azimuthal distribution of
convection for each of the condition-groups, the
percent of observations occurring in each quadrant
of the storm is calculated. All quadrants are placed
relative to storm motion save for the examination
of data according to vertical wind shear; for this
condition, quadrants are placed according to the
direction of the shear vector.
To determine which of the six factors is most
strongly associated with large areas of convec-
tion, a second dataset is constructed so that each
observation time has only one entry. For each
observation time, the areas of all 40 dBZ polygons
over 500 km2 in size are summed. Variables repre-
senting each factor are then added for each entry.
For this dataset, the distance from the coastline
variable represents the distance of the circulation
center of each TC from the nearest point on the
coastline as calculated within the GIS. Spearman
rank correlation coefficients are then calculated to
determine if the largest areal extent of convection
corresponds to the largest or smallest values of the
six factors. The observations are then placed into
the same sub-groups for each condition as those
described for the centroid distance comparisons
save for the distance to the coastline. For this
condition, observations are placed into one of
four groups: 1) within 50 km, 2) 50-100 km, 3)
100-200 km, or 4) greater than 200 km from the
coastline. For each condition-group, the average
area covered by convection is calculated to de-
termine which variables are associated with the
largest area covered by convection, which might
indicate the largest potential for flooding to occur.
RESULTS AND DISCUSSION
The 43 TC landfalls examined in this study have
1545 regions of convection larger than 500 km2
during the 387 observation times (Figure 3). On
average, four convective polygons are present at
each observation time, but 36 observation times
contain no areas of convection larger than 500 km2.
Of the 43 landfalls examined, 25 occurred when
the TC was a tropical storm rather than a hurri-
cane. By twelve hours post-landfall, only Danny
(1997) and Frances (2004) remained hurricanes,
and most TCs were at tropical storm intensity.
Therefore, the majority of convective polygons
examined in the study occur when the TC is at
tropical storm intensity. Also, in nearly half of
the landfalls examined, TCs became extratropical
within 72 hours of landfall.
The discussion of each factor is presented in
order according to the strength of its association
with the location and areal extent of convection.
As storm motion exhibits the strongest relation-
ship to both the location and areal extent of con-
vection, this factor is discussed first in both
subsections. For the location of convection rela-
tive to storm center subsection, the factors are
ordered as follows: forward velocity, time until
extratropical transition, vertical wind shear, in-
tensity, time after landfall, and distance to coast-
line. In the areal extent of convection subsection,
the factors are presented as follows: forward
velocity, time until extratropical transition, inten-
sity, distance from the coastline, vertical wind
shear, and time after landfall.
Location of Convection
Relative to Storm Center
Results suggest that the forward velocity of the TC
has the largest effect on the location of convection
both in the azimuthal and radial directions during
the 24 hours after landfall. Overall, more convec-
tion occurs on the right side of the storm rather
than the left (Figure 4a), which agrees with the
1075
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Figure 3. Positions of the circulation centers of the 43 TC landfalls examined in the current study cat-
egorized by intensity, and the location of the centroid of areas of convection larger than 500 km2

findings of previous research (Frank & Ritchie,
1999; Corbosiero & Molinari, 2003; Lonfat et al.,
2004; Chen et al., 2006). Shapiro (1983) explains
that storm motion induces asymmetries in the
boundary layer frictional drag due to the stronger
winds on the right side of the vortex that result
when translation speed is added to the rotational
speed of the tangential winds. The current study
finds that in TCs moving at velocities under 5 m
s-1, convection is evenly split between the right
front and rear quadrants of the storm. When TCs
are moving at 5-10 m s-1, 60% of convective ob-
servations occur in the right front quadrant and
only 21% are located in the rear of the storm.
Almost 55% of convective observations for TCs
with forward velocities exceeding 10 m s-1 occur
in the left front quadrant, and 95% occur in the
forward half of the storm, indicating a counter-
clockwise shift in the location of convection as
forward velocity increases. This is similar to the
pattern described by Kimball (2008).
Forward velocity is also important when con-
sidering the distance from the circulation center
at which convection tends to occur. The correla-
tion between centroid distance and storm forward
velocity demonstrates that storm motion has the
strongest association with the radial distribution
of convection out of all six factors tested (Table
1). TCs moving at a fast forward velocity have
17% of convective observations located more
than 500 km from the circulation center, which
is the highest percentage of all other sub-groups
of factors examined (Table 2). The distribution
of convection is fairly similar among TCs with
forward velocities below 5 m s-1 and 5-10 m s-1.
For these storms, approximately 70-75% of ob-
servations occur within 300 km of the circulation
center. However, 52% of observations in TCs
moving at speeds greater than 10 m s-1 occur
beyond 300 km. This has important implications
for the findings of previous research as studies
examining convection only within 400 km of the
circulation center (e.g., Chen et al., 2006; Cecil,

1076
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Figure 4. Locations of polygon centroids relative to the circulation center of the storm based on a) storm
motion, b) time until becoming extratropical, c) vertical wind shear, d) intensity, e) hours since landfall,
and f) location relative to the coastline. All figures save c are oriented towards the direction of storm
motion with the right front quadrant at the top right of the figure. Panel c is oriented in the direction
of the vertical wind shear so that downshear right is at the top right of the figure. Only centroids for
convective regions in TCs that became extratropical are depicted in panel b. Annular rings are shown
every 100 km out to 800 km.

2007) may have missed as much as one third of
the convection associated with fast-moving TCs.
Even more importantly, studies examining data
within only 300 km of the circulation center (e.g.,
Molinari et al., 1999; Corbosiero & Molinari,
2002; Corbosiero & Molinari, 2003) may have
missed more than half of the convection that ex-
ists in fast-moving TCs.
The number of hours until a TC becomes
extratropical also exhibits an association with the
location of convection in TCs after landfall (Fig-
ure 4b). When a TC is within 48 hours of becom-

1077
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Table 1. Spearman Rank correlation coefficients

within left front quadrant increases only slightly.
for each convective region’s centroid distance,
However, in the next 24 hours, the amount of con-
and for the total areal extent of all convective
vection in the right (left) front quadrant decreases
regions at each observation time; Bold values
(increases) to 49 (42) %, completing the counter-
are statistically significant at α = 0.01
clockwise shift. Ritchie and Elsberry (2001) and
Centroid Distance Total Areal Extent Atallah and Bosart (2003) explain that this shift
(n=1545)a,b (n=387)c in the azimuthal location of convection occurs
Motion 0.235 0.524 when relatively cool and dry air that diminishes
Intensity -0.143 0.461 convection advects counterclockwise around the
Time until -0.189 a
0.274 circulation center beginning southwest of the TC.
Extratropical Meanwhile, convective precipitation is enhanced
Hour since 0.180 -0.252 to the north of the circulation center as the warm
Landfall
and moist air of tropical origin on the eastern side
Shear -0.057b -0.018c
of the TC is uplifted.
Distance to 0.034 0.013
Coastline The speed of the vertical wind shear appears
a
n = 908
to exert a stronger influence on the azimuthal
b
n = 1152 rather than the radial distribution of convection
c
n = 273 after landfall (Figure 4c). When all observations
are considered, most of the convection is located
in the downshear left and right quadrants, which
ing extratropical, convection decreases markedly
is in agreement with the findings of Corbosiero
in the inner 100 km, while 24 hours earlier, this
and Molinari (2002), Chen et al. (2006), and Cecil
innermost region contains the highest percentage
(2007). However, a clockwise shift in the position
of observations (Table 2). In the outer region of
of the convective regions is observed as vertical
the storm beyond 300 km, the number of convec-
wind shear increases. For TCs experiencing verti-
tive polygons increases from 14% to 43% of the
cal wind shear speeds less than 5 m s-1, 39 (28)% of
total number of observations during the same time
the observations occur in the downshear (upshear)
period (Table 2). These results suggest that a
left quadrant. When the velocity of the vertical
spreading out of convection occurs as a storm
wind shear increases to 5-10 m s-1, the number of
becomes extratropical. The negative correlation
observations increases (decreases) to 53 (17) %
between centroid distance and hours until the TC
in the downshear (upshear) left quadrant. When
becomes extratropical (Table 1) supports this
strong vertical wind shear is present, only 6% of
finding.
observations are located in both upshear quadrants,
Although the majority of the convective poly-
while 53 (36) % are located in the downshear left
gons remain in the right front quadrant regardless
(right) quadrants. This clockwise shift in convec-
of when or if an extratropical transition happens,
tion with increasing vertical wind shear has not
a shift in the counterclockwise direction towards
been discussed by previous researchers. Figures
the left front quadrant occurs as TCs approach the
from Corbosiero and Molinari’s (2002) study only
time that they become extratropical (Figure 4b).
indicate that the downshear quadrants experience
At 48-72 hours before becoming extratropical,
the most lightning flashes in both the inner 100
nearly 50 (25) % of convection is in the right front
km and outer 100-300 km regions of the storm
(rear) quadrant. Twenty-four hours later, the right
regardless of the strength of the vertical wind shear.
front (rear) quadrant contains 61 (10) % of the
While the speed of the vertical wind shear
convection, while the percentage of observations
appears to influence the azimuthal distribution

1078
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Table 2. Percent of polygon centroids representing convective regions contained within each 100 km
annulus outward from the storm center, and the total number of centroids per condition-group. Bold
values indicate the distances with the highest percentage of observations.

Analysis n 0-100km 100-200km 200-300km 300-400km 400-500km 500+ km

Motion
< 5 ms-1 566 21 33 24 14 5 3
5-10 ms-1 736 17 29 23 16 8 8
> 10 ms -1
234 12 16 21 20 15 17
Time until Extratropical
0 - 24 hrs 466 12 23 23 23 12 9
24 - 48 hrs 297 13 26 27 16 7 11
48 - 72 hrs 145 34 28 23 10 1 3
None 527 22 34 21 12 6 6
Shear
< 5 ms-1 132 18 33 23 12 8 6
5-10 ms -1
306 16 31 29 12 7 6
> 10 ms -1
714 21 31 21 14 6 7
Intensity
Hurricane 288 28 26 20 8 8 9
Trop. Storm 870 17 31 23 16 6 7
Trop. Dep. 387 10 24 25 21 12 8
Hours since Landfall
0 200 24 32 26 9 4 7
3 203 27 32 16 12 5 8
6 193 21 34 20 12 5 8
9 200 17 24 26 20 8 5
12 180 17 26 27 17 8 5
15 161 11 26 22 21 14 6
18 157 12 29 20 12 15 11
21 134 12 26 28 19 4 10
24 117 11 21 24 26 9 9
Coastline Proximity
Offshore 299 7 28 26 17 10 12
Near Coast 291 24 27 22 15 7 4
Inland 933 19 29 23 16 7 6

of convection, it has a much weaker associa-
tion with the radial distribution of convection as
compared to storm motion. Approximately 30%
of convective observations are located 100-200
km from the circulation center, and less than 20%
are located in either the inner core, or beyond 400
km regardless of the velocity of the vertical wind
shear. This suggests that stronger wind shear values
do not result in the development or displacement
of convection farther from the circulation center.

1079
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
The lack of a statistically significant correlation
between the velocity of the vertical wind shear
and the distance of the convective observations
relative to the storm center (Table 1) supports
this finding.
Convection is present in both the core and outer
rainbands of hurricanes, while weaker TCs have
little convection present within their core (Table
2). Hurricanes have nearly the same number of
convective regions within 100 km of the circula-
tion center, which includes convection found in the
eyewall, as they do in the region 100-200 km from
the circulation center (Table 2), where the edge of
the principal rainband is often located (Willoughby
et al., 1984; Molinari et al., 1999). Tropical storms
have a clear maximum in convection 100-200 km
from the circulation center. Tropical storms have
more convection in the storm core than do tropi-
cal depressions, but much less convection near
the core of the storm than do hurricanes (Table
2). Twice as many convective observations are
located 300-400 km from the circulation centers
of tropical depressions than within the innermost
100 km, indicating a lack of convection near the
core of tropical depressions (Table 2). The negative
correlation coefficient between intensity and the
distance of the observations from the circulation
center indicates that higher intensity correlates to
a closer centroid distance (Table 1). This result
suggests that convection spreads outwards as the
TCs weaken. As the majority of the observations
occur on the right side of the storm (Figure 4d)
in all three sub-groups, intensity does not appear
have as strong of an association with the quadrant
in which convection is located as does storm mo-
tion and vertical wind shear.
When the locations of convection are examined
based upon the time that has elapsed since landfall,
the spreading out of convection with time that is
discussed by Kimball (2008) is evident. However,
the outward spread of convection does not appear
to happen steadily over time. The radial distribu-
tion of convective polygons is similar during the
first six hours post-landfall, when approximately
1080
55% are located within 200 km of the circulation
center and only 12% are located beyond 400 km
(Table 2). From 9 to 12 hours post-landfall, con-
vection diminishes in storm core, but remains the
same in the region beyond 400 km. Over the next
6 hours, the percentage of observations located
over 400 km away from the storm center doubles.
Towards the end of the 24-hour period, the amount
of convection in the farthest distances diminishes,
but the distance of most convective observations
shifts outwards from 200-300 km to 300-400 km.
A counterclockwise shift in the location
of convection is not observed when time after
landfall is the only variable considered (Figure
4e). The number of convective regions in each
motion-relative quadrant remains fairly consistent
for the first 21 hours post-landfall. The percent
of observations occurring in the right front, right
rear, left rear, and left front quadrants are approxi-
mately 50%, 20%, 10%, and 20% respectively.
However, at 24 hours post-landfall, the maximum
shifts to the right rear quadrant, with 32% of the
observations. Most of these observations belong
to slow-moving tropical depressions, and slow-
moving TCs tend to have more convection present
in the right rear quadrant of the storm than in the
left front quadrant, possibly due to the advection
of moisture from the ocean (Bluestein & Hazen,
1989). Analysis of additional hours post-landfall
is required to determine if a maximum in the left
front quadrant emerges at a later time.
The distance at which convection forms relative
to the coastline appears to have some association
with the azimuthal distribution of convection rela-
tive to the circulation center of the storm. Figure
4a shows that convection exists in the left rear
quadrants of some TCs despite this area being
the most unfavorable location for the formation
of convection given that downward vertical mo-
tion is enhanced on the left side of a TC (Powell,
1990). While many TCs track inland after landfall,
some travel across the peninsula of Florida (e.g.,
Jeanne (2004) and Erin (1995)) or track parallel
to the U.S. coastline (Floyd (1999) and Hanna
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Table 3. The average areal extent of convection

(2008)). Thus, convection located more than 25
for each condition-group and the number of ob-
km offshore may occur in any motion-relative
servations contained within each condition-group;
quadrant of the storm. In this study, over 50%
Bold values indicate group with the largest areal
of the offshore convection is located in the rear
coverage of convection for each condition
quadrants of the TCs (Figure 4f). The 45% of
offshore observations located in the right front Analysis n Average Area (km2)
quadrant are associated with TCs making land- Motion
fall along the east coast of the U.S. and moving < 5 ms-1 175 5842
towards the north, for example Floyd (1999) and 5-10 ms-1 149 10340
Hanna (2008). Offshore observations located the
> 10 ms -1
32 29975
farthest from the circulation center in the right
Time until Extratropical
front quadrant exist in TCs that are nearing the
0 - 24 hrs 87 18994
time at which they become extratropical.
24 - 48 hrs 51 13407
Results from the analysis of the near-shore
48 - 72 hrs 33 6307
observations suggest that the coastline does exert
none 176 5253
an influence on the azimuthal location of con-
Intensity
vection as TCs make landfall and move inland.
Hurricane 56 13039
Previous studies have discussed how frictional
convergence produced along the coastline along Trop. Storm 178 11676

the right side of the storm enhances the develop- Trop. Dep. 121 5849

ment of convection as a TC makes landfall (e.g., Coastline Proximity

Jones, 1987; Frank & Ritchie, 1999). However, < 50 km 149 14773

more than 35% of the observations occur in the 50-100 km 67 6989

two rear quadrants of the storm, suggesting that 100-200 km 69 5787
the coastline influences convective precipitation > 200 km 65 6281
even after the circulation center has moved in- Shear
land. Additionally, since convection occurs near < 5 ms-1 45 4568
the coastline more than 300 km away from the 5-10 ms -1
110 4781
circulation center, coastal areas not affected by > 10 ms-1 118 12273
the strongest winds in a landfalling TC may still Hours since Landfall
experience heavy precipitation. 0 43 9056
3 43 9140
Areal Extent of Convection 6 43 10248
9 43 9973
The largest regions of convection are associated 12 40 11611
with fast-moving TCs that are nearing the time
15 38 10816
of transition to extratropical and located within
18 37 11119
50 km of the coastline. Of these three factors,
21 33 10676
storm motion has the strongest relationship with
24 34 6356
the areal coverage of convection (Table 1). TCs
that are moving at speeds exceeding 10 m s-1 have
larger areas covered by convection on average
than any of the other sub-groups examined (Table
3). The nearly 30,000 km2 average areal coverage

1081
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Figure 5. The total area covered by convection at each observation time compared to: (a) storm motion,
(b) time until becoming extratropical, (c) intensity, (d) location of storm circulation center relative to
the coastline, (e) vertical wind shear, and f) time after landfall.

of convection for fast-moving TCs is 20,000 km2
more than for TCs moving at 5-10 m s-1, and ap-
proximately 10,000 km2 more than for TCs that
are within 24 hours of becoming extratropical.
The top ten largest convective areas occur in TCs
with forward velocities greater than 8 m s-1, and
very few TCs that are moving at this speed have
small areas of convection (Figure 5a).
Interaction with a middle latitude trough is
what typically causes storm motion to increase,
and is also what causes the TC to become extra-
tropical (Jones et al., 2003). All of the observation
times that feature TCs moving at speeds greater
than 10 m s-1 also occur when TCs are within 24
hours of becoming extratropical. However, all
TCs within 24 hours of becoming extratropical
do not necessarily move at fast forward velocities.
Nearly 60% of the observations from this group
occur when forward velocity is below 10 m s-1.
As a TC becomes extratropical, the surrounding
environment becomes baroclinic and the isentro-
pic ascent of the moist air mass associated with
the TC enhances precipitation and causes the
region within which heavy rainfall occurs to in-
crease in area (Atallah & Bosart, 2003; Jones et
al., 2003). As a group, TCs that are within 24
hours of being extratropical have the second larg-
est areas of convection on average (Table 3). The
ten largest areas of convection are associated with
Floyd (1999), Hanna (2008), and Bertha (1996),
which moved towards the north parallel to the
east coast of the U.S. (Figure 3) at speeds greater
than 9 m s-1 as they transitioned into extratropical
storms. The average area of convection increases
for TCs as the time until their extratropical tran-
sition (Table 3). The statistically significant cor-

1082
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
relation coefficient between total areal extent of
convection and time until becoming extratropical
(Table 1) also supports the finding that extra-
tropical transition results in larger regions of
convection. For TCs that do not experience an
extratropical transition, convection develops over
an area of approximately 22,000 km2 in the larg-
est case (Figure 5b), but convection is more com-
monly confined to an area measuring approxi-
mately 5000 km2 (Table 3).
The second largest correlation with convective
area is the maximum sustained wind speed (Table
1), indicating that the area covered by convection
is largest for hurricanes and decreases as TCs
weaken. Although the five largest regions of
convection belong to tropical storms (Figure 5c),
numerous regions smaller than 1000 km2 exist for
both tropical storms and tropical depressions. In
hurricanes, however, the smallest region is more
than 2100 km2 in size. Hurricanes have convec-
tion in both their core and in their outer rainbands
so that greater total area is covered by convec-
tive rainfall than for weaker TCs. This finding
is consistent with that of Shepherd et al. (2007).
Also contributing to the large spatial coverage of
convection for hurricanes is their location close
to the coastline. When the TC is located inland
and the tangential winds are weaker, it is more
difficult to advect moisture from the ocean into
the circulation to sustain heavy rainfall (Bluestein
& Hazen, 1989). The exception to the trend of
diminishing convection as maximum sustained
winds weaken after landfall occurs when TCs
become extratropical. Many of these TCs are at
tropical storm intensity as the size of their con-
vective area grows.
As a group, TCs located within 50 km of the
coastline have the third largest areal extent of
convection out of the 26 sub-groups examined
(Table 3). Once a TC is more than 100 km inland,
no areas of convection larger than 25,000 km2 are
present (Figure 4d). This is likely because less
moisture is available over the land surface for
latent heat flux (Tuleya, 1994; Kimball, 2008),
and continental air masses surrounding the TC
have less available moisture to enhance rainfall
(Bluestein & Hazen, 1989; Cubukcu et al., 2000).
As previously discussed, many of the TCs that
remain within 100 km of the coastline several
hours post-landfall experience an extratropical
transition that helps to increase the area over
which convective rainfall occurs. Despite the clear
tendency for inland TCs to have smaller convec-
tive rainfall regions, a statistically significant
correlation does not exist between a TC’s distance
from the coastline and the area of its convection
(Table 1). When TCs are close to the coastline,
areas of strong convection present in the eyewalls
of hurricanes as well as convection enhanced by
friction along the coastline in the outer rainbands
have a limited spatial extent, which accounts for
many of the regions of convection that occupy
less than 10,000 km2 visible in Figure 5d.
TCs experiencing vertical wind shear velocities
above 10 m s-1 have a larger area of convection
on average than do TCs experiencing less shear
(Table 3, Figure 5e). Nine of the ten largest con-
vective areas for highly-sheared TCs are associ-
ated with Floyd (1999) and Bertha (1996) when
these TCs were approximately 20-35 hours from
becoming extratropical and were located within
50 km of the coastline. Vertical wind shear typi-
cally increases during an extratropical transition
(Jones et al., 2003), and moisture advection from
the nearby ocean also contributes to the large area
of rainfall associated with these highly-sheared
TCs. However, strong vertical wind shear is not
solely associated with an extratropical transition,
as nearly 40% of the observations for highly-
sheared storms occur in TCs that do not become
extratropical within 72 hours of landfall. Many
of these highly-sheared TCs that do not become
extratropical have relatively small areas of convec-
tion surrounding their circulation centers as they
track inland. Thus, a significant relationship does
not exist between the speed of vertical wind shear
and the area covered by convection (Table 1).
1083
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
A negative correlation between the areal extent
of convection and the time elapsed since landfall
indicates that convection decreases in area over
time (Table 1). The two highest averaged areas of
convection occur at 12 and 18 hours post-landfall
(Figure 5f), rather than at the time of landfall, which
has the second smallest average area of convec-
tion (Table 3). At the time of landfall, convection
tends to be organized into smaller regions in the
eyewalls of hurricanes and/or the outer rainbands
of hurricanes and weaker TCs. It is the dynamics
associated with an extratropical transition that
organizes these small regions of convection into
larger regions and expands the overall size of the
rain fields of the TCs. However, the overall size of
the convective areas tends to decrease beginning
approximately 12 hours prior to the TC becoming
fully extratropical (Figure 5b), and this marked
decrease in convection likely accounts for the
negative correlation between the areal extent of
convection and time elapsed since landfall.
CONCLUSION
A GIS was utilized to perform a spatial analysis
of radar reflectivity returns to determine where
regions of convective rainfall are located relative
to the circulation centers of TCs, and how large an
area was covered by these regions. The analysis
was performed for 3-hourly periods during the first
24 hours after 43 TC landfalls. The radial and azi-
muthal positions of each convective region greater
than 500 km2 in area, and the total areal extent of
convection at each observation time, were then
examined relative to six factors. These factors were
storm intensity, vertical wind shear, storm motion,
whether the TC became extratropical, the number
of hours elapsed since landfall, and the distance
of the convective areas as well as the circulation
center of the TC from the coastline.
This study supports the results of previous
modeling and observational studies that suggest
convection shifts counterclockwise and spreads
1084
outwards from the circulation center of TCs after
landfall. Three factors are associated with a coun-
terclockwise shift in convection from the right side
to the front of the storm: a) increasing forward
velocity, b) decreasing time until extratropical
transition, and c) convection forming inland rather
than offshore. However, increasing vertical wind
shear shifts convection in the clockwise direction.
The spreading outwards of convection is associated
with a) decreasing storm intensity, b) increasing
storm forward velocity, and c) decreasing time
until a TC becomes extratropical.
Among the six factors examined in the current
study, the forward velocity of the TC in conjunction
with an extratropical transition had the strongest
association with the location of convection both in
the radial and azimuthal directions relative to the
circulation center of the storm, and also the larg-
est influence on the area over which convection
extends. As the forward velocity of TCs increases
due to interaction with a middle latitude trough,
areas of convection grow in size, are located farther
from the circulation center of the storm, and shift
counterclockwise from the right side of the storm
when motion is slow to the front of the storm when
motion is fast. The dynamical changes in the struc-
ture of a TC experiencing an extratropical transition
act to decrease convection behind and in the core
of the storm while increasing convection ahead of
the storm so that the same pattern of shifting and
growing convection is observed whether examin-
ing TCs according to their forward velocity or the
time until they become extratropical.
Previous studies suggested that vertical wind
shear is the dominant influence on the location of
convection within TCs, but the current study found
that shear did not have as strong of an association
with the location of and area covered by convection
as did storm motion and extratropical transition. As
found by previous researchers, a clear maximum of
convection in the downshear left quadrant convec-
tion was observed in the current study. However,
a clockwise shift in the region where convection
formed from upshear left to downshear right was
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
observed as the velocity of the shear increased,
which is a pattern that has not been discussed by
previous researchers. Although the areas of convec-
tion were larger on average when TCs experienced
high shear rather than shear with a velocity less
than 10 m s-1, the speed of the vertical wind shear
did not significantly alter the radial position of the
convection. These results suggest that the tendency
to base the forecasted location of convective rainfall
primarily on the direction and speed of the vertical
wind shear may need to be reassessed.
This study analyzes a large number of TCs in
a GIS framework. To more accurately determine
specific locations on the ground that may experi-
ence convective rainfall from a TC, and to more
effectively link the location of convective rainfall
to environmental variables, future work should also
utilize the GIS to quantify the shape and orienta-
tion of the convective regions. Adding these spatial
attributes to the size and location data presented
in the current study will allow a complete set of
spatial attributes to be cataloged for each convec-
tive region. Analyzing these shapes at hourly or
half-hourly intervals would facilitate the tracking
of each convective region as it develops so that
changes in its spatial attributes can be calculated
as the TC tracks inland. Additional factors that may
affect the development and displacement of convec-
tion such as the angle at which a TC crosses the
coastline, topography, and moisture present within
the atmosphere and soil, will also be considered so
that these observational results can be compared to
modeling studies that examine similar variables.
The results of these future analyses could then be
utilized to validate the spatial representation of
convective rainfall regions produced by dynami-
cal models such as the HWRF (Davis et al., 2008)
and GFDL (Marchok et al., 2007), and statistical
predictions of TC rainfall produced by models such
as R-CLIPER (Marchok et al., 2007), TRaP (Kid-
der et al., 2005), and PHRaM (Lonfat et al., 2007).
Future research should also consider the con-
tribution of stratiform rainfall to the total rainfall
produced by TCs as they move over land. As
convection diminishes, stratiform rainfall should
comprise a larger portion of the rain fields as the
TC moves inland. The techniques utilized in the
current study should be applied to radar reflectiv-
ity regions having reflectivity values less than 40
dBZ to quantify the position and areal coverage
of stratiform rainfall within the TC rain fields.
The relative contribution of both stratiform and
convective rainfall to the overall areal coverage of
TC rain fields could then be calculated.
ACKNOWLEDGMENT
Three anonymous reviewers are thanked for their
helpful comments that improved the quality of
this manuscript.
REFERENCES
AMS. (2007). Hurricane forecasting in the United
States. Bulletin of the American Meteorological
Society, 88, 87–92.
Anagnostou, E. N. (2004). A convective/stratiform
precipitation classification algorithm for volume
scanning weather radar observations. Meteoro-
logical Applications, 11, 291–300. doi:10.1017/
S1350482704001409
Atallah, E. H., & Bosart, L. R. (2003). The extra-
tropical transition and precipitation distribution
of Hurricane Floyd (1999). Monthly Weather
Review, 131, 1063–1081. doi:10.1175/1520-
0493(2003)131<1063:TETAPD>2.0.CO;2
Bluestein, H. B., & Hazen, D. S. (1989). Dop-
pler-radar analysis of a tropical cyclone over
land - Hurricane Alicia (1983) in Oklahoma.
Monthly Weather Review, 117, 2594–2611.
doi:10.1175/1520-0493(1989)117<2594:DRAO
AT>2.0.CO;2
1085
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
Burpee, R. W., & Black, M. L. (1989). Temporal
and spatial variations of rainfall near the centers
of two tropical cyclones. Monthly Weather
Review, 117, 2204–2218. doi:10.1175/1520-
0493(1989)117<2204:TASVOR>2.0.CO;2
Cecil, D. J. (2007). Satellite-derived rain rates in
vertically sheared tropical cyclones. Geophysical
Research Letters, 342), L02811-L02821.
Chen, S. Y. S., Knaff, J. A., & Marks, F. D. (2006).
Effects of vertical wind shear and storm motion
on tropical cyclone rainfall asymmetries deduced
from TRMM. Monthly Weather Review, 134,
3190–3208. doi:10.1175/MWR3245.1
Corbosiero, K. L., & Molinari, J. (2002). The ef-
fects of vertical wind shear on the distribution of
convection in tropical cyclones. Monthly Weather
Review, 130, 2110–2123. doi:10.1175/1520-
0493(2002)130<2110:TEOVWS>2.0.CO;2
Corbosiero, K. L., & Molinari, J. (2003). The rela-
tionship between storm motion, vertical wind shear,
and convective asymmetries in tropical cyclones.
Journal of the Atmospheric Sciences, 60, 366–376.
doi:10.1175/1520-0469(2003)060<0366:TRBS
MV>2.0.CO;2
Cubukcu, N., Pfeffer, R. L., & Dietrich, D. E.
(2000). Simulation of the effects of bathymetry
and land-sea contrasts on hurricane development
using a coupled ocean-atmosphere model. Jour-
nal of the Atmospheric Sciences, 57, 481–492.
doi:10.1175/1520-0469(2000)057<0481:SOTE
OB>2.0.CO;2
Davis, C., Wang, W., Chen, S. S., Chen, Y. S.,
Corbosiero, K., & DeMaria, M. (2008). Predic-
tion of landfalling hurricanes with the Advanced
Hurricane WRF model. Monthly Weather Review,
136, 1990–2005. doi:10.1175/2007MWR2085.1
DeMaria, M., & Kaplan, J. (1994). A statistical hur-
ricane intensity prediction scheme (SHIPS) for the
Atlantic basin. Weather and Forecasting, 9, 209–
220. doi:10.1175/1520-0434(1994)009<0209:AS
HIPS>2.0.CO;2
1086
DeMaria, M., Mainelli, M., Shay, L. K., Knaff, J.
A., & Kaplan, J. (2005). Further improvements
to the statistical hurricane intensity prediction
scheme (SHIPS). Weather and Forecasting, 20,
531–543. doi:10.1175/WAF862.1
Elsberry, R. L. (2002). Predicting hurricane land-
fall precipitation: Optimistic and pessimistic views
from the symposium on precipitation extremes.
Bulletin of the American Meteorological Society,
83, 1333–1339.
Frank, W. M., & Ritchie, E. A. (1999). Effects of
environmental flow upon tropical cyclone struc-
ture. Monthly Weather Review, 127, 2044–2061.
doi:10.1175/1520-0493(1999)127<2044:EOEFU
T>2.0.CO;2
Geerts, B., Heymsfield, G. M., Tian, L., Halv-
erson, J. B., Guillory, A., & Mejia, M. I. (2000).
Hurricane Georges’s landfall in the Dominican
Republic: Detailed airborne Doppler radar
imagery. Bulletin of the American Meteoro-
logical Society, 81, 999–1018. doi:10.1175/1520-
0477(2000)081<0999:HGLITD>2.3.CO;2
Hill, K., & Lackmann, G. M. (2009). Influence
of environmental humidity on tropical cyclone
size. Monthly Weather Review, 137, 3294–3315.
doi:10.1175/2009MWR2679.1
Jones, R. W. (1987). A simulation of hurricane
landfall with a numerical-model featuring latent
heating by the resolvable scales. Monthly Weather
Review, 115, 2279–2297. doi:10.1175/1520-
0493(1987)115<2279:ASOHLW>2.0.CO;2
Jones, S. C., Harr, P. A., Abraham, J., Bosart, L. F.,
Bowyer, P. J., & Evans, J. L. (2003). The extratropi-
cal transition of tropical cyclones: Forecast chal-
lenges, current understanding, and future direc-
tions. Weather and Forecasting, 18, 1052–1092.
doi:10.1175/1520-0434(2003)018<1052:TETOT
C>2.0.CO;2
A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall
Jorgensen, D. P. (1984). Mesoscale and convec-
tive-scale characteristics of mature hurricanes. Part
I: General observations by research aircraft. Jour-
nal of the Atmospheric Sciences, 41, 1268–1285.
doi:10.1175/1520-0469(1984)041<1268:MACS
CO>2.0.CO;2
Kidder, S. Q., Kusselson, S. J., Knaff, J. A., Ferraro,
R. R., Kuligowski, R. J., & Turk, M. (2005). The
tropical rainfall potential (TRaP) technique. Part
I: Description and examples. Weather and Fore-
casting, 20, 456–464. doi:10.1175/WAF860.1
Kimball, S. K. (2008). Structure and evolution of
rainfall in numerically simulated landfalling hurri-
canes. Monthly Weather Review, 136, 3822–3847.
doi:10.1175/2008MWR2304.1
Konrad, C. E., Meaux, M. F., & Meaux, D. A.
(2002). Relationships between tropical cyclone
attributes and precipitation totals: Considerations
of scale. International Journal of Climatology,
22, 237–247. doi:10.1002/joc.721
Lawrence, M. B., Avila, L. A., Beven, J. L., Frank-
lin, J. L., Guiney, J. L., & Pasch, R. J. (2001). At-
lantic hurricane season of 1999. Monthly Weather
Review, 129, 3057–3084. doi:10.1175/1520-
0493(2001)129<3057:AHSO>2.0.CO;2
Lonfat, M., Marks, F. D., & Chen, S. Y. S. (2004).
Precipitation distribution in tropical cyclones
using the Tropical Rainfall Measuring Mission
(TRMM) Microwave Imager: A global perspec-
tive. Monthly Weather Review, 132, 1645–1660.
doi:10.1175/1520-0493(2004)132<1645:PDITC
U>2.0.CO;2
Lonfat, M., Rogers, R., Marchok, T., & Marks,
F. D. (2007). A parametric model for predicting
hurricane rainfall. Monthly Weather Review, 135,
3086–3097. doi:10.1175/MWR3433.1
Marchok, T., Rogers, R., & Tuleya, R. (2007).
Validation schemes for tropical cyclone quantita-
tive precipitation forecasts: evaluation of opera-
tional models for US Landfalling cases. Weather
and Forecasting, 22, 726–746. doi:10.1175/
WAF1024.1
Matyas, C. J. (2008). Shape measures of rain
shields as indicators of changing environmental
conditions in a landfalling tropical storm. Meteo-
rological Applications, 15, 259–271. doi:10.1002/
met.70
Medlin, J. M., Kimball, S. K., & Blackwell,
K. G. (2007). Radar and rain gauge analysis of
the extreme rainfall during Hurricane Danny’s
(1997) landfall. Monthly Weather Review, 135,
1869–1888. doi:10.1175/MWR3368.1
Molinari, J., Moore, P., & Idone, V. (1999). Convec-
tive structure of hurricanes as revealed by lightning
locations. Monthly Weather Review, 127, 520–
534. doi:10.1175/1520-0493(1999)127<0520:CS
OHAR>2.0.CO;2
Office of the Federal Coordinator of Meteorology.
(2006). Federal Meteorological Handbook, No. 11:
Doppler Radar Meteorological Observations- Part
C: WSR-88D Products and Algorithms. Washing-
ton, DC: U.S. Department of Commerce/ National
Oceanic and Atmospheric Administration.
Parrish, J. R., Burpee, R. W., Marks, F. D., & Grebe,
R. (1982). Rainfall patterns observed by digitized
radar during the landfall of Hurricane Frederic
(1979). Monthly Weather Review, 110, 1933–1944.
doi:10.1175/1520-0493(1982)110<1933:RPOBD
R>2.0.CO;2
Powell, M. D. (1990). Boundary-layer structure
and dynamics in outer hurricane rainbands. 1. Me-
soscale rainfall and kinematic structure. Monthly
Weather Review, 118, 891–917. doi:10.1175/1520-
0493(1990)118<0891:BLSADI>2.0.CO;2
1087
 A Geospatial Analysis of Convective Rainfall Regions within Tropical Cyclones after Landfall

Rappaport, E. N. (2000). Loss of life in the United
States associated with recent Atlantic tropical
cyclones. Bulletin of the American Meteorologi-
cal Society, 81, 2065–2073. doi:10.1175/1520-
0477(2000)081<2065:LOLITU>2.3.CO;2
Rodgers, E. B., Baik, J. J., & Pierce, H. F.
(1994). The environmental influence on tropi-
cal cyclone precipitation. Journal of Applied
Meteorology, 33, 573–593. doi:10.1175/1520-
0450(1994)033<0573:TEIOTC>2.0.CO;2
Rosenfeld, D., Wolff, D. B., & Atlas, D. (1993).
General probability-matched relations between
radar reflectivity and rain rate. Journal of Ap-
plied Meteorology, 32, 50–72. doi:10.1175/1520-
0450(1993)032<0050:GPMRBR>2.0.CO;2
Schwarz, F. K. (1970). Unprecedented rains in
Virginia associated with remnants of Hurricane
Camille. Monthly Weather Review, 98, 851–859.
doi:10.1175/1520-0493(1970)098<0851:TURIV
A>2.3.CO;2
Shapiro, L. J. (1983). The asymmetric boundary-
layer flow under a translating hurricane. Journal
of the Atmospheric Sciences, 40, 1984–1998.
doi:10.1175/1520-0469(1983)040<1984:TABL
FU>2.0.CO;2
Shepherd, J. M., Grundstein, A., & Mote, T. L.
(2007). Quantifying the contribution of tropical
cyclones to extreme rainfall along the coastal
southeastern United States. Geophysical Research
Letters, 34, L23810. doi:10.1029/2007GL031694
Tuleya, R. E. (1994). Tropical storm development
and decay: Sensitivity to surface boundary-con-
ditions. Monthly Weather Review, 122, 291–304.
doi:10.1175/1520-0493(1994)122<0291:TSDA
DS>2.0.CO;2
Wilks, D. S. (1995). Statistical methods in the
atmospheric sciences. San Diego, CA: Academic
Press.
Willoughby, H. E., Marks, F. D., & Feinberg,
R. J. (1984). Stationary and moving convective
bands in hurricanes. Journal of the Atmospheric
Sciences, 41, 3189–3211. doi:10.1175/1520-
0469(1984)041<3189:SAMCBI>2.0.CO;2

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 1, Issue 2,
edited by Donald Patrick Albert, pp. 71-91, copyright 2010 by IGI Publishing (an imprint of IGI Global).

1088
 Section 5
Organizational and Social
Implications

This section includes a wide range of research pertaining to the social and behavioral impact of Geographic In-
formation Systems around the world. Chapters introducing this section critically analyze and discuss trends in
Geographic Information Systems, such as historical GIS, volunteered GIS, and healthcare applications of GIS.
Additional chapters included in this section look at collaborative mapping, agriculture, and conservation. Also
investigating a concern within the field of Geographic Information Systems is research which discusses the effect
of trust and security on Geographic Information Systems. With 19 chapters, the discussions presented in this sec-
tion offer research into the integration of global Geographic Information Systems as well as implementation of
ethical and workflow considerations for all organizations.
1090

Chapter 65
The Digital Geography Lab
at Salem State University:
The Evolution of One of the Oldest
Educational Digital Geospatial Labs

Kym Pappathanasi
Salem State University, USA

Stephen Young
Salem State University, USA

ABSTRACT
Established in 1983, the Digital Geography Lab (DGL) at Salem State University (SSU) is one of the
oldest higher-education digital spatial labs. This paper details the evolution of the DGL as well as its
current status as one of the best computer labs at Salem State University. This paper describes the chang-
ing technology of the DGL over the past 28 years. The DGL not only provides SSU students with cutting
edge geospatial technology, but it is a research lab for faculty and is extensively used in community out-
reach programs. The DGL is being used more and more by other departments such as Geology, Criminal
Justice, Computer Science and Inter-disciplinary Studies which reflects the expansion of the geo-spatial
sciences beyond geography. The success of the DGL lies in part with the University’s management of
the facility where major decisions are made in consultation with the Department and DGL staff. The
configuration and technology of the DGL has changed considerably through the years, reflecting the
transformations in technology and educational philosophies, but the core mission of the DGL has not,
which is to serve the students, faculty and the local community with cutting edge geospatial technology.

INTRODUCTION Laboratory (DGL) (Figure 1). The DGL is one of

Located on the third floor of Meier Hall is one
of Salem State University’s most outstanding
technology resources – the Digital Geography
DOI: 10.4018/978-1-4666-2038-4.ch065
the oldest geo-spatial university labs in the United
States. It was officially established in 1983 and a
few years later in 1986 it was proclaimed a Center
of Educational Excellence by then Massachusetts
Governor Dukakis (RSSG, 2003). As a Center of

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
The Digital Geography Lab at Salem State University
Figure 1. Frederick A. Meier Arts and Sciences
Hall, Salem State University, Salem, Massachu-
setts
Educational Excellence it received special govern-
ment funding for its development of geo-spatial
computing technology for the education of Salem
State students. Salem State University has also
continued to support the lab financially and has
allowed the policy and configuration decisions to
be made unilaterally by the geography department.
This hands-off policy has allowed the lab to grow
and service its students with the primary goals of
geo-computing and geographic education.
The DGL is an exceptional resource for educa-
tion and research at Salem State University (SSU).
All geography majors, minors, and students who
are enrolled in geography courses, are provided
accounts in the DGL. In addition, most geography
classes require students to perform certain labs
and exercises which are only available through
the software provided by the DGL. For example,
Introduction to Geography students map the dis-
tribution of different ethnic groups across the
United States, analyze population growth in India
and run development scenarios for urban growth
in Colorado. Remote sensing students download
satellite data from NASA to analyze deforestation,
global vegetation change and brush fires in Aus-
tralia as well as the melting of glaciers on Mount
Kilimanjaro.
In addition to undergraduate students in the
College of Arts and Sciences, the DGL services
students in the Master of Geo-Information Sci-
ences (GIS) program. Salem State University’s
Master of Geo-Information Sciences is also one
of the oldest in the country with the first gradu-
ate students arriving in 1993. Some DGL related
projects undertaken by masters students include:
archeological mapping, GPS accuracy assessment
(global positioning systems), GPS for nature re-
serve mapping, remote sensing for bird migration,
remote sensing for wind power development, and
environmental justice analysis of power generation
in Massachusetts.
Our graduates use their geo-spatial technical
expertise to secure positions in a variety of govern-
ment, higher-education, not-for-profit and private
sector jobs throughout Massachusetts and beyond.
Federal government agencies where our alumni
work include the FBI, Secret Service, National
Geospatial-Intelligence Agency, NASA and the
Census Bureau. Private companies include ESRI,
Tactician, Applied Geographics, Trimble, and Pic-
tometry. For a more extensive list of our alumni’s
employment please visit our DGL Alumni (2011)
web page at dgl.salemstate.edu/alumni.shtml.
In addition to class-related work, the DGL
provides state-of-the-art facilities for researchers
at Salem State University. Faculty, graduate stu-
dents and undergraduates all use the DGL for their
research needs. Research by faculty encompasses
a wide breadth of areas from estimating floral
intensity using high resolution aerial imagery, to
demographic analysis of voter participation. Other
research includes emergency response mapping,
sea-level rise mapping, transportation model-
ing and land cover change. (Luna, 2010, 2008a,
2008b; Ratner, 2009a, 2009b, 2008, 2007a, 2007b;
Young, 2011a, 2011b, 2004; Young & Harris,
2005). Each year geography faculty and students
attend regional and national conferences where
they present their research. In the past both under-
graduate and master’s students have won national
research competitions held by the Association of
1091
 The Digital Geography Lab at Salem State University
American Geographers (Remote Sensing, GIS,
and Cartography Specialty Groups Illustrated
Poster Competition: Graduate First Place: 2010:
Moumita Duttagupta; Undergraduate First Place
2005: Michelle Roy; 2000: Kevin Coughlin,
Honorable Mentions: 2004, 2006, 2009).
OUTREACH WITHIN AND OUTSIDE
THE DEPARTMENT OF GEOGRAPHY
The DGL is an academic departmental lab which
operates specifically to support the geography
department’s curriculum. The lab has always had
a university-wide open door policy, and in recent
years we have seen increased usage by the geo-
logical sciences department due to the need for
GIS in their field. In addition, other departments
are starting to use GIS applications for research,
such as biology and nursing. More academic
departments are utilizing the DGL which reflects
the expansion of the geo-spatial sciences beyond
geography. This semester (spring 2011) the DGL
will support 58 course sections (873 course reg-
istrations). These courses are offered through the
Departments of Geography, Geological Sciences,
Inter-disciplinary Studies, and Computer Science
at undergraduate and graduate course levels, and
includes courses ranging from Introduction to
Geography, Weather and Climate, and Hydrology
to Advanced GIS, Digital Image Processing and
Spatial Database Design/Analysis.
Through the outreach of the department’s
faculty, the DGL is also involved in community
programs. With the recent expansion of the lab,
it has been possible to invite regional middle and
high school students, as well as the Massachusetts
branch of Women in Science and Engineering
(WISE) program, to use the facility with the
intent of introducing pre-college students to the
geo-sciences. The DGL is also part of Salem State
University’s Center for Economic Development
and Sustainability, running workshops and provid-
1092
ing research support as necessary to the ongoing
projects of the center.
While department enrollments in recent years
have recently declined, the need for access to
specialized geo-spatial software applications
has remained steady due to more classes using
specialized software in their coursework (Fig-
ure 2). Many of the geography-based software
programs are beyond the means of most college
students, and so through the DGL, Salem State
students can learn to use the most advanced and
cutting-edge technology in their field. Starting in
the fall of 2005 the university started requiring
all incoming students to purchase a laptop. This
has not reduced the overall use of the lab, as we
have found that many of our students would rather
not carry their laptop around campus every day.
Therefore in addition to using the specialized
geo-spatial software for classwork, the lab is used
frequently for e-mail, research, printing, scanning
and for checking the campus online portal.
In addition to providing students with state-
of-the-art technology, the DGL provides an envi-
ronment where students can work on projects
together and build a sense of community, which
is often missing in many parts of our commuter-
dominated university (Hickey, 2010). Alumni
often speak of how those long nights together in
the DGL forged deep friendships and a fondness
towards their years at Salem State.
FROM THE PAST TO THE PRESENT
From the early 1980’s to the late 1990’s, primarily
smaller upper division classes used the lab and
the need for physical space was not paramount.
In 1984, the lab consisted of two rooms; the open
lab (a small room of 10 seats, created by splitting
a traditional classroom in half) and a second class-
room which was modified to house the computer
(DEC VAX 11/780), its peripherals, two digitizers
and the staff office. Many of the peripherals were
donated by the leading high-technology companies
The Digital Geography Lab at Salem State University
Figure 2. Number of students utilizing the DGL by year
of that time (Tektronix, Synercom, UNIRAS and
Digital Equipment Corporation). The lab also
quickly developed alliances with many federal
agencies including the National Park Service and
the University Center for Atmospheric Research
(UCAR). As the computer with the peripherals
were the size of a small car (6’H x 4’W x10’L)
and produced excessive heat, it required its own
full-size air conditioning unit built in the room with
it. In 1991 the lab upgraded to the VAX 4300. As
hardware became more efficient and smaller (the
new VAX was the size of an oven), this allowed the
building of a smaller room to house the computer
and the still very large air-conditioner it needed.
The vacant room left over from the absence of the
larger VAX allowed for more lab space for students
and the addition of more computer terminals. In
1994 the lab purchased a DEC AlphaServer 2100
for its geo-spatial applications as well as to be
used as a Gopher and Email server. At this time,
PCs were also becoming more affordable and
the DGL acquired 3 desktop computers to run
the newly introduced ArcView for Windows, to
supplement the AlphaServer’s OpenVMS version.
With the introduction of desktop computers into the
DGL, the lab started to see more underclassmen
using the facility for more than geo-spatial based
applications and Web-based research. Students
were coming by to write and print papers more
frequently, as typewriters were now becoming part
of the past, and as a result the DGL was being used
by all levels of students in the program. In 1998
the DGL transitioned to a Windows NT Server,
and by 2000 the DGL had completely phased
out the DEC OpenVMS OS and had transitioned
over to the Windows operating system for all its
servers and PC clients.
By 2002, with the addition of a large format
plotter, oversize flatbed scanner and laser print-
ers, the DGL was bursting at the seams, sign-up
sheets for computer usage time was necessary and
had to be enforced. The need for a room where
professors could hold lab classes had also become
a necessity as well as a dedicated area for input
and output devices. It was also important to have
an environment for the students that encouraged
group collaboration, which had always been a
large part of the success of the facility. Discus-
sions started as to how to go about expanding
the facility to meet the needs of the students. The
DGL worked with the university’s administration
and eventually came up with a plan to move and
expand the DGL while also addressing other
facility issues in the process. As a result, new
electrical and networking closets were built and
the DGL was able to move into a completely new
1093
 The Digital Geography Lab at Salem State University
space specifically designed as a computing lab.
In the fall of 2007 the DGL moved into the space
traditionally held by the manual cartography lab,
and darkrooms (see DGL’s Public Photo Gallery
at picasaweb.google.com/dgl.salemstate).
The DGL is now comprised of two rooms
joined by a sliding pocket door, allowing for a
spacious feeling while it’s open. The primary
room, the Open Lab, is used on a daily basis by
those with accounts in the DGL. The Open Lab
also has dedicated space to house the printers,
plotters and scanner. The Teaching Lab is a non-
scheduled classroom that is available for faculty
to use during their class time. Faculty sign for the
Teaching Lab as needed and bring their classes in
to provide general instruction on coursework as
it pertains to computer use in the DGL.
The Digital Geography Lab has two Dell
PowerEdge Servers serving thirty-seven clients
(Dell Optiplex 745). Twenty clients are available
for use in the Open Lab (Figure 3) with multiple
desk layouts to accommodate different research
group sizes. In the Teaching Lab (Figure 4), there
are 16 student machines and one machine con-
nected to an overhead digital projector for the
instructor. All seventeen clients in the Teaching
Lab are available for student use when the room
is not reserved for classes.
Figure 3. Open lab, Salem State University’s
digital geography laboratory
1094
Hardware resources include high resolution
color scanning on an 11x17 flatbed scanner. Out-
put devices include high speed HP color laser
printers, and large format HP plotters. Discipline
specific software holdings of the DGL include
ArcGIS (ESRI), ERDAS Imagine, IDRISI (Clark
Labs), Surfer (Golden Software), RockWorks
(Rockware), SAS, and MapInfo (Pitney Bowes).
The DGL promotes equal access for students
of all academic levels. All geography majors,
minors and students enrolled in geography
classes have equal access and rights to the lab’s
facilities. This has been achieved by having full
time staff that are responsible for all hardware,
software, maintenance, and facility operations. It
is important to note that the lab has always been
staffed by degree holders in geography, therefore
having an understanding of cartography and spatial
analysis as well as having a lot in common with
the students and faculty. Due to the university’s
fiscal constraints beginning in late 2008, a real-
locating of staff resulted in the loss of one of the
two full time positions. To supplement this loss,
between 5 and 7 student lab assistants are hired to
assist the DGL systems manager during the year.
The lab assistants monitor the DGL, aid fellow
students with basic computer and coursework
questions, assist the staff with hardware and soft-
Figure 4. Teaching lab, Salem State University’s
digital geography laboratory
The Digital Geography Lab at Salem State University
ware troubleshooting, and keep the lab open at
night and on weekends. All lab assistants must be
majors in the Geography or Geological Sciences
Departments and must have had some previous
DGL experience through their course work. The
success of the day to day operations of the lab
strongly depends on the quality and competence
of the lab assistants. Not only does this give the
students an opportunity to work on-campus, it
provides them with good work experience in a
geo-computing environment and exposes them
to a wide range of software that may be outside
their concentrations.
The amount of hours that the lab is open each
semester varies according to the student employee
budget and eligible students awarded work-study
funding. On average the DGL is open 75 hours per
week during the fall and spring semesters. During
the summer semesters and academic breaks, the
lab is usually open Monday through Friday during
the day. In order for the lab to be open at night or
on weekends the systems manager, a lab assistant,
or faculty member must be present. Although this
does impact the amount of time students have
access to the lab, 24-hour unsupervised access
is not an option due to the extensive amount of
hardware and software in the lab.
CONCLUSION
While the hardware, software and location of the
lab have changed over the years, its mission has
not. With technology education becoming more
and more important at Salem State University,
the DGL will continue to service the faculty and
student body in the College of Arts and Sciences
as well as in the Graduate School with cutting-
edge geo-spatial computing technologies. As the
geo-spatial sciences become more mainstream
in society (i.e., GPS and GIS technologies), labs
such as the DGL play a crucial role in providing a
well educated and essential geo-spatial workforce.
REFERENCES
Hickey, R. (2010). Beyond class hours: The GIS
lab as a center for geographic education. Inter-
national Journal of Applied Geospatial Research,
1(3), 1–16. doi:10.4018/jagr.2010070101
Luna, M. (2008a). Out of sight, out of mind: Dis-
tancing and the geographic relationship between
electricity consumption and production in Mas-
sachusetts. Social Science Quarterly, 89(5), 578.
doi:10.1111/j.1540-6237.2008.00578.x
Luna, M. (2008b). Estimate of limited English
speaking eligible voters in select communities of
the Boston Metropolitan Area. A Report to Greater
Boston Legal Services. Salem, MA: Salem State
College.
Luna, M. (2010). Voter participation and demo-
graphics in Lowell, MA: A GIS study. A report to
ONE Lowell and the City of Lowell. Retrieved
from http://dgl.salemstate.edu/geography/Profs/
luna/projects/2010_01Spring/GPH903/Lowell-
VoterReport2010reduced.pdf
Ratner, K. (2007a). Use of orthophoto imagery
with integrated elevation values to determine
roadway lengths. Salem, MA: FTA Transportation
Research Program.
Ratner, K. (2007b). Why we need to measure our
roadways. Paper presented at the North East St.
Lawrence Valley Division, Association of Ameri-
can Geographers Meeting, New Haven, CT.
Ratner, K. (2008). The changing world of roadway
mapping and measurement. Paper presented at
the Annual Meeting of Association of American
Geographers, Boston, MA.
Ratner, K. (2009a). The reshaping of land use and
urban form in Denver, Colorado through transit-
oriented development. Denver, CO: National Cen-
ter for Intermodal Transportation (NCIT) Grant.
1095
 The Digital Geography Lab at Salem State University
Ratner, K. (2009b). Was it going to happen any-
way? Examining the link between transit and land
use development in Denver. Paper presented at
the Annual Meeting of Association of American
Geographers, Las Vegas, NV.
Remote Sensing Specialty Group of Association
of American Geographers (RSSG-AAG). (2003).
Digital Geography Laboratory. The RSSG News-
letter, 3(1), 5.
Young, S. S. (2004). Satellite detected broad-scale
vegetation change in China, 1982 – 1999. Asian
Geographer, 22(1-2), 123–142. doi:10.1080/102
25706.2003.9684103
1096
Young, S. S. (2011a). Geoinformatics and the
mapping of lands vulnerable to sea level rise. In
P. K. Joshi & T. P. Singh (Eds.), Geoinformatics
for climate change studies. New Delhi, India: The
Energy and Resources Institute (TERI).
Young, S. S. (2011b). The Earth exposed: How
geographers use art and science in their exploration
of the Earth from space. In Dear, M., Ketchum, J.,
Luria, S., & Richardson, D. (Eds.), GeoHuman-
ities: Art, history, text at the edge of place (pp.
183–186). New York, NY: Routledge.
Young, S. S., & Harris, R. (2005). Changing
patterns of global-scale vegetation photo-
synthesis, 1982-1999. International Jour-
nal of Remote Sensing, 26(20), 4537–4563.
doi:10.1080/01431160500239198
The Digital Geography Lab at Salem State University

APPENDIX: FACULTY AND STAFF CREDENTIALS, EXPERTISE

Richard T. Anderson, Professor Emeritus

B.S., Salem State College; M.Ed., Boston State College; M.A., Ed.D., Boston University.
Areas of specialty: World Regions, Europe, Geographic Education
Profile: His major area of scholarship is in Geographic Education. The National Geographic Society
selected him as the founding coordinator of the Massachusetts Geographic Alliance. He is also
the co-director of the Northeast Global Education Center and coordinates numerous geographic
workshops for K-12 teachers. He served as the National Council for Geographic Education rep-
resentative on the Geography Education National Implementation Project. In 1996 he received an
Outstanding Geography Educator Award and then in 1999 a Life-Time Achievement, both from
the National Council for Geographic Education (NCGE) with a Mentor Fund in his name. In 2001
he received an Achievement Award from the NESTVAL Geographic Society. Professor Anderson
has brought in more than $6,000,000 in grant money to Salem State University.

William L. Hamilton, Professor

B.S., Worcester State College; M.S., Southern Illinois University; Ed.M., Worcester State College;
Ph.D., Oregon State University.
Areas of specialty: Analytic Cartography, GIS, Quantitative Application to Geo Spatial Intelligence
(GEOINT), Photogrammetric and Digital Image Processing, Applied Meteorology and Climatol-
ogy, Applied Geosciences Curriculum Analysis and Development
Profile: Over 35 years teaching experience spanning elementary- primary, middle school, special educa-
tion, secondary school to post doctoral education. Held high level positions in and currently consults
with Fortune 100 companies in GEOINT applications. While at SSC developed seventeen new
courses, acquired funding and created the Digital Geography Laboratory, created the Masters in
Geo-Information Science, and the PhD in Geo-Information Science. Widely published in books,
proceedings, journals and software applications and has earned several awards for GEOINT curri-
cula development and integration. Current research endeavors are time-space modeling and system
development for anti-terrorism bio and chemical warfare initiatives. Currently Dr. Hamilton is the
co-director of the SSU Center for Economic Development and Sustainability.

John T. Hayes, Associate Professor

B.A., Bowling Green State University; M.A., Ph.D., University of California-Los Angeles
Areas of specialty: Physical Geography, Climatology, Meteorology, Natural Resource Conservation
and Management, Soils, GIS, Water Resources Management, Environmental Impact Assessment,
Global Warming, Climate Change
Profile: He has authored or co-authored over 20 scientific papers based on mathematical modeling of crop
yield, crop water use, and climate impacts on crop productivity. He has taught in Xian, China for 5
½ weeks (Xibei University) and participated on a 2-week research expedition to the Loess Plateau
region of China. He was co-editor for two years of the new peer-reviewed journal in geography,
The Northeastern Geographer. He has co-chaired the campus’ Earth Day Planning Committee
for the past 6 years which is responsible for putting on a week-long program of speakers, films,
juried art exhibit, and judged student poster competition each April to help the campus celebrate
environmental problems, issues, and successes. Prof. Hayes has supervised student interns over

1097
 The Digital Geography Lab at Salem State University

the past decade for the city of Salem doing energy use audits and climate protection work for the
city and serves on Salem’s Renewable Energy Task Force. John is currently Vice-President of the
New England-St. Lawrence Valley Geographical Society, a regional division of the national As-
sociation of American Geographers (AAG); the position is a President-elect position. John’s recent
research (which he has presented at AAG’s annual meetings) has reviewed the potential carbon
sequestration potential of soils that are used for farming and/or ranching as a component of efforts
to mitigate against rising greenhouse gas concentrations.

Lorri K. Krebs, Associate Professor

M.A., Wilfrid Laurier University; Ph.D., University of Waterloo-Canada.
Areas of specialty: Tourism, Hospitality, Economic Geography, Canada, Caribbean
Profile: Dr. Krebs’ Ph.D. was awarded from the University of Waterloo, Canada in Geography with an
emphasis on marketing tools, tourism and decision-making. She also holds a Masters Degree in
Geography with a focus on ecosystem approaches, parks and protected areas. Her research in these
areas span from trans-boundary issues in Yukon and Alaska, to sustainability and tourism develop-
ment in National Parks. She has worked in various aspects of tourism in Switzerland, Mexico and
Canada. Current research interests include special events, Indigenous people and tourism, market-
ing, and economic development. Dr. Krebs is also Executive Director of Salem State University’s
Center for Economic Development and Sustainability.

Marcos J. Luna, Associate Professor

M.A. California State University, Los Angeles; Ph.D., University of Delaware.
Area of specialty: Environmental Justice, Political Ecology, Energy Policy and Planning, Environmental
Policy, Environmental History, Social Justice and Inequality, GIS Analysis
Profile: Dr. Luna works with a variety of non-profit organizations in the greater Boston metropolitan
area on issues ranging from voter participation to environmental change to urban history. Regularly
works with both undergraduate and graduate students on service learning GIS projects to engage
with the community and to provide students with the opportunity for real-world experience. Before
coming to Salem, Dr. Luna worked as an Environmental Analyst for the Massachusetts Department
of Environmental Protection and as a GIS consultant for Native American tribes in the Southwest
and Northwest. He has held research assistantships in grant-funded remote sensing and spatial
analysis projects and has participated in health and environmental public service projects.

Stephen Matchak, Professor and Department Chairperson

B.A., University of Hartford; M.A., University of Wales; M.A., Ph.D., University of North Carolina.
Areas of specialty: Tourism, Cultural Geography, Folklore
Profile: Dr. Matchak is a cultural geographer especially interested in landscapes, tourism, New England,
Europe, and folklore. His degrees include a Ph.D. from the University of North Carolina as well as
masters degrees from the University of North Carolina and the University of Wales-Aberystwyth.
For several years, Dr. Matchak has been interested in international education and has been a faculty
member on over two dozen international educational trips.

Kym Pappathanasi, Staff Assistant

B.A., University of Vermont; Graduate Certificate in the Cartographic Sciences, Salem State College

1098
The Digital Geography Lab at Salem State University

Areas of specialty: Cartography, GIS, Geo-Computing Systems Management

Profile: Ms. Pappathanasi received her degree in geography at the University of Vermont. After complet-
ing a Graduate Certificate in the Cartographic Sciences at Salem State, she became a Cartography
Instructor at Salem State College. The following year she accepted the Systems Manager position,
and for the last 19 years she has been supervising and managing the Digital Geography Lab at
Salem State University.

Keith Adam Ratner, Professor

B.S., Michigan State University; M.S., Pennsylvania State University; Ph.D., University of Denver.
Area of specialty: Urban and Transportation Planning, and GIS Implementation and Development
Profile: Prior to coming to SSU Dr. Ratner worked for eleven years as a professional planner in New
England at the local, regional and state levels of government. Research projects at SSU have in-
cluded an investigation into the regionalization of transportation planning north of Boston, the use
of three dimensional data for the determination of road lengths for federal reporting requirements,
and the analysis of regional land use changes related to transit oriented development in Denver,
Colorado. Student projects at SSU have included participation in the development of the Boston
Enterprise GIS Website, the development of a Service Area Management GIS for the Boston Centers
for Youth and Families, the development a GIS implementation plan for North Shore Community
Development Corporation.

Steven E. Silvern, Associate Professor

B.A., Clark University; M.A., University of Illinois; Ph.D. University of Wisconsin.
Area of specialty: Cultural Geography, Political Geography, Historical Geography, Indigenous
Profile: Dr. Steven Silvern is a cultural and political geographer who teaches a wide variety of human
geography courses including: World Regions, Sustainable Development, Geographic Research, Politi-
cal Geography, Population Geography and Native American Environments. He has been extensively
involved in using the internet and information technology to enhance geographical education and is
working with the Association of American Geographers to develop on-line instructional materials
on global climate change. Dr. Silvern’s research has been on the political and cultural geography
of Native Americans with current research on right-wing anti-Indian sovereignty movement and
how history museums represent Native Americans. He is also researching the reasons for success
or failure of community supported agriculture in Massachusetts.

Stephen S. Young, Professor

B.A., University of Vermont; M.E.S., Yale University; Ph.D., Clark University
Area of specialty: Biogeography, Global Change, Remote Sensing, China, Asia.
Profile: Through field work Dr. Young has worked in: Southwestern China, England, Taiwan, Japan,
Hong Kong, Australia, Montana and Alaska. He has published over a dozen scientific papers
based on his field work. Currently he is working with the Chinese Academy of Sciences analyzing
land-cover change in and around China’s nature reserves. He also has put together a traveling art
exhibition which demonstrates how Geographers study the Earth from Space. The exhibition has
been shown in over a dozen locations including the Headquarters of the National Science Founda-
tion, Philadelphia, Australia and North Africa.
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 3,
edited by Donald Patrick Albert, pp. 86-96, copyright 2012 by IGI Publishing (an imprint of IGI Global).

1099
1100

Chapter 66
Analysis of Tourist Behavior
Based on Tracking Data
Collected by GPS
Oriol Bernadó
Science and Technology Park for Tourism and Leisure, Spain

Albert Bigorra
Science and Technology Park for Tourism and Leisure, Spain

Yolanda Pérez
University Rovira i Virgili, Spain

Antonio P. Russo
University Rovira i Virgili, Spain

Salvador Anton Clave

University Rovira i Virgili, Spain

ABSTRACT
The research field of tourism management has been witnessing in the last ten years an important de-
velopment in methods based on the analysis of the spatial behavior of visitors, mainly driven by the
progressive enhancement and economic accessibility of satellite-based tracking techniques. However,
the practical operations of destination management systems based on tracking technologies still suffer
from the relative complexity of obtaining and mastering this information. This chapter illustrates the
development of Track-Tour, a user-friendly system that integrates the tracking of visitor spatial activity,
the elaboration of such data according to structured formats, and their visualization through a web-
based viewer. It is believed that this system has the potential to allow an easier uptake by destination
management and planning organizations in their quest for more sustainable tourism impacts.

The chapter presents the design and validation of this comprehensive methodology, and the early results
from a test application in the historical centre of the World Heritage city of Tarragona, a semi-enclosed
area with unpredictable patterns of tourism mobility that typically suffer from congestion and flow
management problems.

DOI: 10.4018/978-1-4666-2038-4.ch066

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
INTRODUCTION
It is clear that mobility is a core aspect of contem-
porary life (Drummond, et al., 2006), and tourism
is by no means indifferent to this. The increased
possibilities and means of transport (reducing
functional distances), as well as the supposed
“normality” of the individual’s occupational,
social and leisure mobility (Urry, 2008) have
engendered a cultural and economic revolution
and has led to the progressive standardization of
tourism as a social commodity. This evolution
modifies the traditional approach to tourism and
the topics being researched or investigated in this
field addressing research into aspects such as
the choice of destinations and forms of tourism
(where to travel), and aspects of performativity
and spatial behavior at the destination (what to do
there, what to consume, and where). This article
focuses particularly on the second aspect, defining
a systematic method for the study of tourist mobil-
ity at destinations or tourist areas that will help
destinations to plan and manage actions geared
towards improving the competitiveness and the
sustainability of tourist activity. Hitherto, this area
of research has aroused scant interest, resulting
in a limited practical usefulness of the methods
employed and the results obtained for policies and
the management of visitable spaces (Dietvorst,
1995; Thornton, et al., 1997; Shaw, et al., 2000).
Even in places where visitor surveys are carried
out regularly, visitor mobility has continued to
merit little attention; the spatial or geographic
aspect of visits in general is neglected, since it is
assumed and estimated to be impossible to amass
such complex information through direct empirical
research (Meng, et al., 2005).
Thornton et al. (1997), among others, point out
that understanding the spatio-temporal behavior of
tourists at destinations could generate different im-
provements such as transport management, avoid
the saturation of places of interest, optimize routes
or, and beyond the actual destination, extend the
geographic distribution of visitors and their spend-
ing in the ensemble of the regions and/or sights
at the destination. These potential actions should
be complemented by better marketing strategies
(Thornton, et al., 1997) and several other aspects,
based on a better segmentation of the target and the
visitable space. Knowing exactly what routes have
been taken, by how many people or vehicles, for
how long and with what effort, what facilities or
shows they went to, who visited and photographed
the monuments or when, as well as what types of
shops and places interested visitors more or less,
could lead to a very significant improvement in
the performance of destinations. This reality will
allow us to establish new solutions designed with
precision on the elements, such as signs, opening
times, the setting of license costs, public transport
engineering, etc. In summary, it leads us to a new
tourist site management scenario on any spatio-
temporal scale.
STATE OF THE ART
This potential improved management of tourist
destinations was unthinkable until only recently;
however, in the last decade, technological devel-
opment (particularly in the new Information and
Communication Technologies (ICT) and satel-
lite communication technology) has permitted a
change of paradigm in the consideration of the
aforementioned mobility. Until barely a few years
ago, research in tourism had generated only gen-
eral information on the activity of visitors through
traditional or “low-tech” methods. While it is true
that general information on tourist activity can be
obtained through methods such as direct surveys
or travelogues, these methods present different
shortcomings in terms of availability of detailed
and real or exact data. This is due to the complex-
ity of the environment: a visitor in an unknown
setting will find it difficult, if not impossible, to
remember the route, and then there is the need to
secure the tourist’s close collaboration and dedi-
cation, which means biasing the sample to those
1101
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
who are very interested in the study and have a
methodical nature which leads them to keep a
record of all their activities. The development
of other methods to ascertain mobility, such as
tracking and recording visitor activity time, either
personally or by analysis of closed circuit televi-
sion cameras (Hartmann, 1988; Gali & Donaire,
2006), encounters different relevant problems such
as violation of privacy, and is barely profitable
due to the limited sampling size it permits, or due
to the impossibility of segmenting tracking data
according to the population’s socio-demographic
characteristics.
The need to know visitor mobility in the dif-
ferent tourist environments calls for the devel-
opment of a specific, effective, and sustainable
methodology (in terms of time and economics).
This is linked to different and new methods that
are provided by different structural improve-
ments in technology, and which are detailed in
several documents (Shoval & Isaacson, 2006;
Shoval & Isaacson, 2007; Shoval, 2008; Spek,
2008) and in recently published books (Schaick,
2008; Shoval & Isaacson, 2010) which permit
a revision of the methodology. This has made a
decisive contribution to reviving these studies for
the development of a more or less homogeneous
series of methods for the analysis of visitors’
spatial behavior in tourist areas. Several studies
have been implemented in different cultural and
physical environments, as well as on different
spatial scales, using different methodologies and
technologies. Particular mention may be made of
projects conducted in urban environments such as
in Akko, Israel (Shoval, 2008) or in Bilbao, San
Sebastian, and Vitoria (Euskadi, Spain) (Alzua,
et al., 2010): two of them use GPS devices but
the second uses smart phones. In the first case,
the use of the devices implied that there must be
a contact and an active collaboration of visitors,
which may be interfering with the experiment;;
in the second, the methodological bias in con-
stituted by the relative exclusivity of the device.
The experience in Euskadi has a strong point in
1102
the analysis part, as they have a webmap and the
potentiality of developing ad-hoc tools for this
purpose. In open natural areas; we refer to Loch
Lomond and Trossachs National Park, Scotland
(Connel, et al., 2008) and Twelve Apostles Na-
tional Park, Victoria, Australia (O’Connor, et
al., 2005). These projects are slightly different
as regards space. The former analyzes car-based
tourism through a survey and latter registers tour-
ist movement through pressure sensors and maps
their route with the aid of an agent-based model
for recreation behavior simulation. This project
shows us another way to get data and a multitude
of analysis possibilities. Another kind of space is
in enclosed areas. For this, we participated in a
project at the Port Aventura theme park (Russo,
et al., 2010). What we learned in this project is
what has helped improve the present stage, as is
described in this chapter.
This ensemble of technologies coincides in
the possibility of data collection on a very finely
honed scale, in real time and at low cost, but it
should be complemented by management and
exploitation tools adapted to needs and to all
types of users. It is in this dimension where the
parallel development of Geographic Information
Systems (GIS), technologies related to territorial
analysis and the study methods of human spatial
behavior render it possible to rise to this challenge.
Like the different tracking technologies, the use
of these GIS tools has become more widespread
and prices have fallen considerably.
Our contribution fits into this new line of re-
search, proposing some improvements to the data
capture, management, and analysis procedures that
substantially increase their efficacy as destination
management instruments.
The fundamental aspects of the projects listed
here are:
• Detailed knowledge of tourists’ spatial
activity (use of space, routes, timing), so
as not to interfere in their spatio-temporal
decisions through the collaboration of visi-
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
tors who voluntarily carry a GPS device
that stores data on their routes.
• The preparation of these data through a
Geographic Information System that in-
cludes, moreover, an analysis module,
both of them developed specifically for
this project to obtain information on fun-
damental profiles in terms of tourist spatial
mobility, “risk” and “opportunity” space in
flow management, and the main determin-
ing elements of these patterns, such as the
attraction capacity of certain elements at
the destination, the effect of natural events
and conditions (e.g. the weather, time-
tables), or systems signaling routes and
accesses.
The implementation of this project will provide
different work phases that include the design of
fieldwork campaigns from the technological and
socio-demographic standpoint and the develop-
ment of the interface used to input and analyze and
view the data, named Track-tour in this project.
TECHNOLOGICAL BACKGROUND
The current technological scenario is ideal for the
comprehensive development of methodologies
and tools that allow us to learn more about visitors’
spatial behavior. The potential and the track record
of these tools in research into tourism, for example
Global Navigation Satellite Systems (GNSS), are
very broad, although no really major effort has
been made. Until only recently, most research
has focused solely on transport analysis, motor
vehicle tracking studies, while the monitoring
of pedestrians has remained virtually untapped.
This may be due to different factors; one possible
explanation is that compiling data on pedestrians
using GPS is far more complicated than doing so
on motor vehicles. If we consider that the tracking
system of a car is merely a device which, once
installed, neither affects nor alters the nature of
the data collected, with pedestrians the tracking
system has to be discreet and passive, so as not to
distort or affect the study participants, and these
requirements are often difficult to fulfill.
Knowledge of pedestrian mobility will pivot
around a combination of technologies for the de-
velopment of two vectors: one based on tracking
tools, and one for the monitoring and manage-
ment of this tracking. This global system, since
it includes the whole process, should interfere as
little as possible with visitors, so that their route
is not affected through their participation in the
research. Thus, the generalization and cheapening
of both technologies makes it possible to develop
a comprehensive methodological corpus to study
visitor mobility in any setting. This methodology
is based on the use of Tracking Technologies (TT)
for collecting visitors’ spatial data. These data will
be integrated in management systems that will
facilitate data treatment and cleansing, as well as
the automated cartographic representation of the
visitors’ behavior in tourist settings. The current
technological reality provides different alterna-
tives for these two tracking and display vectors.
In TT we can choose between different op-
tions. To focus these different possibilities, TT
can be classified, for our purposes, into two major
groups: short- and long-range or in combination.
Both work according to a specific radiofrequency
protocol for wireless data transmission and follow
a synthesis of these. In short-range and potentially
useful TT in this research, we can include differ-
ent communication protocols such as: Radiofre-
quency Identification (RFID), sensor networks
such as the Wireless Sensor Network (WSN), or
Wireless Wide Area Networks (WWAN), such
as Global System for Mobile Communications
(GSM), General Packet Radio Service (GPRS),
and Universal Mobile Telecommunications Sys-
tem (UMTS). The different protocols they use
permit communication between points (in linear
or grid form), providing information to the whole
network from nodes to other nodes, generating
new information.
1103
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Long-range TT can include different GNSS,
either active or under development, including
the current standard, GPS, although the cur-
rent GLONASS, the future GALILEO and the
potential developments of BEIDOU and IRNSS
should also be considered. These systems permit
exact positioning on a terrestrial surface on the
basis of the calculation of distances from at least
three satellites (of known position) and a fourth
satellite that provides the altitude; in practice it
is based on the synchronization of the satellites
and the exact time the signal was transmitted.
Receiver errors are minimized through filters and
augmentation systems that may be Satellite-Based
(SBAS) or Land-Based (GBAS). At conceptual
level, it is estimated that all these communica-
tion protocols will render it possible to ascertain
visitor mobility at an affordable cost for use in
tourist management and be considered beneficial
or profitable. The precision of GNSS systems is
currently undoubted, but positioning depends
exclusively on the visibility of the transmitting
satellites for the receivers; thus, this system is
not effective indoors or in narrow streets, a key
factor in this type of project. In this chapter, we
focus our interest on this type of TT for reasons
explained later on in this paper.
The other technological cornerstone of the
project corresponds to data management tools,
and it is here that GIS and web technology take
on special importance. The characteristics of both
render it possible to develop this functionality as a
comprehensive and multi-user management tool.
The objective is to design a web interface, Track-
Tour, that makes it possible to display data in a
simple and user-friendly way, while also adding
a toolset for the treatment and analysis of data
according to the needs of tourism managers and
the characteristics of each tourist area.
Data display and management in a web envi-
ronment was regarded as the most suitable option
due to universality of use and access; moreover,
in the future, real time data may be loaded into
the system. This decision was based on the con-
1104
solidation of web cartography and Spatial Data
Infrastructure (SDI) viewers, due to the maturity
of different standards, technologies and data man-
agement software. This technological environment
is complemented by improving data storage op-
tions in spatial databases (both proprietary, such
as IBM DB2, Oracle, Microsoft SQL Server, and
Open Source, such as PostgreSQL, MySql,…),
making it possible to display them through dif-
ferent standards (WMS, Web Map Service; GML,
Geography Markup Language; KML, Keyhole
Markup Language,…) and through map servers
(mapserver, GeoServer,…).
TRACK-TOUR: A TOURIST
MOBILITY ANALYSIS SYSTEM
Two sequential work phases for the tourist mobility
analysis may be established: (1) data collection
and (2) data analysis. Each one establishes a clear
relationship with a specific technology, the first
does so with the TT, while the second is based on
WebGIS technology. In this project, both have been
integrated in a tool that permits the comprehensive
management of the whole process, and whose
internal architecture is described in this section.
Architecture
The comprehensive tourist mobility data man-
agement tool envisages two capital phases: data
capture and data management. The main system
architecture is a database complemented by a
specific module: it permits spatial data display,
management, and web publishing. An Object-Re-
lational Database (ORD) was chosen to implement
this project; i.e. a hybrid database that contains
both storage technologies: relational and object.
This ORD delivers responses to users through the
web interface, i.e. which receives data and subse-
quently serves them for analysis. This structure
makes it possible to add data to the system after
sampling, and analyze them following cleansing.
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Both tasks will be implemented within a web
environment (multi-user and multi-platform).
Data display and analysis are organized around
the database but are complemented by a map web
server (which makes it possible to load data from
the ORD as an image in a web environment) and
an API (Application Programming Interface) for
using base maps from external sources (such as
Google, Yahoo, Bing Maps, or Openstreetmap).
The ORD chosen to support the core of the
architecture of this system is a PostgreSQL data-
base with the PostGis extension for spatial data.
This ORD was selected for the following reasons:
• Open Source, this project was started up
for research and integration purposes; the
development freedom and flexibility it of-
fers is appreciated. Moreover, for potential
commercial operation, cost needs to be
minimal.
• Scalability, it can adapt optimally to large
data structures and the amount of system
memory, enabling it to support a greater
number of simultaneous requests.
• In technical terms it implements the use
of rollbacks, subqueries, and transactions,
making for more efficient operation.
• It can verify referential integrity and store
procedures in the actual database, just as
high-level databases managers, such as
Oracle, do.
• Compatibility; the Open Geospatial
Consortium (OGC) awarded a certificate
of compatibility that permits connectivity
with its different standards. Moreover, at
GIS desktop software level (QuantumGis,
Udig, Kosmo, gvSig, Openjump, or the
like) and map server level (GeoServer,
mapserver, …) it can function as a direct-
access data source, thus facilitating view-
ing and edition.
This PostgreSQL-based architecture has
been used to store the information on the study
participants, and the PostGis extension as a
complement to store the spatial objects. The
display of data stored in PostgreSQL/PostGis is
implemented by means of PHP and Open Lay-
ers through a GeoServer map server. The use of
PHP (Hypertext Pre-Processor) is due to several
reasons, although the main one is its standardiza-
tion and the knowledge of it. It is an interpreted
programming language that is currently almost
standard in dynamic web development; moreover,
it permits highly flexible development (and can
be consequently modulated to suit specific needs)
while also delivering reliable connectivity with
database servers.
For spatial data, Open Layers was chosen as
a solution to display cartographic data due to the
stability of the project and the proven results ob-
tained in other studies. This Javascript library is
an OGC-endorsed display standard, which permits
connectivity with different data and lean develop-
ment. The data were delivered to the web interface
through different Web Map Service (WMS) for
the base cartography and GeoServer as WMS
server for own data. The latter was chosen for its
interoperability and scalability.
The system works sequentially. First of all, the
spatial data collected by the devices are download
by means of a PHP form that includes raw data and
which, after cleansing, saves the information and
the waypoints to the PostGis database. Secondly,
the data is analyzed by means of a specific tool
developed in a web environment, using functions
written in PHP, which present the different analysis
options available. Functions developed in PHP
and SQL are used in combination with existing
PostGis functions to perform the calculations of
the chosen analysis. The results of these analysis
calculations are displayed on a web interface by
Open Layers, which, in turn, queries the database
to represent them on a map through Goeserver.
The functions developed for both processes are
addressed in detail later (see Figure 1).
1105
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Figure 1. System architecture
Data Collection
The database is the core of the application, but it
is the data that gives it meaning, hence the reli-
ability of the results obtained depends on the data.
From the outset, we sought to collect the data by
means of a TT, with visitors being given a device
to log their route and then upload the information
to the system. Thus, a series of requirements were
established to be fulfilled by this technology in
order to optimize the process:
• Non-interference: the device to be deliv-
ered had to detect the position of the visi-
tors at all times, without affecting their
route or impinging on their privacy. Above
all, this constrains the size and the weight
of the device.
• Reliability of the signal: the signal must
be continuous, without interruptions or al-
terations by any external or visitor-derived
factor.
• Signal integrity: the resulting data must be
treated as generically and as interoperably
as possible for integration in any informa-
tion system.
1106
• The cost, which must be kept as low as
possible so that high price does not prevent
it from being added to tourist management
processes effectively.
The TT selected was the GPS signal, for dif-
ferent reasons. GPS devices have become popular
in recent years as an almost de facto standard in
location and positioning; they are integrated in
cars, peripherals, telephones, and are proliferating
in everyday life continually, so they are devices
that tourists are familiar with and are not distrust-
ful of. This technological consolidation makes it
possible to estimate the timeliness of the choice.
With regard to the initial assumptions, it is clear
that it fulfills the non-interference requirement, as
there are several small-sized, light models. Regard-
ing the signal, generally speaking the reliability
and integrity requirements are met, albeit with
nuances, which will be addressed in the spatial
adjustment section. As for price, they are low-cost,
around 100 Euros per unit, and moreover, market
evolution causes prices to drop quite quickly or
remain the same, while functions are enhanced.
In fact, one important issue in this project
was to choose the right device to use. Based on
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Table 1. Optimal models
Model Producer Waypoints
BT-Q1000XT Qstarz 200,000
i-gotU GT-200e Mobile action 262,000
V-900 Columbus MicroSD Storage
RBT-2300 Royatek 650,000
GPS-4043 Blumax 150,000
AGL3080 Amod 1,040,000
the devices used in a previous study in the Port
Aventura theme park (Russo, et al., 2010) a study
was carried out of the different choices in the GPS
market to record position continuously. The key
factors that determined the choice were GPS chip
reliability, battery life and storage capacity. Table
1 shows the different models considered optimal.
The models chosen were the BT-Q1000XT
and i-gotU and GT-200e, as their specifications
seemed the most suitable. The next step was to
conduct a field trial to assess the signal quality
and usability of the devices. In the field test phase,
we noted that signal quality, storage capacity and
battery life were quite similar. The devices have
their strengths and weaknesses but overall gener-
ated similar heat output. We decided on the first
device due to its greater usability. The shape of
the device was a positive element, differing in
this respect to others.
GPS Signal and Multipath Error
Further to the above, strictly speaking the GPS
signal does not fulfill the precepts of integrity
and reliability, although this is a relative problem.
Structurally, the capture and reliability of the signal
generated by the GNSS depends on the visibility
of the satellite for the receivers. This conditioning
factor renders the devices vulnerable, for example,
to atmospheric alteration, which are corrected by
augmentation systems such as WASS, EGNOS,
etc. (satellite systems based on other satellite or-
bits that make the relevant corrections). Another
Battery power WASS/EGNOS
42h. Yes
80 h. Yes
24h. Yes
9 h. Yes
25h. Yes
20 h. Yes
problem with GPS devices is direct visibility in
dense urban environments, as is the case of most
European cities. Positioning depends on at least
three satellites in the direct line of sight, which, in
the case of high buildings and/or narrow streets,
is not always feasible, generating positioning
problems. This is known as the multipath error.
The multipath error appears when the receiver
detects parts of the same signal at two different
times. The first part received comes directly from
a satellite, which is the real signal, the other one
is a reflection of the same signal generated when
it bounces off a surface (for example high build-
ings), and is the distorted signal. This second part,
the distorted signal, is received later due to the
delay caused when travel distance increases. The
magnitude of the multipath error depends on the
following factors:
• Position and types of surface near the re-
ceiver where the signal can bounce
• The height of the antenna with regard to
the Earth’s surface
• GPS wave distance signals. Dimension of
the signal.
Analyzing the factors and the scope of the
different studies, we can see the effect of this
structural condition of the GPS signal. If we study
data from the routes through the historic centre of
Tarragona, a semi-enclosed space, we can see that
the multipath error is magnified. This is due, on
the one hand, to the fact that in this part of Tarra-
1107
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
gona, the receiver is surrounded by high buildings
where the signal is reflected continually, and on
the other, to the fact that Tarragona is at very low
altitude. These two factors lead the multipath error
to be augmented and therefore degrade the quality
of the signal in this spatial context. In summary,
urban morphology generates, on the one hand,
signal discontinuity, derived from the loss of
satellite visibility, and on the other, the linking
of correct data to other incorrect data, generated
by the signal bouncing off buildings.
The worst issue is that this error is random and
from the reading, it is impossible to know, with
certainty, the time or place of impact or intensity
of deviation. The error arises from satellite posi-
tion and it is not feasible to establish a system
to fix it previously. These anomalies prompted
the search for a solution: the development of an
automated cleansing tool.
Snap Tool
As was described in the preceding section, and
due to the multipath error, the signal received by
the devices, particularly in the case of Tarragona,
was produced at source. To attempt to solve this
difficulty, an automatic and realistic tool was
designed and implemented to adjust waypoints.
The urban fabric is used to adjust these points
and the different objective premises are taken into
account with regard to the coordinates collected
by the devices and the visitors’ real movement
possibilities. These are the premises addressed:
• Speed: It is assumed that people only
move through the old town of Tarragona
on foot. This conditions speed of move-
ment and, in this case, the maximum speed
estimated was 1.12 m/s. This speed may be
modified, but in dense urban settings with
nearby tourist sites, no means of transport
is likely to reach higher speeds.
• Distance: Since the average speed estimat-
ed is 1.12 m/s and the GPS records a co-
1108
ordinate every 10 seconds, the maximum
distance between two recorded points can-
not surpass 11.27 m.
• Direction: In an urban tourist environ-
ment, it is evident that movement may
only be made in the street and in or over
elements adapted for this purpose, such as
walls, promenades, beaches … Thus, visi-
tor flows can only move through an exist-
ing, and therefore known, urban fabric.
• Connectivity: An urban route must fol-
low the connectivity relationship, hence to
move from one point to another you have
to follow streets that are connected to each
other.
The premises described above were used as the
bases for the development of a spatial adjustment
tool of the points collected by the GPS. These
conditions limit visitor movement possibilities in
time and in space, but also permit correction; even
if the devices generate erroneous signals they may
be corrected in most cases. The generalization of
studies and tools like this one will render it pos-
sible to improve the ruggedness and reliability of
the spatial adjustment.
At a technical level, a series of steps are fol-
lowed; the first is the elimination of incorrect
positions. The positioning of the points on different
GPS receivers will be performed according to the
intensity of the signal received, for example, if
direct satellite view is not received they estimate
the position. Thus, this first step eliminates these
points (in most cases they are at the beginning
of the route or the exit of an indoor space). The
second premise refers to the adjustment of the
points to the speed of pedestrian movement. In
this phase, the tool applies a first algorithm to the
remaining points to discard points that do not fulfill
the first premise. Since each point has an associ-
ated moment in time and a position in space, the
speed between each one of them and the previous
and subsequent point in time is calculated. If the
speed surpasses 1.12 m/s, the point in question
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
is discarded. The following picture shows an ex-
ample with five points ordered in time and each
one of them separated from the previous one by
10 seconds (GPS coordinate recording interval).
The distance between points (p1, p2) and (p3, p4)
is about 10 meters, on the other hand the distance
between (p2, p3) and (p3, p4) is more than 30
meters, therefore this algorithm eliminates point
p3, maintaining p1, p2, p4, and p5, as can be
observed in Figure 2.
To fulfill the third premise, the points resulting
from the first filter are adjusted to the closest
street. This step is necessary in the case of the
historic centre of Tarragona, because the impreci-
sion of the GPS leads some points to fall over
buildings. The following image shows an ex-
ample of this step. The original route taken by the
tourist is shown as a line, the points that simulate
Figure 2. Snapping tool: Operating model
Figure 3. Snapping tool: Wrong result
the positions captured by the GPS, as squares,
and the repositioned points following the meth-
odology depicted as bullets. As can be observed,
the first four points are repositioned correctly, as
are the last two. This is not the case of the rest of
the intermediate coordinates, which are reposi-
tioned in the adjacent streets; hence, the third
premise is not fulfilled (see Figure 3).
To solve this problem and fulfill the fourth
premise, a condition has been added to the repo-
sitioning of the points. This condition refers to
checking that the street where the point is relo-
cated is connected to the street of the previous
point. If this condition is not fulfilled, it would
mean that the tourist had gone through a building,
and as this, in principle, is impossible, another
street was sought. This new street had to be one
closest to the current point and, at the same time,
1109
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Figure 4. Snapping tool: Right result
to the street of the previous point. Figure 4 shows
the result of applying this algorithm and how the
problem of the previous step is solved.
Figure 5 shows a real example of this reposi-
tioning in the streets of the historic centre of
Tarragona.
Fieldwork
The methodology of the fieldwork prepared for
this project was modified according to the study
scenario. Sampling was performed in Tarragona.
The historic centre of Tarragona, from the stand-
point of visitor mobility, is a very peculiar space
since, while access is free, it is a historic walled
Figure 5. Snapping tool: Example
1110
area with a limited number of entry points (9) and
a high density of streets in the form of a practically
regular grid. In any case, this location is no different
from many other historic cities in Europe, such as
Venice, Bruges, Dubrovnik, Toledo, and many oth-
ers, where spatial behavior has been identified as
one of the main aspects determining inefficiencies
in the use of public space and consumption (see
for example Russo [2002] for the case of Venice).
The working methodology basically consisted of
distributing GPS devices to visitors, who took the
route they wanted and returned the device at the
end of the visit or period. In Tarragona, control
points were set up at the different entry points
to the old city so as not to condition the routes.
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
These points were used to deliver and collect the
devices at the beginning and end of the route, re-
spectively. This process was complemented by a
qualitative survey (nationality, age, gender, place
of accommodation, companions, etc.) to comple-
ment the spatial data of each route with a wide
socio-demographic profile generating patterns of
mobility, visits and specific activities. The device
delivery process was prepared on the basis of a
scheduled plan that aimed to cover most of the
possibilities depending on space and time, so that
the routes had different beginnings and endings.
This first sampling process was carried out over
a 14-day period for 11 hours, with a total of 436
surveys, and each device was issued, on average,
1.74 times a day in July, 2010.
Data Analysis
Data collection is one of the project’s key points,
the other cornerstone being data analysis, a pro-
cess which is detailed in this section. The main
objective of the study is to learn about the dis-
tribution of visitors in each one of the reference
scenarios; in spatial and temporal terms. A series
of sub-objectives was established with regard to
this general characterization of tourist behavior:
• The identification of the general patterns
of visitors’ spatial behavior, segmented
by their different socio-demographic
characteristics.
• The identification of spaces with more and
less appeal, segmented in terms of socio-
demographic characteristics.
• The analysis of the influence of visit dura-
tion (time allocation) on the spatial behav-
ior of the visitors.
• The sequence and selection of the visits
to places of interest or different landscape
elements, segmented by time and budgets.
• The analysis of space congestion, either
generically or segmented (by space, time
or socio-demographic characteristics).
• The daily “life cycle” of the tourist space
analyzed, as well as of the different spe-
cific tourist attractions.
• The identification of the existence of mo-
bility clusters between the different socio-
demographic groups, both generally, in
terms of space, and specifically, in terms of
attractions or sites of interest.
The information from the GPS, together with
that of the socio-demographic survey, makes up
an aggregate of georeferenced data. As has already
been mentioned, the spatial data are added to the
PostgreSQL/PostGis database and cleansed with
the space adjustment tool. Parallel to this process,
the alphanumeric information is also inputted into
the database, so that each route is linked to its socio-
demographic data and the activities the tourist had
engaged in. These data, in both scenarios, are the
cornerstone of the current analysis. This process
may be carried out from different perspectives
and with different methodologies. One option,
the most classic, proposes the implementation of
a results report comprised of two documents that
renders it possible a) to learn about all the socio-
demographic characteristics of the sample, and b)
the different trends in visitor mobility, the latter
on cartography. A series of conclusions linked
to this proposal may be obtained that enhance
knowledge of visitor routes as well as of the most
commonly visited points and spaces (generically
or segmented by socio-demographic data). Modern
technology permits a fully integrated and flexible
analysis that adapts to managers’ needs.
This project aims to generate an application,
with the help of web technologies, in which this
analysis is dynamic and adapted to the specific
needs of each tourist area. Intrinsically, mobility
means immediacy, thus it would seem obvious
that it would be better to conduct an analysis that
offers quick results in order to be able to acceler-
ate the decision-making process. On the basis of
this idea, analysis architecture was developed and
it was deemed more convenient to design a web
1111
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
viewer, Track-tour, that provides the generic and
consensus analysis tools according to the tourist
area studied. The objective of this viewer is to
permit greater flexibility of configuration and
adaptation to different environments or to the
specific needs of each tourist space; at the same
time, it should offer suitable management and
generate results integrally and reliably.
On the basis of these premises, the analysis
viewer requires a series of specific and concrete
characteristics to fulfill its requirements properly.
One initial premise is that it should be interoper-
able, i.e., development can be implemented and
replicated in different environments regardless of
the web browser or the operating system. Another
requirement is that development should be flexible
and dynamic so that querying will be agile and
useful. The last and probably the most important
requirement is that the system, the viewer, must
be reliable. The solution to these requirements is
provided by the actual genesis of the system: using
a web environment is what actually affords the
viewer the leanness required. Reliability depends
on the application’s internal source code and on its
interaction with the database, and it also delivers
dynamism, avoiding long response times. Apart
from the code, calls to the database have been
optimized by minimizing the unnecessary data
loading and by pre-calculating core functions.
Interoperability is achieved by implementing
technologies with the highest possible degree of
standardization, thus avoiding the use of additional
plug-ins or specific formats.
The core of this viewer is the PostGis data-
base, which is the data source; the main axis of
display will be PHP language. This will enable
us to generate the responses required by users in
the viewer. In specific terms, the data displayed
will depend on other technologies, such as the
Open Layers library for display and GeoServer
as cartography server.
In structural terms, the viewer will feature a
series of clearly different parts (Figure 6). The
external graphic structure is comprised within a
1112
flexible environment and permits flexible adap-
tation to any corporate web environment. On a
more specific level, the viewer has different parts
or modules: a menu panel, a cartographic display
panel and a tool panel. The first allows the user to
toggle between the different tasks performed by
the viewer, i.e. browse through different submenus
such as data loading, survey edition, and return
to the matrix web... The second, the cartographic
display panel, takes up most of the space on the
browser window with a view to displaying the
data in response to an action on the different
tools provided by this panel. The base map, on
which the tourist mobility data are deployed, is
served through external sources, such as Bing
Maps, Google Maps, or Open Street Map through
GeoServer.
This tool panel is the viewer’s key element
and features two groups of different actions. The
first groups are the actions related to browsing
and data contextualization, while the second
groups, detailed below, are those developed spe-
cifically for the project. The former include the
basic tools of any web geoportal such as zooming
tools and layer management, while the latter in-
clude several developed analysis instruments
linked to a printing module and another data filter,
which make it possible to display the result of the
analysis processes according to a socio-demo-
graphic or temporal parameter.
The analysis tools developed use the accumu-
lation of points from the GPS devices, with the
following being developed for this project and
tourist environment: routes, stop points, sights,
functional polygons (time and space) and concen-
tration. The first generates and enables the display
of visitors’ itineraries in a color range varying
from the maximum value (red) to the minimum
value (green) with dynamically variable intervals
according to the filter established. The two specific
analysis tools, stop and interest points, relate to
the most important points for the visitors in the
sampling space using the same color range. The
points of interest count all the points and establish
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Figure 6. Web interface analysis
absolute maximums linked to the time dedicated
to each one (symbol size is proportional to the
length of the visit). The stop points relate the
spaces with the concentration of visitors so that
visitor frequency can be assessed. The polygonal
tools show the use and occupation of the space ac-
cording to a predefined spatial segmentation with
regard to functional polygons that correspond to
a combined parameter of urban morphology and
uses of space. This tool displays data according
to concentration by number of visitors or by time
of visit. The last tool, the concentration function,
is used to display the evolution of the use of the
urban space. This provides for a grid of small
regular polygons to which the visitor value for
each one is associated; the greater the value the
greater the color hue.
The tools developed as the central part of the
project allow us to identify mobility values and
patterns for the whole data set, although filtering
is also possible. The complement to this tool panel
is another specific filtering tool that can segment
data according to the qualitative variables gener-
ated by the surveys and also time patterns.
PRELIMINARY RESULTS AND
PENDING RESEARCH CHALLENGES
Preliminary results of Track-tour are being ana-
lyzed. As a sample of possibilities that we get with
1113
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Figure 7. Main focus attractions by time and space in Tarragona
this tool we can show a map with the difference
in intensity of visits in Tarragona according to
time concentration, left side image, and space
concentration, right side image (Figure 7).
In any case, data analysis ability of Track-tour
has room for improvement both at specific level,
modifying the PHP language, which generates
the output, and even by changing the tools. The
consolidation of this comprehensive tool as a
tourist management standard may make it pos-
sible to target development according to users’
specific needs; in this case, fewer tools may mean
greater functionality. In this regard, data display
can be both formally and technically improved.
In fact, there is the issue of improving the
methodology designed as a result of the combined
development of the TT and the display technolo-
gies associated with the data, which is still open
to testing, and the inclusion of new technologies
that appear on the market. Likewise, one task that
remains to be addressed is to explore the different
options for the analysis of the information further
with a view to achieving greater adaptation to the
needs of managers of tourist destinations.
As for the future involving TT in this devel-
opment, there are several potential lines of de-
velopment, although they converge in two main
possibilities. This study chose a development
linked to GPS devices, but there are other options
1114
that should be studied. In closed or semi-enclosed
spaces, a possible alternative would be to estab-
lish a system of digital monitoring of the space
and link it to specific qualitative data produced
indirectly through surveys. There are a great deal
of technologies and standards to develop and
test, which could fit the terms described in our
technological background section. If GPS devices
will continue to be used, and in the line of future
TT-linked developments, data cleansing is another
key area for improvement, especially as far as
error-measuring devices and spatial adjustment
are concerned, including a more finely honed
definition of the method used for this adjustment.
The capture of spatial data depends on the
TT, but what data should be linked to these posi-
tions? The qualitative information linked to the
routes is another element to be validated. Socio-
demographic surveys tend to be very extensive,
although only a few variables are used for spatial
analysis. It would be advisable to determine, in
collaboration with the stakeholders involved,
which are the key aspects for spatio-temporal
analysis and socio-demographic segmentation.
A new sampling campaign would be needed
to consolidate the tools developed and to validate
the functionalities defined and applied. In this
way, it would be possible to certify the automa-
tion of the cleansing tool. Another interesting
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
possibility would be to replicate this experiment
in a different geographic environment to evaluate
the system’s adaptability and its flexibility with
regard to territorial characteristics. Another field
of future development stems from the possibility
of automating the data collection environment yet
further. The idea is that this process should enable
two tasks: (a) that the operators that prepare the
surveys can do so via a user-friendly editor and that
the data from these studies can be copied directly
to the database, and (b) that the spatial data can
be validated by relating them to the thematic data
from the surveys.
Finally, in another sense we can assess how
new mobile information technologies and devices
such as smart phones among others, with position-
ing and localization services can be adapted for
tourism research. Another specific example of this
could be runkeeper (http://runkeeper.com/). This
product works as a sports community; to share,
track and measure fitness. This approach could
satisfy both consumers’ needs and expectations
and provide information for research. In fact, from
them we can glean much important knowledge
about tourist and personal space mobility.
CONCLUSION
Tourism is a structural economic activity in many
places and it is expected to continue to be so. Some
tourist destinations, as economic products, have
reached a period of maturity, which calls for a
rethink in the way that they work, are maintained
and are sold. The major challenge faced by these
destinations is to hold their position in the market
and, if possible, expand it in what is an increasingly
competitive environment. In this regard, this topic
heralds an incomparable opportunity for work,
research, and development, where information
and communication technologies should play a
key role. The development of ICT and associated
methodologies to improve the management of
tourist destinations can be addressed in many ways,
one of which is the proposal described herein.
In this economic activity, support to ICT may
be expected to be visible, real and useful on differ-
ent scales, in a short period of time and in all its
facets, such as promotion, destination management
and the planning of tourist activities. The use of
tools and methods for automating, facilitating, and
harmonizing processes is almost an imperative.
The virtually infinite development possibilities
in tourism linked to information and communi-
cation technologies are being demonstrated with
each passing day, but they call for comprehensive
development, implementation, and management
between the different authors and the different
technicians.
This chapter demonstrates how the techno-
logical tracking of visitors in tourist spaces can
be organized and implemented as a fundamental
tool to improve destination management. At a
conceptual level, it has enabled us to move from
an unknown element, namely visitors’ routes, to
one that is known, i.e. the spatio-temporal behav-
ior of tourists filtered by means of meaningful
socio-economic variables. In this regard, the tool
presented may become an important application
that helps to take decisions in tourist areas since
it makes for easy adaptation of the fieldwork to
the specific characteristics of each destination and
for a flexible adjustment of the different system
processes.
The methodology defined, and the develop-
ment and implementation of the system, are rated
as positive, surpassing the expected results. The
assessment of the technologies and the devices
was suitable, as was the process related to the
fieldwork and the development of the computing
application.
In general, the conclusion of this chapter is
that ICT generate many and good opportunities
for all of the economic activities linked to tour-
ism, and that they should continue to do so in
order to consolidate research in the development
framework.
1115
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
ACKNOWLEDGMENT
This project was funded by the FEDER European
Regional Funds and Tarragona Provincial Council.
REFERENCES
Alzua, A., Aramburo, I., Gerrikagoitia, J. K.,
Peralta, M., & Espinosa, N. (2010). eGIStour:
Sistema de medición de flujo de visitantes. TU-
RITEC, 289-300.
Dietvorst, A. G. J. (1995). Tourist behaviour and
the importance of time-space analysis. In Ash-
worth, G. J., & Dietvorst, A. G. J. (Eds.), Tourism
and Spatial Transformations. Wallingford, CT:
CAB International.
Drummond, J., João, E., & Billen, R. (2006).
Current and future trends in dynamic and mobile
GIS. In Drummond, J., Billen, R., João, E., &
Forrest, D. (Eds.), Dynamic and Mobile GIS:
Investigating Changes in Space and Time (pp.
289–300). London, UK: Taylor and Francis.
doi:10.1201/9781420008609.pt1
Galí Espelt, N., & Donaire Benito, J. A. (2006).
Visitors’ behavior in heritage cities: The case
of Girona. Journal of Travel Research, 44(4),
442–448. doi:10.1177/0047287505282956
Hartmann, R. (1988). Combining field methods
in tourism research. Annals of Tourism Research,
15, 88–105. doi:10.1016/0160-7383(88)90073-4
Meng, L., Zipf, A., & Reichenbacher, T. (Eds.).
(2004). Map-based mobile services: Theories,
methods and implementations. Vienna, Austria:
Springer-Verlag.
O’Connor, A., Zerger, A., & Itami, B. (2005).
Geo-temporal tracking and analysis of tourist
movement. Mathematics and Computers in
Simulation, 69(1-2), 135–150. doi:10.1016/j.
matcom.2005.02.036
1116
Russo, A. P. (2002). The vicious circle of tourism
development in heritage cities. Annals of Tourism
Research, 29(1), 165–182. doi:10.1016/S0160-
7383(01)00029-9
Russo, A. P., Anton Clavé, S., & Shoval, N. (2010).
Advanced visitor tracking analysis in practice:
Explorations in the Port Aventura theme park and
insights for a future research agenda. In Gretzel,
U., Law, R., & Fuchs, M. (Eds.), Information and
Communication Technologies in Tourism 2010
(pp. 159–170). Vienna, Austria: Springer-Verlag.
doi:10.1007/978-3-211-99407-8_14
Shaw, G., Agarwal, S., & Bull, P. (2000). Tour-
ism consumption and tourist behaviour: A British
perspective. Tourism Geographies, 2(3), 264–289.
doi:10.1080/14616680050082526
Shoval, N. (2008). Tracking technologies and
urban analysis. Cities (London, England), 25(1),
21–28. doi:10.1016/j.cities.2007.07.005
Shoval, N., & Isaacson, M. (2006). The appli-
cation of tracking technologies to the study of
pedestrian spatial behavior. The Professional
Geographer, 58(2), 172–183. doi:10.1111/j.1467-
9272.2006.00524.x
Shoval, N., & Isaacson, M. (2007). Tracking
tourist in the digital age. Annals of Tourism
Research, 34(2), 141–159. doi:10.1016/j.an-
nals.2006.07.007
Shoval, N., & Isaacson, M. (2010). Tourist mobil-
ity and advanced tracking technologies. London,
UK: Routledge.
Thornton, P. R., Williams, A. M., & Shaw, G.
(1997). Revisiting time-space diaries: An explor-
atory case study of tourist behavior in Cornwall,
England. Environment & Planning A, 29(10),
1847–1867. doi:10.1068/a291847
Urry, J. (2008). Preface. In Burns, P., & Novelli,
M. (Eds.), Tourism and Mobilities: Local-Global
Connections (pp. xiv–xv). Wallingford, CT: CABI.
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Van der Spek, S. (2008). Spatial metro: Track-
ing pedestrians in historic city centres. In van
Schaick, J. (Ed.), Urbanism on Track: Applica-
tion of Tracking Technologies in Urbanism (pp.
79–102). Amsterdam, The Netherlands: IOS Press.
Van Schaick, J. (Ed.). (2008). Urbanism on track:
Application of tracking technologies in urbanism.
Amsterdam, The Netherlands: IOS Press.
ADDITIONAL READING
Ahas, R., Aasa, A., Mark, Ü., Pae, T., & Kull,
A. (2007). Seasonal tourism spaces in Estonia:
Case study with mobile positioning data. Tour-
ism Management, 28(3), 898–910. doi:10.1016/j.
tourman.2006.05.010
Ahas, R., Aasa, A., Roose, A., Mark, Ü., & Silm,
S. (2008). Evaluating passive mobile position-
ing data for tourism surveys: An Estonian case
study. Tourism Management, 29(3), 469–486.
doi:10.1016/j.tourman.2007.05.014
Asakura, Y., & Iryo, T. (2007). Analysis of tourist
behavior based on the tracking data collected using
a mobile communication instrument. Transporta-
tion Research Part A, Policy and Practice, 41(7),
684–690. doi:10.1016/j.tra.2006.07.003
Chardonnel, S., & Van Der Knaap, W. (2004).
Managing tourist time-space movements in recre-
ational areas: A comparative study of a protected
natural park in the French Alps and the “De Hoge
Veluwe” Dutch national park using the same
methodology. Revue de Geographie Alpine, 90(1),
37–48. doi:10.3406/rga.2002.3071
Chhetri, P., Arrowsmith, C., & Jackson, M. (2004).
Determining hiking experiences in nature-based
tourist destinations. Tourism Management, 2(1),
31–43. doi:10.1016/S0261-5177(03)00057-8
Connell, J., & Page, S. J. (2008). Exploring the
spatial patterns of car-based tourist travel in Loch
Lomond and Trossachs national park, Scotland.
Tourism Management, 29, 561–580. doi:10.1016/j.
tourman.2007.03.019
Donaire Benito, J. A., & Galí Espelt, N. (2008).
Modeling tourist itineraries in heritages cities:
Routes around the old district of Girona. Pasos.
Revista de Turismo y Patrimonio Cultural, 5(3),
435–449.
Donggen, W., & Tao, C. (2001). A spatio-temporal
data model for activity-based transport demand
modelling. International Journal of Geo-
graphical Information Science, 15(6), 561–585.
doi:10.1080/13658810110046934
Fennell, D. A. (1996). A tourist space-time bud-
get in the Shetland Islands. Annals of Tourism
Research, 23(4), 811–829. doi:10.1016/0160-
7383(96)00008-4
Frihida, A., Marceau, D. J., & Thériault, M.
(2004). Development of a temporal extension
to query travel behavior time paths using an
object-oriented GIS. GeoInformatica, 8(3), 211.
doi:10.1023/B:GEIN.0000034819.57376.92
Itami, R. M. (2003). RBSim3: Agent-based simu-
lations of human behavior in GIS environments
using hierarchical spatial reasoning. In Proceed-
ings of the ModSim 2003 Conference. ModSim.
Keul, A., & Kühberger, A. (1997). Tracking
the Salzburg tourist. Annals of Tourism Re-
search, 24(4), 1008–1012. doi:10.1016/S0160-
7383(97)00038-8
Kwan, M., & Casas, I. (2006). Gabriel: GIS
activity-based travel simulator: Activity sched-
uling in the presence of real-time information.
GeoInformatica, 10(4), 469. doi:10.1007/s10707-
006-0343-7
1117
 Analysis of Tourist Behavior Based on Tracking Data Collected by GPS
Lam, T., & Hsu, C. H. C. (2006). Predicting
behavioral intention of choosing a travel desti-
nation. Tourism Management, 27(4), 589–599.
doi:10.1016/j.tourman.2005.02.003
Lau, G., & Mckercher, B. (2007). Understanding
tourist movement patterns in a destination: A GIS
approach. Tourism and Hospitality Research,
7(1), 39–49. doi:10.1057/palgrave.thr.6050027
Lew, A., & Mckercher, B. (2002). Trip destina-
tions, gateways and itineraries: The example
of Hong Kong. Tourism Management, 23(6),
609–621. doi:10.1016/S0261-5177(02)00026-2
Lew, A., & Mckercher, B. (2006). Modeling
tourist movements: A local destination analysis.
Annals of Tourism Research, 33(2), 403–423.
doi:10.1016/j.annals.2005.12.002
Loiterton, D., & Bishop, I. (2005). Virtual en-
vironments and location-based questioning for
understanding visitor movement in urban parks
and gardens. In Proceedings of Trends in Real-
Time Landscape Visualization and Participation
(Conference on Information and Technologies in
Landscape Architecture). Springer.
Mcwilliam, S, E. G., & Crompton, J. L. (1997).
An expanded framework for measuring the ef-
fectiveness of destination advertising. Tourism
Management, 18(3), 127. doi:10.1016/S0261-
5177(96)00114-8
Modsching, M., Kramer, R., Gretzel, U., & Ten
Hagen, K. (2006). Capturing the beaten paths:
A novel method for analysing tourists’ spatial
behavior at an urban destination. In Proceedings
of the International Conference on Information
and Communication Technologies in Tourism
2006. Springer.
Montanari, A., & Muscarà, C. (1995). Evaluating
tourist flows in historic cities: The case of Venice.
Tijdschrift voor Economische en Sociale Geografie,
86(1), 80–87. doi:10.1111/j.1467-9663.1995.
tb01831.x
1118
Mukai, N., Honma, H., & Kosugi, M. (2007). Gen-
eration of dynamic moving path for autonomous
human model. In Proceedings of MODSIM07:
Land, Water and Environmental Management:
Integrated Systems for Sustainability. MODSIM.
Oppermann, M. (1997). First-time and repeat visi-
tors to New Zealand. Tourism Management, 18(3),
177–181. doi:10.1016/S0261-5177(96)00119-7
Van Der Knaap, W. G. M. (1999). Research re-
port: GIS-oriented analysis of tourist time-space
patterns to support sustainable tourism devel-
opment. Tourism Geographies, 1(1), 56–69.
doi:10.1080/14616689908721294
Xia, J., & Arrowsmith, C. (2005). Managing scale
issues in tourist spatio-temporal movement model-
ling. In A. Zerger & R. M. Argent (Eds.), Proceed-
ings of the Modsim 2005 International Congress
on Modelling and Simulation. MODSIM.
Xia, J., Arrowsmith, C., Jackson, M., & Cart-
wright, W. (2008). The wayfinding process rela-
tionships between decision-making and landmark
utility. Tourism Management, 29(3), 445–457.
doi:10.1016/j.tourman.2007.05.010
KEY TERMS AND DEFINITIONS
AVITA: Is the methodological body based on
the joint use of this new generation of tracking
technologies (TT) for the collection of spatial data
on visitors, possibly to be integrated by surveying
activities, and the use of GIS for the elaboration
and representation of visitor spatial behaviour.
Budget-Time Analysis: Is the systematic re-
cord of a person’s use of time over a given period.
It describes the sequence, timing and duration of
the person’s activity.
Global Navigation Satellite Systems:
(GNSS): Refers to different positioning techniques
using satellites, which may operate in isolation
or in combination.
Analysis of Tourist Behavior Based on Tracking Data Collected by GPS

GPS: Is a planet-wide navigation system
that can determine position and time. Accuracy
depends on satellite view and the use of augmenta-
tion systems that vary from meters to centimeters.
Multipath Error: Is a general error in GPS
signal receivers and appears when the receiver
detects parts of the same signal at two different
times. The first is received directly from a satel-
lite and the other part is a reflection of this signal
because it has bounced off a surface (for example
high buildings or the same surface). The error
depends on the receiver, antennas, the reflecting
material.
Spatial Analysis: Is the process of extracting or
creating new information about a set of geographic
features; it is useful for evaluating suitability and
capability, for estimating and predicting, and for
interpreting and understanding spatial behavior
of any feature.
Tracking Technologies: Group of wireless
technologies that allow collecting any kind of
spatial data, to deliver location-based service, such
as GPS, GIS, WLAN, RFID and others.

This work was previously published in Discovery of Geospatial Resources: Methodologies, Technologies, and Emergent Ap-
plications, edited by Laura Díaz, Carlos Granell and Joaquín Huerta, pp. 241-260, copyright 2012 by Information Science
Reference (an imprint of IGI Global).

1119
1120

Chapter 67
The RiskCity Training
Package on Multi-Hazard
Risk Assessment
Cees van Westen
University of Twente, The Netherlands

Tsehaie Woldai
University of Twente, The Netherlands

ABSTRACT
As part of the capacity-building activities of the United Nations University – ITC School on Disaster Geo-
Information Management (UNU-ITC DGIM) the International Institute for Geo-information Science and
Earth Observation (ITC) has developed a training package on the application of GIS for multi-hazard risk
assessment. The package, called RiskCity comprises a complete suite of exercise descriptions, together
with GIS data and presentation materials on the various steps required to collect and analyze relevant
spatial data for hazard, vulnerability and risk assessment in an urban environment. The package has
been developed in collaboration with several partner organizations on different continents, and is used
as the basis for a series of courses. Currently it is available as a distance education course.

INTRODUCTION 2009). The possible impacts of hazardous events

The world is confronted with a rapidly growing
impact of disasters, due to many factors that cause
an increase in the vulnerability of society combined
with an increase in (hydrometeorological) hazard
events related to climate change (Blaikie et al.,
1996; IPCC, 2007; EM-DAT, 2009; Munich Re,
DOI: 10.4018/978-1-4666-2038-4.ch067
are large, especially in developing countries and
governments have to incorporate risk reduction
strategies in development planning at different
levels (UN-ISDR, 2004). The evaluation of the
expected losses due to hazardous events requires
a spatial analysis, as all components of a risk as-
sessment differ in space and time (Van Westen,
2010). Therefore risk assessment can only be car-
ried out effectively when it is based on extensive,

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
The RiskCity Training Package on Multi-Hazard Risk Assessment
multidisciplinary studies on the basis of spatial
information, derived from Remote Sensing and
other sources. There is an urgent need to include the
concepts of disaster geo-information management
into emergency preparedness planning, spatial
planning and environmental impact assessment.
This requires capacity building and training of di-
saster management experts and professionals, such
as planners, engineers, architects, geographers,
environmental specialists, university teachers
etc. (Van Westen, 2010). The Hyogo framework
of action 2005-2015 of the UN-ISDR indicates
risk assessment and education as two of the key
areas for the development of action in the coming
years (UN-ISDR, 2005).
Worldwide, a number of organizations are
specialized in providing short training courses on
disaster risk management-related issues (ADPC,
2005). Some organizations have also prepared
training materials that are accessible through the
internet; for example the Disaster Management
Training Programme (DMTP), or the International
Federation of Red Cross and Red Crescent Societ-
ies (IFRC). Most of these however concentrate on
community-based methods (Peters Guarin, 2008).
Disaster risk management courses at BSc or MSc
level are now available in many Universities on
all continents
Relatively few training materials are avail-
able on multi-hazard risk assessment. A good
textbook on the subject is made by Smith and
Petley (2009). Online training materials can be
obtained for example from the websites of the US
Federal Emergency Management Agency (FEMA,
2010a) and Australian Emergency Management
Agency (EMA, 2002). There is a very useful guide
prepared by FEMA called “Understanding your
risks” that guides readers through the various
phases of a risk assessment (FEMA, 2010c). The
development of innovative forms of learning and
teaching oriented towards building new curricula
in the field of natural risk has attracted attention
in European initiatives such as DEBRIS (2006)
and NAHRIS (2006). As far as GIS-related ma-
terial related to multi-hazard risk assessment is
concerned, the HAZUS methodology developed
in the US can be considered the standard. This
comprehensive loss estimation software which
runs under ArcGIS is a very good tool for carrying
out loss estimations for earthquakes, flooding and
windstorms (FEMA, 2010b), but is restricted to
use in the USA, due to constraints in the data and
the classifications used for elements at risk and
fragility curves. The manuals of Hazus, however,
provide a very good overview of the entire process
of multi-hazard risk assessment. Courses on the
use of HAZUS can be followed online from the
ESRI Virtual Campus (ESRI, 2010). Another very
promising initiative is the CAPRA project (World
Bank, 2010). The Central American Probabilistic
Risk Assessment (CAPRA) initiative aims to
strengthen the regional capacity for assessing,
understanding and communicating disaster risk in
Central America. The project has developed tools
for multi-hazard probabilistic risk assessment,
using Open Source software tools (World Bank,
2010). However, complete GIS based training
packages on spatial hazard and risk assessment
using low-cost or free GIS software are still very
scarce, to the knowledge of the authors. One ex-
ample is a training package in English and Spanish
developed for Central America in the framework
of the UNESCO RAPCA project (ITC, 2004).
This paper describes the main aspects of a
GIS-based training package on multi-hazard risk
assessment, which has been developed by the
United Nation University (UNU)– ITC School
for Disaster Geoinformation Management. The
International Institute of Geo-Information Science
and Earth Observation (ITC), has recently been
integrated in the University of Twente after a period
of 60 years as an institute for postgraduate train-
ing and research in the field of geo-information
directed to capacity building and institutional
development of professional and academic or-
ganizations from developing countries. In 2005,
ITC and the United Nations University established
a collaborative programme on the use of spatial
1121
 The RiskCity Training Package on Multi-Hazard Risk Assessment
information for disaster management, which re-
sulted in the formation of the UNU-ITC School for
Disaster Geo-Information Management. The main
activities of the DGIM School focus on training,
education, curriculum development, knowledge
development and research collaboration. This
is done through the establishment of Univer-
sity networks in Asia, Africa and Latin America,
where the member Universities exchange spatial
information, course materials and jointly carry out
training and research projects. The DGIM School
develops training packages and courses that are
given jointly with the partners of the networks
in various countries. The materials are uniform,
and have been developed in different languages,
and the support is given by local University staff
who have followed earlier training and by staff
from the UNU-ITC DGIM School. One of these
courses is on Multi-hazard risk assessment, which
is centered around a case study on the use of
Geographic Information Systems, and Remote
Sensing for the assessment of hazard, vulnerability
and risk in a typical urban area representative of
situations in many developing countries. Rapid
urbanization, combined with a lack of planning,
often leads to the spreading of squatter areas
located in hazardous areas, such as steep slopes,
flood prone areas etc.
RISKCITY TRAINING PACKAGE
The package, called RiskCity, comprises a com-
plete suite of exercise descriptions, together with
GIS data and presentation materials on the vari-
ous steps required to collect and analyze relevant
spatial data for hazard, vulnerability and risk
assessment in an urban environment. Figure 1
gives a schematic overview of the various steps
involved in the process. The RiskCity training
package focuses on demonstrating the procedures
of risk assessment for natural and human-induced
hazardous phenomena in an urban environment
within a developing country. We have selected an
1122
urban area, because the elements at risk have a
much higher density, the study areas are generally
smaller and the scale of analysis larger as compared
to a rural setting. This allows us to demonstrate
which tools can be used for generating hazard as
well as elements at risk databases, even in data
poor environments, and to show how qualitative
and quantitative techniques for risk assessment can
be used, and in which situation. Also the combined
effect of different hazard on the overall risk can
be better demonstrated, as well as the effect of
risk reduction measures.
One important consideration in designing the
exercises is that people from developing countries
should not be restricted in using it due to financial
burdens for software acquisition. Therefore the
aim was to use Open Source software as a basis.
The exercises are written for the Integrated Land
and Water Information System (ILWIS, 2010). It
is a Geographic Information System (GIS) with
integrated image processing capabilities. It also
has its own attribute data analysis, spatial data
entry and conversion modules. New modules have
been recently added for Spatial Multi Criteria
Evaluation, analysis of Digital Elevation Models
and for digital stereo image interpretation (ILWIS,
2010). The strongest point of the software is the
map calculation module that allows extensive
modelling with raster maps, also using scripts.
ILWIS, is very user-friendly software, and allows
the participants to concentrate on the risk assess-
ment application rather than on the specifics of
the software.
Each of the exercises in RiskCity has its own
dataset. The data are all provided in separate
directories, including the results of the previous
exercises that are needed to make a subsequent
one. So it is possible to carry out each of the
exercises separately. Result files and PowerPoint
presentations with instructions are also included.
The overall structure of the RiskCity training
package is given in Figure 2, and an overview
of the various components is presented in Table
1. Four different types of hazards are evaluated:
The RiskCity Training Package on Multi-Hazard Risk Assessment
Figure 1. Schematic representation of multi-hazard risk assessment
landslides, floods, earthquakes and technological
hazards. The training package starts with introduc-
tory exercises dealing with the software and with
the study area, where students learn the various
hazard problems by evaluating high resolution
images. An important component of RiskCity is
the generation of a database of elements at risk in
order to evaluate the vulnerability of buildings and
population. Here two options are considered. The
first is that there is no spatial data available, except
for a high resolution image, and the students have
to generate mapping units with homogeneous types
of buildings by stereo interpretation and digitiz-
ing on the image. Sampling is then carried out to
define the number of buildings and population
per mapping unit. The second option for generat-
ing the elements at risk database assumes there
are digital data available in the form of building
footprint maps, census information and detailed
elevation data from a Lidar survey. These are used
to calculate the number of buildings per mapping
unit and land use type, and to characterize the
buildings, for instance by calculating their height
and floorspace using Lidar data. The floorspace
is used then to distribute the census population
over the mapping units, and population estimates
are made for day- and night-time scenarios. The
elements at risk database contains information on
the buildings, with important attributes such as
urban land use type, construction type, floorspace
and height, as well as on population for a daytime
and night-time scenario.
There are a wide range of hazard assessment
exercises, not only those dealing with the four
types of hazard mentioned before, which are based
on the data of RiskCity, but also those which use
data from other areas for tsunami, cyclone, vol-
canic, forest fires and land degradation hazard
assessment. They use a variety of approaches,
such as inventory-based, heuristic, statistical and
deterministic. The vulnerability assessment in-
cludes exercises on the use of vulnerability curves
1123
 The RiskCity Training Package on Multi-Hazard Risk Assessment
Figure 2. General structure of the RiskCity training package
for assessing physical vulnerability, as well as the
use of expert-based Spatial Multi Criteria Evalu-
ation for the evaluation of social vulnerability and
capacity.
RISK CONCEPTS
The Risk Assessment approach adopted in
RiskCity is based on the definition of risk from
UN-ISDR (2009) “The potential disaster losses,
in lives, health status, livelihoods, assets and
services, which could occur to a particular com-
munity or a society over some specified future time
period”. In other words risk can be described as
“the probability of losses”. Risk can be presented
conceptually with the following basic equation
(see also Table 1 and Figure 1):
Risk = Hazard * Vulnerability * Amount of ele-
ments-at-risk (1)
1124
and the more conceptual equation:
Risk = Hazard * Vulnerability / Capacity
(2)
In the RiskCity training package both equations
are used. Equation [2] is only conceptual, but al-
lows incorporating the multi-dimensional aspects
of vulnerability, and capacity, which are often
integrated with hazard indicators using Spatial
Multi-Criteria Evaluation. Equation [1], given
above, is not only a conceptual one, but can also
be actually calculated with spatial data in a GIS
to quantify risk, with a focus on (direct) physical,
population and economic losses. As illustrated
in Figure 3 there are two important components,
which also should be spatially represented: hazards
and elements at risk. They are characterized by
both spatial and non-spatial attributes. Session 2 of
the RiskCity training package explains the spatial
data requirements for hazard and elements-at-risk
The RiskCity Training Package on Multi-Hazard Risk Assessment
Table 1. Overview of the exercise structure of RiskCity
Session Theory
1. Introduction to the course, objectives,
Introduction to risk structure, and set-up; Explanation of the
assessment available materials
2. Presentation of data requirements for
Spatial data the various types of hazards; Sources of
for risk assessment spatial data
3. Hazard types; Main concepts of hazard
Hazard assessment assessment; Frequency magnitude –
relationships; Methods for use of GIS in
hazard assessment
4. Types of elements at risk;
Elements at risk Classification of buildings, infrastruc-
ture, lifelines, critical facilities; Popula-
tion information;
Collection of elements at risk informa-
tion
5. Types; social vulnerability; physical
Vulnerability as- vulnerability; Methods for vulnerability
sessment assessment; Participatory GIS; Spatial
Multi Criteria Evaluation
6. Loss estimation models; HAZUS; quali-
Risk tative risk assessment; QRA; Basics of
assessment flood risk, seismic risk, landslide and
technological risk assessment
7. Risk evaluation; risk governance; risk
Risk management communication; cost benefit analysis;
Using risk information for emergency
planning; spatial planning, and environ-
mental impact assessment
8. Each participant makes it own final
Final project project and does a small research
data, and how available data from the internet
can be used.
Hazards are characterized by their temporal
probability and magnitude or intensity derived
from frequency magnitude analysis (this is
treated in Session 3 of the training package). In
this respect magnitude and intensity can be con-
sidered as synonymous terms that express the
severity of the hazard. For instance flood depth,
flow velocity, and duration in the case of flooding.
RiskCity Exercise
Introduction to ILWIS and the RiskCity dataset
Obtaining free data from the Internet;
Generating a DEM using SRTM data;
Download and georeferencing Google Earth images;
Creating and interpreting multi-temporal images
Frequency assessment;
Flood modeling 2D flood propagation model, Philippines;
Flood hazard monitoring with SPOT imagery, Bangladesh;
Statistical landslide hazard assessment, Colombia;
Deterministic landslide hazard assessment, India;
Modeling pyroclastic flow deposits, Philippines;
Earthquake hazard assessment: Kathmandu, Nepal;
Hazard analysis of cyclone flooding in Bangladesh;
Monitoring coastal erosion and sedimentation, China;
Modeling land subsidence and sea level rise, Indonesia
Generating a database of elements at risk from scratch;
Generating a database of elements at risk using existing data; Partici-
patory GIS for risk assessment
Generating vulnerability curves;
Spatial Multi-Criteria Evaluation for vulnerability and qualitative risk
assessment
Flood risk assessment;
Landslide risk assessment;
Seismic risk assessment;
Technological risk assessment;
Multi-hazard risk assessment
Risk information for emergency preparedness & Response; Simula-
tion exercise; Analysis of costs & benefits of risk reduction scenarios;
Using risk information in environmental assessments
Select a topic from a list, analyze a problem and generate a report and
presentation
For earthquakes the terms magnitude and inten-
sity do have a different meaning, with magnitude
expressing the energy level of the earthquake (on
the Richter scale) and intensity expressing the
local effects of the earthquake, that vary over a
distance, becoming less further from the epicen-
ter (and expressed in qualitative classes such as
the Modified Mercalli Intensity). The hazard
component in equation [1] actually refers to the
probability of occurrence of a hazardous phenom-
1125
 The RiskCity Training Package on Multi-Hazard Risk Assessment

Figure 3. Basic function of risk, which can be divided into the components of hazard, the vulnerability,
and the amount of elements at risk that are exposed to the hazard

enon with a given intensity within a specified
period of time (e.g., annual probability). Hazards
also have an important spatial component, both
related to the initiation of the hazard (e.g., a vol-
cano) and the spreading of the hazardous phe-
nomena (e.g., the areas affected by volcanic
products such as lava flows). Session 3 gives an
overview of the approaches that can be used for
the analysis of the temporal and spatial components
of hazards. Table 2 gives a summary of definitions
used in the GIS-based risk assessment.
Elements at risk are the population, properties,
economic activities, including public services, or
any other defined values exposed to hazards in a
given area. They are also referred to as “assets”.
Elements at risk also have spatial and non-spatial
characteristics. First of all there are many differ-
ent types of elements at risk (which are treated in
session 4 of the training package) and they can
be classified in various ways. The way in which
the amount of elements-at-risk are characterized
(e.g., as number of buildings, number of people,
economic value or the area of qualitative classes
of importance) also defines the way in which the
risk is presented. The interaction of elements at
risk and hazard defines the exposure and the
vulnerability of the elements-at-risk. Exposure
refers to people, property, systems, or other ele-
ments present in hazard zones that are thereby
subject to potential losses. The spatial interaction
between elements-at-risk and hazard footprints,
which is often referred to as “exposure” in other
risk formulas, is an integral component of GIS-
based risk assessment, and therefore the term

1126
The RiskCity Training Package on Multi-Hazard Risk Assessment

Table 2. Summary of definitions used in the GIS-based risk assessment

Term Definition
Natural hazard A potentially damaging physical event, phenomenon or human activity that may cause the loss of life or injury,
property damage, social and economic disruption or environmental degradation. This event has a probability of
occurrence within a specified period of time and within a given area, and has a given intensity.
Elements-at-risk Population, properties, economic activities, including public services, or any other defined values exposed to
hazards in a given area”. Also referred to as “assets”. The amount of elements at risk can be quantified either in
numbers (of buildings, people etc), in monetary value (replacement costs, market costs etc), area or perception
(importance of elements-at-risk).
Exposure Exposure indicates the degree to which the elements at risk are exposed to a particular hazard. The spatial inter-
action between the elements at risk and the hazard footprints are depicted in a GIS by simple map overlaying of
the hazard map with the elements at risk map.
Vulnerability The conditions determined by physical, social, economic and environmental factors or processes, which increase
the susceptibility of a community to the impact of hazards. Can be subdivided in physical, social, economical,
and environmental vulnerability.
Capacity The positive managerial capabilities of individuals, households and communities to confront the threat of disas-
ters (e.g., through awareness raising, early warning and preparedness planning).
Consequence The expected losses in a given area as a result of a given hazard scenario.
Risk The probability of harmful consequences, or expected losses (deaths, injuries, property, livelihoods, economic
activity disrupted or environment damaged) resulting from interactions between (natural, human-induced or man-
made) hazards and vulnerable conditions in a given area and time period.

exposure is not used as such in the risk equation.
When we calculate the risk equation using a
Geographic Information System (GIS) the ele-
ments at risk that are exposed to the hazards are
automatically obtained using map overlaying
techniques
Vulnerability refers to the conditions deter-
mined by physical, social, economic and envi-
ronmental factors or processes, which increase
the susceptibility of a community to the impact
of hazards. Vulnerability can be subdivided in
physical, social, economical, and environmental
vulnerability. The vulnerability of communities
and households can be based on a number of cri-
teria, such as age, gender, source of income etc.
which are analyzed using equation [2]. However,
according to equation [1] vulnerability is evalu-
ated as the interaction between the intensity of the
hazard and the type of element-at-risk, making use
of so-called vulnerability curves. The concept of
vulnerability and the generation of vulnerability
curves are treated in session 5 of the training pack-
age, including the use of participatory methods
for community-based risk assessment.
The risk curves form the basis of subsequent
cost-benefit analysis, in which for each hazard
type, a number of risk reduction measures is
evaluated. The investments to implement certain
measures (e.g., relocation of houses, flood con-
trol) are estimated and compared to the reduc-
tion in annual losses that would result if they are
implemented. Based on this, the most appropriate
methods for risk reduction are selected. The last
part of the RiskCity exercises deal with a final
project in which the participants are given a par-
ticular problem they have to solve with the risk
information obtained earlier.
ABOUT THE STUDY AREA
The exercises deal with a hypothetical case study,
which is originally based on an actual study area:
Tegucigalpa in Honduras. Tegucigalpa suffered
severe damage from landslides and flooding dur-
ing Hurricane Mitch in October 1998 when the
city received 281 mm of rain in 3 days (Mastin &
Olsen, 2002). Due to river flooding, an old land-

1127
 The RiskCity Training Package on Multi-Hazard Risk Assessment

slide was reactivated and an entire neighborhood
on top of it was destroyed. The landslide caused
the damming of the river and resulted in severe
flooding in large parts of the city centre for several
weeks (Harp et al., 2002). These events are easily
identifiable on the high resolution image which
serves as the basis for the exercises (Figure 4).
After Hurricane Mitch, USGS and JICA carried
out extensive work in Honduras and produced
extensive datasets.
Only part of the exercises is based on the ac-
tual situation in Tegucigalpa. In order to be able
to reach the learning objectives, modifications
and additions were made to the original data. It
is very difficult to have a dataset for a particular
area where all aspects of multi-hazard risk assess-
ment can be properly demonstrated, either because
particular hazard types do not happen in the city
or because particular data sets are incomplete,
restricted or erroneous. The time required to
prepare a suitable dataset and related exercises is
very large. Therefore we have decided to make a
generic case study that is representative for many
cities in developing countries. When the training

Figure 4. High-resolution image of the center of RiskCity with some of the hazard and vulnerability
features indicated

1128
The RiskCity Training Package on Multi-Hazard Risk Assessment
package is used in local course, often also local
data sets are used to illustrate part of the procedure.
In the case of RiskCity several types of remote
sensing data were used. Aerial photographs for
several periods, including the period of the major
disaster event in 1998, and two sets of satellite
data from 2001 and 2006 were the basis for land-
slide mapping. High resolution satellite data was
used for mapping elements-at-risk, and medium
resolution Aster data for generating a land use
map of the area. Elements-at-risk data can be ob-
tained at different levels of detail. In the RiskCity
case study this is done at the urban level, where
information needs to be as detailed as possible,
preferably at the individual building level, or at a
slightly more aggregated level of mapping units
or building blocks with homogenous land use and
building type. Figure 4 gives an example of the
high resolution image used with some information
on the areas affected in 1998.
CONCLUSION
The course is designed in such a way that even
non-GIS specialists can follow the course, since
the instructions are describing the steps in a cook-
book manner, at least in the initial phases. In the
later part of the course, when participants are
more used to the ILWIS software the exercises
becoming more demanding, and participants have
to design the steps to solve problems themselves.
The course is also designed in such a way that
participants can also follow it without hands-on
exercises with the ILWIS GIS for the exercises.
This may be the case for courses at decision
makers’ level, when the participants are not that
much interested in the particular steps to follow
in a risk assessment, but want to know more about
the overall procedure and what can be done with
the (intermediate) results. Therefore we have
made the course also in a WebGIS version. The
WebGIS version allows participants to evaluate
the individual steps of the methodology without
actually doing GIS analysis. They will not use
ILWIS, but will use the WebGIS version. The
WebGIS exercises will take much less time than
the GIS version, and have separate exercise de-
scriptions. The WebGIS version can be accessed
at: http://geoserver.itc.nl:8181/cartoweb3/htdocs/
RISKCITY.php
The RiskCity training package is constantly
being updated and further improved. The plan is
to incorporate more Participatory GIS approaches
in the training package, as well as to include
more examples of the use of risk information in
environmental assessment and spatial planning.
There is also a plan to make a separate version
focusing on risk occurring in rural areas, taking
into account flooding, forest fires, drought, and
land degradation as the main types of hazards.
The training package is used regularly in courses.
The duration of this course can vary between 2
days and 3 months. The standard duration is 6
weeks (distance education version) or 3-4 weeks
(fulltime course), depending on the option for
doing the final project in session 8. The course
materials are available in English, with major parts
also translated in Spanish, Vietnamese, Chinese,
and Georgian languages. The RiskCity training
package is given annually as a distance education
course on Multi-hazard risk assessment. Courses
on multi-hazard risk assessment are offered an-
nually in the Netherlands, Mexico, Bolivia, and
Thailand, and frequently also in India and China.
We are currently exploring the option to organize
an annual course on multi-hazard risk assessment
with partner organizations from Africa. Informa-
tion can be found on http://www.itc.nl/Course-
Finder (Select the option disaster management
for an overview of all courses).
ACKNOWLEDGMENT
We would like to thank Gonzalo Funes from
Honduras for providing the initial data sets. The
Digital Surface Model and flood information
1129
 The RiskCity Training Package on Multi-Hazard Risk Assessment
was obtained from a study by the United States
Geological Survey. The high resolution image
was obtained from a project funded by JICA.
Ruben Vargas Franco, Dinand Alkema, Lorena
Montoya, Michiel Damen, Nanette Kingma,
Antonio Naverette, Jean Pascal Iannacone,
Manzul Hazarika and Norman Kerle are thanked
for their contributions on various aspects of this
case study. Colleagues from ADPC, AIT, CDUT,
ICIMOD, UNAM, CLAS, IIRS and UGM are
thanked for the friendly collaboration and the
testing of the training package in various train-
ing courses. The following persons are thanked
for the translation into Spanish: Carlos Saavedra,
Jose Antonio Navarrete, Ruben Vargas, Edward
Gonzalez, Estuardo Lira, and Manolo Barillas.
Fan Xuanmei is thanked for translating part of it
in Chinese. Kakha Bhattak This work is part of
the United Nations University – ITC School for
Disaster Geo-Information Management (http://
www.itc.nl/unu/dgim).
REFERENCES
ADPC. (2005). Knowledge development, educa-
tion, public awareness training and information
sharing: A primer of disaster risk management in
Asia. Retrieved from http://www.adpc.net
Blaikie, P., Cannon, T., Davis, I., & Wisner, B.
(1996). At risk: Natural hazards, people’s vulner-
ability, and disasters. London, UK: Routledge.
DEBRIS. (2006). Innovative education for risk
management: Development of innovative forms of
learning and teaching oriented towards building
a family of new curricula in the field of natural
risk. Retrieved from http://www.e-debris.net
EM-DAT. (2009). The OFDA/CRED international
disaster database. Brussels, Belgium: Université
Catholique de Louvain.
1130
EMA. (2002). Disaster loss assessment guidelines.
Retrieved from http://oneresponse.info/resources/
NeedsAssessment/publicdocuments/EMA%20
-DisasterLossAssessmentGuidelines.pdf
ESRI. (2010). Virtual campus courses: HAZUS-
MH (multi-hazards) for decision makers. Re-
trieved from http://training.esri.com/Courses/
FEMA. (2010a). Mitigation planning: “how to”
guides. Retrieved from http://www.fema.gov/
plan/mitplanning/risk.shtm
FEMA. (2010b). Hazus, FEMA’s software for es-
timating potential losses from disasters. Retrieved
from http://www.fema.gov/plan/prevent/hazus/
FEMA. (2010c). Understanding your risks. Iden-
tifying hazards and estimating losses. Retrieved
from http://www.fema.gov/library/viewRecord.
do?id=1880
Harp, E. L., Castaneda, M., & Held, M. D. (2002).
Landslides triggered by Hurricane Mitch in Te-
gucigalpa, Honduras (Tech. Rep. No. 02-0033).
Retrieved from http://pubs.usgs.gov/of/2002/
ofr-02-0033/
ILWIS. (2010). The integrated land and water
information system. Retrieved from http://52north.
org/
International Institute for Geoinformation Science
and Earth Observation (ITC). (2004). Módulo de
capacitación: Aplicación de Sistemas de Infor-
mación Geográfica y Sensores Remotos para el
Análisis de Amenazas, Vulnerabilidad y Riesgo.
Retrieved from http://www.itc.nl/external/unesco-
rapca/start.html
IPCC. (2007). Fourth assessment report: Climate
change 2007: Impacts, adaptation and vulner-
ability. Retrieved from http://www.ipcc.ch/ipc-
creports/ar4-wg2.htm
The RiskCity Training Package on Multi-Hazard Risk Assessment

Mastin, M. C., & Olsen, T. D. (2002). Fifty-year UN-ISDR. (2005). Hyogo framework for action
flood-inundation maps for Tegucigalpa, Honduras 2005-2015: Building the resilience of nations and
(Tech. Rep. No. 02-261). Retrieved from http:// communities to disasters. Retrieved from http://
pubs.usgs.gov/of/2002/ofr02261/ www.unisdr.org/wcdr/intergover/official-doc/L-
docs/Hyogo-framework-for-action-english.pdf
Munich Re. (2009). Topics GEO: Natural catastro-
phes. Retrieved from http://www.munichre.com UN-ISDR. (2009). Terminology on disaster risk
reduction. Retrieved from http://www.unisdr.
NAHRIS. (2006). Dealing with natural hazards
org/eng/terminology/terminology-2009-eng.html
and risks. Retrieved from http://www.nahris.ch/
van Westen, C. J. (2010). GIS for the assessment of
Peters Guarin, G. (2008). Integrating local knowl-
risk from geomorphological hazards. In Alcantara-
edge into GIS based flood risk assessment, Naga
Ayala, I., & Goudie, A. (Eds.), Geomorphological
city, The Philippines (Doctoral dissertation, Wa-
hazards and disaster prevention (pp. 205–219).
geningen University). ITC Dissertations, 157, 352.
Cambridge, UK: Cambridge University Press.
Smith, K., & Petley, D. N. (2009). Environmental doi:10.1017/CBO9780511807527.017
hazards: Assessing risk and reducing disaster.
World Bank. (2010). Central American probabi-
London, UK: Routledge.
listic risk assessment (CAPRA). Retrieved from
UN-ISDR. (2004). Living with risk: A global http://www.ecapra.org
review of disaster reduction initiatives. Retrieved
from http://www.unisdr.org/eng/about_isdr/bd-
lwr-2004-eng.htm

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 3, Issue 1,
edited by Donald Patrick Albert, pp. 41-52, copyright 2012 by IGI Publishing (an imprint of IGI Global).

1131
1132

Chapter 68
Automatic Classification of
Decorative Patterns in the
Minoan Pottery of Kamares Style
Filippo Stanco
University of Catania, Italy

Davide Tanasi
Arcadia University, USA

Giuseppe Claudio Guarnera

University of Catania, Italy

Giovanni Gallo
University of Catania, Italy

ABSTRACT
An important feature of the Minoan culture is the pottery of Kamares style, that documents the Cretan
cultural production between the first half of the 2nd millennium BC. This high level painted production,
characterized by the combination of several diverse motifs, presents an enormous decorative repertoire.
The extraordinary variety of combinations between elementary motifs according to a complex visual
syntax makes interesting the automatic identification of the motifs, particularly upon potsherds. A com-
plete pipeline to accomplish this task is still a challenge to Computer Vision and Pattern Recognition.
Starting from a digital image ROI identification, motif extraction, robust contour detection should be
performed to obtain a bag of digital shapes. In a second phase each of the extracted shapes has to be
classified according to prototypes in a database produced by an expert. The co-occurrence of the dif-
ferent shapes in a specimen will, in turn, be used to help the archaeologists in the cultural and even
chronological setting.

DOI: 10.4018/978-1-4666-2038-4.ch068

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
INTRODUCTION
In the past decades, the application of the com-
puter science in the archaeological research, and
especially in the field of prehistory, turned out
from a simple auxiliary technology into a cog-
nitive strategy influencing the approach to the
ancient artifacts. Although the introduction of
the database enhanced the possibility of dealing
with large amount of text data, the problem of
taking into account huge groups of visual data
still remains unsolved. One common case that
makes hard the initial steps of an archaeological
study is for example the analysis of decorative
repertoires of some prehistoric pottery classes
that are characterized by the exuberant use of
a multiplicity of motifs. The most complicated
artistic production of the Mediterranean prehis-
tory is certainly the Kamares style pottery (Figure
1), flourished in Crete in the first half of the 2nd
millennium, which main feature is to present a
complex system of polychrome painted decoration
with a rich decorative alphabet aimed to produce
an endless variation of visual results. The highest
number of vessels and potsherds available of this
class all over Cretan territory and the fact that a
complete framework for the Kamares decorative
grammar and the associations ‘motif/vessel/site/
Figure 1. Examples of Kamares style vessels (this image has been obtained as a collage from several
public sources)
chronological layer’ is far from being understood,
determined a stop in the cognitive progress of the
Minoan civilization.
In this perspective Computer Vision and Pat-
tern Recognition could provide a great support in
automatically assisting the archaeologists in clas-
sification of Kamares pottery fragments (Figure2),
especially considering that in many cases, the
visual informations available are of several kinds,
like watercolors, black and white and color pho-
tographs of unequal resolution and quality, pencil
sketches and high quality digital photographs.
Within the research program Archeomatica Proj-
ect (IPLab, 2010) of the Catania University, de-
voted to the application of Computer Graphics
and Image Processing techniques in the field of
prehistoric archaeology (Stanco, Battiato, Gallo,
2011), a specific study for the developing of an
automatic classification system of the Kamares
simple decoration elements (Farinella, Stanco,
Tanasi, 2008), (Guarnera, Stanco, Tanasi, Gallo,
2010) (Gallo, Stanco, Tanasi, in press) started
under the inspiration of the Center of Cretan
Archaeology (2010) of the Catania University,
dealing with the Kamares pottery from Phaistos
(Levi, Carinci, 1988), (Carinci, 1997).
The application of this research strategy on
the available visual corpus of image data is the
1133
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style

Figure 2. Selection of Kamares pottery sherds from Phaistos (courtesy of Prof. V. La Rosa)

object of the research that we present in this
chapter. The chapter is structured as follow: in
Sections 1.2 we introduce the Kamares pottery.
Section 1.3 shows the proposed technique. In
Section 1.4 a review of existing shape similarity
techniques
Background: The Kamares Pottery
The Kamares style pottery represents the main
artistic feature of the Minoan civilization between
the 20th and 17th century BC and its technical
and stylistic level achieved was never equaled in
the Aegean Bronze Age (Walberg, 1978), (Wal-
berg, 1983), (Walberg, 1987), (Walberg, 2001),
(Knappett, 1999), (Zois, 1998). This class, with
a large lifespan that covers the Protopalatial and
Neopalatial periods, can be divided in four main
phases related to the local chronology arranged
by A. Evans (1921): Pre-Kamares or 1 (Middle
Minoan IA), Early Kamares or 2 (Middle Minoan
IB/IIA), Classical Kamares or 3 (Middle Minoan
IIA/IIB/IIIA), Post-Kamares or 4 (Middle Minoan
IIIA/IIIB). Kamares ware is named for finds first
identified at the end of 19th century in the Kamares
cave sanctuary in Mount Ida, and its diffusion on
the Cretan territory can be distinguished between
the elegant production of the palatial centers of
Knossos and Phaistos (Palatial Kamares) (Day,
Wilson, 1998) and the less impressive creations

Figure 3. Map of Crete showing principal sites with Kamares style pottery (Walberg, 1983)

1134
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
of the provincial workshops of Eastern, Western,
Central and Eastern-Central Crete (Provincial
Kamares areas 1-4) as in Figure 3.
Fine fabric, thrown on the wheel, sometimes
with very thin bodies (2-3 mm like in the egg shell
varieties), painted with polychrome pigments on
dark background, the Kamares style has a vast
assemblage of shapes and its distinctive feature
is a huge repertoire of exuberant decorative mo-
tifs often resembling naturalistic atmospheres
(Figure 1). About the colors, the black slip which
covers most of Kamares vessels has been identi-
fied as a clayish paint layer containing iron oxides
of spinel type with a high content of potassium.
It is produced by exposure to high temperatures
in a reducing atmosphere, thereby turning in into
a black sintered layer. The red color consists of
red ochre, which in contrast with the black slip
does not contain potassium at all. As a result of
the lower content of potassium, the red paint
layer does not melt and turn into a sintered layer
in a reducing atmosphere but it remains porous
and turns bright red when exposed to oxidation.
The white pigment is composed by calcium sili-
cate or talc depending on the chronological pe-
riod of the production and of the production
centre itself. In the decoration of large vessels of
domestic type also a less elaborated and much
lasting dark on light decoration system is used as
Figure 4. Schematic shape taxonomy of the Kamares pottery repertoire (Walberg, 1978)
it was found appropriate for vases which had to
be strong and practical rather than attractive.
In a preliminary survey, seventy-four different
shapes were identified including both the deco-
rated and fine ware and domestic and specialized
forms. During the four chronological phases, open
and closed shapes are equally preferred and even
if there is a general conservativeness in the use
of the same formal typologies some specific ten-
dencies for each phase can be defined (Figure 4).
For what concerns the decoration, the Kamares
style is basically characterized by the application
of thirty-one different core motifs (abstract, rec-
tilinear, pictorial, pictorialized motifs) and patterns
(stone, rock, sponge and metal surface patterns),
each of them with several levels of varieties,
composed together in order to obtain complex
decorative syntaxes joined to create much more
complicated composite designs.
In general the repertoire of core motifs main-
tains itself the same in both the main production
districts, the Palatial and Provincial, and basically
remains unaltered during all the chronological
development, just with few elements peculiar for
some phases. But it must be also considered that a
core motif, even simple, can deeply change shape
when transformed from a 2D sketch to a 3D design
on the vessel body, when combined with copies
of the same motif or joined with other motifs,
1135
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
Figure 5. Schematic taxonomy and possible elaborations of some Kamares decorative motifs (Walberg,
1978)
when elaborated or pictorialized and in relation to
different chronological phases. As a result all this
makes very hard, in some cases, the classification
of each element (Figure 5).
About the composition of decorative elements
and its relation with the tectonic of the vessel,
two different method can be identified. First is
the unity decoration that depends on the shape of
the vase and that can be circumcurrent, that means
planned with regard to the whole vessel body and
composed by encircling zones, facial, that means
planned with regard only to one part of the vessel
body as seen from a special angle, and zonal, that
means composed by a system of decorated zones
related to each other in order to create coherent
designs. Second is the structural decoration that
consists in separation of the vase into different
parts and accentuates the horizontal and vertical
axes. In the both type of decoration a field division
for the location of the motifs is applied, selecting
different parts of the body vessels for the defini-
tion of the limits of the decorated area. In the field
division the choice of the decoration strategy for
the accessory parts of the vase, like rim, lip,
handle, spout, is very peculiar of each period and
production centre. It must be also considered that
both the shapes and the decoration contain dy-
namic effects and the decoration cannot merely
be described as a movement across a surface, as
in the case of a static background. It must be
described as a movement in the same direction
1136
or opposed to the movement inherent in the ves-
sel shape. The dynamic effects of the vase and
the decoration together create the overall effect,
that can give, for example, the illusion of the
contraction or expansion of some vessel parts. In
the selection of the composition strategies it’s
possible to find difference between the Palatial
and Provincial production and to define tendencies
from the Pre-Kamares to the Post-Kamares
phases (Macgillivray, 1998).
The overall effect of a Kamares vessel is
sometimes completed by the application of an
accessory decoration represented by incision,
relief and applied plastic details or by the use of
complementary embellishment techniques like
the ripple, barbotine and the trickle decoration.
Finally, it must be noticed that differences can
be find also within the same centre and in the
same period because, besides the guidelines of
this production, the essence of the style itself was
strongly influenced by the creativity and aesthetic
concept of the single artisan, that constantly elabo-
rated the core motifs and experimented new ways
of expression. As a result, the Kamares style is
more than the simple combination of syntaxes of
elementary motifs with different vessel type. It is
a complicated interplay, carried out to the small-
est detail, where every elements has an important
part, especially in the Palatial production. The
fact that each Kamares style product depends by
many quantifiable variables (time, place, vessel
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
shape, high number of probable composition of
core decorative elements, colors and accessory
decoration techniques) and just by one variable
not quantifiable at all that is the unpredictable
will of the artisan, it made quite impossible for
the archaeologists specialized in the Middle
Minoan pottery to carry out a exhaustive study
of the main features of this style, also because of
the huge amount of data represented by millions
of decorated potsherds.
The first study of classification of the Kamares
pottery was carried out by G. Walberg in 1978 and it
was dedicated to the Palatial production of Phaistos
and Knossos (Walberg, 1978), (Walberg, 1987)
(Figure 3). In this work, the role of Phaistos, as the
most important production centre and also the only
site with a clear chronological progression in the
development of the production itself, was pointed
out and a corpus of core motifs and patterns found
there was published. Furthermore, the features
characterizing the climax of the Kamares style
were find between Middle Minoan IIA and Middle
Minoan IIIA, within the phase named Classical
Kamares. In 1983 a second reassessment of the
Kamares materials from Provincial districts was
carried out by the same author (Walberg, 1983),
emphasizing analogies and differences between
Palatial and Provincial production and trying to
rebuild the decorative repertoire, as whole as
possible, in order to enrich the main assemblage
and obtain a complete study and classification of
the Kamares pottery.
After the Walberg’s corpora, a large quantity
of new Kamares vessels were found and published
together with groups of pottery coming from
old excavations, in both cases often fragmented
potsherds. This re-opened the problem for the
archaeologists of dealing with a class of materials
so variable present in huge amount of specimens.
In particular the interpretation of the decorative
motifs and syntaxes partially preserved on the
sherds, in order to ascribe them to the repertoire
of a specific workshop, trying to match the frag-
mentary information with a standard collection
of visual references arranged by the scholars, has
become the hardest part of the research.
Although the goal of an exhaustive classifica-
tion of Kamares pottery’s production is likely to
be unobtainable with the present state of the art,
Computer Vision and Pattern Recognition could
provide a great support in automatically assisting
the archaeologists in the classification task. Our
contribution illustrates a complete pipeline to
automatically process these data.
The processing starts with the extractions of
a clean representation of the decorative designs.
Our goal is to create a standardized database of
“shapes” that could be successively automatically
investigated with Pattern Recognition methods.
Iconographical
Documentation Available
Besides the large amount of specimens available,
another significant problem in an exhaustive study
of the Kamares style pottery is dealing with an
heterogeneous group of iconographical data.
For the two most important Palatial production
centers, Phaistos and Knossos, the documenta-
tion is basically composed by watercolors, black
and white pictures, technical drawings in scale
with front view and section and reconstructive
unscaled perspective drawings. The best preserved
vessels representing the highest stylistic level of
the Kamares style, in the editions of the two sites,
were illustrated by unscaled watercolors carried
out by E. Stafani, R. Oliva and Th. Fanourakis
for Phaistos and D. Mackenzie for Knossos. In
the reproduction of the complex Kamares painted
decoration, each modern artist or technician at
work was influenced by his own personal taste
and this caused, in many cases, alteration and
distortion of the original shape of the motifs and
of the original scheme of the syntaxes.
It must be also considered that the draftsmen
working in Crete in the first half of the XX century,
had different formation and artistic education.
For the reproduction of the Phaestian Kamares,
1137
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
E. Stefani, chosen by the director L. Pernier for
the documentation of the excavation, was an ar-
chitect. Decades later, D. Levi, new head of the
expedition, chose R. Oliva and Th. Fanourakis, two
painters strongly influenced by his archaeological
point of view. Otherwise at Knossos, a large part
of the graphical documentation, including the
watercolors, was carried out by D. Mackenzie
(Momigliano, 1999), that was an archaeologist,
second in charge after A. J. Evans, head of the
mission. In this case he demonstrated a more
straight and scientific method of depicting artistic
features (Mackenzie, 1906).
Furthermore, the fact that the documentation
of Kamares vessels from Knossos dated back to
1921-1935 (Evans, 1921) while that of Phaistos
was performed in 1935-1951 and 1976 (Pernier,
1935), (Pernier, Banti, 1935), (Levi, 1976), it
determined different choices in the reproduction
strategies, like different line thickness or colors,
due to the changing taste of the time (Figure6).
Finally, another misleading problem in the inter-
pretation of the Kamares pottery from Knossos
come also from the reconstructive drawings, where
in many cases the hypothesis of the missing part
of a motif or of a decorative outline was denied
by the subsequent findings.
In this paper the results achieved working
basically on the visual archive provided by Wal-
Figure 6. Watercolors of Kamares style pottery and potsherds: (a) Phaistos 1935 drawings by E. Stefani
(Pernier, 1935); (b) Phaistos 1976 drawings by R. Oliva (Levi, 1976); (c) Knossos 1935 drawings by
D. Mackenzie (Mackenzie, 1906)
1138
berg’s corpora (Walberg, 1983), (Walberg, 1987)
will be discussed.
KAMARES SHAPE ANALYSIS:
SYSTEM OVERVIEW
The automated system that we are building to
assist the archaeologist in assessing similarities
among the Kamares findings follows a general
pipeline that could be easily adapted to other
similar case studies. As in most Computer Vision
applications the idea is to go from the raw data
(pixels) to a symbolic representation of the image
content. These information may, in turn, be fed
to an intelligent system that assists the experts in
formulating and checking working hypotheses
about the scenes captured in the original images.
More specifically for the present case study the
system starts with digitalized format of a pictorial
representation of a vase or of a sherd. Figure 7
shows the successive steps performed. The digital
data are the input of a contour extraction algorithm.
Please, notice that in this way we are disregarding
any 3D information: perspective deformation, and
occlusions together with noise and artifacts due
to low resolution are simply ignored at this stage.
Contour extraction produces a collection of
contours: some of them are not relevant for our
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style

Figure 7. The complete pipeline of the proposed technique

application and should be filtered out (Gonzalez,
Woods, 2006), (Gonzalez, Woods, Eddins, 2009),
(Moore, 1968). We choose to ask the interventa-
tion of the expert for this Region Of Interest (ROI)
identification task because of the occurrence of
cases like those reported in Figure 8. In the Figure
several occurrences of the “raquet” decorative
pattern are shown. It is clear that the isolation of
this complex motif is the product of an informed
semantic choice and that this expert choice is not
easily mimicked with a fully automatic approach.
ROI identification most commonly reduces the
set of contour shapes to examine. Even so Kamares
decorative motifs are reduced by this step into a
smaller but yet complex set of elementary shapes.
Figure 9 shows several instances of the pattern
that are obtained after ROI selection. Observe that
in some case it is more convenient to refer to the
“flattened” version of the motifs whenever these,
manually produced by an expert draftsman, are
available. These lucky cases reduce the problems
due to 3D distortion and occlusion but introduce
in the pipeline the bias of the draftsman.
The proposed system deals with simpler shapes
one at each time. They are analyzed by the shape
analyzer and reduced to a set of numerical and
geometrical invariants. These invariants are the
final features that the matching module of the
system considers for search and similarity re-
trieval into a reference shape database.
The reason to deal with the elementary simpler
shapes instead that with the set of contours in a
ROI has the following motivations:
• Algorithms that deal with simple contours
and extract suitable invariants from them
are much better understood and robust at
the actual state of art;

Figure 8. Examples of occurrences of the “raquet” motif in different vases

1139
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
Figure 9. Kamares potteries and relative decorations
• A ROI represents a complex figurative
“proposition” composed of simpler el-
ementary shapes. Unfortunately, the “order
of reading” of such complex visual struc-
tures is not linear. Indeed it is not easy to
“read” those drawing in a canonical un-
ambiguous way.
Shape Similarity Measures
To apply the technique presented in this paper
is necessary a contour line representation of the
motifs. Hence, a decorative pattern is previously
translated into a digital sequence of consecutive
points on a raster plane.
Several issues about resolution and standard-
ization of these rasterized contours arise. Going
from digital images to digital shapes is far from
being an easy task: different media (water colors,
photos and hand drawn schematic lines) require
Figure 10. (a) Figure interpretation by the archaeologist. (b) Figure interpretation by the computer
scientist. The same color indicates the same symbol.
1140
different methods of contour extraction and pres-
ent different algorithmic challenges. Once the
pictorial data have been translated into a more
abstract “contour shape” representation it is pos-
sible to apply to them the shape similarity measures
presented below. The idea is to present the shape
database with a “query” which represents a
decorative element that the archaeologist wishes
to recognize. The answer to the query is an estimate
of the similarity distance between the query shape
and the shapes in the database. In this chapter we
present different shape similarity algorithms:
contour flexibility technique (Xu, Liu, Tang,
2008), Shape Context (Belongie, Malik, Puzicha,
2002), logo recognition into document images
(Shen, Jin,Chang, Wu, 2005), Procrustes methods
(McNeill, Vijayakumar, 2006), and Circular
Blurred Shape Model descriptor (Escalera, Fornes,
Pujol, Escudero, Radeva, 2010).
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style

Notice that even if the published ideas in shape The shape context considers the set of n-1 vec-
recognition are overabundant in the present scien- tors originating from a point to all other sample
tific literature most of them are relative to complex points on a shape, which expresses the configura-
vision tasks where one wishes to recognize real tion of the whole shape with respect to the reference
objects that move and deform within a complex point. Since shapes may vary from an instance to
real environment. This is not the case for our ap- another in the same category, the full set of vectors
plication where the most common deformations is too detailed and does not guarantee robustness.
observed are affine distortions, noise, irregulari- The distribution over relative position is a more
ties, lacunae, and non linear scaling. robust and compact description, hence for each
point pi on the shape, a coarse histogram hi of the
Shape Matching using relative n-1 coordinates is computed:
Shape Context
hi (k ) = #{q ≠ pi : (q − pi ) ∈ bin(k )} (1.1)
Belongie et al. in 2002 propose a correspondence-
based shape matching method using shape contexts
The length r and orientation θ of the vectors
(Belongie et al., 2002). Shape matching using
(q- pi) are quantized to create a histogram map
shape contexts is an improvement to traditional
which is the shape context used to represent the
Hausdorff distance based methods. It extracts
point pi. To make the histogram more sensitive to
a global feature, called shape context, for each
the location of nearby points than to the location
corresponding point. The matching between cor-
of points farther away, these vectors are mapped
responding points is then the matching between
into a log-polar space representation (Figure11).
the context features. In this approach an object is
Shape contexts are distribution represented as
treated as point set and it is assumed that the shape
histograms: it is hence possible to use the χ2 test
of the object can be captured by a finite subset of
statistic to define the cost of matching two points.
its points, for example sampled uniformly from
Consider two shapes P and Q. A point pi on the
the external contour. These points do not need
first shape and a point qj on the second shape, Ci j
to be key-points such as inflection points, cusps,
= C(pi, qj) denotes the cost of matching the two
extrema of curvature, etc., and can be derived
points in exam, where
from edge pixels found by a simple edge detec-
tor, whose output is a set P of n points P = {p1,
p2, …, pn}, pi ∈ ℜ2.

Figure 11. The log polar maps of the contest of two different points in the contour of a shape

1141
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
2
K h (k ) − h (k )
C (pi , q j ) = 1 / 2∑ 
 i j  ing regions of shapes often appear at slightly differ-
h (k ) + h (k )
k =1 i j
and hi(k), hj(k) denote the K - bin normalized
histogram respectively at pi and qj. Given the set
of costs Ci j between all pairs of points pi of P and
q j . of Q, the total cost of matching
H (π) = ∑ i C (pi , q π(i ) ) has to be minimized,
with the constraint of one-to-one matching. The
matching of two shapes then is done by matching
two context maps of the shapes, which is a matrix-
based matching. It minimizes the total cost of
matching between one context matrix and all the
permutations of another context matrix. To reduce
the matching overhead, the shortest augmenting
path algorithm for the matrix matching is used.
When the number of sample points is not equal
on the two shapes, the matching is done adding
dummy nodes to the smaller point set, with a
constant matching cost of εi. Since all measures
are taken with respect to points on the object,
invariance to translation is intrinsic to the defini-
tion of the descriptor. The use of histograms makes
shape context insensitive to small deformations,
non linear transformation occlusions and presence
of outliers. Additional robustness to outlier can
be added not allowing points labeled as outlier to
contribute to any histogram. The above definition
of shape context is not invariant to rotation, which
can be obtained using a relative frame instead of
the absolute frame described, e.g., treating the
tangent vector at each point as the positive x-
axis.
Hierarchical Procrustes Matching
Hierarchical Procrustes Matching (HPM) (Mc-
Neill, Vijayakumar, 2006) is a segment-based
shape matching algorithm that approximates
the perceptual similarity between two shapes by
matching progressively smaller boundary seg-
ments. The denomination makes reference to a
mythological character that used to stretch or cut
1142
his guests to make them fit to his bed. Correspond-
ent scales and positions and matching algorithms
must be flexible to adjust these variations. Simple
global point methods (McNeill, Vijayakumar,
2005), (Zhang, Zhang, Krim, Walter, 2003) and
non linear approaches (Bookstein,1997), (Sebas-
tian, Klein, Kimia, 2003) suffer from presence
of smooth deformation and may produce a large
discrepancy between a perceived difference and
its numerical estimation. HPM avoids this by
matching in a global to local direction. Longer
segments that have been matched provide initial
matches for the shorter segments, which can then
slide and stretch/contract in order to find the best
matches at this smaller scale.
Given two objects represented by n 2D points:
U = (U1, …, UN)N and V = (V1, …, VN)N ∈ ℜN ×2
this approach normalizes the boundary length of
the polygon associated with V and then transforms
U to match V. Each point is represented as a com-
plex number Vn = (xn,yn) → xn + i yn = wn ∈  .
Then, V→ w and U → z ∈ N . After centering w
and z so that ∑ n =1 wn = ∑ n =1 z n = 0 , the Pro-
N N
crustes Distances (PD) between U and V is given
by
d p (U ,V ) = (w * w − (w * zz * w ) / z * z ) / N
(1.2)
In order to compute HPM is necessary to
compute the Global Procrustes Matching (GPM)
that find the best correspondence between U
and V with the smallest PD among the N valid
correspondences that respect the cyclic order
of the indices. This initial correspondence is in
turn refined using a vector that is one half of V
and finding the smallest PD with each “half” of
U. The matched segments at this length provide
initial estimates for matching at the successive
scale that can be performed recursively. The shape
similarity of U and V is the weighted sum of scores
at each length, where the weight is related to the
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
segments length. Let Sl denote the sum of PDs
over the matched segments at a fixed length l.
The asymmetric similarity of U and V is given by
dF (U ,V ) ≡ w100d p (U ,V ) + ∑ wl sl contour flexibility is defined as ω(p, r) = min(ω+(p,
l =50,25,12.5
(1.3)
where the wl are constant weights, 100 is the
boundary length, and 50, 25 and 12.5 are length
of segmentswhen we divide the boundary. Since
dF(U, V) ≠ dF(V, U), the symmetric similarity is
given by
DF (U ,V ) ≡ dF (U ,V ) + dF (V ,U ) (1.4)
In order to consider the average confidence
with which each segment match is selected, the
PDs are normalized by making the average PD
equal to 1, and then apply HPM as normal. This is
also known in literature as Normalized Procrustes
Distances.
Shape Matching by
Contour Flexibility
It this Section we briefly report the main ideas
about contour flexibility. The reader is urged
to find the details of this algorithm in (Xu, Liu,
Tang, 2008). Vision science has developed as
an interdisciplinary research field, frequently
involving concepts and tools from computer sci-
ence, image processing, biology, psychology and
cognitive science, for example the importance
of high curvature points in shape perception has
been already described in (Atteave, 1954) and
(Biderman,1985). From this theories the important
features for shape recognition come from parts of
the object where changes occur (e.g., corners). A
recent descriptor, named contour flexibility (Xu,
Liu, Tang, 2008), depicts the deformable potential
at each point along a closed curve, extracting both
global and local features, with proper trade-off
between them. It can be observed that in most
cases articulated high flexible parts correspond
to high curvature points. Let ℓ be a closed simple
contour, surrounding a bounded domain D. For
a point p on a contour ℓ and a given radius r, the
r), ω-(p, r)), where ω+ is the interior flexibility
and ω- is the exterior flexibility:
∫C p+,r
k+ (x )dx
ω+ (p, r ) = (1.5)
∫ C p+,r
dx
∫C p−,r
k− (x )dx
ω− (p, r ) = (1.6)
∫ C p−,r
dx
C p+,r and C p−,r respectively are the connected
components containing p in the sets {x ∈ D:║ x- p
║≤ r} and {x ∈ ℜ2 -D:║ x- p ║≤ r}. k+ and k- are
two function of distance transform on ℜ2 ,
k+ (x ) = d (x , ℜ2 \ D ) and
k− (x ) = d (x , D ), x ∈ ℜ2 (1.7)
d(⋅,⋅) is the minimum Euclidean distance between
two sets. The radius r is called bendable size and
should be tuned taking into account the width of
the limb-like parts of an object. The bendable size
r and the contour flexibility are proportional to the
scale of the contour but invariant to translation,
rotation and the choice of starting point for the
parameterization of the contour. Let
z(t) = (x(t), y(t)); 0 ≤ t < 1 (1.8)
be the arc-length parameterization of a contour
ℓ. Since it is difficult to match flexible parts of
two contours (e.g., using uniform sampling of the
contour), a better strategy is to give large weights
to inflexible landmarks and smaller weights to
more flexible landmarks, using more samples
1143
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
on the segments of a contour which are more
inflexible. Let
t
∑ ω(z (u), r )du
γ(t ) = 0
1
(1.9)
∑ ω(z (u), r )du
0
the optimal sampling can be obtained with sam-
pling speed dγ/dt. Considering each landmark a
complex number zk =xk+ jyk, the contour can be
treated as a complex vector Φ=(z1, z2,…, zn)T and
the Procrustean Distance between two sequences
can be used for global matching of Φ1 and Φ2 :
(
dˆ(Φ1, Φ2 ) = cos−1 max S∈N Φ1, σ s (Φ2 ) ) regions. The inverse of these distances are com-
(1.10)
where σ s (Φ2 ) is a cyclic permutation of Φ2 , to
achieve independence from the starting point and
“ ” denotes the inner product of two complex
vectors. The matching score between the two
shapes represented by Φ1 and Φ2 is determined
by
αˆ
Md =
π
( )
d (Φ1, Φ2 ) + D Ω1, Ω2 (1.11)
where D(⋅,⋅)is the warping distance between the
two sequences of the contour flexibility values
extracted from the contours; Ω1, Ω2 are the se-
quences of the values of the contour flexibility at
each landmark of uniformly sampled sequences,
starting form the leading landmark ŝ found by
(1.10), α weighting factor, π normalization factor.
Circular Blurred Shape Model is a recent rota-
tionally invariant descriptor, which makes use of
a correlogram structure to capture spatial arrange-
ment of object parts, shared among regions defined
by circles and sections (Escalera et al., 2010). The
correlogram structure can be defined as follows:
1144
given a number of circles C, number of sections
S, and an image region I, a correlogram B = {b{1,1},
…, b{C, S}} is a radial distribution of sub-regions
of the image. In each region b of the correlogram
the centroid b* can be located by its coordinates
and used as reference point. The regions around b
define the neighborhood of b, and the number of
neighbors depends on the spatial location b (e.g.
inner circle, middle circle or extern circle).
The descriptor makes use of the information
obtained from the contour of the object, which
can be extracted for example by means of an edge
detector. Every point x in the contour map is taken
into account, calculating first the distance to the
corresponding centroid of the region in which x
lies, then the distances to the centroids of neighbor
puted and normalized by the sum of total distances
and the obtained values are then added to the cor-
responding positions of the descriptor vector n,
which as dimension C×S. As in the case of shape
context, the use of histograms makes description
tolerant to irregular deformations. As for the com-
plexity, for a map of k relevant contour points, the
computation of the descriptor requires O(k) simple
operations. Parameters C and S defines the degree
of spatial information taken into account in the
description process (i.e., as the number of regions
increase the description becomes more local) and
should be tuned for each particular application.
See also Figure 12. To obtain a rotationally invari-
ant descriptor a second step needs to be included.
The main diagonal Gi of correlogram B, with the
highest density, is searched. This diagonal is then
used as a reference to rotate the descriptor. The
orientation of the rotational process, so that Gi is
aligned with the x-axis, is that corresponding to
the highest description density at both sides of Gi.
Once obtained the rotationally invariant CBSM de-
scriptor, it can be used to design a symbol spotting
methodology (e.g., using Adaboost to distinguish
between foreground and background) or a multi-
class classifier by embedding binary classifiers,
for instance using Adaboost to define a classifier
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
Figure 12. Circular Blurred Shape Models of the same shape with different choices of the parameters
C and S.
based on the features that best discriminate one
class against another, combining then binary clas-
sifiers in a Error Correcting Code Framework.
EXPERIMENTAL RESULTS
To test our system we have built an experimental
image database. The database has been created as
follows. First of all, a selection of images from
(Walberg, 1978) have been inserted in the col-
lection. A total of 36 motifs have been chosen.
The selection of these motifs among the much
larger catalogue in (Walberg, 1978) has been done
randomly, although in this initial stage of our
research we choose to sort out the most complex
motifs. Instances of these chosen patterns show a
high-class variability in terms of scale, rotation,
rigid and elastic deformations when observed in
the pictorial reference corpus. Some of the shapes
exhibits a low inter-class variability especially
for simpler shapes. This explains for example
some resulting ambiguities of the classification.
The images selected are in a way the “canonical”
reference for each of the decorative motifs to rec-
Figure 13.
ognize. To use as a reference only the clean and
canonical drawing prepared by an expert is too
demanding for any shape recognition algorithm.
Following a general praxis in the shape recogni-
tion community we hence enriched the database
with three variations for each decorative motif.
These variations have been artificially obtained
applying the following transformations: rotation in
clockwise sense of about 33 degrees, perspective
camera distortion where camera orientation has
been assigned at random and random warping.
Examples of the shape included in the database
are reported in Figure 13, an example of varia-
tions over a given shape is reported in Figure 14.
Eventually the database is made of 144 shapes.
Those shapes have been stored in binary images
of 500×500 pixels. Those images have been pro-
cessed for contour extraction in Matlab (Lorensen,
Cline, 1987), (Maple, 2003). The resulting con-
tours have been resampled and for each shape a
vector of 100 points have been obtained. Using
a higher sampling rate does not improve the per-
formance of the system, perhaps because a
denser sampling of a contour tends to preserve
some of the noise of the original image making
1145
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
Figure 14.
recognition and matching a harder task. The
down-sampling has been performed in two dif-
ferent ways, and the results of both selections
have been stored in the reference database. The
first way to sample 100 points from a closed
contour is to compute 100 equally spaced pints
along the contour. Notice that the sample points
are obtained with cubic interpolations of close by
points on the contour. This uniform sampling
procedure is applied to be able to compute CBSM
distance and to estimate local contour flexibility
values. A second non-uniform sampling has been
also performed. This second sampling takes 100
points according to contour flexibility values:
more points are allocated in the most flexible
segments of the shape. This non-uniform sampling
procedure is applied to be able to compute contour
flexibility distance. Contour flexibility value in
each of the sampled point is also stored. In sum-
mary our system refers to this collection of 144×2
point vectors and 144 contour flexibility values
to compute similarity distances of queries from
the database. Observe that this mathematical
representation is quite compact and easily allows
the scaling up of the system up to thousands of
shapes.
Table 1. Performance of the shape matching procedures
% of perfect matches within
the four closest shapes
Circular Blurred Shape Model (CBSM) 48%
Contour Flexibility (CF) 75%
1146
We choose to check the performance only of
the CSBM algorithm and of the Contour Flexibility
approach. Indeed CSBM is a refinement of the
initial Belongie’s proposal and it is safe to assume
that it will perform better. A similar reasoning
justifies the choice of Contour Flexibility over
Hierarchical Procrustes Distance for our tests.
The experiments have been carried out as
follows. Each shape in the database has been, in
turn, considered as a query and the best matching
shapes (other than the query shapes itself) have
been considered as the results of the retrieval opera-
tion. In Table 1.1 we reported the results obtained
with this set of experiments. In particular the first
column of the Table indicates the percentage of
total queries when the algorithms found, within
the four most similar shapes, the query shape itself
and the three variations of it in the database. The
second column in the table indicates the percent-
age of total queries when the algorithms found,
within the three most similar shapes (other than
the query shape itself) the majority of shapes of
the correct class.
A closer observation of the results reveals that
most of the errors are “near misses”. An example
of these is given in Figure15.
% of majority of correct
matches within the four closest
shapes
72%
99%
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
Figure 15. Instance of false positives produced by the system, (a) CF; (b) CBSM
The results are very encouraging, although the
system is far from being error free it provides a
valuable help to the expert. When trying to match
a novel query an archaeologist, in almost all of
the cases, has to look only at a very small (4 to
6) reference shapes provided as an answer from
the system. This greatly reduces the time that he
should otherwise spend to look among a larger
set of candidate reference patterns. On the other
hand at this stage of development of the system
the human expert intervention is still necessary.
Real World Examples
To further support our claim about the robustness
and the usefulness of our approach we present two
real world examples. The queries and the results
are shown in Figure16. The application of the
algorithm for the automatic classification of the
Kamares style pottery revealed itself particularly
efficient for the immediate discrimination of the
motifs on the chronological and geographical
scales. Basing upon the Walberg’s corpora it’s
possible to offer two examples of the procedure
working. In the first case in Figure16(a), we have
a potsherd with a spiral type motif partially pre-
served. From the image the algorithm automati-
cally extracts the contour and checking for matches
in the available database it suggests three hypoth-
esis of interpretation, the simple j-spiral (Walberg
2.i.1), the running j-spiral (Walberg 2.i.2), even
possibly deformed, and the j-spiral with filled
angle (Walberg 6.2), again with a certain degree of
alteration. The immediate restriction of the range
of possible motifs gives to the archaeologists the
chance to easily check the available data on the
geographical and chronological distribution of
the motifs themselves and obtain confirmation for
their initial hypothesis. In fact, while the simple
J- spirals has a wide diffusion in both Provincial
and Palatial areas and in the four chronological
phases, the j-spiral with filled angle is present
in the Palatial area just in the phase 3 and in the
Provincial district exclusively in the phases 2 and
3 and just in the East and East-Central Crete (in
(Walberg, 1983) pp. 38-39, 41, pll. 28, 31; (Wal-
berg, 1987), pp. 48-49, 51, 180, 183). Again in the
second example in Figure16(b), in the potsherd
is visible a series of teardrops motifs. After the
extraction of the contour, the algorithm suggests
two different matches, the spiral derivatives (Wal-
berg 8.33), both the original motif and its altered
versions, and the petaloid loops (Walberg 12.i.1).
Also in this case, the two motifs have exclusive
distribution features. The spiral derivatives are
present in the Palatial area in phases 2-4 and are
1147
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
Figure 16. Examples of real queries to the database
absent in phase 1, while in the Provincial district
they occur in all the chronological phases but
restricted just to East, East-Central and Central
Crete. The petaloid loops have same distribution
in the Palaces as the above mentioned derivatives
but in the Provincial areas they are exclusive of
phase 1 and 3 also having the same geographical
diffusion ((Walberg, 1983), pp. 43-44, 48-49,
pll. 35, 41; (Walberg, 1987), pp. 52-53, 57-58,
183, 188).
CONCLUSION AND FUTURE WORK
In this chapter a pipeline to automatically classify
simple Kamares decoration has been presented. In
particular, we focused on the shape matching of the
pipeline showing how contour flexibility together
with Procustean distance, circular blurred shape
model or shape context may solve this problem.
The proposed system will be of great support in
automatically assisting the archaeologists in clas-
sification operations according time and place of
production of Kamares pottery fragments.
This first example of application of Computer
Vision and Pattern Recognition techniques to a
specific topic of the prehistoric archaeology as
the Minoan Kamares pottery opened a completely
new field of investigation that in the future can
deeply change the approaching of the scholars
to iconographical problems. The satisfactory
results obtained with the classification of single
motifs, even deformed by the preservation of the
1148
specimens or altered by the chronological and
geographical variability of the production and by
the creativity of the Minoan artisans, let expect the
extension of this technique also to the decorative
syntax of the Kamares style, in order to allow the
archaeologist to look forward for the long waited
exhaustive study of this so significant feature of
the Mediterranean prehistory.
ACKNOWLEDGMENT
The authors wish to thank Professors V. La Rosa
and F. Carinci for their precious support and the
many useful advices that they have provided dur-
ing this research.
REFERENCES
Attneave, F. (1954). Some informational aspects
of visual perception. Psychological Review, 61,
183–193. doi:10.1037/h0054663
Belongie, S., Malik, J., & Puzicha, J. (2002). Shape
Matching and Object Recognition Using Shape
Contexts. IEEE Transactions on Pattern Analysis
and Machine Intelligence, 24, 509-522. doi:http://
doi.ieeecomputersociety.org/10.1109/34.993558
Biderman, I. (1985). Human image understanding:
Recent research and a theory. Computer Vision
Graphics and Image Processing, 32, 29–73.
doi:10.1016/0734-189X(85)90002-7
Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style
Bookstein, F. (1997). Landmark methods for
forms without landmarks: morphometrics of
group differences in outline shape. Medical Im-
age Analysis, 1(3), 225–243. doi:10.1016/S1361-
8415(97)85012-8
Carinci, F. (1997). Pottery workshops at Phaestos
and Haghia Triada in the protopalatial period, in
Aegaeum (Vol. 16). Center of Cretan Archaeology.
(S.d.). Center of Cretan Archaeology. Retrieved
July 31, 2010 from http://www.cac.unit.it
Dawkins, R. M., & Laistner, M. (1912)... The
Excavation of the Kamares Cave in Crete. in
Annual of the British School at Athens, 19, 1–34.
Day, P. M., & Wilson, D. E. (1998). Consuming
power. Kamares ware in protopalatial Knossos.
Antiquity, 72, 350–358.
Escalera, S., Fornes, A., Pujol, O., Escudero, A.,
& Radeva, P. I. (2009). Circular Blurred Shape
Model for symbol spotting in documents. In
ICIP09 (pp. 2005-2008).
Evans, A. (1921). The Palace of Minos at Knos-
sos. London.
Farinella, G. M., Stanco, F., & Tanasi, D. (2008).
Digital-Kamares: un sistema automatizzato per
riconoscere e classificare la ceramica minoica di
stile Kamares. In AIAr 2008, V Congresso Nazio-
nale di Archeometria, Scienza e Beni Culturali
(pp. 547-556).
Gallo, G., Stanco, F., & Tanasi, D. (in press).
Riconoscimento automatico di motivi decorativi
nelle ceramiche minoiche di stile Kamares da
Festós. In 13a Giornata di Archeometria della
Ceramica: Nuove metodologie analitiche e di
trattazione dei dati nello studio della ceramica
in archeologia ed in architettura.
Gonzalez, R. C., & Woods, R. E. (2006). Digital
Image Processing (3rd ed.). Upper Saddle River,
NJ: Prentice-Hall, Inc.
Gonzalez, R. C., Woods, R. E., & Eddins, S. L.
(2009). Digital Image Processing Using MATLAB.
Upper Saddle River, NJ: Gatesmark Publishing.
Guarnera, G. C., Stanco, F., Tanasi, D., & Gallo,
G. (2010). Classification of decorative patterns
in Kamares pottery. In In proceedings of SCCG
26yh Spring Conference on Computer Graphics
(pp. 20-23). Comenius University, Bratislava.
IPLab. (S.d.). Archeomatica Project. Retrived July
31, 2010 from http://www.archeomatica.unict.it
Knappett, C. (1999). Tradition and innovation
in pottery forming technology. Wheel-throwing
at Middle Minoan Knossos. BSA, 94, 101–129.
Levi, D. (1976). Festós e la Civiltá Minoica I,
Roma.
Levi, D., & Carinci, F. (1988). Festós e la Civiltá
Minoica.
Lorensen, W. E., & Cline, H. E. (1987). Marching
cubes: A high resolution 3D surface construc-
tion algorithm. In SIGGRAPH ‘87: Proceed-
ings of the 14th annual conference on Com-
puter graphics and interactive techniques (pp.
163–169). New York: ACM. doi:http://doi.acm.
org/10.1145/37401.37422
Macgillivray, J. A. (1998). Knossos. Pottery
groups of the Old Palace period. London.
Mackenzie, D. (1906). The Middle Minoan pottery
of Knossos. Journal of Hellenic Studies.
Maple, C. (2003). Geometric design and space
planning using the marching squares and march-
ing cube algorithms. In Geometric Modeling
and Graphics, 2003. Proceedings. 2003 Interna-
tional Conference on (pp. 90 - 95). doi:10.1109/
GMAG.2003.1219671
McNeill, G., & Vijayakumar, S. (2005). 2D shape
classification and retrieval. In IJCAI’05: Proceed-
ings of the 19th international joint conference
on Artificial intelligence (pp. 1483–1488). San
Francisco: Morgan Kaufmann Publishers Inc.
1149
 Automatic Classification of Decorative Patterns in the Minoan Pottery of Kamares Style

McNeill, G., & Vijayakumar, S. (2006). Hierarchi-
cal Procrustes Matching for Shape Retrieval. In
CVPR ‘06: Proceedings of the 2006 IEEE Com-
puter Society Conference on Computer Vision and
Pattern Recognition (pp. 885–894). Washington,
DC, USA: IEEE Computer Society. doi:http://
dx.doi.org/10.1109/CVPR.2006.133
Momigliano, N. (1999). Duncan Mackenzie:
a Cautious Canny Highlander and the Palace
of Minos at Knossos. (Vol. 72). Bulletin of the
Institute of Classical Studies, London.
Moore, G. A. (1968). Automatic Scanning and
Computer Processes for the Quantitative Analy-
sis of Micrographs and Equivalent Subjects. In
PPR68 (pp. 275-326).
Pernier, L. (1935). Il palazzo minoico di Festós.
Pernier, L., & Banti, L. (1951). Il palazzo minoico
di Festós.
Sebastian, T. B., Klein, P. N., & Kimia, B. B.
(2003). On Aligning Curves. IEEE Trans. Pattern
Anal. Mach. Intell., 25(1), 116–125. doi:http://
dx.doi.org/10.1109/TPAMI.2003.1159951
Shen, D. F., Jin, L., Chang, H. T., & Wu, H. H.
(2005). Trademark Retrieval Based On Block
Feature Index Code. In ICIP05 (pp. III: 177-180).
Stanco, F., Battiato, S., & Gallo, G. (2011). Digi-
tal Imaging for Cultural Heritage Preservation.
Analysis, Restoration ad Reconstruction of Ancient
Artworks. CRC Press.
Walberg, G. (1978). The Kamares style. Overall
effects (Vol. 10). Uppsala: Acta Universitatis
Upsaliensis. Boreas.
Walberg, G. (1983). Provincial Middle Minoan
pottery. Mainz.
Walberg, G. (1987). Kamares. A study of the
character of Palatial Middle Minoan pottery.
Göteborg.
Walberg, G. (2001). The role and individuality of
Kamares ware. in Aegean Archeaology, 5, 9-18.
Xu, C., Liu, J., & Tang, X. (2008). 2D Shape
Matching by Contour Flexibility. IEEE Trans-
actions on Pattern Analysis and Machine Intel-
ligence, 31, 180-186. doi:http://doi.ieeecomput-
ersociety.org/10.1109/TPAMI.2008.199
Zhang, J., Zhang, X., Krim, H., & Walter, G. G.
(2003). Object representation and recognition
in shape spaces. Pattern Recognition, 36(5),
1143–1154. doi:10.1016/S0031-3203(02)00226-1
Zois, A. A. (1968). Der Kamares-Stil. Werden
und Wesen (Diss.), Tübingen.

This work was previously published in Pattern Recognition and Signal Processing in Archaeometry: Mathematical and Com-
putational Solutions for Archaeology, edited by Constantin Papaodysseus, pp. 186-211, copyright 2012 by Information Science
Reference (an imprint of IGI Global).

1150
 1151

Chapter 69
Cultural Dasymetric Population
Mapping with Historical GIS:
A Case Study from the
Southern Appalachians

George Towers
Concord University, USA

ABSTRACT
There has been a recent flurry of interest in dasymetric population mapping. However, the ancillary
coverages that underlie current dasymetric methods are unconnected to cultural context. The resulting
regions may indicate density patterns, but not necessarily the boundaries known to inhabitants. Dasymetric
population mapping is capable of capturing the cultural commonality and community interaction that
define social spaces. Dasymetric mapping may be improved with methodologies that reflect the ways in
which social spaces are established. This research applies a historical GIS methodology for identifying
early 20th Century agricultural neighborhoods in southern Appalachia. The case study is intended to
encourage discovery of additional methods for mapping population on the scale of lived experience.

INTRODUCTION landscape. The resulting dasymetric spaces ground

Human geographers’ use of dasymetric mapping
has increased with advances in geospatial data.
By intersecting land use imagery or property
parcel data with census geography, gradations
in population density are revealed that refine the
cartographic basis for investigating the cultural
DOI: 10.4018/978-1-4666-2038-4.ch069
spatial analysis of social issues like environmental
justice. Most dasymetric analysis, however, is
predicated on the problematic assumption that
discontinuities in population density align with the
boundaries delimiting the micro-regional cultural
landscape of neighborhoods and communities.
Cultural dasymetric mapping, cartography that
conforms dasymetric space to cultural contours,
is a rarity. The following review of the dasymetric

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Cultural Dasymetric Population Mapping with Historical GIS
literature chronicles this inconsistency, calling
attention to the need for culturally informed dasy-
metric mapping. The methodology presented in the
case study section offers an example of cultural
dasymetric mapping and invites exploration of the
diversity of techniques for infusing dasymetric
mapping with cultural significance.
DASYMETRIC MAPPING
Dasymetric mapping is designed to improve upon
choropleth maps that employ arbitrarily bounded
enumeration areas. Choropleth mapping is inher-
ently problematic in two fundamental ways. First,
choropleth maps suffer from the modifiable areal
unit problem (MAUP) (Langford & Unwin, 1994;
Openshaw, 1984): “Any observed pattern in the
mapped data may be to a large extent due to the
particular configuration of zonal boundaries used,
and all relationships observed between variables
will only hold for this particular aggregation”
(Martin, 1996, p. 974).
Second, areal boundaries drawn without
reference to the choropleth map subject invite
the “ecological fallacy,” that is, areal units im-
ply regional homogeneity and mask internal
heterogeneity. Regional homogeneity is defined
by the human geography textbook trichotomy
of functional, formal, and perceptual regions.
Functional regions are determined by spatially
conscribed interaction. Fellow residents of func-
tional regions - school districts, for example – are
compelled to cooperate. Formal and perceptual
regions are based on identity. In the case of formal
regions, inhabitants may be objectively ascribed a
shared identity through common socioeconomic
or cultural characteristics. Perceptual regions are
self-designated. Residents share an attachment
to place and regionally define their community.
As the basis for much spatial analysis, census
geography offers an important example of these
two ills of choropleth mapping (Openshaw, 1996;
Schuurman, Leszczynski, Fielder, Grund, & Bell,
1152
2006). The foundational US Census unit, the cen-
sus block, often endorses the ecological fallacy.
Outside high density city neighborhoods, census
blocks expand to capture sufficient population and
their internal heterogeneity increases accordingly
(Crandall & Weber, 2005; Goodchild, Anselin, &
Deichmann, 1993; Mennis, 2003). The Census
Bureau’s use of roads to bound census blocks
contributes not only to the ecological fallacy
but shows how arbitrary boundaries create the
MAUP. Rural roads focus residential settlement,
community interaction, and identity (Daniels,
1999; Theobald, 2005). Serving as census block
boundaries, roads split rural neighborhoods. With
rural population concentrated along roadways,
rural census blocks are doughnut-like, consisting
of peopled peripheries and empty centers. Road
boundaries heighten heterogeneity by lumping
together fragments of disparate neighborhoods
linked by census block circumferences.
Dasymetric mapping promises to resolve the
MAUP and the ecological fallacy by incorporat-
ing appropriate ancillary information. In keeping
with the connotation of “dasymetric” in its original
Russian and in Wright’s 1936 translational title,
“A method of mapping densities of population,”
the most common method of dasymetric map-
ping forms boundaries with population density
discontinuities. By adjusting thresholds separating
density categories, dasymetric boundaries enclose
areas of homogenous density (Mennis, 2003).
The application of density based dasymetric
mapping to socio-spatial analysis is based on the
longstanding assumption that density disconti-
nuities demarcate socially meaningful regions.
For example, Goodchild et al. argue that “… the
problem of basis change for socioeconomic data
can be seen as one of estimating one or more
underlying density surfaces” (1993, p. 388),
Martin supports “modelling the same informa-
tion [population characteristics] in the form of
density surfaces” (1996, p. 974), and Moon and
Farmer assert that “Communities of place could
Cultural Dasymetric Population Mapping with Historical GIS
be operationalized by measures of population
density …” (2001, p. 46).
Land use / land cover (LULC) maps are the
most widely used ancillary data in the construc-
tion of population density surfaces. The simplest
methodology is binary, coding LULC classes
for the presence or absence of population. The
resulting presence / absence layer is combined
with census geography to reassign population
from the uninhabited to the inhabited parts of
census enumeration units. For example, the binary
method has recently been employed to analyze
environmental justice and homelessness risk
(Boone, 2008; Fielder, Schuurman, & Hyndman,
2004). Expanding upon this approach is the reclas-
sification of LULC codes into several categories
weighted for their estimated relative population
densities. Density surfaces are calculated by mul-
tiplying the weighted coverage with population
counts for census districts (Holt, Lo, & Hodler,
2004; Mennis, 2003, 2009). Mennis (2002) has
used the weighted dasymetric method to measure
environmental justice.
Residential property and structures are mapped
in a variety of forms, each of which is conducive
to dasymetric population mapping. For example,
Maantay, Maroko, and their colleagues have de-
veloped their Cadastral-Based Expert Dasymetric
System to assess a variety of environmental justice
issues in New York City including flood hazards
and disease (Maantay & Maroko, 2009; Maan-
tay, Maroko, & Porter-Morgan, 2008; Maantay,
Maroko, & Culp 2009). Parcel based population
density surfaces serve to identify and measure
suburban sprawl (Hasse & Lathrop, 2003; Irwin
& Bockstael, 2002). Similarly, researchers have
used house locations in conjunction with popula-
tion maps to generate density surfaces for socio-
cultural analysis (Kirk, 2009; Moon & Farmer,
2001; Theobald, 2001). Finally, postal coded
address points may be integrated into density
surface generation (Mesev & McKenzie, 2008).
The volume of research devoted to develop-
ing and implementing density surfaces for social
analysis reflects the utility of this method in rep-
resenting settlement patterns. Density surfaces are
essential for subjects like emergency evacuation
planning that focus on where, not who, people
are (Chen, McAneney, Blong, Leigh, Hunter,
& Magill, 2004; Garb, Cromley, & Wait, 2007;
Khomarudin, Strunz, Post, Zobeder, & Ludwig,
2008). Density discontinuities, however, are not
necessarily relevant to distinguishing between
neighborhoods. For example, coherent communi-
ties may be centered on open spaces like parks
or, conversely, may bind a low density periphery
to a high density core.
To demarcate culturally meaningful regions,
dasymetric methodologies must model identity
and interaction. For example, Trodd and Gearey
(2006) incorporate participatory GIS in dasymetric
mapping to delineate perceptual regions for the
provision of early childhood services. Finding a
close match between census geography and local
leaders’ perceptions of neighborhood boundar-
ies, they suggest that infusing local geographic
identity with census information will help target
social service delivery.
Grannis (2005) uses dasymetric methods to
define functional and formal regions within cit-
ies. Drawing upon the consensus that “walkable
neighborhoods” evoke “socio-physical homoge-
neity, a shared sense of place, connection, and
access” (Moudon et al., 2006, p. 102) and upon
his own survey research, Grannis postulates that
networks of pedestrian streets bounded by heav-
ily trafficked thoroughfares comprise meaningful
urban neighborhoods. He automates neighborhood
boundary identification with GIS and finds that
the resulting spaces generally coincide with urban
census block boundaries. Grafting census data into
his walkable neighborhoods reveals residential
racial segregation at a scale significant to resi-
dents. The functionally derived neighborhoods’
internal homogeneity suggests that they are also
formal regions.
To accommodate standardized LULC and
parcel data offering detailed information over
1153
 Cultural Dasymetric Population Mapping with Historical GIS
vast expanses, dasymetric mapping has distanced
itself from localized data collection. Contextually
sensitive research methods are, however, more
appropriate for dasymetric mapping because
neighborhood and community boundaries are
grounded in local culture. The examples provided
by Trodd and Geary and Grannis show that par-
ticipant observation and in-depth interviewing
merge with GIS analysis to dasymetrically reveal
the boundaries of regional interaction and identity.
The following case study presents another qualita-
tive methodology useful for cultural dasymetric
mapping, historical ethnography.
CASE STUDY: CULTURAL
DASYMETRIC MAPPING OF
HISTORIC SOUTHERN APPALACHIA
Methodology
In an earlier article, I incorporated the collective
observations of Appalachian ethnographers into
GIS analysis to systematically map the boundar-
ies of the rural southern Appalachia communi-
ties of a century ago (Towers, 2010). Beginning
in the 1930s, anthropologists, sociologists, and
geographers produced case studies of the shared
culture underlying the region’s farm family neigh-
borhoods between the Civil War and World War
Two. On the typical semi-subsistence farms of the
era, families diversified production from grains
to fruits to vegetables and raised a menagerie
of livestock including cattle, sheep, and pigs.
Neighboring farmers depended on social and
economic reciprocity. Farm neighborhoods were
close-knit, kin-based homogenous communities
that shaped residents’ identity. In other words,
they were simultaneously functional, formal, and
perceptual regions.
Like Grannis, I used walking-based social
interaction as a starting point for regional identi-
fication. This pedestrian premise is supported by
Appalachian ethnography which called attention
1154
to how topography circumscribed community.
According to James S. Brown, a pioneering so-
ciologist of rural Appalachia, “Neighborhoods
develop among people who have frequent and
regular contacts, and in this region topography
has helped to determine social relationships of
the inhabitants and to form neighborhood group-
ings” (Brown, 1988, p. 52). Approximating the
boundaries of these functional neighborhoods
is a three part process: creating a terrain-based
travel cost surface that modeled the mountains’
socially limiting effects, identifying nodal points
centering each neighborhood, and, assigning land
to each neighborhood based on the cost of travel
to its node.
I chose Summers County, West Virginia, a
county that was exceptionally statistically repre-
sentative of southern Appalachian farming from
1880 through 1940, as the pilot area for my initial
study. To make a travel cost surface for Summers
County, I used the 30 meter US Geological Sur-
vey (USGS) National Elevation Dataset (NED).
I began by calculating a slope surface from the
NED. I then modified an equation from applied
physiology to convert the slope surface into a
caloric cost surface for walking at slope (Minetti,
Moia, Roi, Susta, & Ferretti, 2002).
k = (((280.5 * s5) – (58.7 * s4) – (76.8 * s3) +
(51.9 * s2) + (19.6 * s) + 2.5) * w * d) / 4184
k represents kilocalories expended
s represents percent slope
w represents body weight in kilograms
d represents meters traveled
In other words, the cost surface makes foot
travel between any pair of NED cells measurable
in terms of its caloric expenditure.
Central neighborhood locations in early 20th
century rural Appalachia were post offices,
country schools, and churches. Post offices were
established in the tiny hamlets that dotted the
countryside and were named on the 1912 USGS
Cultural Dasymetric Population Mapping with Historical GIS
maps available for Summers County. A typical
hamlet was a low order central place whose ser-
vices typically included a post office, a general
store, a mill, and a blacksmith shop (Hart, 1975).
Interspersed between hamlets were farm neighbor-
hoods supporting one-room country schools. The
local schools were not only community centers
hosting educational, social, and political functions
(Gulliford, 1996), they were also deliberately
evenly spaced to ensure that children’s walk to
school was minimized (Holy, 1929).
With neighborhood nodes placed on the walk-
ing cost surface, territory was assigned to each
node based on ease of access. The NED cells
sharing the same lowest cost destination formed
my hypothetical agricultural neighborhoods.
Hypothesis confirmation consisted of con-
firming that settlement patterns within the pro-
spective neighborhoods matched ethnographic
descriptions. This was predicated on verifying
that the houses shown on the 1912 maps were a
comprehensive inventory of the residential land-
scape. Adjusting the number of mapped houses for
estimated vacancy rates and the household counts
from the 1910 manuscript census for undercount
produced nearly identical figures for magisterial
districts and supported using the 1912 USGS maps
as a primary data source.
I then reviewed ethnographic descriptions
of neighborhood size and settlement patterns to
establish expectations for the spatial patterns of
house locations contained in the hypothetical
neighborhoods. According to secondary sources,
Appalachian agricultural neighborhoods con-
tained between one dozen and four dozen farm-
houses regularly dispersed along streams flowing
through hollows or along narrow ridgetop plateaus
(Wilhelm, 1978). Spatial analysis of house loca-
tions within the hypothetical agricultural neigh-
borhoods resulted in house counts, distributions,
and densities that closely fit these expectations.
The representativeness of the study area, the
legitimacy of the map data, and the faithfulness
of the GIS-generated neighborhood settlement
patterns to those documented by ethnographic
and geographic fieldwork combine to confirm
that my methodology offers a replicable and re-
liable means of bounding the social spaces that
structured the regional cultural landscape in the
decades surrounding the turn of the last century.
Study Areas
To show the utility of this social dasymetric
technique to map historical population, I examine
two West Virginia study areas: Wyoming County
in the state’s southwestern coalfields, and, the
farmlands of the lower Greenbrier River Valley
in Greenbrier and Monroe Counties (Figure 1).
Structure locations from 1920s USGS 15 minute
topographic maps were digitized and are the pri-
mary data for the GIS analysis. The topographic
maps for Wyoming County date from 1924 to 1929
except for the 1914 Eccles quadrangle covering
the far northeastern corner of the county. All of the
Lower Greenbrier Valley study area quadrangles
were surveyed in 1921.
Wyoming County
Settled in the early years of the 19th century,
Wyoming County supported a pre-industrial agri-
cultural economy based on subsistence and barter.
The agricultural neighborhoods described above
organized community life in Wyoming County
through the 1800s (Cook, 2000). As elsewhere in
the region, these farm neighborhoods found focal
points in country schools. As of 1895, there were
at least 95 small school buildings in the county
(Bowman, 1965).
Wyoming’s timber and coal attracted land
speculators in the late 1800s. In advance of the
railroad, the timber industry came to Wyoming
County in the 1890s. William Ritter, southern West
Virginia’s leading timber baron, owned 30,000
acres by 1906 and had built a company town at
Estell, later renamed Maben. Complete with an
immigrant workforce and charges of peonage,
1155
 Cultural Dasymetric Population Mapping with Historical GIS

Figure 1. Study areas

Maben resembled the coal camps that were to narrow and rough. Wagons and horses could
follow (Feller, 2001; Lewis, 1998). hardly get over them. There were no bridges across
Delayed by the county’s rugged topography the streams” (Feller, 2001, p. 22). For example,
(Table 1), physical isolation, and fractious politics, Keystone in neighboring McDowell County is 20
the railroad finally arrived in Wyoming County kilometers across the mountains from Mullens.
in 1906 (Cook, 2000). The Virginian Railway Prior to the railroad, shipping between the two
ran through eastern Wyoming County on its way locales took three days (Feller, 2001). These
south from Deepwater on the Kanawha River to primitive conditions assert the topographical
Princeton in Mercer County. Western Wyoming tyranny reflected in the pedestrian cost surface
County was not linked to the railroad until the modeling employed in this research.
late 1920s (Bowman, 1965). With the railroad came the coal industry. The
The railroad “meant the end of our way of life” county’s first mine was dug near Mullens in
remembered resident Howard Mullins. Kinsman 1912. As World War One grew demand, another
Foley Mullins reported that prior to the railroad, 14 mines opened in eastern Wyoming County.
“Transportation was very difficult as roads were County coal production, an incipient 1,848 tons
in 1913, surpassed one hundred thousand tons by
Table 1. Slope 1915, one million tons by 1918, and three million
tons by 1925 (Bowman, 1965).
Slope categories Wyoming Lower Greenbrier Population grew with the coal industry, much
County Valley of it imported from distant labor reservoirs. By
Flat (0-15% slope) 10% 49% 1925, there were 2,513 coal miners in the county,
Steep (15-35%) 31% 39% 36% of whom were African Americans from the
Very steep (35-100%) 59% 11% South and 7% were European immigrants (Trotter,

1156
Cultural Dasymetric Population Mapping with Historical GIS
1990).The county’s eastern coal districts, Barker’s
Ridge and Slab Fork, more than doubled from a
combined population of 3,353 in 1910 to 7,343
in 1920. Mullens, straddling the border of the two
districts, incorporated to become a town of 1,425
in 1920 (Figure 2).
The cultural landscape of the World War One
coal boom in Wyoming County was of a piece
with the regional experience. Colorfully named
coal camps like Bud, Caloric, and Montecarlo
dotted the Virginian’s route. For example, Itmann,
named for mine operator I.T. Mann, was con-
structed of 100 pre-fabricated houses over a three
month period in 1918 (Feller, 2001).
Despite rural industrialization’s territorial
encroachments, county farmers found a market
in the timber and coal camps. Indeed, Lewis finds
that instead of an “antithetical” relationship, farms
coexisted “harmoniously” with lumber camps
and mines in Wyoming County (Lewis, 1998, p.
257). Farms and farmland increased from 1880
and peaked in 1910 with the new commercial
opportunities (Cook, 2000). Table 2 shows that
while the number of farms and the amount of
Figure 2. Wyoming County
farmland declined slightly through the 1920s,
farming occupied a dramatically smaller propor-
tion of Wyoming Countians as mining expanded.
The 1920s topographic maps display the as-
cendance of coal in eastern Wyoming County in
contrast with the county’s agrarian western por-
tion. Settlement patterns on the two sides of the
county diverge accordingly. As it winds through
narrow valleys, the railroad connects dense clus-
ters of housing in eastern Wyoming’s company
towns. The agricultural neighborhoods to the west
share a dispersed pattern of farmsteads.
Lower Greenbrier River Valley
The Lower Greenbrier River Valley study area
represents the long-established farm country of
southern Appalachia’s ridge and valley physio-
graphic province. Agricultural productivity is
insured by broad stream valleys underlain by
limestone formations (Table 1). As the mainstream
of settlers flowed down the Shenandoah Valley
in the mid 1700s, some branched off into south-
eastern West Virginia. Disrupted by the French
1157

Table 2. Wyoming County agriculture, 1880-1930
Date County Farms
Population
1880 4,322 608
1890 6,247 862
1900 8,380 1,202
1910 10,392 1,462
1920 15,180 1,181
1930 20,926 1,154
Source: University of Virginia, 2004.
and Indian War, permanent settlement followed
independence. By the 1820s, the lower Greenbrier
was well-integrated into the national economy by
the James River and Kanawha turnpike. Livestock
were the leading regional export and local farms
also sold corn and small grains. The region also
boasted an early tourism attraction in mineral
springs resorts. Of these, White Sulphur Springs
remains the most prominent (Rice, 1986).
A contested Civil War theater, the Greenbrier
Valley quickly recovered to become one of the
new State of West Virginia’s leading stock-raising
centers. Facilitated by the construction of the
Chesapeake and Ohio Railroad (C&O) along the
Greenbrier River in 1869, cattle, sheep, horses, and
hogs were exported from valley farms to eastern
cities. In the late 19th and early 20th centuries,
local farmers continued to support themselves
with mixed farming that included corn, wheat,
oats, vegetables and fruit. The region’s agrarian
prominence is celebrated by the state fairgrounds
in Fairlea (Rice, 1986).
The timber industry came to the region’s rich
hardwood forests in the 1870s. As in Wyoming
County, investors acquired large tracts of tim-
berlands and initiated a natural resource boom
cycle. Conveyed at first by the river, then by the
rails, logs were processed in sawmills that sprang
up along the transportation corridor formed by
the Greenbrier River and the C&O. Ronceverte,
built in the 1870s, more than doubled in size from
1900 to 1910 to over 2,000 residents and was the
1158
Cultural Dasymetric Population Mapping with Historical GIS
Estimated pct. of county Average farm size Pct. of County
population on farms (hectares) Territory in Farmland
77 54 36
76 56 35
79 45 37
77 44 41
43 40 36
34 36 32
center of the valley’s lumber trade. Its huge state
of the art mill stayed in operation until the timber
industry slowed in the 1920s. By 1910, some 25
other large sawmills had been built in the lower
Greenbrier region at towns including Alderson,
Fort Spring, and Union (Lewis, 1998; Rice, 1986).
Early 20th Century rural population growth
followed the route of the Greenbrier River and
C&O through the Greenbrier County section of
the study area and bypassed the Monroe County
section lying to the south (Figure 3). The rural
Greenbrier County portion grew from a population
of 4,912 in 1900 to 6,380 in 1920 while the rural
Monroe County portion fell from 4,820 in 1900 to
4,434 in 1920. The economic development contrast
between Monroe and Greenbrier counties is clearly
shown in Tables 3 and 4. Investment in timber and
railroading grew Greenbrier’s population steadily.
The 1920s growth spike came with the opening
of the far western portion of the county to coal
and timber. While agricultural figures hold fairly
constant, farming occupies a smaller proportion
of the population. Monroe County, however,
remained exclusively agricultural throughout the
period and Table 4 shows that farming expanded
to the county’s capacity by 1900 and remained
consistent thereafter.
Study Area Comparison
Density analysis reflects the dominance of agri-
culture in the Lower Greenbrier Valley compared
Cultural Dasymetric Population Mapping with Historical GIS

Figure 3. Lower Greenbrier Valley

to Wyoming County. In the flatter, more fertile,
and longer-established farmland of the Lower
Greenbrier Valley, population density was half
again as high as in rural Wyoming County (Table
5). To assess the distribution of density, I apply
four settlement categories derived from historic
Appalachian farm sizes: vacant, archetypal agri-
cultural, agricultural, and commercial (Towers,
2010). Vacant land is far enough from settlement
that it is unlikely to have been part of a farm. I
set the vacancy threshold at 800 meters from
the nearest structure on the basis of regional
farm size. One hundred years ago, few southern
Appalachian farms exceeded 200 hectares. For
example, in Summers County, 90 percent of farms
were smaller than 200 hectares. In a landscape

Table 3. Greenbrier County agriculture, 1880-1930

Date County Farms Estimated pct. of county Average farm size Pct. of County
Population population on farms (hectares) Territory in Farmland
1880 15,060 1,506 55 131 74
1890 18,034 1,872 57 82 58
1900 20,683 2,434 65 68 62
1910 24,833 2,551 56 53 50
1920 26,242 2,439 51 52 48
1930 35,878 2,242 38 53 44
Source: University of Virginia, 2004.

1159
 Cultural Dasymetric Population Mapping with Historical GIS

Table 4. Monroe County agriculture, 1880-1930

Date County Farms Estimated pct. of county Average farm size Pct. of County
Population population on farms (hectares) Territory in Farmland
1880 11,501 1,248 60 78 79
1890 12,429 1,521 67 66 81
1900 13,130 1,794 75 56 81
1910 13,055 1,918 81 51 79
1920 13,141 1,834 77 52 78
1930 11,949 1,826 76 49 73
Source: University of Virginia, 2004

occupied by evenly spaced 200 hectare farms
centered on farmhouses, the houses will be 800
meters from each other.
Archetypal agricultural land would have been
covered by average sized farms. Late 19th and
early 20th southern Appalachian farms averaged
40 hectares in size. Allowing for these farms to
be slightly unevenly spaced suggests that zero,
one or two farmhouses should be found within a
40 hectare area. Therefore, I chose 40 hectares as
the search area surrounding each cell on the den-
sity surface. As ethnographic reports concur that
southern Appalachian farm households averaged
about five members (Brown, 1988; Martin, 1984),
archetypal agricultural population density is rep-
resented by non-vacant land containing fewer
than 25 people per square kilometer.
Agricultural land extends beyond archetypal
agricultural densities to settlement patterns too
crowded to be consistent with Appalachian farm-
ing. Eight hectares was the low end of the smallest
census farm size category, implying that properties
of less than 8 hectares were not used as farms.
Using the 40 hectares surrounding each cell for
density calculations means that five very small
farms could coexist within this space and suggests
that six houses within the 40 hectare search area
serve as the threshold between agricultural and
commercial land. In terms of population density,
this translates into an upper limit of 75 people per
square kilometer on agricultural land. Figures 4
and 5 and Table 5 confirm that population densi-
ties associated with agricultural settlement were
much more prevalent in the Lower Greenbrier
Valley than in Wyoming County.

Table 5. Population density

DISCUSSION
Population Density Indicators Wyoming Lower
County Greenbrier Ethnography documents that early 20th Century
Valley
southern Appalachian agricultural neighborhoods
1920 Study Area Population 15,270 16,174 provided a spatial structure for community interac-
Study Area (sq. km) 1,299 908 tion and identity. In previous work, I demonstrated
Population per sq. km 11.8 17.8 a GIS methodology that reliably reconstructs the
Vacant 29% 6% boundaries of these miniature culture regions
Archetypal agri- 61% 80% and is a basis for cultural dasymetric population
Density cultural
categories mapping (Towers, 2010).
Agricultural 8% 12%
Commercial 3% 2%

1160
Cultural Dasymetric Population Mapping with Historical GIS

Figure 4. Population density map of Wyoming County

Figure 5. Population Density map of the Lower Greenbrier Valley

1161
 Cultural Dasymetric Population Mapping with Historical GIS
Applying the GIS methodology to the two
study areas generated boundaries around rural
communities and towns. Rural communities were
organized around a single node – a post office,
school, or church - by the cost allocation process.
Towns were represented by several nodes on the
town’s periphery. The territories allocated to the
nodes were combined to represent the town’s ef-
fective vicinity, not its jurisdictional boundaries.
Calculating populations for GIS-generated
neighborhoods is a straightforward process. 1920
Census populations for enumeration units - incor-
porated towns and the unincorporated portions of
magisterial districts - were divided by the number
of houses mapped within each enumeration area
to provide a population estimate for each house.
The estimated population of the houses in each
GIS neighborhood is then summed to produce
neighborhood population estimates. Tobler’s pyc-
nophylactic requirement that dasymetric divisions
combine to reconstitute original population totals
is easily met with this method (Tobler, 1979).
Because the original census population is entirely
distributed across the mapped houses, it is recover-
able from any regional recombination of houses.
Figure 6. Dasymetric map of Wyoming County showing population
1162
Figures 6 through 9 are dasymetric population
maps based on the GIS-generated communities.
In Wyoming County, the rural portions of the
seven magisterial districts averaged 5.8 people
per mapped structure with only two of the districts
more than one person from the mean: Barkers
Ridge at 4.5 and Clear Fork at 7.4 people per
house. In the Lower Greenbrier Valley’s seven
rural magisterial district areas, the average was
4.6 people per house with all of the districts be-
tween 4.0 and 5.1. Within city limits, population
per house averaged 5.7 in Wyoming County and
6.1 in the Lower Greenbrier Valley.
The communicative value of the dasymetric
population maps is readily compared with cho-
ropleth maps and density surfaces. Choropleth
population mapping is shown in Figures 10 and 11.
Drawn from 1920 data for magisterial districts and
incorporated towns, the choropleth maps hint at the
population shift to eastern Wyoming’s coalfields
and at the even population distribution across the
Lower Greenbrier Valley. As enumeration units
imposed for administrative purposes, the magiste-
rial districts carried limited functional significance
but were not an explicit basis for social identity.
Cultural Dasymetric Population Mapping with Historical GIS

Figure 7. Dasymetric map of Wyoming County showing population density

Figure 8. Dasymetric map of the Lower Greenbrier Valley showing population

1163
 Cultural Dasymetric Population Mapping with Historical GIS

Figure 9. Dasymetric map of the Lower Greenbrier Valley showing population density

Figure 10. Choropleth Population map of Wyoming County

1164
Cultural Dasymetric Population Mapping with Historical GIS

Furthermore, at the county scale, the large size of
the magisterial districts makes them vulnerable to
the ecological fallacy. The obscuration of variation
within the relatively large districts is demonstrated
by the density surfaces and dasymetric maps.
The density surfaces (Figures 4 and 5) are
constructed from house locations assigned esti-
mated populations and are calculated at the fine
grain of 30 meter raster cells using a 40 hectare
search area. Discontinuities between the four
density categories defined above are informative.
For example, the knotted string of coal camps
lining the Virginian Railway stands out on the
Wyoming County map. The uneven rural settle-
ment of agriculturally marginal, yet unindustrial-
ized western Wyoming County is also manifest.
In contrast, the even population densities of the
Lower Greenbrier Valley reflect the agrarian
landscape of a premier farming region. Density
discontinuities, however, cannot distinguish be-
tween distinct coal camps nor draw out the neigh-
borhoods that comprised the Lower Greenbrier
farm country.
Informed by ethnographic ancillary data, the
dasymetric maps (Figures 6 through 9) display
population at the intermediary scale of lived
communities. For example, descriptive statisti-
cal analysis of the dasymetric neighborhoods
identifies three groups of rural communities:
Wyoming coal camps, Wyoming agricultural
neighborhoods, and Lower Greenbrier Valley
agricultural neighborhoods. Tables 6 and 7 show

Figure 11. Choropleth Population map of the Lower Greenbrier Valley

1165
 Cultural Dasymetric Population Mapping with Historical GIS

Table 6. Wyoming County communities

Community statistics Agricultural neighborhoods Coal camps Mullens Pineville

Number 105 20 1 1
Average size (sq. km) 10.6 8.1 8.0 16.6
Standard deviation of size 7.6 5.6 NA NA
Average population 74 281 1,446 383
Standard deviation of pop. 49 209 NA NA
Population per sq. km 7.0 34.7 173.7 23.1

Table 7. Lower Greenbrier Valley communities

Community statistics Agricultural neighborhoods Towns1

Number 153 4
Average size (sq. km) 5.6 11.4
Standard deviation of size 3.7 NA
Average population 65 1,550
Standard deviation of pop. 42 NA
Population per sq. km 11.6 134.7
1. Towns are Alderson, Lewisburg, Ronceverte, and Union.

that average density in Wyoming’s coal camps is
almost five times that in neighboring agricultural
neighborhoods. A further contrast may be drawn
between agricultural neighborhoods in the two
study areas. Lower Greenbrier Valley agricultural
neighborhoods are smaller and denser, reflecting
the region’s superior agricultural setting.
mapping, in contrast, does not model the identity
and interaction that socialize space. Like Trodd
and Gearey’s generation of perceptual regions and
Grannis’ analysis of social networks, my use of
historical ethnography suggests a diffuse yet shared
direction for further cultural dasymetric mapping.

CONCLUSION
The study presented above is intended to advance
the case for cultural dasymetric mapping. As in the
southern Appalachian example, cultural dasymetric
mapping involves merging contextually sensitive
qualitative research with census geography or other
generic data. Cultural dasymetric methodologies
are not necessarily transferable across space and
time but demonstrate the productivity of deepening
contextual knowledge to choose the combination
of research resources that best maps the scale of
lived experience. Contemporary density dasymetric
ACKNOWLEDGMENT
The author thanks the editor of this special edition
of the IJAGR, Rebecca Dobbs, for her interest in his
work. He also thanks the anonymous reviewers for
their very insightful and productive suggestions.
REFERENCES
Boone, C. G. (2008). Improving resolution of cen-
sus data in metropolitan areas using a dasymetric
approach: Applications for the Baltimore ecosys-
tem study. Cities and the Environment, 1(1), 3.

1166
Cultural Dasymetric Population Mapping with Historical GIS
Bowman, M. K. (1965). Reference book of
Wyoming County history. Parsons, WV: McClain
Printing Company.
Brown, J. S. (1988). Beech Creek: A study of a
Kentucky mountain neighborhood. Berea, KY:
Berea College Press.
Chen, K., McAneney, J., Blong, R., Leigh, R.,
Hunter, L., & Magill, C. (2004). Defining area
at risk and its effect in catastrophe loss estima-
tion: A dasymetric mapping approach. Applied
Geography (Sevenoaks, England), 24(2), 97–117.
doi:10.1016/j.apgeog.2004.03.005
Cook, S. R. (2000). Monacans and miners: Na-
tive American and coal mining communities in
Appalachia. Lincoln, NB: University of Nebraska.
Crandall, M., & Weber, B. (2005). Defining rural
Oregon: An exploration. Corvallis, OR: Oregon
State University.
Daniels, T. (1999). What to do about rural sprawl?
Paper presented at the American Planning As-
sociation Conference, Seattle, WA.
Feller, J. W. (2001). Memories and photos of
“Mullins,” West Virginia: Vol. 1. 1894 to 1925.
St. Albans, WV: Harless Printing Company.
Fielder, R., Schuurman, N., & Hyndman, J. (2004).
Improving census-based socioeconomic GIS for
public policy: Recent immigrants, spatially con-
centrated poverty and housing need in Vancouver.
ACME: An International E-Journal for Critical
Geographies, 4(1), 145–171.
Garb, J. L., Cromley, R. G., & Wait, R. B. (2007).
Estimating populations at risk for disaster pre-
paredness and response. Journal of Homeland
Security and Emergency Management, 4(1), 3.
doi:10.2202/1547-7355.1280
Goodchild, M. F., Anselin, L., & Deichmann, U.
(1993). A framework for the areal interpolation
of socioeconomic data. Environment & Planning
A, 25, 383–397. doi:10.1068/a250383
Grannis, R. (2005). T-communities: Pedestrian
street networks and residential segregation in
Chicago, Los Angeles, and New York. City &
Community, 4(3), 295–322. doi:10.1111/j.1540-
6040.2005.00118.x
Gulliford, A. (1996). America’s country schools
(3rd ed.). Niwot, CO: University Press of Colo-
rado.
Hart, J. F. (1975). The look of the land. Upper
Saddle River, NJ: Prentice Hall.
Hasse, J., & Lathrop, R. G. (2003). A housing-
unit-level approach to characterizing residential
sprawl. Photogrammetric Engineering and Re-
mote Sensing, 69(9), 1021–1030.
Holt, J. B., Lo, C. P., & Hodler, T. W. (2004).
Dasymetric estimation of population density and
areal interpolation of census data. Cartography
and Geographic Information Science, 31(2),
103–121. doi:10.1559/1523040041649407
Holy, T. C. (1929). Survey of education in West
Virginia: Vol. III. School buildings. Charleston,
WV: West Virginia State Department of Education.
Irwin, E. G., & Bockstael, N. E. (2002). Interacting
agents, spatial externalities, and the evolution of
residential land use patterns. Journal of Economic
Geography, 2, 31–54. doi:10.1093/jeg/2.1.31
Khomarudin, M. R., Strunz, G., Post, J., Zobeder,
K., & Ludwig, R. (2008, November). Spatial im-
provement of information on population distribu-
tion using GIS approaches: An input for tsunami
people exposure assessment. Paper presented at
the International Conference on Tsunami Warning
(ICTW), Bali, Indonesia.
Kirk, R. W. (2009). Land use and terrestrial
carbon storage in Western North Carolina from
1850-2030: A historical reconstruction and simu-
lation study. Unpublished doctoral dissertation,
University of Minnesota, Duluth, MN.
1167
 Cultural Dasymetric Population Mapping with Historical GIS

Langford, M., & Unwin, D. J. (1994). Generating
and mapping population density surfaces within
a geographical information system. The Carto-
graphic Journal, 31(1), 21–26.
Lewis, R. L. (1998). Transforming the Appa-
lachian countryside: Railroads, deforestation,
and social change, 1880-1920. Chapel Hill, NC:
University of North Carolina.
Maantay, J., & Maroko, A. (2009). Mapping ur-
ban risk: Flood hazards, race, and environmental
justice in New York. Applied Geography (Sev-
enoaks, England), 29(1), 111–124. doi:10.1016/j.
apgeog.2008.08.002
Maantay, J., Maroko, A., & Culp, G. (2009). Us-
ing geographic information science to estimate
vulnerable urban populations for flood hazard and
risk assessment in New York City. In Showalter,
P. S., & Lu, Y. (Eds.), Geospatial techniques in
urban hazard and disaster analysis (pp. 71–97).
New York, NY: Springer. doi:10.1007/978-90-
481-2238-7_5
Maantay, J. A., Maroko, A. R., & Porter-Morgan,
H. (2008). A new method for mapping popula-
tion and understanding the spatial dynamics of
disease in urban areas: Asthma in the Bronx,
New York. Urban Geography, 29(7), 724–738.
doi:10.2747/0272-3638.29.7.724
Martin, C. E. (1984). Hollybush: Folk building
and social change in an Appalachian community.
Knoxville, TN: University of Tennessee Press.
Martin, D. (1996). An assessment of surface
and zonal models of population. International
Journal of Geographical Information Systems,
10(8), 973–989. doi:10.1080/026937996137684
Mennis, J. (2002). Using geographic information
systems to create and analyze statistical surfaces
of population and risk for environmental jus-
tice analysis. Social Science Quarterly, 83(1),
281–297. doi:10.1111/1540-6237.00083
Mennis, J. (2003). Generating surface models
of population using dasymetric mapping. The
Professional Geographer, 55(1), 31–42.
Mennis, J. (2009). Dasymetric mapping for es-
timating population in small areas. Geography
Compass, 3(2), 727–745. doi:10.1111/j.1749-
8198.2009.00220.x
Mesev, V., & McKenzie, P. (2008). Urban neigh-
borhood pattern recognition using high spatial
resolution remotely sensed data and point-based
GIS data sources. In Wise, S., & Craglia, M.
(Eds.), GIS and evidence-based policy making
(pp. 19–42). Boca Raton, FL: CRC Press.
Minetti, A. E., Moia, C., Roi, G. S., Susta, D., &
Ferretti, G. (2002). Energy cost of walking and
running at extreme uphill and downhill slopes.
Journal of Applied Physiology, 93, 1039–1046.
Moon, Z. K., & Farmer, F. L. (2001). Population
density surface: A new approach to an old problem.
Society & Natural Resources, 14, 39–49.
Moudon, A. V., Lee, C., Cheadle, A. D., Garvin, C.,
Johnson, D., & Schmid, T. L. (2006). Operational
definitions of walkable neighborhood: Theoreti-
cal and empirical insights. Journal of Physical
Activity & Health, 3(1), 99–117.
Openshaw, S. (1984). The modifiable areal unit
problem: Concepts and techniques in modern
geography. Norwich, UK: Geo Books.
Openshaw, S. (1996). Developing GIS-relevant
zone-based spatial analysis methods. In Longley,
P., & Batty, M. (Eds.), Spatial analysis: Modelling
in a GIS environment (pp. 55–74). New York, NY:
John Wiley & Sons.
Rice, O. K. (1986). A history of Greenbrier County.
Lewisburg, WV: Greenbrier Historical Society.

1168
Cultural Dasymetric Population Mapping with Historical GIS
Schuurman, N., Leszczynski, A., Fielder, R., Gr-
und, D., & Bell, N. (2006). Building an integrated
cadastral fabric for higher resolution socioeco-
nomic spatial data analysis. In Proceedings of
the 12th International Symposium on Progress
in Spatial Data Handling (pp. 897-920).
Theobald, D. M. (2005). Landscape patterns of
exurban growth in the USA from 1980 to 2020.
Ecology and Society, 10(1), 32.
Tobler, W. (1979). Smooth pycnophylactic inter-
polation for geographical regions. Journal of the
American Statistical Association, 74, 519–530.
doi:10.2307/2286968
Towers, G. (2010). Rediscovering rural Appala-
chian communities with historical GIS. South-
eastern Geographer, 50(1), 58–82. doi:10.1353/
sgo.0.0065
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 4,
edited by Donald Patrick Albert, pp. 38-56, copyright 2011 by IGI Publishing (an imprint of IGI Global).
Trodd, N., & Gearey, W. M., Jr. (2006). A participa-
tory GIS-based method to construct output areas
for neighbourhood analysis and social planning.
Retrieved from http://www.urisa.org/gearey
Trotter, J. W. Jr. (1990). Coal, class, and color:
Blacks in Southern West Virginia, 1915-32. Ur-
bana, IL: University of Illinois.
University of Virginia, Geospatial and Statistical
Data Center. (2004). Historical census browser.
Retrieved from http://fisher.lib.virginia.edu/col-
lections/stats/histcensus/index.html
Wilhelm, G. Jr. (1978). Folk settlements in the
Blue Ridge Mountains. Appalachian Journal,
5(2), 204–245.
Wright, J. K. (1936). A method of mapping
densities of population with Cape Cod as an
example. Geographical Review, 26, 103–110.
doi:10.2307/209467
1169
1170

Chapter 70
Using Volunteered Geographic
Information to Assess the
Spatial Distribution of West
Nile Virus in Detroit, Michigan
Kevin P. McKnight Ashton M. Shortridge
Michigan Department of Transportation & Michigan State University, USA
Michigan State University, USA
Meghan D. Burns
Joseph P. Messina Montana Natural Heritage Program, USA
Michigan State University, USA
Bruce W. Pigozzi
Michigan State University, USA

ABSTRACT
West Nile Virus is a vector-borne flavivirus that affects mainly birds, horses, and humans. The disease
emerged in the United States in 1999 and by 2001 had reached Michigan. In clinical human cases, the
most common symptoms are fever, weakness, nausea, headache, and changes in mental state. The crow is
the most common wildlife host in the life cycle of the virus. The state of Michigan, through the Michigan
Department of Community Health, collected the spatial locations of over 8,000 dead birds (Corvidae),
statewide, during 2002. The large number of samples made spatial and temporal hotspot detection pos-
sible. However, the volunteer reporting method produced a dataset with a direct correlation between
the numbers and locations of the dead birds and human population density and accurately identifying
hotspots remains a challenge. Geographic variation in dead bird intensity was modeled using both
global and local spatial clustering algorithms. Statistical models identified overall spatial structure and
local clustering. Identification of hotspots was confounded by limited information about the collection
procedures, data availability and quality, and the limitations of each method.

DOI: 10.4018/978-1-4666-2038-4.ch070

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
INTRODUCTION
West Nile Virus (WNV) is a vector-borne RNA
flavivirus that affects mainly birds, horses, and
humans (Anderson, Vossbrinck, Andreadis, Iton,
Beckwith, 2001). This paper considers the spatial
distribution of West Nile Virus across a major
metropolitan region from a medical geography
perspective, which is concerned with the relation-
ship between diseases and their spatial contexts.
Examination of the timing and geographic locations
of disease cases, as well as their environmental
and social settings, can lead to identification of
patterns and associations among the individual
events together with key covariates. These as-
sociations may reflect underlying environmental
influences and lead to greater understanding about
the factors and drivers affecting the transmission
and geographic diffusion of a disease, greatly
facilitating the response of public health officials.
An important innovation in medical geography is
the development of statistically-driven hot spot
detection methods (Anselin, 1995; Fotheringham
& Zhan, 1996; Openshaw, 1996). Such methods
are typically employed when extensive and often
exhaustive, high-accuracy spatial data on disease
are available (Ozdenerol, Bialkowska-Jelinska,
Taff, 2008). Frequently, secondary demographic
datasets are employed to model the population at
risk, thereby accounting for spatial variation in that
background population and helping identify anoma-
lous hot spots (Cooke, Katarzyna, Wallis, 2006).
In contrast to the contexts of previous stud-
ies with highly accurate and extensive data, this
paper is concerned with the employment of vol-
unteered geographic information in the analysis
and detection of hot spots. The term, “volunteered
geographic information,” (VGI) is used to indicate
data that are measured and reported by the public.
The accuracy and completeness of such data may
be quite variable; in some cases, it may be more
reliable than authoritative sources (Goodchild,
2008), especially in periods of rapid and critical
change such as natural or human disasters. For
example, during a wildfire, a cell phone equipped
person on the spot may be in a much better posi-
tion to inform the public about local conditions.
Important advantages of VGI in the context of
this paper include cost-efficiency, timeliness,
and citizen engagement with public health is-
sues. However, its utility for subsequent spatial
analysis may be questionable due to biases and
data quality concerns.
Focusing on data collected during the 2002
WNV epidemic in southeastern Michigan, USA,
this paper explores two related research questions.
1. Are volunteered geographic health data
suitable for use in statistical spatial hot spot
detection applications?
2. Assuming that the previous question is true,
then during the 2002 WNV epidemic in
Michigan, did spatial or temporal clusters
emerge that might be indicators of future
risks?
The remainder of this introduction provides
important background on WNV, its spread in
southeastern Michigan, and the surveillance meth-
ods employed there. Then it considers the nature
of available VGI for the epidemic in this region.
It identifies critical uncertainties in the available
information that limit the kinds of questions that
can be asked and conclusions that can be drawn.
The paper continues by discussing data processing,
including geocoding, which is a critical operation
for using VGI, and describes the spatial hotspot
detection methods used in the research. We then
report our findings, and discuss what those findings
actually indicate about the spatial and temporal
spread of WNV across the Detroit metropolitan
area in 2002.
WEST NILE VIRUS
In 1999, WNV became an endemic disease in the
Western Hemisphere. The disease was first iden-
1171
 Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
tified in New York City in 1999, resulting in 67
human cases and seven deaths (Crave & Roehrig,
2001). Over the next few years, the disease spread
across the United States through a bird-mosquito-
bird cycle in which the birds serve as the viral
reservoir and mosquitoes spread the virus through-
out susceptible populations. When environmental
factors such as weather or host immunity boost
viral amplification, an increase in the number of
infected “bridge vector” mosquitoes (mosquitoes
that bite both humans and birds) results (Petersen
& Marfin, 2002). Currently 43 species of mos-
quito have tested positive for WNV (Marra et al.,
2004), though the Culex species of mosquito is
the most common bridge vector transferring the
virus from birds to humans. Among birds, the
American crow (Corvus brachyrhynchos) and
the blue jay (Cyanocitta cristata) in the family
Corvidae have seen the highest mortality rates.
In the northeastern United States increases in
crow deaths have been quickly followed by the
emergence of WNV cases in humans (Petersen &
Marfin, 2002). In its first five years in the United
States, the disease had infected at least 400,000
humans and killed countless birds, mammals, and
reptiles (Dodd, 2003; Marra et al., 2004).
Beginning in 2001, disease surveillance in
Michigan was conducted in part via a state man-
aged toll-free hotline which citizens could call
to report dead bird sightings. The location and
condition of each bird was recorded and the bird
was tested for the virus. By August 2001, WNV
was confirmed in dead crows in Michigan. Of the
142 dead crows and 79 blue jays tested in 2001,
63 crows and 2 blue jays tested WNV positive. In
that first year of monitoring, WNV was detected
in corvids in ten counties (Wayne, Oakland, Ma-
comb, Washtenaw, Jackson, Calhoun, Ingham,
Barry, Ottawa, and Muskegon) as well as in three
mosquito pools (two in Oakland County and one in
Macomb County). The 2002 transmission cycle in
Michigan was the most significant. With continu-
ous surveillance, the first human WNV cases were
reported on August 16, 2002 with a final epidemic
1172
tally of 614 confirmed human cases. Of the nine
counties surveyed in the Lower Peninsula, seven
produced WNV-positive mosquito pools resulting
in a total of 58 positive pools. The majority of
the WNV-positive mosquitoes were Culex, but
a Coquillettidia perturbans pool tested positive
as well (State of Michigan, 2006). At the time of
writing the program continues, with the State of
Michigan providing web-based services to report
dead birds and mammals while the USGS main-
tains disease maps for reported corvid, human,
mosquito, sentinel, and veterinary cases.
This paper focuses on the 2002 dead bird
data set for the Greater Detroit metropolitan area
(Wayne, Oakland, and Macomb counties), Figure
1, for three reasons. First, Metro Detroit, with
about 4,043,000 people and covering about 2,150
square miles, is the largest population center in
Michigan and one of the major urban regions in
the United States. The potential hazard offered by
WNV is greatest in this region. Second, this area
contains a majority of the 2002 dead bird records.
This abundance makes it particularly tractable for
conducting high spatio-temporal resolution hot
spot analysis. Third, geographic, demographic,
land cover and road framework, data are available
and relatively accurate, making both geocoding
and multivariate analysis particularly feasible.
The combination of widespread human popula-
tion density, a large number of dead bird records,
mostly developed landcover, and accurate spatial
referencing information makes Greater Detroit
the overall best site for this research in the state
of Michigan.
VOLUNTEERED GEOGRAPHIC
INFORMATION
What kinds of questions could we ask of this
dataset, and what assumptions must be made?
One approach is to assume that the likelihood of
reporting a dead bird varies by spatial location,
in part, as a function of the human population
Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
Figure 1. Detroit Metropolitan Area study site with dead bird (crows) GAM clusters by month
density around that location. The logic behind
this assumption is that the more human eyes there
are in an area, the more likely a dead bird is to be
reported. By controlling for human population, we
can identify locations with anomalous concentra-
tions of dead birds. Anomalous concentrations in a
particular area may be due to one or a combination
of several factors:
• Elevated rate of WNV among corvids
• Elevated mortality rate due to non-WNV
causes
• Elevated population of corvids
• Higher propensity of volunteers to call in
birds
A second and potentially much more serious
problem is the interpretation of areas with anoma-
lously low concentrations of reported dead birds,
after controlling for human population density.
These could be due to errors of omission rather
than an actual lack of bird mortality:
• People did not see dead birds
• People did not report dead birds
These errors of omission could be due to
geographic variability in human propensity to
identify dead birds because of environmental or
social barriers to interaction (dense vegetation,
neighborhood design or demographics leading to
fewer people spending time out of doors). Under-
reporting might also be due to spatial variation
in the public’s awareness or interest in the WNV
surveillance program. Finally, low numbers of
dead birds at a particular location might simply
be due to smaller concentrations of corvids in that
area. The overall infection or fatality rate might
be just as high as in other places, but the location
would not be flagged.
1173
 Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
Consequently, the utility of volunteered geo-
graphic information for identifying anomalous
concentrations of WNV-infected corvids appears
very limited, despite an intensive data processing
effort that resulted in exceptionally clean geocoded
information and the use of modern spatial statistics.
These data seem to be more useful for what they
suggest about the likelihood for human interac-
tion with WNV. Each record is in fact a report
about a close encounter between a human and a
dead bird, and may therefore serve as an indicator
of exposure risk. In this case we simply have to
assume that the bird was infected, and, in turn,
infected more mosquitoes in close proximity to
its place of death, and that the chance of a human
being bitten by an infected mosquito is therefore
elevated in that area. This approach avoids prob-
lems associated with several, though not all, of
the factors listed above. In the discussion later in
the paper, we describe what steps are possible to
assess the impact of some of those factors.
DATA PREPROCESSING
After the research team obtained the dead bird
data from the state, substantial preprocessing was
required prior to analysis. The street addresses of
dead birds had to first be geocoded. Geocoding, or
address matching, is a process by which addresses
are given geographic coordinates (e.g., latitude
and longitude) based on a reference layer or a
street database (Bonner, Daikwon, Nie, Rogerson,
Vena, & Freudenheim, 2003). Where no explicit
coordinates at the land parcel level exist within
the reference layer, address range data for street
segments are employed to interpolate coordi-
nates for the address. The process of geocoding
is indispensable in spatial epidemiology given
that it is a relatively inexpensive and efficient
mechanism for locating cases. For volunteered
geographic information, where citizens gener-
ally identify location with addresses rather than
spatial coordinates, establishing a cost basis is
1174
less clear. Given the combined costs associated
with large numbers of cases and patient privacy
regulations, the collection and use of disease
surrogates as the dead birds, in this case, is very
cost effective (Bonner, Daikwon, Nie, Rogerson,
Vena, & Freudenheim, 2003).
From the original statewide database only the
crows and blue jays were selected for geocoding.
Since the MDCH recorded the address only once
and then recorded the number of birds of each
species, a separate record had to be created for
each bird. Misspellings or capitalization variants of
crow and blue jay were also queried and replaced.
The geocoding process was completed in ArcGIS
(ESRI, 2007). The reference data layer utilized
was the Michigan Framework All Roads data, last
updated May 1, 2009, and provided by the state of
Michigan; available for download from the Michi-
gan Geographic Data Library (Michigan Center
for Geographic Information). Since this database
contains address range information for both the
left and right side of each street, the dead bird
addresses could be geocoded to their approximate
positions along the length and the correct side of
the street segment. For the matching options the
spelling sensitivity was set at 80, the minimum
candidate score at 10, and the minimum match
score at 60. These parameter settings improved
the completeness of the geocoding operation by
relaxing the constraints on spelling accuracy and
the numeric address inputs. Resulting geographic
coordinates for each dead bird record were added to
the dataset. Each address was checked individually
to make sure it was located in the proper county
or city as the other attributes collected. In the case
of potentially misspelled street names the street
name was changed if it matched street address
records in the data collection. In some cases the
standardized street address had to be changed to
match the roads database.
An extensive manual procedure was conducted
for records with street addresses that could not
be matched. First, the address was searched us-
ing the Internet mapping engine MapQuest™.
Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus

If the street address could not be located using
MapQuest™, the nearest street intersection was
inspected to determine whether the street address
numbers had changed, more houses had been built
since the roads file was last updated, or the street
name was misspelled. If the address still could not
be located then a rematch was attempted using
the nearest street intersection. If this final effort
failed, or if more than one possible location was
identified, then the address remained unmatched
and was discarded.
Once the addresses were geocoded, an analysis
of the matched addresses was conducted. Only
those corvid entries that contained a bird type
and sighting data were used. For computational
efficiency, the sighting dates were converted to
the day of the year (ordinal date). For example,
May 1, 2002 would be converted to the 121st day
several models can account for associated popu-
lation density around dead bird locations. Popula-
tion data were obtained from the 2000 US Census
for this purpose and census block group data, the
finest spatial resolution dataset available for the
entire study area, were employed.
SPATIAL STATISTICS
After preprocessing was completed, three broadly
distinctive detection models were used: Moran’s
I, Local Indicators of Spatial Association (LISA),
Table 1. Crows by county
County Dead Bird Population Ratio
Records

of the year. Additionally, the week number was Oakland 887 1,194,156 .00074278

assigned to each dead corvid sighting. Macomb 502 787,625 .00063736

The dataset was clipped to the spatial extent Wayne 1,108 2,061,162 .00053756

of the study area (Oakland, Macomb, and Wayne TOTAL 2,497 4,042,943 .00061762
counties) (Figure 1). This reduced the original
8,249 statewide records to a study area total of Table 2. Crows by month
2,533 bird records, which were subjected to further
quality assessment. With the dataset cleaned of Month Records
unreliable records, 2,497 dead bird locations, all April 2
of which were crows, remained. Crow locations May 32
were attributed with spatial boundary informa- April and May 34
tion. Using a tagging function, township, city, and June 59
block group information were appended to each June 1-15 34
data point. This final data preparation process was June 16-30 25
completed using TransCAD™ (Caliper, 2006). At July 396
this point, the dataset was considered finished. As July 1-10 20
Table 1 illustrates, most of the dead birds were July 11-20 108
found in Wayne County, with Oakland and Ma-
July 21-31 268
comb Counties having comparable ratios of dead
August 1,826
birds to people. The data were then partitioned by
August 1-10 875
month and within the month over the course of
August 11-20 461
the transmission season (Table 2).
August 21-31 490
Given the method of dead bird data collection,
September 166
it appears likely that the underlying human
October 16
population influenced the bird data simply due to
TOTAL 2,497
opportunity. By incorporating human population,

1175
 Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus

and the Geographic Analysis Machine. The first
two tests were completed with the ClusterSeer™
software package and the third was coded using
the statistical programming language R. These
approaches were chosen because they allowed a
progression from global cluster detection using
solely the dead bird data, through neighborhood-
Where N equals the number of regions, wij is
a weight representing the connectivity between
areas i and j, zi is the rate in zone i centered around
the mean rate (using zi = xi – x ; xi is the rate in
zone i); and S0 is the sum of the weights.
N N

scale metrics for hot-spot identification, to an

S0 = ∑ ∑w ij
,i ≠ j (2)
i= 1 j= 1
adaptive model assessing local dead bird con-
centrations while accounting for the underlying
human population. Global metrics only consider The expectation of I for a spatially random
whether the pattern is non-spatially random and process is:
not where anomalous concentrations are located.
Local models assess the spatial dependence of −1
E (I ) = (3)
deviations in attribute values from their mean: that (N − 1)
is, the second order spatial properties (Anselin,
1995; Bailey & Gatrell, 1995). Locally elevated
The expectation of I gets closer to 0 as N
concentrations can then be identified and mapped
increases, and can range from -1 to 1. The vari-
as clusters. By investigating patterns for different
ance of I for a spatially random process depends
time periods during 2002, the pattern and intensity
on the assumptions of the research question. In
of a spreading virus epidemic can be estimated.
this work, we employ a random permutation
The Moran Contiguity Coefficient, called
hypothesis, in which the observed I is presumed
Moran’s I, is a weighted correlation coefficient
to be drawn at random from the distribution of
used to detect departures from spatial random-
I metrics calculated on the attribute set. Monte
ness (Moran, 1950). It is typically calculated for
Carlo simulations are used to construct a ranked
aggregated variables (here, dead bird counts)
distribution of I values; the empirical I is compared
stored in geographical zones; in this research
with this distribution to determine significance.
these zones were generated by dividing the study
Moran’s I is a global statistic, a single number
area into a 10 km by 10 km grid. Departures from
representing the overall degree of spatial autocor-
randomness may suggest the presence of patterns
relation in a variable measured in spatial zones.
such as clusters, but may also identify geographic
It does not provide any information about where
trends. Positive spatial autocorrelation means that
elevated values might be grouped within the study
nearby areas tend to have similar values; in this
area. Such areas may be identified through the use
case, the numeric value for Moran’s I will be large
of local indicators of spatial association, or LISA
and positive. A value close to 1 is indicative of
statistics. Local Moran’s I is one such metric;
high positive spatial autocorrelation or a clustered
it is literally a disaggregated global Moran’s I,
spatial structure. Moran’s I is defined as:
defined as:
N N
N ∑ ∑ wij z i z j N

i= 1 j= 1
N ∑ wij z i z j
I= (1) j= 1
N Ii = (4)
S0 ∑ z 2 N

i −1
i
S0 ∑ z 2
i
i −1

1176
Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
Where the terms are as defined for Equation
3. The Moran metric I is calculated for each loca-
tion i in the region, and can be mapped. A high
positive value for location i indicates the presence
of clusters of similar attribute values around that
location. A value near zero indicates that the local
area under consideration or the average of the im-
mediate neighbors of the cell, or both, is near zero.
The Geographic Analysis Machine (GAM) is
a multi-scale cluster location tool used to analyze
spatial point distributions to find evidence of lo-
calized geographical clustering while accounting
for background population at risk. It was initially
designed to identify disease clusters such as child-
hood leukemia (Openshaw, Craft, Charlton, &
Birch, 1988; Openshaw, 1996). The process is a
computationally intensive approach for automated
cluster identification. It incorporates a technique
that compares the intensity of events within
circles of varying radii; the circles are centered
on a fine grid imposed over the area of interest.
The observed number of events (e.g., dead birds)
within each circle is compared to an expected
number of events based on a derived background
rate, which controls for population, via a Poisson
test. Circles with significantly elevated numbers
of observed cases are then flagged and retained.
An important user decision in the GAM model
is the rejection criterion for the significance test.
The level of significance serves as a filter to reject
circles with lower rates. We chose to set this level
high, (Openshaw, Craft, Charlton, & Birch, 1988),
to focus on areas of greatest cluster intensity. The
models were all initially completed with a 0.01
rejection criterion; however this caused large
areas to be identified as significant. Therefore,
smaller thresholds were employed, as low as 1 x
10-9, to filter out large numbers of circles. Due to
the number and locational variability of the data
records in each time period and the underlying
population in these locations, different confidence
intervals were required to achieve the greatest
focus of the cluster detection.
The result of the filtering process is a set of
circles containing significantly elevated numbers
of cases. These circles are then smoothed with a
kernel density function to generate maps of hot
spots (Openshaw, Charlton, Wymer, & Craft,
1987). As GAM employs a variety of circle radii, it
does not require the specification of a single spatial
scale for cluster presence. Although it incorporates
statistical testing to identify cluster significance,
it is considered an exploratory method.
RESULTS
The data were aggregated to a 10 by 10 grid over
the study area and partitioned by month due to the
density of the data points. This enabled the calcu-
lation of global and local Moran statistics on the
spatial pattern of aggregated bird counts for time
periods throughout the transmission season. The
Moran’s I result for all records was 0.63, showing
strong, significant positive spatial autocorrelation
with substantial temporal variation. Early in the
season the Moran’s I value showed a small peak
in May rising to 0.27. There is a brief drop-off
during June, with a sharp increase toward the latter
half of the month. The Moran’s I value increases
through July showing increased autocorrelation
with the seasonal peak in the first part of August
with a value of 0.64. The model for August showed
the greatest positive autocorrelation.
Maps of Local Moran’s I suggest some tempo-
ral and spatial pattern consistency in the aggregated
dead bird data. For all data, clusters appeared to
be concentrated near the center of the study area
(Birmingham/ Royal Oak areas) and along the
eastern edge (Grosse Pointes area). Monthly totals
provide a perspective on changes through time.
Insufficient numbers of records for April forced
their inclusion with the records from May for
modeling. Results for April/May showed cluster-
ing in areas with greater human population. For
June, higher concentrations shifted to the Livonia/
Westland area. The July results demonstrated a
1177
 Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
cluster pattern in the inner suburbs surrounding
the city of Detroit, with moderate intensity in the
cities of Royal Oak and Warren. The August data,
including the majority of the records, suggested a
shift in cluster intensity from areas immediately
around the city of Detroit to smaller cities in the
metro area. September showed very moderate
clustering throughout the middle of the study area.
Results for October presented some clustering
along the eastern side of Oakland County.
The results of GAM for all records (Figure 1)
in this model showed a very similar pattern to the
local Moran map. It identified a large central clus-
ter over the Birmingham/ Royal Oak/ Southfield
areas, with a smaller and separate cluster over
Grosse Pointe. GAM results for the month-by-
month results were variable, however. The April/
May results showed two small clusters. The larger
of the two clusters was centered over Pontiac
Township and the smaller cluster was located over
the Grosse Pointe communities. Results for June
initially showed one large cluster over the southern
part of the study area. However after adjusting the
significance level most of the intensity was over
the Westland area, with a thin band to the north
reaching into Livonia. For July, GAM identified
a large cluster stretching from Birmingham to St.
Clair Shores, and another large cluster reaching
from Livonia to Westland. Results for August
show similar patterns to the GAM results for all
records. Two distinct clusters for the month of
August were identified. The larger more intense
cluster centralized over the Birmingham area,
and the second cluster covered Grosse Pointe.
For September, clustering was identified over the
northern half of the area. At the initial significance
level clustering was widespread; however after
changing the level clusters appeared to be focused
to two locations near the village of Clarkston.
With few data points in October, GAM clusters
moved even further into the rural areas of the
study area. The main cluster was located in west
Oakland County; this is consistent with the local
Moran’s I results.
1178
DISCUSSION
The preparation process and subsequent analyses
of the geocoded dead crow data resulted in some
interesting findings. We consider data processing
results and their implications for similar stud-
ies first, and then evaluate the results from the
clustering study. Through the combined aid of
an internet mapping website and ancillary data,
an address matching rate of 83.5% was achieved,
with 72.1% of the records scoring a quality rating
of 75 or above. This indicates a relatively com-
plete and accurate conversion of the volunteered
data. Intensive manual validation of records was
critical in achieving these high rates. With the
address matching rate of commercial geocoding
companies found to be 60 – 80% (Kreiger, Water-
man, Lemieux, Zierler, & Hogan, 2001), it may
be advantageous for researchers to conduct their
own geocoding depending on time and resources.
The positional accuracy of address matching
is an important consideration, as errors increase
result uncertainty. Sources of error include the
positional accuracy of the road network, the cor-
rectness of the recorded address; attribute accuracy
of the address ranges stored in the roads data, and
the fidelity of the interpolation along the segments
to the actual address locations (Karimi & Durcik,
2004). One study found that the majority of ad-
dress derived geocoded points fall within 100 m
of GPS points with an overall median distance of
38 m (Bonner, Daikwon, Nie, Rogerson, Vena, &
Freudenheim, 2003). At the metropolitan scale of
this application, errors of this magnitude appear
inconsequential. For neighborhood-scale research
projects, this degree of positional error might be
unacceptable.
A more fundamental source of uncertainty in
this analysis is due to the volunteered nature of
the data; this is asserted geographic information,
not the result of a systematic census or a designed
sample (Goodchild, 2008). Use of such data raises
interesting questions that impact the results re-
ported here, and have important implications for
Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
studies employing volunteered data with similar
characteristics to those employed in the present
study. Often surveillance of emerging infectious
diseases demands an opportunistic approach to
data collection.
It is useful to contrast the data in hand with
the information we would have liked. In a hot
spot analysis, we would like to know the spatial
distribution of the population at risk: here, the
corvid population in Metro Detroit during the
weeks of the study. We would also like to have an
exhaustive, or mapped, set of all WNV-positive
date-referenced dead corvid locations. A second-
best corvid dataset would be a random sample
of all dead corvid locations. Instead we have no
information about the geographic distribution of
the underlying bird population, and the sample
of dead birds is biased: only birds identified and
reported by volunteers are in the database. At best,
this is a dataset of the geographic intersection of
informed human volunteers and dead birds.
The effects of this are clear from both the spatial
and attribute information obtained. The geographi-
cal distribution of dead birds is positively and
strongly associated with human populations. This
association is not uniform, however, as certain
urban areas consistently report dead birds, and
other urban areas consistently do not. Patterns
in the attribute data also suggest reporting bias.
Particularly notable is the striking difference in
quantities of American crow and blue jay reports.
There were nearly five times as many American
crow sightings than blue jay sightings in the
statewide database and none in our subset. This
may be due to the larger physical size of the crow,
its identifiability, or its natural behavior to roost
and to scavenge for food in urban environments.
Alternatively, the impact of WNV on the American
crow may have been reported more in the media
and thus the public may have been more interested
or aware of American crow deaths.
Both Local I and GAM methods consistently
identified persistent clusters in inner-ring suburban
communities. We raise two competing explana-
tions for clusters in these areas: the environmental
hypothesis and the reporting bias hypothesis.
The environmental hypothesis is that these areas
possess qualities that are amenable to crows and/
or WNV-positive mosquitoes. If one accepts this
hypothesis, the hot spots observed in these areas
are important indicators of WNV infection, and
their variation through the warm months of 2002
can provide insight into the diffusion of the dis-
ease across metropolitan Detroit. In contrast, the
reporting bias hypothesis is that higher reporting
levels in certain communities are due to greater
information about the surveillance campaign and
greater willingness to participate. Geographic
variation in both the effectiveness of advertis-
ing methods used to inform the public about the
reporting program and social and cultural char-
acteristics related to willingness to participate in
the program have a direct effect on the patterns
seen in these data. In support of this hypothesis we
note that the communities associated with the most
pronounced clusters are some of the most affluent
in the study area and may enjoy improved access
to media information about WNV and the citizen
surveillance program, increased willingness to
report sightings, and greater interaction with the
outdoors. All of these factors could explain higher
reporting levels. In contrast, no clusters were
observed within the Detroit City limits and lower
income communities. People in these areas may
not have been informed of the program or might
have been less interested in participating. Under
this hypothesis, observed dead bird clusters are
no less “real”, but they may not reflect the spatio-
temporal patterns of WNV impact so much as
the socio-economic geography of Metro Detroit.
CONCLUSION
This research was concerned with the utility of
volunteered geographic information reported by
private citizens to track the progress of an infec-
tious, vector-borne disease across a metropolitan
1179
 Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus
region. By employing this volunteered, asserted
information, many well-known problems in de-
signed, authoritative data collection efforts are
exacerbated. Of particular concern are errors in
data collection and entry, post-processing accuracy
assessment and the general credibility of the as-
serted data. The problem of false negatives – of
dead birds that weren’t reported – is acute. Data
records are biased to areas where informed, will-
ing people are located. We analyzed the impact
of these problems on an exploratory investigation
into geographic variability in a database of dead
bird sightings in the warm months of 2002 in
metropolitan Detroit. Standard global and local
autocorrelation tests as well as a multiscale clus-
tering algorithm were implemented. Despite the
challenges, similar results were found using each
of the models. All local models identified clusters
in similar areas. One large cluster was present in
all three of the local models for the entire metro-
politan Detroit dataset. Modest temporal variation
in local clustering is apparent. However, areas
with clusters of dead bird records also supported
large human populations. We posit two competing
hypotheses for the nature of these clusters: that
they are due to actual variation in corvid mortal-
ity, or that they are due to variation in knowledge
and willingness of the local human population to
report dead bird sightings. Further information
is needed to determine which hypothesis is more
valid. Further research on this topic could follow
two different tracks. First, from the clustering per-
spective, additional years of dead bird data, along
with alternative models and methods to identify
spatial and temporal clustering in the dead bird
data would be useful. The second would examine
the collection methods, environmental, and bio-
logical aspects related to the spread of WNV. By
obtaining more complete information about the
local etiology of WNV, a more compelling model
for the underlying geographic distributions of bird
and mosquito populations could be constructed.
There are many facets to disease cluster analy-
sis that go beyond the information found at the
1180
case-location level. Aspects of the environment
such as land cover and land use should be consid-
ered. Mosquito breeding micro-environments need
to be identified to determine points where the vec-
tor might be initially located. Also, the proximity
of birds and humans to these locations should be
known, and information regarding the flying range
of the vector would identify areas where humans
and birds would be at risk. Examining the breed-
ing cycle of the vector would help in identifying
possible periodic blooms in mosquito activity. This
research did not examine the roosting patterns
of the bird population; this information would
identify the range of the host, and combined with
the mosquito information, would identify the host
population at risk. Research into the case fatality
rate of WNV on the bird population, as well as
the effects of the virus on the movement, and the
length of time before mortality would also assist
in the temporal study. One of the large drawbacks
to the integrity of the dead bird data was that
there was not 100% testing of the reported birds;
complete testing of the birds would remove any
false positive errors in the sample. Also obtain-
ing a GPS (Global Positioning System) location
of the dead bird would remove the location error
associated with address matching and the error
in the base GIS layers. Finally an examination of
the WNV public information campaign would be
valuable for correcting reporting biases, and for
improving future efforts to obtain volunteered
geographic information.
REFERENCES
Anderson, J. F., Vossbrinck, C. R., Andreadis,
T. G., Iton, A., & Beckwith, W. H. (2001). A
phylogenetic approach to following West Nile
virus in Connecticut. Proceedings of the Na-
tional Academy of Sciences of the United States
of America, 98(23), 12885–12889. doi:10.1073/
pnas.241472398
Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus

Anselin, L. (1995). Local indicators of spatial Jelinski, D., & Wu, J. (1996). The modifiable
association-LISA. Geographical Analysis, 27, areal unit problem and implications for landscape
93–115. doi:10.1111/j.1538-4632.1995.tb00338.x ecology. Landscape Ecology, 11, 129–140.
doi:10.1007/BF02447512
Bailey, T. C., & Gatrell, A. C. (1995). Interactive
spatial data analysis. Essex, UK: Longman. Karimi, H., & Durcik, M. (2004). Evaluation of
uncertainties associated with geocoding tech-
Bonner, M., Daikwon, H., Nie, J., Rogerson,
niques. Computer-Aided Civil and Infrastructure
P., Vena, J., & Freudenheim, J. (2003). Po-
Engineering, 19, 170–185. doi:10.1111/j.1467-
sitional accuracy of geocoded addresses in
8667.2004.00346.x
epidemiologic research. Epidemiology (Cam-
bridge, Mass.), 14, 408–412. doi:10.1097/01. Kreiger, N., Waterman, P., Lemieux, K., Zierler,
EDE.0000073121.63254.c5 S., & Hogan, J. (2001). On the wrong side of
the tracts? Evaluating the accuracy of geocod-
Caliper (2006). Mapping software, GIS, and
ing in public health research. American Journal
transportation planning software. Newton, MA:
of Public Health, 91, 1114–1116. doi:10.2105/
TransCAD.
AJPH.91.7.1114
Cooke, W., Katarzyna, G., & Wallis, R. C. (2006).
Lawson, A. B., & Kleinman, K. (2005). Spa-
Avian GIS models signal human risk for West
tial and syndromic surveillance for public
Nile virus in Mississippi. International Journal
health. Chichester, UK: John Wiley & Sons.
of Health Geographics, 5(36).
doi:10.1002/0470092505
Craven, R. B., & Roehrig, J. T. (2001). West
Marra, P., Griffing, S., Caffrey, C., Kilpatrick,
Nile Virus. Journal of the American Medical
A., McLean, R., & Brand, C. (2004). West Nile
Association, 286(6), 651–653. doi:10.1001/
virus and wildlife. Bioscience, 54, 393–402.
jama.286.6.651
doi:10.1641/0006-3568(2004)054[0393:WNVA
Dodd, R. Y. (2003). Emerging infections, transfu- W]2.0.CO;2
sion safety, and epidemiology. The New England
Moran, P. A. P. (1950). Notes on continuous
Journal of Medicine, 349(13), 1205. doi:10.1056/
stochastic phenomena. Biometrika, 37, 17–23.
NEJMp038138
Openshaw, S. (1996). Using a geographical analy-
ESRI. (2007). ArcGIS 9.2 desktop help. Retrieved
sis machine to detect the presence of spatial clusters
from http://webhelp.esri.com/arcgisdesktop/9.2/
and the location of clusters in synthetic data. In
index.cfm?TopicName=welcome
Alexander, F. E., & Boyle, P. (Eds.), Methods for
Fotheringham, A. S., & Zhan, F. B. (1996). investigating localised clustering of disease (pp.
A comparison of three exploratory methods 68–87). Lyon, France: IARC Scientific.
for cluster detection in spatial point pat-
Openshaw, S., Charlton, M., Craft, A., & Birch,
terns. Geographical Analysis, 28(3), 200–218.
J. (1988). Investigation of leukaemia clusters by
doi:10.1111/j.1538-4632.1996.tb00931.x
use of a geographical analysis machine. Lan-
Goodchild, M. F. (2008). Assertion and authority: cet, 331(8580), 272–273. doi:10.1016/S0140-
The science of user-generated geographic content. 6736(88)90352-2
In Proceedings of the GeoInfo Colloquium for
Andrew U. Frank’s 60th Birthday.

1181
 Using Volunteered Geographic Information to Assess the Spatial Distribution of West Nile Virus

Openshaw, S., Charlton, M., Wymer, C., & Petersen, L. R., & Marfin, A. A. (2002). West
Craft, A. W. (1987). A mark I geographical Nile virus: A primer for the clinician. Annals of
analysis machine for the automated analysis of Internal Medicine, 137, 173–179.
point data sets. International Journal of Geo-
State of Michigan. (2006). Emerging dis-
graphical Information Science, 1(4), 335–358.
ease issues: West Nile virus. Retrieved from
doi:10.1080/02693798708927821
http://www.michigan.gov/emergingdiseas-
Ozdenerol, E., Bialkowska-Jelinska, E., & Taff, es/0,1607,7-186-25805---,00.html
G. N. (2008). Locating suitable habitats for West
Nile Virus-infected mosquitoes through associa-
tion of environmental characteristics with infected
mosquito locations: A case study in Shelby County,
Tennessee. International Journal of Health Geo-
graphics, 7(12), 200.

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 3,
edited by Donald Patrick Albert, pp. 72-85, copyright 2011 by IGI Publishing (an imprint of IGI Global).

1182
 1183

Chapter 71
GIS Application of Healthcare
Data for Advancing
Epidemiological Studies
Joseph M. Woodside
Cleveland State University, USA

Iftikhar U. Sikder
Cleveland State University, USA

ABSTRACT
Healthcare practices increasingly rely on advanced technologies to improve analysis capabilities for
decision making. In particular, spatial epidemiological approach to healthcare studies provides signifi-
cant insight in evaluating health intervention and decisions through Geographic Information Systems
(GIS) applications. This chapter illustrates a space-time cluster analysis using Kulldorff’s Scan Statistics
(1999), local indicators of spatial autocorrelation, and local G-statistics involving routine clinical
service data as part of a limited data set collected by a Northeast Ohio healthcare organization over a
period 1994 – 2006. The objective is to find excess space and space-time variations of lung cancer and
to identify potential monitoring and healthcare management capabilities. The results were compared
with earlier research (Tyczynski & Berkel, 2005); similarities were noted in patient demographics for
the targeted study area. The findings also provide evidence that diagnosis data collected as a result of
rendered health services can be used in detecting potential disease patterns and/or utilization patterns,
with the overall objective of improving health outcomes.

INTRODUCTION systems. In healthcare settings application of such

The increasing demand for health data analysis in
spatial and temporal scale has made emerging tech-
nologies such as Geographic Information Systems
(GIS) an essential tool for healthcare information
DOI: 10.4018/978-1-4666-2038-4.ch071
new technology are proving useful in the analysis
of health data and planning of healthcare services
(Pfeiffer, Robinson, Stevenson, Stevens, Rogers,
& Clements, 2008). The ability of GIS to manage
and retrieve georeference data has demonstrated
its value in the integration of complex epidemio-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 GIS Application of Healthcare Data for Advancing Epidemiological Studies
logical models through visualization of spatial and
temporal relationships. This has been recognized
by the World Health Organization (WHO):
Geographical information systems (GIS) provide
ideal platforms for the convergence of disease-
specific information and their analyses in relation
to population settlements, surrounding social and
health services and the natural environment. They
are highly suitable for analyzing epidemiological
data, revealing trends and interrelationships that
would be difficult to discover in tabular format.
Moreover GIS allows policy makers to easily
visualize problems in relation to existing health
and social services and the natural environment
and so more effectively target resources. (World
Health Organization, 2008)
Geographical analysis is not only important
for the identification of patterns of healthcare
outcomes it also offers insight into understanding
the association or linkage to political processes
and policy makers (Cromley, 2002; Gatrell, 2002).
Health data from managed health care organiza-
tions offers the opportunity to analyze unusual
geographical patterns of disease. Routine, aggre-
gated healthcare data stored in health systems can
be utilized to identify disease clusters or utilization
patterns. Recently methods have been sought to
further improve identification within case and
disease management programs.
The real world clinical service data stored
in healthcare information systems provides op-
portunity to analyze spatio-temporal patterns at
finer granularity. The investigation of space and
space-time epidemiological patterns often gives
rise to the explanation of factors that might cre-
ate an adverse health condition. This study uses
routine, aggregated service data to find excess
space and time variations in rendered services
where the primary diagnosis was lung cancer.
From the health care management point of view,
if clusters are detected and explanatory factors
linked, this understanding allows for better patient
1184
care, i.e. serving a particular population with tar-
geted specialists, and preventing spread of disease
amongst populations. This research aims to study
different clustering methods of the spatial and
spatio-temporal patterns of lung cancer particu-
larly for routine clinical service data collected by
a Northeast Ohio healthcare organization over a
period from 1994 – 2006.
GIS
Geographic information systems (GIS) integrate
computer applications and data for capturing, stor-
ing, querying, analyzing, viewing and modeling
geographic and spatial information for improved
decision making. GIS are distinguished from
other information systems, based on their ability
to utilize geographic data (ESRI, 2010; Chang,
2006). GIS originated in the 1960s and 1970s,
and were introduced to mainstream use during
the 1990s with the addition of a graphical user
interface, reduced software and hardware costs,
and available data. GIS have been used in a
wide spectrum of applications including: natural
resources planning, hazard management, crime
mapping, transportation, navigation, farming,
environmental monitoring, and epidemiology,
among others (Chang, 2006; Li, 2008).
The application architecture below displays
common components within a healthcare infor-
mation system for epidemiological studies: a web
server, GIS, and database. The web component
includes data entry, exchange, and transfer, along
with a centralized user interface for healthcare
information. The web server also permits viewing
of displayed data either on-premise or remotely,
enabling greater accessibility of information by
the end-user. The GIS maintains geographic and
spatial information, and also incorporates statisti-
cal and data analysis tools, such as a path analysis,
data querying, map data exploration, frequency
distribution, regression, chi-square analysis, and
multivariate analysis to analyze healthcare data
GIS Application of Healthcare Data for Advancing Epidemiological Studies
Figure 1. Application architecture
for decision making. The database contains in-
tegrated forms of healthcare data, and maintains
quality, timeliness, aggregation and other data
management items. Common inputs include raw
data, metadata or description of the data, and map
data (Chang, 2006; Li, 2008).
GIS Healthcare Applications
GIS along with geocoded data is a key objective
within healthcare, in order to improve costs, dis-
ease surveillance, and interventions. In Healthy
People 2010, the GIS goal for use in healthcare
systems is targeted at 90% from the current 45%
baseline (CDC, 2010). GIS applications have
been utilized in a variety of healthcare service
categories including epidemiology, pharmacode-
pidemiology, public health, and managed care,
with specific examples presented below.
In epidemiology GIS applications have in-
cluded disease state detection through clustering,
quality improvement, and cost containment. For
example in Amin et al. (2010), the authors utilize
GIS spatial and space-time clustering to detect
childhood cancer rates in Florida, discovering po-
tential environmental or risk factors in geographic
areas. In Yu et al. (2004), the authors utilize GIS
to identify quality improvement and cost factors
associated with common veteran population
conditions such as chronic heart failure (CHF),
colorectal cancer, diabetes, HIV/AIDS, ischemic
heart disease, psychiatric disorders, spinal-cord
injury (SCI), stroke, and substance abuse. The
GIS output assisted with policy development
with regard to resource allocation, quality con-
siderations, and cost identification. The CDC and
Huang et al. (2010) utilize GIS to identify obesity
prevalence in conjunction with other conditions
such as diabetes. The findings have identified
significant geographic variances at the state level
in both spatial and space-time components (Amin,
2010; CDC, 2010; Huang, 2009; Yu, 2004). In
pharmacoepidemiology, Brownstein et al. 2010
applied GIS to develop risk maps for opoid abuse,
in order to assist with health intervention planning.
Significant clusters were found, and could be
utilized to determine availability of prescriptions,
for public health agencies to use in planning, and
treatment surveillance (Brownstein et al., 2010).
Environmental disease monitoring is a key
component of healthcare monitoring. Conditions
such as cancer, birth defects, asthma, bird flu,
and respiratory problems have been linked to
environmental factors. Environmental data has
been collected for many decades by government
agencies, though is often contained in disparate
data locations. In Li et al. (2008) the authors pro-
pose a GIS integrated environmental monitoring
system, and utilize asthma condition as a GIS
mapping against air emissions. In Stelling et al.
(2010), outbreak detection is employed through
GIS surveillance. Statistical events were identified
and referred to health departments, for appropri-
ate response. These systems are able to monitor
in real-time to improve public health (Li et al.,
2008; Stelling et al., 2010).
1185
 GIS Application of Healthcare Data for Advancing Epidemiological Studies

Table 1. GIS healthcare applications

Healthcare Services Category Healthcare GIS Application Supporting Works

Epidemiology Health Care Cost Containment Yu et al. (2004)
Quality Improvement Amin et al. (2010)
Disease Detection Huang et al. (2010)
Pharmacoepidemiology Substance Abuse Brownstein et al. (2010)
Cluster Detection
Pharmaceutical Risk Management
Public Health Outbreak Detection Ling et al. (2008)
Environmental Monitoring Stelling et al. (2010)

GIS Application Example: Space-
Time Cluster Analysis
The question whether diseases such lung cancer
or breast cancer are spatially clustered is an ac-
tive research area (Lawson, Biggeri, Böhning,
Lesaffre, Viel, & Bertollini, 1999; Lawson, A., &
Denison, D., 2002; Marshall, R. J., 1991; Tango,
T., & Takahashi, K., 2005). Since the detection of
spatial and temporal patterns of clusters of lung
cancer is sensitive to the clustering algorithm, it is
difficult to evaluate results from a single method
(Jacquez, G. M., & Greiling, D. A., 2003). Cur-
rently, most of the comparative analysis of disease
clusters depends on simulated data (Ozonoff, A.,
Bonetti, M., Forsberg, L., & Pagano, M., 2005).
Tycznski developed a broad atlas of cancer in
Ohio which involved a “smoothing” method where
weighted averages of cancer per county were
calculated versus geographic location of patients
with cancer at the time of diagnosis. However, the
clusters are generated by considering only spatial
aspect. The temporal characteristics of the cluster
are not reported (Tyczynski, J. E., et al., 2005).
In recognition of the usual epidemiological
definition of cluster, this study adopts the formal
definition of cluster which refers to the patterns
of location of disease cases, relative to the pat-
tern of non-cases (Wakefield, J. C., Kelsall, J.
E., & Morris, S. E., 2000). In principle, since the
cases are more clumped than non-cases, the dif-
ference between the two patterns is statistically
recognizable. Intuitively, a cluster is an excess
value which exceeds the normal value for the
space and/or time. The closer a cluster popula-
tion is defined, the excess value will be greater
for the cluster population, and the significance
will be greater. The closer a cluster population is
defined, the greater the excess value will be for
the cluster population, and the significance will
be greater. Initial assessments of clusters include
reviews of cases, boundaries of space and time,
estimated number of cases, estimates of standard-
ized mortality ratios, statistical significance, and
public communication. Cluster analysis has been
frequently used to identify occurrence of morbid-
ity or unusual localized trends in disease patterns
(Alexander, F. F., 1992).
A considerable amount of research in tempo-
ral and spatial context in ‘scan’ statistics has been
invested in identifying disease clusters. The
theory has been successfully applied in a wide
variety of epidemiological studies for cluster
detection (Viel, J. F., Arveux, P., Baverel, J., &
Cahn, J. Y., 2000; Perez, A. M., Ward, M. P., Tor-
res, P., & Ritacco, V., 2002; Sankoh, O. A., Ye,
Y., Sauerborn, R., Muller, O., & Becher, H., 2001).
It has been identified as the most powerful
method for detecting local clusters (Kulldorff,
M., Tango, T., & Park, P. J., 2003; Song, C., &
Kulldorff, M., 2003). Ideally, this method is suit-
able where one needs to scan for clusters in space
that vary over time. Since the method for detect-
ing clusters is entirely unsupervised, there is no
need for a priori knowledge for the population
size. The method is based on the concept of ‘win-

1186
GIS Application of Healthcare Data for Advancing Epidemiological Studies
dows’ which are defined to contain a fixed popu-
lation (N*), and are centered on each area centroid.
The algorithm identifies a significant excess of
cases within a predefined moving window that
exhaustively searches all existing space-time
locations and keeps increasing size in space time
until it reaches a maximum limit. The maximum
number of cases M = maxYj (N * ) across the
j
windows is used as a test statistic. In this case j
represents the indexes of the areas defended in
N*. It is also possible to use a fixed number for
the population by introducing predefined con-
straints (e.g., the number population with a circle
should be less than a specified fraction of the
total population of the study area). Hjalmars et
al. (1996) and Kulldorff et al. (1997) applied a
similar method to detect childhood leukemia and
breast cancer incidence. The test statistic is based
on maximum likelihood ratio statistic across all
circles.
Yj Y −Yj
Y  Y −Y  +
  +
L = max  j  
j 
 I (Yj > E j )
j  E j  Y+ − E j 
(1)
Here, Yj and Ej represent the observed and
expected number of cases within the window j.
The indicator function I(.) becomes 1 when the
observed number exceeds the expected number
of cases within the window; otherwise the value
is 0. When the window with greatest exceedance
is encountered, the sampling distribution of like-
lihood ratio is determined using a Monte Carlo
test of cases across windows under a random
distribution assumption. Thus under the repeated
permutation, the distribution of likelihood statis-
tic, the null hypothesis is developed. The result
is significant at 0.05 levels if the likelihood ratio
is among the top 5% of all the values. It is also
possible to determine secondary cluster with a
lower significance level.
We applied the SaTScan method for a spatial
and space-time analysis for detecting local clusters
(Kulldorff, M., 1997). Due to temporal trends,
clusters may be generated for a ramp up or down
in data trends. For this reason, the space-time
permutation model automatically adjusts for these
temporal data trends. In the study we used case
data, with the spatial location represented by zip
code centroid latitude and longitude, and with
time represented by service month. The actual
number of cases in a cluster is compared with the
expected count if the spatial and temporal loca-
tions of all records were independent. A cluster
is determined to be present in a spatial location,
if, during the time period, there are excess cases
or recess within the surrounding areas. Using a
cylindrical window with a spatial base and time
as height, the space-time statistic is defined. The
window is moved in space and time, and a cylin-
der is created for each possibility. The algorithm
accounts for multiple testing by calculating the
maximum likelihood of occurrence for all pos-
sible cluster locations and sizes (Kulldorff, M.,
1997, 2001). In this study, retrospective analysis
was performed in terms of months, with periods
representing January 1994 through May 2006. In
each window, the alternative hypothesis concludes
that there is heightened risk.
The Poisson model is used for the space-time
permutation probability model as this allows for
covariate adjustments, in this case age and gender.
This likelihood function is maximized over all
windows, and the maximum likelihood window
describes the prevalent cluster. The test statistic
is calculated by generating a large random sample
from the data generated under the null hypothesis.
Monte Carlo testing is used to obtain the predicted
value. In this study, Monte Carlo replications were
generated to produce a P-value to 0.001. Covari-
ates were used since clustering can occur due to
covariates. Covariates are adjusted to prevent this
false clustering. The time precision was monthly
and ranged from January 1994 through May 2006
based on available data at time of collection. The
1187
 GIS Application of Healthcare Data for Advancing Epidemiological Studies
maximum spatial and temporal cluster size was
the default 50% of cases (Kulldorff, M., 2005).
The scant statistic generated both primary and
secondary clusters and the output was reported
in ASCII format, which contains a log likelihood
ratio and the significance level for the study area.
The output file was finally imported in standard
GIS environment of ArcGIS to visualize cluster
location for further spatial analysis. The Pois-
son model was also used for the purely spatial
probability model, and follows closely with the
space-time permutation model. However this
model utilizes a population file which includes
information regarding the at risk population, and
was taken from the 1999 US Census Bureau Zip
Code file, with regard to total 1999 population for
each zip code. As this file did not include additional
population attributes such as age and gender, the
case file along with the population file excluded
these for this model (Kulldorff, M., 1997; U.S.
Census Bureau, 1999).
Spatial autocorrelation analysis provides both
global and local clusters which can be detected
by Moran’s I statistic. The global pattern can be
detected from Moran’s Scatterplot, where the
slope of the regression line represents Moran’s
I. We applied the local indicators of spatial au-
tocorrelation (LISA) method using sample cases
in each zip code. The local Moran statistic for
location i is given as follows (Wong, D. W. S., &
Lee, J., 1999):
I i = z i ∑ wij z j (2)
j  j
Where zi, zj are the deviations from the mean
for associated x, and where zi is the z-score of
xi. A high Moran’s I indicates associated values,
whereas a low value indicates non-associated
values. The row-standardized matrix, used to
estimate weights for each of the unit’s neighbors,
is defined as wij (Wong, D. W. S., et al., 1999).
When the study involves the measurement of
1188
Moran’s I for rates, the underlying assumption
of stationarity may be violated due to intrinsic
variance instability of the rates. Since the popula-
tion at risk in the study area varies significantly
across patient zip codes, variance instability may
lead to spurious inference for Moran’s I. To ac-
count for this effect, the Empirical Bayes (EB)
standardization was performed (Assuncao, R.,
& Reis, E. A.) using zip code population as the
base variable. The standardized rate was used to
calculate the univariate LISA. The spatial autocor-
relation analysis utilized local tests for Moran’s I
statistic, with significance maps generated to the
P-value of 0.05 (Anselin, L., Syabri, I., & Kho,
Y., 2004). Sensitivity analysis was done by chang-
ing the number of permutations (9999 times) for
different significance cutoff values.
While Moran’s I is effective in identifying
presences of clustering of similar values cluster-
ing, it cannot differentiate between high and low
values. Another spatial autocorrelation statistic,
the general G-statistic (Getis, A., Ord., J.K.,
1992), is able to detect hot and cold spots. The
G-Statistic also uses cross-product statistics to
measure spatial association, similar to Moran’s
I. The local G-statistic is the local version of the
general G-statistic, and it indicates how the value
of each unit is associated with surrounding units
within distance d (Wong, D. W. S., et al., 1999).
 ∑ w (d )x 
 ij j 

Gi (d ) =  j  j ≠i (3)
 ∑ x j 

Here, wij(d) is the weight, with distance d,
and the weight is 1 if j is within d of i; otherwise
weight is 0. The cross product of the points i and
j are represented by xixj. A high Gi(d) indicates a
spatial association of similar high values; a low
Gi(d) indicates low and below-average values.
A z-score near 0 indicates no spatial pattern; a
highly negative z-score indicates low values; and
GIS Application of Healthcare Data for Advancing Epidemiological Studies
a highly positive z-score indicates high values
(Wong, D. W. S., et al., 1999). We applied the
local G-statistic using samples cases in each zip
code. The standardized Empirical Bayes rate was
used to estimate the G-statistics.
Data Compilation and Analysis
The study included the area of Northeast Ohio.
The service data was collected from northeast
ohio health plan with appropriate approval from
the Internal Review Board. It includes the routine
clinical service data of over the service period of
January 1994 through May 2006. The case file
contains information about the cases where each
record represented an individual service (Kull-
dorff, M., 2005). The dataset was anonymised
to avoid disclosure of individual information.
In addition, a limited number of attributes was
authorized for addressing the specific research
objective in finding excess space-time variations
of lung cancer. Among the attributes available in
the case file are primary diagnosis, patient zip
code, service month, age, and gender. Service
month was the month of service/diagnosis where
data was collected and available for use beginning
in January 1994 through May 2006; the attribute
age represents the age at date of service. In cases
where multiple services/diagnoses existed, only
the first occurrence for each individual was used
to avoid cluster creation through repeat services.
The dataset was generated during 07/2006 and
included all services rendered or received to date.
For services rendered at external locations, several
months may have elapsed before service infor-
mation was received, and may be excluded from
this study for this reason. It would be possible to
periodically re-run past results to ensure the latest
dataset available. Due to system memory require-
ments (32-bit Windows memory allocation size),
and software limitations, individual diagnoses
were required to be selected and scanned (Kull-
dorff, M., 2005). International Classification of
Diseases, (Ninth Revision) Clinical Modification
(ICD-9-CM) Code 162.XX was used to represent
the diagnosis for lung cancer. The aggregation
unit of geographic location used is the five-digit
US zip code. The coordinates were then mapped
in ESRI’s ArcGIS 9.0. Geocoding was performed
through assignment of coordinates (latitude and
longitude) to each patient zip code. The coordinate
file contains the geographic coordinates for each
location id specified in the case file. Coordinates
were specified using latitude and longitude of
each 5-digit patient zip code entered in decimal
degrees, and, where identical, one or more coor-
dinates were combined for a single location. Each
patient zip code was geocoded to the centroid
using a US census 1999 zip code file containing
the latitude and longitude. For graphical analysis
and to reflect the majority of sample records, a
fourteen-county region with contiguous zip codes
was used to represent the Northeast Ohio region.
The dataset used for the analysis included a total
of 2,364 records or unique initial service claims
for patients having included the ICD-9-CM Code
162.XX (lung cancer) as the primary diagnosis,
and 152 unique patient zip codes.
To account for the inherent variance instability
of rates (Bailey, T. C., & Gatrell, A. C., 1995) of
lung cancer incidence, empirical Bayes smoothing
was performed (Clayton, D., & Kaldor, J., 1987),
whereby the raw rates were adjusted towards the
overall average of the study area. The technique
consists of computing the weighted average
between the raw rates for each zip code and the
study area average with weights proportional to
the underlying population at risk. In other words,
small zip codes (i.e., with a small population at
risk) will tend to have their rates adjusted consid-
erably, whereas for larger zip codes, the rates will
barely change (Clayton, D., & Kaldor, J., 2005).
The empirical Bayes (EB) smoothed box map in
Figure 2 shows that 5 zip codes are in the upper
outlier and as many as 34 zip codes are within
75 percentiles.
1189
 GIS Application of Healthcare Data for Advancing Epidemiological Studies
Figure 2. Empirical Bayes (EB) smoothed box map of lung cancer in Northeast Ohio during 1994-2006
RESEARCH RESULT
The spatial scan statistic result shows that there
exists a cluster of lung cancer in the Northeast
Ohio region. Areas of excess were detected with
statistical significance using a spatial Poisson
probability model for services with a diagnosis of
lung cancer. The results show that a statistically
significant cluster exists with a relative risk (RR)
of 4.164 at P-value of 0.001 that includes 72 zip
codes. The results also listed another statistically
significant cluster with RR of 0.186 and P-value of
0.001, and contained 28 zip codes within the con-
tiguous sample area. As expected, the significance
levels match closely with the geographic sample
distribution. The space-time scan statistic result
also shows that there exists a cluster of lung cancer
in the region for the space-time scan statistic. With
spatial-temporal data and covariates considered,
statistically significant areas of excess were also
detected using the space-time scan statistic for
services with a diagnosis of lung cancer. Clusters
were scanned first only for location of patient zip
code, and then scanned with the addition of an
attribute (age, gender). Furthermore, clusters were
scanned using five-year age brackets. The results
show that a statistically significant cluster (P-value
1190
<= 0.01) exists. The primary cluster detected is a
geographically contiguous area in Northeastern
Ohio, with a relative risk (RR) of 1.784 and sig-
nificant P-value of 0.001. The cluster contained 44
zip codes within the contiguous sample area and
spanned a time period of 7/1/1999 – 8/31/2002.
The location of the cluster changes when includ-
ing time, as compared with the purely spatial
model. The primary cluster shows no geographic
change when adjusted for attributes. It was found
that adjustment of attributes is not required as the
cluster location does not change when attributes
are introduced Secondary clusters were identified,
but were not statistically significant.
The LISA analysis indicates a cluster of high
incidence of lung in the region with statistically
significant clusters (P-value <= 0.05), containing
64 zip codes within the study area. The local G-
statistic result shows that there exists a hot spot of
lung cancer in the region. The results also show
that statistically significant clusters (P-value <=
0.05) consist of 89 zip codes. Figure 3 shows the
spatial statistic significance graphs for the SatScan,
LISA, and local G-statistic. The figure also shows
the space-time scan statistic significance graphs.
Table 2 shows the significance level measured
by the P-value at the 0.001, 0.01, and 0.05 levels.
GIS Application of Healthcare Data for Advancing Epidemiological Studies
The corresponding percentage area, (where the
total area of significance was divided by the total
area for which sample cases existed), the number
of sample cases, number of zip codes contained,
average age, and county by gender are included,
with each method listed as a column.
The scan statistic result shows that there exists
a cluster of lung cancer in the region for the
purely spatial statistic. Areas of excess were de-
tected with statistical significance using a purely
spatial Poisson probability model for services
with a diagnosis of lung cancer. The results show
that a statistically significant high value cluster
(P-value <= 0.05) exists and contains 72 zip codes.
The results also listed another statistically sig-
nificant low value cluster containing 28 zip codes
within the contiguous sample area. The results
list 54 zip codes as not significant. The scan sta-
tistic result also shows that there exists a cluster
of lung cancer in the region for the space-time
scan statistic. The results show that a statistically
significant high value cluster (P-value <= 0.01)
exists. The cluster contained 52 zip codes and
spanned a time period of 7/1/1999 – 8/31/2002.
Two of the zip codes contained similar coordinates,
which left 50 unique zip code locations. No low
value clusters were identified. The LISA result
shows statistically significant cluster of lung
cancer in the region for the purely spatial statistic.
The results show that statistically significant
clusters (P-value <= 0.05) exist, and contained
87 not significant, 31 low-low, 2 low-high, 4
high-low, and 28 high-high value zip codes.
The local G-statistic result shows that there
exists a cluster of lung cancer in the region for
the purely spatial statistic. The results show that
statistically significant clusters (P-value <= 0.05)
exist, and contain 63 not significant, 31 low-low,
0 low-high, 16 high-low, and 42 high-high value
zip codes. Figure 3 shows the purely spatial sta-
tistic cluster graphs for the scan statistic, LISA,
and local G-statistic. It also shows the space-time
statistic significance graphs for the scan statistic.
DISCUSSION
This study detected several high and low asso-
ciation clusters using the spatial and space-time
methods. When comparing the various methods
of the scan statistic, LISA, and local G-statistic,
spatial association areas are somewhat similar
geographically. The local G-statistic returned the
highest number of records (2002) with regard to
significance at the 0.05 level. The scan statistic
method also returned a similarly high number of
associated records (2000) as the local G-statistic.
The LISA method returned a comparatively
smaller number of significant records (1242);
however when comparing LISA geographically
to the scan statistic and local G-statistic, LISA
appears to exclude bordering zip codes included
in the other methods. Conversely, LISA had the
highest number of records (1122) that were not
significant, followed by the scan statistic (364)
and local G-statistic (362). The scan statistic
contained the highest number of high-high cluster
records (1896), followed by the local G-statistic
(1377) and LISA (1069). The local G-statistic had
the highest number of high-low spatial outlier
records (419), followed by LISA (31) and scan
statistic (N/A). Only LISA contained low-high
spatial outlier records (31). The local G-statistic
contained the highest number of low-low records
(206), followed by scan statistic (104), which the
output indicated low only, and LISA (100).
When accounting for temporal trends using the
scan statistic, this study detected only one area
of excess lung cancer spanning the three-year pe-
riod of the study using a space-time method. The
SatScan space-time statistic returned the fewest
number of records (103) and geographic location
in terms of significance and clustering, and it was
different geographically from the purely spatial
methods in that the majority of high value asso-
ciation cases and area was not contained within
Cuyahoga County, but rather in Stark and Summit
Counties. In addition, no low value association
clusters were detected, as with the purely spatial
1191
 GIS Application of Healthcare Data for Advancing Epidemiological Studies

Figure 3. The area of statistically significant (p<=0.05) lung cancer using a spatial autocorrelation and
scan statistic for space and space-time model

1192
GIS Application of Healthcare Data for Advancing Epidemiological Studies

Table 2. P-value distribution of area, sample cases, zip codes, average age and gender attributes of the
cluster(s) for each method

P-Value SpaceTime Scan Spatial Scan G-statistic LISA

Statistic Statistic
0.05 – 1
% Area 74.30% 43.00% 63.20% 63.00%
Cases 2261 364 362 1122
Zip # 108 54 63 88
Avg Age 67.7 66.4 65.7 67.1
Male 1276 218 207 637
Female 985 146 155 485
0.01 - 0.05
% Area - - 14.90% 25.20%
Cases - - 252 640
Zip # - - 22 37
Avg Age - - 69.6 67
Male - - 138 346
Female - - 114 294
0.001 – 0.01
% Area - - 8.40% 8.70%
Cases - - 195 483
Zip # - - 17 21
Avg Age - - 66.7 69.2
Male - - 100 281
Female - 95 202
0 - 0.001
% Area 25.70% 57.00% 13.50% 3.10%
Cases 103 2000 1555 119
Zip # 44 98 50 6
Avg Age 65.6 67.8 67.9 69.2
Male 55 1113 886 67
Female 48 887 669 52

methods. The introduction of time allows for a
more focused area and localized study region.
Since the health data from managed health care
organizations are aggregated at the zip code level,
the smallest mapping unit discernible in a GIS
map is the zip code polygon of the study area.
Compared to long temporal range (1994-2006)
of the study, the temporal occurrence of cluster
was found significant only within the three years
of 1999-2002. The narrow temporal concentra-
tion and high relative risk suggests that further
investigation is required to understand the causal
factor within the time period.
The application of scan statistic is specifically
useful when there is a plausibility of a single
hotspot. Since the likelihood ratio is derived
from a single hot/cold spot, the possibility of
detecting multiple hot/cold is ruled out in the

1193
 GIS Application of Healthcare Data for Advancing Epidemiological Studies
scan statistic algorithm. Moreover, scan statistic
detects space-time clusters by using cylindrical
windows; hence the result needs to be interpreted
with respect to the choice a specific shape of the
window. While the cluster is detected by rejecting
the null hypothesis, the circular spatial window
tends to detect a larger cluster than the true cluster
by absorbing surrounding regions where there
is no elevated risk (Tango, T., & Takahashi, K.,
2005). The relatively large cluster area detected
in this study could be interpreted as the result of
the boundary effect of the circular window. The
effect could be confirmed by comparing the result
with a flexibly shaped scanning window which
allows for irregular shapes (Tango, T., et al., 2005).
Moreover, the results need to be interpreted by
comparison with other spatial cluster algorithms
(e.g., spatial filtering [34], generalized additive
models (GAM) (Ozdenerol, E., Williams, B., &
Kang, S. Y., & Magsumbol, M., 2005), and Bayes-
ian disease mapping (BYM)) (Besag, J. E., York,
J. C., & Mollie, A., 1991). When utilizing spatial
autocorrelation (LISA), the number of significant
records is reduced due to the generation of larger
P-values. The spatial autocorrelation of simulated
rates creates large variances in LISA values,
leading to insignificant P-values, as neighboring
values are more likely to be jointly low or high
(Goovaerts, P., & Jacquez G. M., 2004). The lo-
cal G-statistic produces the highest number of
significant records, but also produces the highest
number of spatial outlier records, which, when
accounted for, produce a more similar, but still
higher records count than LISA. The local G-
statistic uses a z-score scale around 0 to indicate
clustering situations, which leads to a classification
for all non-zero significant records.
In comparison with Tycznski’s study (2005),
this study reviewed spatio-temporal clusters of
lung cancer cases to identify areas of interest,
whereas the former study considered overall
mortality rates of all cancer types within Ohio to
determine time trends. Several key areas can be
synthesized between the two studies to provide
1194
further topics of interest. Tycznski notes that
lung cancer was the leading cause of death across
categories; this fact coincides with the focused
interest of this study on the leading cause of cancer
death. Tycznski also notes higher mortality rates in
Blacks as compared with Whites, and recommends
focused efforts on this demographic. In this study
a spatial cluster is found surrounding Cuyahoga
County, which, when compared with surrounding
Northeast Ohio counties, has a greater percent-
age of the Black population. When incorporating
temporal trends into the study, an additional area
of interest is identified within Northeast Ohio.
PRACTITIONER IMPLICATIONS
Practitioner-based usage of this study falls into
two general categories: identification and preven-
tion. Detection of lung cancer, or other chronic
diseases identified through spatial and temporal
clusters can discover population areas of interest,
can gauge the effectiveness of disease reduction
methods within a given population area, and is
considered an important public health measure.
This study compares the results of three different
methods to gain greater insight and better reflect
spatial and temporal variations, as related to
identification and/or monitoring current programs.
Increasing disease rates may trigger increased
focus or re-direction of current efforts. Decreas-
ing rates may be utilized as validation for current
efforts, improvement value, or as a component of
a shared knowledge base for other efforts. Early
detection programs along with case and disease
management programs can also utilize geographic
modeling information to determine the overall pro-
gram design and success factor achievement. The
program design may vary by geographic location,
or may be used to identify the optimal program
methodology for the particular locality (2005).
As an important step in enhanced treatment,
efforts have been made to materialize best practices
of disease therapies. Identification is made through
GIS Application of Healthcare Data for Advancing Epidemiological Studies
geographical variances, and can trigger alerts to
dedicated specialists for discovery of previously
unknown opportunities. Activities following the
discovery of identified areas may include patient/
provider education, therapy identification, local-
ity factors, and identification for early treatment,
among others. This type of geographical evalua-
tion has occurred with early-stage breast cancer.
Geographic information systems and associated
analysis can instruct physicians, patients, and other
medical personnel in resource distribution, costs,
health outcomes, and patient satisfaction. The
location, service provider, and services performed
can be analyzed by geographic area to identify
effective methods. Once identified, these methods
can be standardized to improve consistency across
many locations, in order to gain effectiveness.
Further granularity can be provided to allow for
reduced area identification, in which specific
disease control strategies can be implemented to
supplement broader programs (Gregorio, D. I.,
Kulldorff, M., Barry, L., Samocuik, H., & Zarfos,
K., 2001).
Other studies have sought to determine vari-
ances between local and global populations. Find-
ings have suggested that differences in survival
rates are not a result of biological factors, but
rather treatment and prevention factors. These
include disease stage at diagnosis, effects of
compound diseases, and general treatment fac-
tors. Temporal trends are required in addition to
spatial trends to locate areas whose intervention
would prove most beneficial. Prior studies have
examined geographic locations to determine
physician and medical shortages. The findings
suggest that while the existing supply is in fact
adequate, the distribution is not optimal. This cre-
ates access restrictions across previously unseen
urban, suburban, and rural areas. Literature sug-
gests that the physician specialty combination as
well as distribution of non-physician clinicians is
vital for health outcomes. Accessibility measures
require accuracy to appropriately capture the flows
within service areas. Managed care plans have
the ability to improve access and utilization of
preventative care. These managed care plans gen-
erally provide more preventative services which
result in improved outcomes over fee or service
based plans (Mobley, L. R., Root, E., Anselin,
L., Lozano-Gracia, N., & Koschinsky, J., 2006).
Whether a singular method or rather a com-
bination of factors prove effective when adapted
to the patient set, it is imperative to provide tools
to successfully identify patients and program out-
comes through monitoring, and enact additional
programs to prevent future diseases and complica-
tions. This contributes to the overall management
goal of improving quality of care, while reducing
overall costs, thus ensuring high-quality, afford-
able healthcare.
CONCLUSION AND
FUTURE DIRECTIONS
The spatial scan statistic, LISA, and local G-sta-
tistic were able to detect several areas of high and
low value clusters. The space-time scan statistic
identified one area of excess lung cancer spanning
the three-year period of the study and produced a
different result than a purely spatial scan statistic.
The cluster remained unchanged and statistically
significant even after covariate adjustment. The
findings also provide evidence that diagnosis data
collected as a result of rendered health services can
be used in detecting potential disease patterns and/
or utilization patterns throughout space and time.
The methods shown are also useful for identify-
ing which areas have the highest occurrence of a
particular disease for the placement of facilities or
specialists. Time can also be a factor for identify-
ing more specific areas of interest, given a large
geographic region. The possible boundary effect
of circular window and space time-interaction
resulting from geographical population shifts
in the study area needs to be investigated. The
popularly used Knox method (1964) and other
tests for space-time interaction are required to
1195
 GIS Application of Healthcare Data for Advancing Epidemiological Studies
confirm if the space-time clustering of lung
cancer in the study area is a result of underlying
geographical population distribution (Kulldorff,
M., 1999). Unbiased space-time interaction tests
and prospective space-time permutation analysis
may also be performed as additional data is made
available (Kulldorff, M., 2005; Hjalmars, U., &
Kulldorff, M., 1999).
As managed care programs are initiated in ar-
eas of interest, additional localized studies can be
utilized to track improvements in overall mortality
rates as a result of the initiatives. By providing
granularity to the zip code level in this study,
focused geographic programs can be developed
at strategic facility sites and can be modified
based on ongoing spatio-temporal surveillance
of the population. As health care entities seek
to improve healthcare outcomes, these methods
provide a useful tool in identifying geographic
areas of interest.
While the research objective is not to identify
a specific cause-and-effect relationship between
lung cancer and environmental conditions such as
air pollution or individual smoking habit, further
research is needed to establish a space-time spe-
cific causal relationship and a latency period of
cancer. Given the scope and objective of this study,
it was not deemed necessary to collect additional
detailed case data. For future study, additional
attributes which are not available for the current
dataset should be generated from a localized study
that would include things such as ethnicity, socio-
economic status of individual cases, exposure to
tobacco smoke, and population migration. These
attributes can be adjusted for in the model in order
to determine effects. Localized studies can also
determine utilization of services in which lung
cancer diagnosis appears. This identification can
be useful in managing the care for those with an
identified diagnosis and determining frequency
of services for those with an identified diagnosis.
1196
REFERENCES
Alexander, F. F. (1992). Methods for the assess-
ment of disease clusters. In Geographical & envi-
ronmental epidemiology: Methods for small-area
studies (pp. 238-247).
Amin, R. (2010). Epidemiologic mapping of
Florida childhood cancer clusters. Pediatric Blood
& Cancer, 54, 511–518.
Anselin, L. (2005). Exploring spatial data with
GeoDa: A workbook. Center for Spatially Inte-
grated Social Science.
Anselin, L., Syabri, I., & Kho, Y. (2004). GeoDA:
An introduction to spatial data analysis.
Assuncao, R., & Reis, E. A. (n.d.). A new pro-
posal to adjust Moran’s I for population density.
Statistics in Medicine, 18, 2147-2161.
Bailey, T. C., & Gatrell, A. C. (1995). Interactive
spatial data analysis. New York, NY: John Wiley
and Sons.
Besag, J. E., York, J. C., & Mollie, A. (1991).
Bayesian image restoration with two applications
in spatial statistics. Annals of the Institute of
Statistical Mathematics, 43, 1–59. doi:10.1007/
BF00116466
Brownstein, J., Green, T., Cassidy, T., & Butler,
S. (2010). Geographic Information Systems and
pharmacoepidemiology: Using spatial cluster
detection to monitor local patterns of prescription
opioid abuse. Pharmacoepidemiology and Drug
Safety, 19, 627–637. doi:10.1002/pds.1939
CDC. (2010). U.S. obesity trends.
CDC, HRSA. (2010). Healthy people 2010.
Chang, K. (2006). Introduction to Geogrpahic In-
formation Systems. New York, NY: McGraw-Hill.
GIS Application of Healthcare Data for Advancing Epidemiological Studies
Clayton, D., & Kaldor, J. (1987). Empirical Bayes
estimates of age standardized relative risks for use
in disease mapping. Biometrics, 43, 671–681.
doi:10.2307/2532003
Cromley, E. K. (2003). GIS and disease. Annual
Review of Public Health, 24, 7–24. doi:10.1146/
annurev.publhealth.24.012902.141019
ESRI. (2010). ArcGIS.
Gatrell, A. (2002). Geographies of health. Malden,
MA: Blackwell.
Getis, A., & Ord, J. K. (1992). The analysis
of spatial association by use of distance sta-
tistics. Geographical Analysis, 24, 189–207.
doi:10.1111/j.1538-4632.1992.tb00261.x
Goovaerts, P., & Jacquez, G. M. (2004). Account-
ing for regional background and population size in
the detection of spatial clusters and outliers using
geostatistical filtering and spatial neutral models:
The case of lung cancer in Long Island, New York.
International Journal of Health Geographics,
3(1), 14. doi:10.1186/1476-072X-3-14
Gregorio, D. I., Kulldorff, M., Barry, L., Samocuik,
H., & Zarfos, K. (2001). Geographical differ-
ences in primary therapy for early-stage breast
cancer. Annals of Surgical Oncology, 8, 844–849.
doi:10.1007/s10434-001-0844-4
Hastite, T. J., & Tibshirani, R. J. (1991). General-
ized additive models. New York, NY: Chapman
and Hall.
Hjalmars, U., Kulldorff, M., Gustafsson, G.,
& Nagarwalla, N. (1996). Childhood leu-
kaemia in Sweden: Using GIS and a spatial
scan statistic for cluster detection. Statis-
tics in Medicine, 15, 707–715. doi:10.1002/
(SICI)1097-0258(19960415)15:7/9<707::AID-
SIM242>3.0.CO;2-4
Huang, L., Tiwari, R., Pickle, L., & Zou, Z. (2009).
Covariate adjusted weighted normal spatial scan
statistics with applications to study geographic
clustering of obesity and lung cancer mortality
in the United States. Statistics in Medicine, 29,
2410–2422.
Jacquez, G. M., & Greiling, D. A. (2003). Local
clustering in breast, lung and colorectal cancer in
Long Island, New York. International Journal of
Health Geographics, 2.
Knox, G. (1964). The detection of space-time
interactions. Applied Statistics, 13, 25–29.
doi:10.2307/2985220
Kulldorff, M. (1997). A special scan statistic. Com-
munications in Statistics Theory and Methods, 26,
1481–1496. doi:10.1080/03610929708831995
Kulldorff, M. (1999, June). The Knox method
and other tests for space-time interaction.
Biometrics, 55, 544–552. doi:10.1111/j.0006-
341X.1999.00544.x
Kulldorff, M. (2001). Prospective time periodic
geographical disease surveillance using a scan
statistic. Journal of the Royal Statistical Society.
Series A, (Statistics in Society), 164(1), 61–72.
doi:10.1111/1467-985X.00186
Kulldorff, M. (2005). SaTScanTM v6.0: Software
for the spatial and space-time scan statistics.
Information Management Services, Inc.
Kulldorff, M., Tango, T., & Park, P. J. (2003).
Power comparisons for disease clustering tests.
Computational Statistics & Data Analysis, 42.
Lawson, A., Biggeri, A., Böhning, D., Lesaffre,
E., Viel, J. F., & Bertollini, R. (1999). Disease
mapping and risk assessment for public health. In
Lawson, A., Biggeri, A., Böhning, D., Lesaffre,
E., Viel, J. F., & Bertollini, R. (Eds.), Disease
mapping and risk assessment for public health.
London, UK: John Wiley & Sons.
1197
 GIS Application of Healthcare Data for Advancing Epidemiological Studies
Lawson, A., & Denison, D. (2002). Spatial
cluster modelling. Boca Raton, FL: CRC.
doi:10.1201/9781420035414
Li, L., Xu, L., Jeng, H., Naik, D., Allen, T., &
Frontini, M. (2008). Creation of environmental
health Information System for public health ser-
vice: A pilot study. Information Systems Frontiers,
10, 531–542. doi:10.1007/s10796-008-9108-1
Marshall, R. J. (1991). A review of the statistical
analysis of spatial patterns of disease. Journal of
the Royal Statistical Society. Series A, (Statistics
in Society), 154, 421–441. doi:10.2307/2983152
Mobley, L. R., Root, E., Anselin, L., Lozano-
Gracia, N., & Koschinsky, J. (2006). Spatial
analysis of elderly access to primary care services.
International Journal of Health Geographics, 5.
Ozdenerol, E., Williams, B., Kang, S. Y., &
Magsumbol, M. (2005). Comparison of spatial
scan statistic and spatial filtering in estimating
low birth weight clusters. International Journal
of Health Geographics, 4(19).
Ozonoff, A., Bonetti, M., Forsberg, L., & Pagano,
M. (2005). Power comparisons for an improved
disease clustering test. Computational Statistics
& Data Analysis, 48, 679–684. doi:10.1016/j.
csda.2004.03.012
Perez, A. M., Ward, M. P., Torres, P., & Ritacco,
V. (2002). Use of spatial statistics and monitoring
data to identify clustering of bovine tuberculosis
in Argentina. Preventive Veterinary Medicine,
56, 63–74. doi:10.1016/S0167-5877(02)00124-1
Pfeiffer, D. U., Robinson, T. P., Stevenson, M.,
Stevens, K. B., Rogers, D. J., & Clements, A. C. A.
(2008). Spatial analysis in epidemiology. NewYork,
NY: Oxford University Press Inc. doi:10.1093/ac
prof:oso/9780198509882.001.0001
1198
Sankoh, O. A., Ye, Y., Sauerborn, R., Muller, O., &
Becher, H. (2001). Clustering of childhood mortal-
ity in rural Burkina Faso. International Journal
of Epidemiology, 30, 485–492. doi:10.1093/
ije/30.3.485
Song, C., & Kulldorff, M. (2003). Power
evaluation of disease clustering tests. Interna-
tional Journal of Health Geographics, 2, 1–8.
doi:10.1186/1476-072X-2-9
Stelling, J. (2010). Automated use of WHONET
and SaTScan to detect outbreaks of Shigella spp.
using antimicrobial resistance phenotypes. Epide-
miology and Infection, 138, 873–883. doi:10.1017/
S0950268809990884
Tango, T., & Takahashi, K. (2005). A flexibly
shaped spatial scan statistic for detecting clusters.
International Journal of Health Geographics, 4,
1–15. doi:10.1186/1476-072X-4-11
Tyczynski, J. E., & Berkel, H. J. (2005). Long-
term all-sites cancer mortality time trends in Ohio,
USA, 1970–2001: Differences by race, gender
and age. BMC Cancer, 5.
U.S. Census Bureau. (1999). U.S. Postal Service
ZIP codes.
Viel, J. F., Arveux, P., Baverel, J., & Cahn, J. Y.
(2000). Soft-tissue sarcoma and non-Hodgkin’s
lymphoma clusters and a municipal solid waste
incinerator with high dioxin emission levels.
American Journal of Epidemiology, 152, 13–19.
doi:10.1093/aje/152.1.13
Wakefield, J. C., Kelsall, J. E., & Morris, S. E.
(2000). Clustering, cluster detection, and spatial
variation in risk. In Elliot, P., Wakefield, J. C.,
Best, N. G., & Briggs, D. J. (Eds.), Spatial epi-
demiology: Methods and applications. Oxford,
UK: Oxford University Press.
WHO. (2008). Public health mapping. WHO.
GIS Application of Healthcare Data for Advancing Epidemiological Studies

Wong, D. W. S., & Lee, J. (1999). Statistical Yu, W. (2004). Using GIS to profile healthcare
analysis of geographic information. Hoboken, costs of VA quality-enhancement research initia-
NJ: Wiley. tive diseases. Journal of Medical Systems, 28,
3. doi:10.1023/B:JOMS.0000032844.26308.95

This work was previously published in New Technologies for Advancing Healthcare and Clinical Practices, edited by Joseph
Tan, pp. 362-377, copyright 2011 by Medical Information Science Reference (an imprint of IGI Global).

1199
1200

Chapter 72
Colorado 14ers, Pixel by Pixel
Brandon J. Vogt
University of Colorado at Colorado Springs, USA

ABSTRACT
This document describes a capstone learning exercise designed for undergraduates enrolled in an intro-
ductory geospatial tools course. The overarching theme of the exercise, Colorado 14ers, Pixel by Pixel,
is mountain geography. While immersed in a digital mountainous landscape, students explore topics
of geomorphology and geomorphometry and discover how it is possible for a summit that rises well
above 14,000 feet not to qualify as a true Colorado ‘14er’. To address these topics, students examine
freely-available digital elevation models (DEMs) using common geospatial analysis tools. The exercise
was developed with five purposes in mind: (1) students should be introduced gently to geospatial soft-
ware; (2) students should enter upper-level geography courses armed with basic computer skills and
a diverse geospatial toolkit; (3) students should internalize the fact that invariably digital geographic
explorations are influenced by scale; (4) students of geography should be able to identify natural and
anthropogenic impacts to a landscape, and; (5) to help students conceptualize what geographers do,
learning exercises should include local and interesting case studies. As an exercise intended to synthesize
geographic concepts covered throughout a course, the learning objectives range from grasping broad
theoretical concepts related to scale and measurement to learning specific computer skills related to
directory structures and file naming conventions. Deliverables of the exercise include posting four maps
and answers to nine questions to a website. The outcome of the exercise is a confident, geospatial tool-
savvy student who is eager to further investigate the field of geography.

INTRODUCTION1 ronmental Studies (GES) geospatial tools course.

Colorado 14ers, Pixel by Pixel is a stand-alone,
computer-based exercise that concludes Digital
Earth, an undergraduate University of Colorado at
Colorado Springs (UCCS) Geography and Envi-
DOI: 10.4018/978-1-4666-2038-4.ch072
Digital Earth introduces geographic information
systems (GIS), global positioning systems (GPS),
geovisualization, remote sensing, geomorphom-
etry, virtual globes, and web design. The course
seeks to prepare students for the digital worlds they
will encounter in upper-level geography courses

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Colorado 14ers, Pixel by Pixel
and exposes students, in exciting and interactive
ways, to the breadth and depth of geography. The
course, which is required for all GES majors,
focuses on eight computer-based exercises (labs).
As examples, in one lab, students use the virtual
globe ‘SkylineGlobe’ to compare and contrast
fashion by ‘surfing’ crosswalk cameras in large
cities across the globe. In another lab, students
manipulate 3D models in a GIS to investigate
controls of geology on settlement patterns in
and around Durango, Colorado. Asking students
to solve imaginative problems using interactive
and exploratory approaches inspires interest in
geospatial tools and their practical applications.
Digital Earth has been well received by students
at UCCS as measured by higher than expected
enrollment numbers and strong course evaluations.
Colorado 14ers, pixel by pixel is the capstone
exercise of Digital Earth and as such represents
a summary of many of the concepts and tools
encountered during the semester. Set in a virtual
mountainous Colorado landscape, the exercise
requires students to explore issues related to physi-
cal geography, map scale, spatial resolution, and
digital measurement. The exercise asks students to
demonstrate through various mappings how it is
possible that one peak in a group of three 14,000
foot + peaks does not qualify as a true 14er. A
‘14er’ is one of 54 iconic peaks in Colorado (e.g.,
Pikes Peak, Longs Peak, Mount of the Holy Cross)
that reaches or exceeds an altitude of 14,000 feet
(Blake, 2002).
PURPOSE STATEMENT
The intent of the exercise is to focus student
interest on an interesting geographic problem
that requires a suite of geospatial tools to solve.
To address this intent and the specific learning
expectations of the exercise (described in the
‘Learning Objectives’ section below), five general
purposes were developed:
• Minimize intimidation: Students of ge-
ography should learn to apply powerful
and complex tools for geospatial analysis
in non-intimidating and exciting ways.
Negative or frustrating first encounters
with geospatial tools such as ArcMap
GIS can intimidate and dissuade students.
Similarly, the datasets explored are often
difficult to view, edit, and manage in these
software environments. Avoiding intimida-
tion is addressed in the exercise by provid-
ing students with data and a step-by-step,
cookbook style set of instructions, many
with screen-shot examples.
• Colorado 14ers, Pixel by Pixel: is a cap-
stone exercise for a course that serves is a
prerequisite for all upper-level geospatial
tools courses including GIS, remote sens-
ing, image processing, and internet GIS. A
student familiar with geospatial terminolo-
gy (e.g, raster, vector, interpolation, DEM,
TIGER) who possesses some general com-
puter skills (e.g., copying folders, trans-
ferring files to a web server) and is armed
with a diverse geospatial toolkit (e.g., GIS,
geovisualization, geomorphometry) is po-
sitioned to enter upper-level courses with
heightened confidence.
• Scale is king: With most geographers rely-
ing at some point on GIS, remote sensing,
digital cartography, digital measurements,
and raster and/or vector data analysis, it
is important at early stages in a student’s
academic trajectory to recognize that all-
things-digital are influenced by scale.
• Human impacts: Geologic and atmo-
spheric hazards, sustainability, climate
change, and political stability are topics
currently being addressed by geographers.
To elucidate the connectedness between
humans and environment in these contexts,
students should interact with geographic
data face-to-face to visualize tangible ex-
1201

amples (forms) of natural and anthropo-
genic processes.
• What geographers do: Geographic inves-
tigations should be delivered to students
using local, applied, and interesting case
studies. This approach helps students rec-
ognize what geography is and what geog-
raphers do.
LEARNING OBJECTIVES
The learning objectives of the exercise dictate that
students gain experience using multiple geospatial
tools, different data models, and different file types
delivered at different resolutions. Concurrent with
tool and scale-related objectives, the exercise seeks
to introduce students to the depth and breadth of
the discipline of geography and seeks to inspire
students to take additional upper-level geospatial
tools courses. The specific learning objectives are
for students to:
• Visualize on a map a socially-derived
topographic constraint: following a set
of 14er ‘rules,’ students use GIS-based ras-
ter measurement techniques to determine
which of three nearby 14,000 foot plus
peaks is not a true 14er
• Observe on a DEM patterns of anthro-
pogenic processes, fluvial and paleogla-
cial erosion, and frost-shattering: these
resulting forms are made visible by cre-
ating a pixel diversity map with 32 color
classes (elevation diversity serves as a
proxy for surface roughness)
• Understand issues of map resolution:
students compare elevation value accu-
racies and contour line shapes of a 10 m
DEM to a 30 m DEM
• Gain experience using local operators:
students us a map calculator to convert
measurement units and to show surfaces
above 14,000 feet
1202
Colorado 14ers, Pixel by Pixel
• Visualize topographic relief on DEMs:
students create contour lines, index con-
tour lines, hillshade perspectives, and
slope maps
• Gain experience posting, presenting,
and sharing maps and information shar-
ing: students export maps from GIS soft-
ware, edit images, design web pages, and
securely transfer all files to a publicly ac-
cessible UCCS website
THE EXERCISE
Colorado 14ers, Pixel by Pixel is summarized in
a short presentation that provides an overview
of the exercise and outlines its purpose, learning
objectives, gradable items, and data and software
requirements (Appendix A). The full-length class
exercise exists in an electronic downloadable
format (Appendix B). The exercise requires an
internet connection to view helper images. An
answer key provides guidelines for instructors
(Appendix C). The URL below points to the two
DEMs required to complete the exercise: http://
www.uccs.edu/bvogt/co_14ers (http://www.uccs.
edu/bvogt/co_14ers).
REAL-WORLD CLIENTS
This exercise was developed with undergraduate
geography students in mind as the primary client.
However, in addition to serving as a learning
experience for college students, outcomes of the
exercise could serve real-world customers. Those
who could gain insight from this exercise include
outdoor recreation managers, peak ‘baggers’ (i.e.,
those attempting to summit a group of peaks in a
checklist fashion), and physical / natural scientists.
According to the Colorado Fourteeners Initia-
tive (CFI), a non-profit group whose focus is the
protection and preservation of Colorado 14ers,
as many as 500,000 visitors attempt to summit a
Colorado 14ers, Pixel by Pixel
Colorado 14er each year (CFI, 2010). This non-
profit group, as well as recreation managers with
the United States Forest Service and the National
Park Service, seeks to minimize erosion and trail
degradation while maintaining a challenging
recreation experience for peak baggers. Colorado
14ers, Pixel by Pixel provides recreation managers
with a tool that could help, for example, place a
new or reroute an existing hiking trail that ascends
a peak within a specific slope range.
Other potential clients served by this exercise
include the 500,000 peak baggers who attempt
to summit a Colorado 14er each year. Colorado
14er peak baggers who understand how socially-
defined criteria (e.g., the ‘rules’ for 14er selection)
are extended to a physical landscape may be
enlightened by the peak identification process or
may find peace of mind when bypassing a >14,000
ft. peak that is not classified as a true 14er.
Finally, geomorphologists, environmental
geologists, ecologists, paleoclimatologists, and
other physical and natural scientists could directly
benefit from this exercise: Ten and 30 m DEMs
are commonly used in scientific investigations
to identify, map, and measure various configura-
tions of landforms and landscapes (Pike, 2002).
Knowledge extracted from these measures and
mappings are then used to connect form to physi-
cal and/or human process. The exercise could
serve to train researchers in how to use common
geospatial tools to extract relevant information
from freely-available DEMs.
STUDENT ENGAGEMENT TIME
The exercise is generally completed in four to six
hours. Ideally, one to two hours working on the
exercise should be under the direct supervision
of an instructor. This face-to-face time allows the
instructor to work with students on geographic
concepts or on software-related trouble-shooting.
The exercise can be finished outside of the class-
room provided students are equipped with the
necessary software and the ability to transfer files
to a campus-based web directory.
ACKNOWLEDGMENT
The exercise described in this paper, Colorado
14ers, Pixel by Pixel, won the Applied Geogra-
phy Specialty Group (AGSG) of the Association
of American Geographers’ (AAG) 2010 class
exercise award competition. The AGSG/AAG
competition announcement invited graduate stu-
dents and faculty to submit an innovative applied
geography exercise designed for undergraduate
students. Submission guidelines requested a 250
word abstract; an exercise handout to students; a
narrative describing the intent of the exercise, the
fictional or real-world client, student engagement
time; and five to ten PowerPoint slides summariz-
ing the exercise.
REFERENCES
Blake, K. (2002). Colorado fourteeners and the
nature of place identity. Geographical Review,
92(2), 155–179. doi:10.2307/4140968
Colorado Fourteeners Initiative (CFI). (2010).
A partnership for preservation. Retrieved from
http://www.14ers.org/
Pike, R. (2002). A bibliography of terrain modeling
(geomorphometry), the quantitative representa-
tion of topography (Tech. Rep. No. 02-465).
Washington, DC: United States Geological Survey.
1203
 Colorado 14ers, Pixel by Pixel

APPENDIX A:COLORADO 14ERS, PIXEL BY PIXEL:

An Exercise in Mountain Geography (Slide Presentation)

Figure 1.

1204
Colorado 14ers, Pixel by Pixel

Figure 2.

1205
 Colorado 14ers, Pixel by Pixel

Figure 3.

1206
Colorado 14ers, Pixel by Pixel

Figure 4.

1207
 Colorado 14ers, Pixel by Pixel

APPENDIX B: COLORADO 14ERS, PIXEL BY PIXEL (STUDENT HANDOUT)

Figure 5.

1208
Colorado 14ers, Pixel by Pixel

Figure 6.

1209
 Colorado 14ers, Pixel by Pixel

Figure 7.

1210
Colorado 14ers, Pixel by Pixel

Figure 8.

Figure 9.

1211
 Colorado 14ers, Pixel by Pixel

APPENDIX C: ANSWER KEY TO CLASS EXERCISE ‘COLORADO 14ERS, PIXEL

BY PIXEL’

Students are required to place four images (each with caption) and the answers to the nine questions
below on their class exercise web page. Specifically, the following are graded:

1. One image with caption showing a close-up of the saddle (with contour lines) connecting Mt.
Cameron and Mt. Lincoln

Figure 10.

2. One image with caption showing all elevations above14,000 ft. on the Alma, CO DEM

Figure 11.

1212
Colorado 14ers, Pixel by Pixel

3. One image with caption showing the results of the slope analysis of the Alma, CO DEM

Figure 12.

4. One image with caption showing the results of the variety analysis of the Alma, CO DEM

Figure 13.

1213
 Colorado 14ers, Pixel by Pixel

5. Answer the following set of questions:

5 a. In terms of square meters, exactly how much area are we talking about IN ONE PIXEL on a
10 m DEM and on a 30 m DEM?
Answer: 100 square meters and 900 square meters respectively
5 b. What is the highest elevation value on the 10 m Alma, CO DEM? What is the highest eleva-
tion value on the 30 m Alma, CO DEM and why do you suppose the 10 m and the 30 m show
different maximum elevations?
Answer: The highest elevation on the 10 m DEM is 14,285.4 ft. and the highest elevation on the
30 m DEM is 14,245.4 ft. US Geological Survey 10 m DEMs were made from 30 m DEMs.
Thus, the difference in maximum elevations can be explained by the cartographic interpola-
tion process: To create finer resolution data (10 m) from coarser resolution data (30 m), the
interpolation process used by the USGS created a more smooth surface that in some cases
was forced above and below the known (more coarse) elevation data.
5 c. According to 14er ‘rules,’ one must descend 300 feet between 14ers for an adjacent 14,000 ft. +
peak to be an official 14er (Roach 1999, pg. 287). What is the vertical elevation loss between
Mt. Cameron and Mt. Lincoln? According to 14er ‘rules’ is Mt. Cameron an official 14er?
Answer: About 220 ft. No - a descent of less than 300 ft. means that Mt. Cameron is not an of-
ficial 14er.
5 d. Which areas (landforms) on your DEM show the lowest slope values (the most gentle slopes)?
Answer: The water surface in the reservoir, valley floors, and some high locations that sit well
above the level of former glaciation.
5 e. Which areas (landforms) on your DEM show the highest slope values (the steepest slopes)?
Answer: Slopes adjacent to ridges, cirque walls, and incised stream channel walls.
5 f. Based from the slope analysis, what natural/physical processes do you suppose cause STEEP
slopes and what processes do you supposed generate GENTLE slopes?
Answer: Steep slopes are likely caused by glacial erosion during last two or so glaciation periods
(e.g., Bull Lake and Pinedale), talus (scree) accumulations that reside at or near their angle
of repose, and valley incision by running water. Gentle slopes are likely caused by sediment
fill in valleys, rapid rock weathering at very high elevations (>13,000 ft.), and water surfaces
on reservoirs.
5 g. Based from a slope analysis, hillshade analysis, or variety analysis, provide three examples
of human (anthropogenic) forms visible on the DEM. Describe likely processes responsible
for shaping these features.
Answer: 1) a reservoir; containment of water, 2) narrow flat sections running along ridge tops;
foot traffic / compaction of unstable loose rock, 3) road grades; road construction, 4) mine
tailings; oversteepening of slopes from mining
5 h. From your variety analysis, what generalizations can you make that connect surface ‘variety’
and ‘landforms’?
Answer: High variety shows heterogeneity while low variety shows homogeneity. Generalizations
between variety and landforms include landform age (young = high elevation variety; old
= low elevation variety); glacial processes (glacial erosion on valley walls = high elevation
variety; glacial erosion on valley floor = low elevation variety); fluvial processes (fluvial
erosion = high elevation variety; fluvial deposition = low elevation variety).

1214
Colorado 14ers, Pixel by Pixel

5 i. Why are the contour lines different (e.g., why do the lines follow slightly different paths) for
the two different resolutions?
Answer: Contour lines on the 10 m DEM show higher sinuosity in plain view than do the contour
lines on the 30 m DEM. 10 m contour lines conform more closely to finer scale landsurface
features than do the 30 m contour lines.

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 2,
edited by Donald Patrick Albert pp. 17-32, copyright 2011 by IGI Publishing (an imprint of IGI Global).

1215
1216

Chapter 73
A Component-Based 3D
Geographic Simulation
Framework and its Integration
with a Legacy GIS
Zhen-Sheng Guo
Hokkaido University, Japan

Yuzuru Tanaka
Hokkaido University, Japan

ABSTRACT
There is an increasing demand for 3D geographic simulation systems. Most systems currently available are
closed and based on fixed architectures. Some systems allow us to develop and customize a 3D geographic
simulation system, but this usually requires the writing of extensive program code. Especially in 3D geo-
disaster simulations, for example, we need to dynamically integrate 2D legacy GIS with 3D geographic
simulation systems in order to investigate the details about the damaged areas and the consequences of the
disasters. The authors propose a component-based application framework for 3D geographic simulation
that can integrate a legacy 2D GIS with geographic simulation systems in a 3D visual environment. Their
framework provides a set of 3D visual components required for the development of a new interactive 3D
visual geographic simulation. In their framework, component integrators can construct 3D geographic
simulation systems by composing the 3D visual components. Moreover, the authors’ integration framework
provides two fundamental integration mechanisms, view integration and query integration mechanisms, to
integrate it with legacy 2D systems. The view integration function maps the 2D rendering of a legacy 2D
GIS onto the surface of the 3D geography used in a 3D visual geographical simulator, and then dispatches
every event on the geographic surface to the original 2D GIS. The query integration automatically converts
each 3D simulation result that is shown as a set of highlighted regions on the surface of the geography
to the corresponding regional query to the 2D GIS. The proposed framework is based on their 3D meme
media architecture in which components are represented as meme media objects, and their interoperation
is defined by slot connections between them. As a result, their framework enables users to compose 3D
geographic simulation systems and to integrate a legacy 2D GIS with a 3D geographic simulation system
simply by composing display objects in a 3D visual environment.

DOI: 10.4018/978-1-4666-2038-4.ch073

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
INTRODUCTION
Various geographic simulation systems have
been developed to meet the growing need for
3D geographic simulation of disasters, such as
avalanches, floods, and landslides. In general,
the development of each has been primarily inde-
pendent with closed and fixed architectures, it is
therefore difficult to introduce existing functions
of one existing geographic simulation system
to others. Moreover, although the geographic
simulation systems generally have strong disaster
simulation capability, they often lack the informa-
tion necessary for analysis and evaluation of the
disaster, whereas many legacy geographic infor-
mation systems (GIS) can provide visual map and
region query functions for disaster analysis and
evaluation. For this reason, GIS and geographic
simulation systems often need to be integrated.
However, in general, the development of the 3D
geographic simulation systems and legacy GIS
have been primarily independent and integrative
uses are still at an early stage.
There are many conventional GIS, and simula-
tors have become more advanced. Because these
types of software are usually independent, closed
systems, they are difficult to customize and to
integrate with others. Although some systems,
such as GRASS (GRASS) and ArcGIS (ArcGIS),
can be customized with library packages, they
still require users to write lengthy programs.
Other systems, such as GeoVista (Takatsuka,
2002) and AVS (Upson, 1989), provide a visual
programming environment for the development
or customization of the 3D simulation systems.
These systems allow users to interact only with
their visual programs; users cannot interact with
their visual output to develop a new 3D system or
modify their input parameter values. Component
object model (COM) is a model for linking soft-
ware modules. When such a link is established,
the modules can communicate with each other
through standard communication interfaces. These
links require users to be skilled in programming.
Open Geospatial Consortium (OGC) standards
(OpenGIS) can be used to integrate different Web
services through an OGC-standard Web service in-
terface. This integration demands OGC-compliant
Web services. However, most of the currently
available Web services are not OGC-compliant
(Alves, 2006).
This chapter proposes a component-based 3D
geographic simulation framework. Our frame-
work provides basic composite components for
3D geographic simulation, and two fundamental
legacy GIS integration mechanisms for component
integrators. Our framework enables component
integrators to easily build 3D geographic simula-
tion systems by visually combining the required
components and by defining data flows among
the components. The component integrators need
to have knowledge about the legacy systems, for
example, knowledge about the display and query
APIs and their parameters for the legacy GIS,
the input and output data formats of the legacy
simulators, and the data flows among compo-
nents. The component integrators also need to
know how to combine the required components
together. The end users of this framework only
need to replace some component with another of
the same functionality, by mouse manipulation,
to carry out 3D geographic simulations integrated
with a legacy GIS.
The component-based 3D geographic simula-
tion framework presented here incorporates the
basic components of 3D geographic simulation
systems and two legacy GIS integration mecha-
nisms into an integrated environment. This frame-
work provides a set of 3D visual basic components
required for the development of a new interactive
system that can integrate legacy 2D GIS and legacy
geographical simulators to make 3D geographic
simulation systems. In the integration framework,
construction of a new 3D geographic simulation
system only requires combining the necessary 3D
visual components. To support this framework
of integrating a legacy GIS, we also propose two
fundamental integration mechanisms: view inte-
1217
 A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
gration and query integration. The view integration
maps the 2D rendering of a legacy 2D GIS onto the
surface of the 3D geography used in a 3D visual
geographical simulator, and then dispatches every
event on the geography surface to the original 2D
GIS. The query integration automatically converts
each 3D simulation result that is shown as a set of
highlighted regions on the surface of the geogra-
phy to a corresponding regional query on the 2D
GIS. We used the 3D meme media system called
IntelligentBox (Tanaka, 2003; Okada, 1995) as the
basic platform of our framework. IntelligentBox
is a component-based 3D visual software devel-
opment system. In IntelligentBox, each primitive
component called a box has a special function and
some state holders called slots, as well as a special
3D shape. End users can define a functional and
visual composition using primitive boxes and slot
connections among them. Users can construct 3D
applications just through direct manipulation of
visual components. Moreover, each component,
once created, provides a simple API for its reuse
with other components in different applications.
The remainder of this chapter is structured
as follows: section 2 shows an overview of the
Figure 1. The component-based framework
1218
component-based framework, the detail of the
framework, and its integration mechanism with
a legacy GIS. In section 3, we describe an appli-
cation of our framework. Finally, we make some
concluding remarks.
COMPONENT-BASED FRAMEWORK
This section describes the proposed component-
based 3D geographic simulation framework.
Figure 1 shows the framework, which consists
of three types of 3D visual components used to
develop a 3D geographic simulation system. Our
framework allows component integrators to use
a mouse to combine these visual 3D components
together to construct geographic simulation sys-
tems. Moreover, in our framework, the display and
query functions of a legacy 2D GIS are integrated
with the 3D geographic simulation systems. Our
framework involves three basic technologies: 3D
visual components, component composition, and
integration of a legacy 2D GIS with 3D geographic
simulation systems technology.
A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
Details of the 3D Visual Components
This subsection describes the three basic com-
ponents. The 3D terrain component visualizes
terrain data stored in the elevation database as
a 3D terrain model. It also works as a provider
of the terrain data to other components in the
simulation environment. The geographic simu-
lation component is used for integrating legacy
simulators into our simulation framework. It also
calculates the necessary parameters for the geog-
raphy simulator and for visualizing the results of
simulation. The query component integrates the
query function of the legacy 2D GIS into the 3D
simulation environment. In this research, we used
a type of 3D meme media called IntelligentBox
as a basic system. IntelligentBox offers a series of
interactive 3D components. Each 3D component
has its own function and a standardized data ac-
cess interface called slots. Component integrators
can interactively define the parent-child relation-
ships between the 3D components to transfer
data between them by connecting slots. We will
explain the slots and the data that are kept in each
component.
Figure 2 shows the 3D terrain component.
It has slot #terrain_data and slot #displayAPI,
which hold the terrain data and the map range
parameters. The terrain data include the list of
the elevation data and the coordinate location of
the mesh point. The 3D terrain component can
display a 3D terrain model in the 3D environment
using the terrain data that are saved in slot #ter-
rain_data. This component obtains a map image
from a legacy 2D GIS and maps the map onto
the 3D terrain surface using the necessary data
(position, size, scale) saved in slot #displayAPI.
We call the integration of the display function of
the legacy GIS with the 3D terrain component
view integration. It is explained below in detail.
Figures 3 show the geographic simulation
component. It includes a legacy geographic
simulator (geo-simulator) component and a pa-
rameter calculation component. The legacy geo-
graphic simulator component integrates the
functions of a legacy geography simulator into
the 3D geography simulation environment. The
legacy geographic simulator component holds the
data that is acquired from other components in
slot #geo_data and generates an input file for the
legacy geographical simulator. The data stored in
slot #geo_data includes the terrain data and the
parameter values for the legacy simulator. We can
specify the path of the input and output files for
the legacy geographical simulator with slots #in-
put and #output. The legacy geographical simula-
tor obtains the input file from the path, which is
specified by slot #input, and sends simulation
results to the path, which is specified by slot
#output. The legacy geographical simulator com-
ponent observes the update of the output file of
the legacy geographical simulator (specified by
slot #output) at regular intervals. When the output
file is updated, the legacy geographical simulator
component reads the output file and holds the
result of the simulation in slot #result. The simu-
lation result is generally an array, each element
of which corresponds to a grid cell, and takes as
its value either 0 or 1. Finally, this component
visualizes the result in the same 3D environment.
The parameter calculation component calculates
the necessary parameters for the geographical
simulator and visualizes the calculation result.
The parameter calculation component holds the
data needed to calculate the parameter in slot
Figure 2. 3D terrain component
1219
 A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
Figure 3. Geographic simulation component
#geo_data and saves this calculation result in slot
#result. Moreover, if it is necessary, this compo-
nent can visualize the result data in a 3D environ-
ment.
Figure 4 shows the query component. It has
slot #geo_data, slot #queryAPI, and slot #result,
which hold the simulation results, query param-
eters, and query results respectively. The query
component receives the simulation result and
keeps it in slot #geo_data. The simulation result
includes the terrain data and the simulation result
values (for example, the damage zone data). Next,
the query component uses the data in slot #geo_
data to generate a range parameter and calls the
query function of the legacy 2D GIS with this
parameter through slot #queryAPI. For example,
the range parameter of a grid cell in the damaged
area takes the value 1. Then, the query component
Figure 4. Query component
1220
obtains a query result from the legacy 2D GIS
and holds it in slot #result. Finally, the query
component displays the query result (i.e., the
damaged road IDs) in the 3D environment. We
call the integration of the query functions of the
legacy GIS with the 3D geographic simulations
query integration, described in detail below.
Our framework allows component integrators
to combine these components to construct geo-
graphic simulation systems. The interoperation
of visual components is defined by slot connec-
tions among them. As a result, they can be com-
bined with other components by connecting their
slots. They interoperate by exchanging messages
through slot connections. Our framework also
enables end users to integrate a legacy 2D GIS
with a 3D geographic simulation system simply
through the composition of their display objects.
A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
Component Composition
In our framework, component integrators can com-
bine the 3D components arbitrarily as desired. To
combine more than one component, the processed
data and parameters must be exchanged among
the components. In our approach, the data and
parameters can be exchanged through the slots.
In addition, we use the view integration and query
integration mechanisms to integrate the display
and query functions of the 2D legacy GIS with
the 3D simulation environment.
Figure 5 shows an example of a composition.
In this example, we combine four components:
a 3D terrain component, a parameter calculation
component, a legacy geo-simulator component,
and a query component. The processed data and pa-
rameters are exchanged among these components
through the slots. First, the parameter calculation
component obtains the terrain data from the 3D
terrain component through slot #terrain_data.
The parameter calculation component calculates
the parameters using the terrain data and holds
the parameter values in its slot #result. Next, the
legacy geographical simulator component obtains
Figure 5. An example composition
the parameter data from the slot #result of the
parameter calculation component and holds the
parameter values in its slot #geo_data. The legacy
geo-simulator component uses the parameter data
to create simulation and holds the simulation result
in its slot #result. Finally, the query component
obtains the simulation result from the slot #result
of the legacy geo-simulator component and holds
the simulation result in its slot #geo_data. The
query component uses the simulation result to
generate a query for integrating the simulation
result with a legacy GIS and holds the query in its
slot #queyAPI. The query component integrates
the simulation result with the legacy GIS through
the slot #queyAPI and holds the query result in
slot #result.
Figure 5 shows the basic composition structure
of our framework. In addition, our framework
allows vertical and horizontal extension. Vertical
extension means that we can combine other dif-
ferent legacy geo-simulation components or
legacy GIS components with the topmost query
component in Figure 5 to simulate a secondary
or tertiary disaster caused by the preceding dam-
age. Horizontal extension means that we can
1221
 A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
combine two or more different legacy simulation
components and legacy GIS components with the
same 3D terrain component to comparatively
study different disaster simulations for the same
area. Moreover, we can simply change the input
data of the 3D terrain component to immediately
apply the same simulation system to another area.
Such flexibility and extensibility characterizes
our framework.
Integrating a Legacy 2D GIS with 3D
Geographic Simulation Systems
We integrated the display and query functions of
a legacy 2D GIS with a 3D simulation system.
We call the two integrations view integration and
query integration.
The view integration function maps the 2D
rendering of a legacy 2D GIS onto the surface
of the 3D terrain model used in a 3D visual geo-
graphical simulation, and it dispatches every event
on the terrain surface to the original 2D GIS. We
use the 3D shadow-copy mechanism (Ohigashi,
2005; Ohigashi, 2006) to map the 2D graphical
user interface (GUI) of a legacy GIS onto the
surface of the 3D terrain model. Moreover, the 3D
shadow-copy mechanism dispatches every mouse
event that occurs on the mapped surfaces of the
Figure 6. Integrating the GUI of a legacy 2D GIS with a 3D terrain model
1222
3D terrain component to the original 2D legacy
GIS with an inverse transformation of the texture
mapping process. Therefore, we can easily extend
the GUI of a 2D legacy GIS to a 3D geographic
information system. We use the 3D shadow-copy
mechanism to map the 2D GUI of a legacy GIS
onto the surface of the 3D terrain model.
Figure 6 shows a view integration of a legacy
2D GIS with 3D terrain model. We call this the
3D map component. First, the 3D map component
calculates the position, size, and scale information
from the original terrain data that is kept in slot
#terrain_data, which is used for generating ter-
rain polygons, as shown in Figure 6 (i). Next, this
component displays the appropriate part of the 2D
map from the whole map by calling the display API
of the legacy 2D GIS as shown in Figure 6 (ii).
Then, this component maps the clipped map onto
a 3D terrain using a 3D shadow-copy mechanism
as shown in Figure 6 (iii). Finally, we obtain a 3D
map as shown in Figure 6 (iv).
The query integration function integrates
query functions of a legacy 2D GIS with 3D geo-
graphic simulation systems. We propose a query
integration function that uses the query API of the
legacy 2D GIS. For example, query integration
automatically converts each 3D simulation result
that is shown as a set of highlighted regions on
A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
the surface of the geography to a corresponding
regional query to 2D GIS. In general, to search
for objects in a given area, we usually integrate a
database and use SQL to retrieve the objects from
the database. For instance, for a query searching
for a damaged cell to obtain its population saved
in GisDataList (GD), the GD schema would be
GisDataList (Population: integer, CellID:integer)
Moreover, the damaged cell can be referred to
by its ID. Therefore, this query can be expressed
in SQL as
SELECT GD.Population
FROM GisDataList GD
WHERE GD.CellID=ID
Alternatively, if the query is searching for rail-
ways across damaged cells, we have to query an
object-oriented database with an object-oriented
query. The schema of railway (RAILWAY) and
damaged cell object (DCELL) is shown as
RAILWAY (RouteID: integer, Line:
line)
DCELL (CellID:integer, Cell:area)
Therefore, the query can be expressed in an
object-oriented query as
SELECT x.Line inside (y.Cell)
FROM x in RAILWAY
y in DCELL
WHERE y.Cell=DamagedCell and
x.Line inter (y.Cell),
where x.Line is a line object that belongs to
RAILWAY and y.Cell is a cell object that belongs
to DCELL.
However, these query expressions are long
and complex. Furthermore, it is possible that the
database we want may not exist. On the other
hand, in many cases, the legacy GIS provide
APIs to develop new systems. We can call only
one query API function for retrieving the infor-
mation we want. A query integration mechanism
for integrating the 2D GIS query function with
3D geographic simulation results by calling its
query API is proposed.
Figure 7 shows the integration of a legacy
2D GIS query with a geographic simulation.
The right is a query component, and the left
is a legacy 2D GIS (a road map). The legacy
2D GIS offers a query API of searching for
objects within a specified range from some
layer. For instance, the query function could be
SearchForObjects(Within(range), From(layer)).
If we want to find damaged roads in damaged
zones, which can be obtained from a simulation
result, we can call the query function
SearchForObjects(Within(damaged zones),
From(road layer). The query component receives
the simulation result and keeps it in slot #geo_data
as shown in Figure 7 (i). Next, the query compo-
nent uses the data in slot #geo_data to generate a
range parameter and calls the query function of
the legacy 2D GIS with this parameter through
slot #queryAPI as shown in Figure 7 (ii). Then,
the query component obtains a query result from
the legacy 2D GIS and holds it in the parameter
slot #q_result of slot #queryAPI as shown in
Figure 7 (iii). Figure 7 (iv) shows that slot #result
holds the query result and the terrain data. Fi-
nally, the query component displays the query
result (i.e., the damaged road IDs) in the 3D en-
vironment as shown in Figure 7 (v). The query
result and terrain data are stored in slot #result to
be used by the next simulation or query.
APPLICATIONS
We present two applications of our framework to
demonstrate its effectiveness. The two application
systems were constructed from several compo-
nents. They integrate legacy GIS functions into
3D simulation systems.
1223
 A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
Figure 7. Integrating the query function of a legacy 2D GIS with 3D geographic simulations
An Avalanche Simulation
and its Disaster Query
In this section, we describe an example applica-
tion of our framework. This application shows
an avalanche simulation and its disaster query.
The scenario is an avalanche hazard evaluation.
A risk management expert is analyzing the road
damage caused by an avalanche. The analysis
process consists of 4 phases: (1) visualizing a 3D
map as the simulation base, (2) investigating the
highly dangerous slopes where avalanches could
easily happen, (3) calculating the avalanche dam-
aged zones from a dangerous slope to the point
where sliding snow would stop, and (4) searching
for damaged roads within the damaged zones.
The avalanche simulation and its disaster query
process require four components: a 3D terrain
component, a dangerous-slope search component,
an avalanche simulation component, and a query
component. The 3D terrain component visualizes
the 3D terrain data as a 3D terrain model and dis-
plays a map of the 2D legacy GIS on its surface. It
serves as the basic component for the simulation
and visualization. The dangerous-slope search
component investigates the dangerous slopes in
1224
the specified area and visualizes those slopes on
the 3D terrain surface. The avalanche simulation
component simulates the avalanche and visual-
izes the simulation result (avalanche area) on the
same 3D terrain. The query component searches
for damaged roads within the avalanche area by
integrating the query function of the legacy GIS
and visualizing the query result in the same 3D
simulation environment.
Figures 9 and 10 show an avalanche simula-
tion and its disaster query. Figure 8 (A) shows the
application framework, while Figure 8 (B) shows
a screenshot. The system consists of a 3D terrain
component, a dangerous-slope search component,
an avalanche simulation component, and a query
component. Using these components, the simula-
tion and query are carried out in four steps.
In the first step, we use the 3D terrain compo-
nent to visualize the terrain data held in slot #ter-
rain_data as a 3D terrain model and use view
integration to visualize a legacy GIS map on the
3D terrain surface. The map range parameter is
specified through slot #displayAPI as shown in
Figure 8 (A) (i). This generates a 3D map as shown
in Figure 8 (B) (a).
A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS

Figure 8. An application for an avalanche-simulation system

Figure 9. the avalanche simulation with CONEFALL

1225
 A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS

Figure 10. An application for a flood simulation system

In the second step, we use the dangerous-slope
search component to look for the high-danger
slopes where avalanches could easily happen.
In this process, the dangerous-slope search com-
ponent needs 3D terrain data and a 3D map to
visualize the dangerous slopes. Then, we combine
the dangerous-slope search component with the
3D terrain component (shown in Figure 8 (B) (1)).
The dangerous-slope search component obtains
the terrain data from the 3D terrain component
through slot #terrain_data shown in Figure 8 (A)
(ii). It is generally known that slopes with angles
between 35 and 40 degrees are prone to avalanches
(K. Maeno, 2000). Therefore, the dangerous-slope
search component calculates each slope angle. As
shown in Figure 8 (A) (iii), the grid cells of slope
angles between 35 and 40 degrees were set as 1
and others were set at 0. The dangerous-slope
search component holds the calculation result data
and terrain data in slot #result. This component
visualizes the dangerous slopes as red cells on the
3D map shown in Figure 8 (B)(b).

1226
A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
In the third step, we use the avalanche simu-
lation component to simulate and visualize the
avalanche damaged zones that would occur if the
snow slid from a dangerous slope to a predicted
stop point. Here, we used a legacy simulator
called CONEFALL (Jaboyedoff) for simulating
the damaged zones. CONEFALL is designed to
estimate the zones prone to rock falls. It works on
the same principle as the avalanche simulation. In
this process, the avalanche simulation component
needs to communicate with the legacy simulator
CONEFALL by exchanging files. It also needs
the dangerous slope and 3D terrain data as pa-
rameters. Therefore, we combine the avalanche
simulation component with the dangerous-slope
search component (shown in Figure 8 (B) (2)).
As shown in Figure 8 (A) (iv), the avalanche
simulation component obtains the dangerous
slope and terrain data from the dangerous-slope
search component through slot #geo_data and slot
#result. The files are exchanged with CONEFALL
through slot #input and slot #output. The result
of this simulation is visualized on the 3D map as
shown in Figure 8 (B) (c).
Figure 9 shows the avalanche simulation with
CONEFALL. The simulation is carried out as fol-
lows. First, the avalanche simulation component
obtains the terrain data and the dangerous-slope
list from the dangerous-slope search component
and keeps these data in slot #geo_data as shown
in Figure 9 (A) (i). Then, the avalanche simulation
component generates an input I/O file using the
data held in slot #geo_data and saves this file in a
path specified via slot #input shown in Figure 9 (A)
(ii). After that, CONEFALL simulates the damaged
zones using the file, and sends the result file of
its simulation calculation to a specified file path.
The output of CONEFALL (simulation result) is
a list of 1s and 0s in which the 1s correspond to
the damaged zones. Then, the avalanche simula-
tion component reads in the simulation result file
from the path specified in slot #output, as shown
in Figure 9 (A) (iii). Finally, the simulation result
is held in slot #result and the simulation result is
visualized on the 3D map as shown in Figures
11 (iv) and (v). Moreover, the terrain data is also
held in slot #result and can be used for the next
simulation or query.
In the fourth step, we use the query component
to search for and visualize the damaged roads
within the damaged zones. This query process
needs the damaged zones data and terrain data
to generate query parameters. We combine the
query component with the avalanche simulation
component to find the damaged roads in the dam-
aged zones as shown in Figure 8 (B) (3). The query
component obtains the damaged zones from the
avalanche simulation component shown in Figure
8 (A) (iv), and the damaged zones are used to
generate the query parameters of the legacy GIS.
The query parameters are held in slot #queryAPI.
Using this slot, we integrate the query function
of a legacy 2D GIS with the query integration
mechanism to search for damaged roads in the
damaged zone, as shown in Figure 8 (A) (vi). The
query result (damaged roads) is held in slot #result.
This component displays the query result in the
3D environment as shown in Figure 8 (B) (d).
The simulation result is usually represented as
a set of grid cells (as shown in Figure 9 (B)). The
point (X00min,Y00min) is the lower-left corner
of cell 00, while the value d is the side length of
one cell. The query functions of legacy 2D GIS are
usually specified as a vertex coordinate pair, i.e.,
the lower-left corner and the upper-right corner
of a damaged cell, as a range parameter. We can
calculate the vertex coordinate pair (Xijmin,Yijmin)
and (Xijmax,Yijmax) of any cell ij by the formulae
(Xijmin,Yijmin) = (X00min + i × d, Y00min + j
× d) and (Xijmin,Yijmin) = (X00min +(i+ 1)×
d, Y00min + (j+1) × d). We can convert all the
damaged cells to a coordinate pair set in this way.
The coordinate pair set can be used as the range
parameter of the legacy 2D GIS query function.
Therefore, a query component can use the coor-
dinate pair set as a range parameter for calling the
query function of the legacy 2D GIS to find objects
in them. For instance, we search for damaged
1227
 A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
roads, those roads overlapped with the damaged
cells, by calling the query function SearchForO
bjects(Within((Xijmin, Yijmin),(Xijmax,Yijmax)),
From(road layer)).
A Flood Simulation and
its Disaster Query
Using our framework, we can perform a different
simulation simply by changing the simulator com-
ponents. In this section, we show a flood simulation
application and its disaster query using the same
3D terrain and query components. The scenario
is a flood hazard evaluation. A risk management
expert is analyzing the road damaged that would
be caused by a flood. The analysis consists of three
phases: (1) visualizing a 3D map as the simulation
base, (2) calculating the flooded zone, and (3)
searching for damaged roads within the flooded
zone. The flood simulation and its disaster query
process require three components: a 3D terrain
component, a flood simulation component, and
a query component. The 3D terrain component
visualizes the 3D terrain data as a 3D model and
displays a map of 2D legacy GIS on its surface. It
serves as the basic component for the simulation
and visualization. The flood simulation component
simulates the flood and visualizes the simulation
result (water surface). The query component
searches for damaged roads within the flooded
zone by integrating the query function of a legacy
GIS and visualizing the query result in the same
3D simulation environment.
Figure 10 shows a flood simulation and its di-
saster query. Figure 10 (A) shows the component-
based application framework, while Figure 10 (B)
shows a screenshot of the simulation and the query
evaluation. This system consists of a 3D terrain
component, a flood simulation component, and a
query component. Using these components, the
flood simulation and its disaster query are carried
out in three steps.
In the first step, we use the 3D terrain compo-
nent to visualize the terrain data held in slot #ter-
1228
rain_data as a 3D terrain model and to visualize
a legacy GIS map on the 3D terrain surface with
view integration, as shown in Figure 10 (A) (i).
The map range parameter is specified through slot
#displayAPI. This generates a 3D map, as shown
in Figure 10 (B).
In the second step, we use the flood simula-
tion component to calculate the flooded zone.
In this process, the flood simulation component
needs 3D terrain data. Therefore, we combine
the flood simulation component with the 3D ter-
rain component shown in Figure 10 (B) (1). The
flood simulation component obtains the terrain
data from the 3D terrain component through slot
#terrain_data and keeps it in slot #geo_data as
shown in Figure 10 (A) (ii). The flood simulation
component changes the height of the water surface
based on the simulation results, taking into account
the total amount of rainfall. It enables users to vi-
sually find the submerged areas of the simulation
as shown in Figure 10 (B). The flood simulation
result data is a set of 0s and 1s. The cells under
the water surface are set as 1, as shown in Figure
10 (A) (iii). So that another query or simulation
process can use the same terrain data, the flood
simulation component keeps the simulation result
data and terrain data in its slot #result.
In the third step, we use the query component
to search for damaged roads within the flooded
zone. In this process, the query component needs
the data for the flooded zone and the 3D terrain.
Therefore, we combine the query component
with the flood simulation component as shown
in Figure 10 (B) (2). As shown in Figure 10 (A)
(iv), the query component obtains the data for the
flooded zone and the terrain from the flood simu-
lation component through slot #result and keeps
it in slot #geo_data. When the query component
receives these data, it searches the flooded zone
contained geometrically in its cuboids’ shape, and
the flooded zone is used as a condition of a query
to the legacy GIS. We integrated the query function
of the legacy 2D GIS with the query integration
mechanism as shown in Figure 10 (A) (v). As
A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS
shown in Figure 10 (B), the query component
displays the flooded road IDs (the roads within
the flooded zone as shown in Figure 10 (A) (vi))
in the 3D environment.
In addition, using our framework, we can
perform the following process simply. We can
interactively locate the desired areas by moving
the query component or by changing the size of
the component. To investigate another area of the
3D terrain, we need simply to change the input
data of the 3D terrain component. To search for
different objects in the same area, we can also
combine another query component with the same
simulation component at the same time.
CONCLUSION
In this chapter we have proposed a new component-
based 3D geographic simulation framework. Using
this framework, component integrators can easily
develop 3D geographic simulation systems just
by combining the required components and by
defining data flow among these components. The
end users of this framework only need to replace
some component with another of the same func-
tionality, using mouse manipulation, to carry out a
3D geographic simulation integrated with a legacy
GIS. Our framework incorporates the fundamental
components of 3D geographic simulation systems
and two legacy GIS integration mechanisms into an
integrated environment. The framework provides
a set of 3D visual components required for the
development of a new interactive system that can
integrate legacy 2D GIS and legacy geographical
simulators to make 3D geographic simulation sys-
tems. The visual component technology provides
a set of 3D visual components required for the
development of new interactive 3D visual geo-
graphic simulations integrated with GIS functions.
In the proposed framework, constructing a new 3D
geographic simulation system only requires the
component integrators to combine the necessary
3D visual components. To support the integration
of a legacy GIS with the simulation system, we also
propose two fundamental integration mechanisms:
view integration and query integration mechanisms.
The view integration technology can integrate the
GUI of a legacy 2D GIS on the surface of the 3D
terrain used in a 3D visual geographical simulator
and then dispatch every event on its surface to the
original 2D GIS. The query integration technology
can integrate the query function of a legacy 2D
GIS by automatically converting each 3D simula-
tion result to a corresponding regional query on
a 2D GIS. As a result, our approach enabled us
to construct 3D geographic simulation systems
interactively, moreover, to integrate the GUI and
the query function into a 3D simulation environ-
ment simply through the composition with their 3D
display objects by direct manipulation on a com-
puter screen. This framework also allows users to
integrate a legacy GIS with a legacy geo-simulator
that is independently developed and not assumed
to integrate with it.
In this chapter we have proposed an integra-
tion technique for legacy geographic simulators
whose I/O text file formats are open to the public.
However, in a future, it will become necessary to
propose a new integration technique for various
other simulators that do not satisfy this condition.
To cope with such integration it is necessary to
provide different query components for different
GIS APIs. The method of dynamic API slots may
become necessary. Such a solution should provide
a wizard system to help users to semi-automatically
create the required slots from the API list of the
legacy system.
REFERENCES
Alves, L. L., & Davis, C. A., Jr. (2006) Interoper-
ability through Web services: evaluating OGC
standards in client development for spatial data
infrastructures. VIII Brazilian Symposium on
GeoInformatics, 2006, INPE,193–208. ArcGIS.
Retrieved from http://www.esri.com/products.html
1229
 A Component-Based 3D Geographic Simulation Framework and its Integration with a Legacy GIS

GRASS (Geographic Resources Analysis Support
System). Retrieved from http://grass.itc.it/
Jaboyedoff, M. & Labiouse. V. CONEFALL: a
program for the quick preliminary estimation
of the rock-fall potential of propagation zones,.
Computer & Geosciences.
JCMA.(1988). Protection Against Snow Hand-
book.
Maeno, K. & Fukuda, M. (2000). Avalanche and
Snowstorm.
Ohigashi, M., Guo, Z. S., & Tanaka, Y. (2006).
Integration of a 2D legacy GIS, legacy simula-
tions, and legacy databases into a 3D geographic
simulation, SIGDOC ’06: Proceedings of the 24th
Annual Conference on Design of Communication,
pp.149–156.
Ohigashi, M., & Tanaka, Y. (2005). Shadows on
3D: Hosting 2D legacy applications into a 3D
meme media environment. In Leipziger Informa-
tik-Tage 2005(LNI 72, pp. 401–410).
Okada, Y., & Tanaka, Y. (1995). IntelligentBox: A
constructive visual software development system
for interactive 3D graphic applications. Proc. of
Computer Animation, 95, 114–125.
OpenGIS Standards and Specifications. Retrieved
from http://www.opengeospatial.org/standards
Takatsuka, M., & Gahegan, M. (2002). Ge-
oVISTA Studio: a codeless visual programming
environment for geoscientific data analysis and
visualization. Computers & Geosciences, 28(10),
1131–1144. doi:10.1016/S0098-3004(02)00031-6
Tanaka, Y. (2003). Meme Media and Meme
Market Architectures: Knowledge Media for
Editing, Distributing, and Managing Intellec-
tual Resources. IEEE Press, John Wiley & Sons.
doi:10.1002/047172307X
Upson, C., & Faulhaber, J. Thomas, Kamins,
D., Laidlaw, D.H., Schlegel, D., Vroom, J., Gur-
witz, R. & van Dam, A. (1989). The application
visualization system: A computational environ-
ment for scientific visualization. IEEE Com-
puter Graphics and Applications, 9(4), 30–42.
doi:10.1109/38.31462

This work was previously published in Interdisciplinary Advances in Adaptive and Intelligent Assistant Systems: Concepts,
Techniques, Applications, and Use, edited by Gunther Kreuzberger, Aran Lunzer and Roland Kaschek, pp. 63-82, copyright
2011 by Information Science Reference (an imprint of IGI Global).

1230
 1231

Chapter 74
Collaborative Mapping and GIS:
An Alternative Geographic
Information Framework

Edward Mac Gillavry

Webmapper, The Netherlands

ABSTRACT
The collection and dissemination of geographic information has long been the prerogative of national
mapping agencies. Nowadays, location-aware mobile devices could potentially turn everyone into a
mapmaker. Collaborative mapping is an initiative to collectively produce models of real-world loca-
tions online that people can then access and use to virtually annotate locations in space. This chapter
describes the technical and social developments that underpin this revolution in mapmaking. It presents
a framework for an alternative geographic information infrastructure that draws from collaborative
mapping initiatives and builds on established Web technologies. Storing geographic information in
machine-readable formats and exchanging geographic information through Web services, collaborative
mapping may enable the “napsterisation” of geographic information, thus providing complementary
and alternative geographic information from the products created by national mapping agencies.

INTRODUCTION govern the territory. Away from home, maps were

Since the Enlightenment, mapping and the pro-
duction of geographic information have been
institutionalised: the map is the power. At home,
maps were used as an instrument for nation build-
ing as nation states emerged: a legitimisation
device (McHaffie, 1995). People learned about
their country and administrations needed a tool to
DOI: 10.4018/978-1-4666-2038-4.ch074
an instrument for colonisation, when Africa and
Asia were split among the European nation-states.
During the last few decades, there has been
rapid democratisation of geographic information
and maps. Sawicki and Craig (1996) distinguish
three ways in which this movement is apparent.
First, the locus of computing power and data access
is broadening. Second, the level of skills to turn
raw geospatial data into geographic information
has become less demanding. Third, the locus of

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

applications has moved closer to the citizenry.
Geographic information systems moved from
mainframes and the UNIX operating system onto
personal computers and the Windows operating
system. From research and government, GIS
spread into the business sector. The PARC Xerox
Map Server and Virtual Tourist brought maps to
everyone’s PC in the late 1990s, followed by online
map Web sites such as MapQuest and Multimap.
In 1997, Brandon Plewe noted that “the Internet
holds promise for exponential increases in the
efficiency and effectiveness of the ways in which
we obtain, use and share geographic information
in all its forms” (Plewe, 1997). In July 2002, 7.1
million European users visited one of the many
online map Web sites (Nielsen//NetRatings, 2002).
Google Maps, introduced in February 2005,
reached almost 1.7 million visitors in that month
(Buchwalter, 2005).
Although maps are more widely used than ever,
the production of geographic information, and
especially mapping, is still highly concentrated
among national mapping agencies and the GI in-
dustry. But this oligarchy is soon to be dissolved,
for we see the third aspect of the democratisation
of geographic information–the locus of applica-
tions moving closer to the citizenry–becomes
apparent now that location-aware mobile devices
are coming within everyone’s reach. GPS units
are not only available to surveyors anymore, as
cheaper devices are sold for outdoor recreation.
Also, small GPS antennae can communicate
with other devices over Bluetooth, and there
are already mobile phones and personal digital
assistants (PDAs) for the consumer market that
have GPS-chips built in.
At the same time, digital maps have become
portable. Various mobile phone operators have
started to deliver location-based services to mobile
devices. Mobile phones come with route planning
applications, thus making in-car navigation sys-
tems redundant. Maps are not only delivered to
the desktop, but also to mobile phones and PDAs,
1232
Collaborative Mapping and GIS
requiring new visualisations as the screen size,
resolution, and use patterns differ significantly.
Collaborative mapping is an initiative to col-
lectively create models of real-world locations
online that anyone can access and use to virtually
annotate locations in space (McClellan, 2003). The
value of the annotations is determined by physi-
cal and social proximity, the former expressed
in distance, the latter in “degrees of separation.”
Thus, the informational value and the pertinence
of spatial annotations is not only dependent on
physical distance, but also dependent on the
trust relationship between individuals or groups
of people through social networks: the “Web of
Trust” (Espinoza, Persson, Sandin, Nystrom,
Cacciatore, & Bylund, 2001).
However, there is a discrepancy between physi-
cal and social proximity. Privacy and personal
freedom become highly important issues when
one’s location is related to their social behav-
iour. On the other hand, the fear of surveillance
that accompanies positioning is already gradu-
ally reducing in society (Ahas & Mark, 2005).
Furthermore, this discrepancy can be mediated
by users themselves by storing annotations and
tracks locally, thus creating distributed reposito-
ries, and by explicitly setting the level of privacy
on each of these annotations and tracks. Finally,
users themselves remain in control of their so-
cial identification–their preferences and social
network–while they make use of collaborative
mapping services, whereas, for example, the so-
cial positioning method aggregates these social
characteristics to study the space-time behaviour
of society (Ahas & Mark, 2005). Collaborative
mapping services are therefore less pervasive in
the privacy of their users because users negotiate
the trade-off between the benefits of the service
and their privacy concerns.
Collaborative mapping is to location-based
services (LBS) what blogging is to the Web: a
mechanism for authoring one’s own location-
based experiences (Bleecker, 2005). The notion
of “social proximity” as a measure of relevancy
Collaborative Mapping and GIS
of geographic information sets collaborative
mapping apart from corporate LBS-providers.
Requesting the nearest Italian restaurants, the
service not only takes into account the distance
to the restaurant from the current position of the
user, but it also navigates his/her social network
to find a suitable result. People are willing to
travel farther because the particular restaurant is
a favourite haunt of one of their relations. Fur-
thermore, geographic information is not merely
broadcast to users, but users can actively contribute
to the service: geographic information flows back
to the service. A collaborative mapping service
then works like a spatially enabled notebook or
message board, depending on the privacy settings
people attach to their postings.
Collaborative mapping may even become a
vehicle for community-based participatory plan-
ning. As users share their thoughts, opinions, and
ideas about physical places, local knowledge,
community needs, and specific social histories
may thus be appreciated and incorporated into the
process (Harris, Weiner, Warner, & Levin, 1996).
In this way, the necessary information is produced
in an operation whereby the local knowledge aris-
ing from the social narratives, the mental map, is
converted into data within a GIS for research and
policy formulation.
This chapter aims to describe the various de-
velopments that have led to the necessity and fea-
sibility of collaborative mapping. Based on current
practice, it goes on to describe an approach for an
alternative geographic information infrastructure
that draws from collaborative mapping initiatives
and builds on established Web technologies.
BACKGROUND
Free Geographic Data
Collaborative mapping as a practice to create
geographic data was born out of the lack of af-
fordable and copyright-free geographic data.
Governments spend billions to create large-scale
geographic data to fulfill their civil and military
responsibilities, and only governments can afford
the hardware and software, as well as the labour
costs of trained personnel involved in creating
large-scale geographic data (Harris et al., 1996).
Nevertheless, geographic data created by US
government agencies such as the US Geological
Survey (USGS) and the Census Bureau is available
copyright free and made available at the nominal
cost of copying it, whereas in most other countries,
geographic data has to be paid for and is subject
to strict copyright laws.
Since the market position of national map-
ping agencies (NMAs) is changing rapidly, their
budgets being cut and status changed, their geo-
graphic databases have become their main capital
assets. In the new economy, the digital elevation
model and digital line graphs are the new currency
(McHaffie, 1996). To survive in their new role,
they have to protect their assets with copyright
restrictions and inhibiting pricing structures.
Whereas other government agencies and academic
institutions in these countries obtain geographic
data at a discounted rate, nonprofit organisations,
(small) businesses, and individual citizens have
found these protective measures inhibiting. Not
only pricing and copyright put up barriers. As the
GI industry is evolving rapidly, patents have been
granted to various key players in the market. This
practice is a delicate trade-off between encour-
aging innovation and encouraging monopolies.
Collaborative mapping has therefore become a
practice to create copyright-free geographic data
for free and to bring together existing freely avail-
able sources of geographic data.
Positioning Technology
There are various developments that have fa-
cilitated the viability of collaborative mapping.
First of all, the democratisation of mapping has
not only brought maps to everyone’s desktop or
hand-held device, the availability of position-
1233

Figure 1. Various trends drive the uptake of collaborative mapping
ing technology contributes to this trend as well.
GPS devices used to be the domain of surveyors
and geodetic engineers, but nowadays there are
many cheap GPS devices available for outdoor
enthusiasts, which become smaller and smaller.
As of 1 May 2000, selective availability has been
turned off so public users can locate themselves
using GPS with “military precision.”
With Bluetooth, separate GPS devices commu-
nicate with palmtops and mobile phones. Recently,
there are palmtops and even mobile phones that
come with a built-in GPS antenna. Furthermore,
the Enhanced 911 (E-911) US federal mandate
requires mobile phone operators provide precise
location information of mobile devices within 50
to 300 meters. The ability to establish one’s loca-
tion is unprecedented. The data that is obtained
from location-aware devices indicates the actual
location and movement of people, with increas-
ing precision in real-time (Ahas & Mark, 2005).
These location-aware devices turn everyone
into a producer of geographic information, thus
bridging the cartographic labour process. Ever
since the first differentiation of cartographic
labourers that occurred with the invention and
adoption of the printing press, when compila-
tion and reproduction of the map were separated
(McHaffie, 1996), mappers are “slogging” into
the messy reality of the field again in order to
produce the “map.”
1234
Collaborative Mapping and GIS
Location Awareness
Publishing on the Web has changed over the last
few years. There is no need to have your own
Web server or even know how to code HTML:
nowadays everyone can publish on their Weblog.
These Web based publications consist primarily
of periodic articles in reverse chronological order.
Admittedly, most early Weblogs were updated
manually, but tools to automate the maintenance
of such sites made them accessible to a much
larger population.
Not only Weblog authors (bloggers) writing
about similar topics started to form communities,
but there also started to appear listings of Weblogs
that were from the same country (e.g., the now
defunct Gblogs) or even located in the same city
(BrightonBloggers). The next step came when
NYCBloggers (http://www.nycbloggers.com/)
and London Bloggers (http://londonbloggers.
iamcal.com/) started indexing bloggers by their
nearest subway stop or tube stop respectively.
Mapping the virtual space of the blogosphere on
to the physical one, these listings put a sense of
place back into cyberspace.
In a final step, the positional accuracy increased
from country, town, and subway stop to the level
of geographic coordinates, while the geographic
extent increased to cover the world with GeoURL
(http://geourl.org/). Bloggers are encouraged to
Collaborative Mapping and GIS
add HTML meta tags to their Web pages to inform
visitors about their location:
<meta name=”ICBM”conte
nt=”52.09022,5.08159”>
<meta name=”geo.position”conte
nt=”52.09022;5.08159”>
<meta name=”DC.title”
content=”Webmapper: what the map can
be.”>
GeoURL then spiders the Web and indexes
Weblogs that have these meta tags, and puts pins
on the world map indicating the locations of all
the Weblogs. Some bloggers travel regularly, and
Weblogs can be updated using mobile devices,
so-called moblogging. Location thus becomes
another means to structure Weblog entries. Also,
bloggers frequently write about locations they
are visiting. Developers have created extensions
for popular blogging tools such as Movable Type
for authors to attach a location to their Weblogs
and individual postings, and to keep track of their
travels, nearby Weblogs, and the locations of their
visitors. Other tools such as IndyJunior (http://
www.bryanboyer.com/indyjunior/) and World66
(http://www.world66.com/) allow bloggers to
show their travels and the countries visited. The
World as a Blog (http://brainoff.com/geoblog/)
shows the locations of recently updated We-
blogs. All in all, Weblogs have certainly raised
the awareness of location as a means to structure
information sources.
Online Communities
Online communities have been around since the
start of the Internet. More and more, it is not only
important to advocate that you belong to a certain
community, but also who you know in various
communities. Search engines rate Web sites based
upon the number of hyperlinks that link to a Web
site. Online social networks started appearing in
2002, when the term was used to describe the
means of networking in virtual communities, and
became popular in 2003 with the advent of Web
sites such as Friendster, Tribe.net and LinkedIn.
While Friendster allows members to chart their
network of friends, LinkedIn aims at profession-
als to keep track of their business contacts. The
popularity of these sites rapidly grew, and major
companies have entered the Internet social net-
working space.
Whereas these online social networks require
visitors to subscribe to a particular network before
using the service, and people end up subscribing
to all these Web sites, thus becoming a member
of all separate networks. XFN and FOAF are
distributed networks. People define their own
networks and can identify their online friends
and the nature of their relationship. As FOAF is
based on RDF, it can be mixed with other RDF
vocabularies, so resources can be shared among
networks of people. Thus, sharing information
among a group of people based on the nature of
relationships is a feasible thing.
Open Source GIS
The final trend contributing to the birth of col-
laborative mapping is the adoption of open source
in the field of GIS. The IT industry has a long
history of open source projects, from operating
systems, Web servers, and programming languages
to applications. Only recently, the GI industry
is catching up, with open source databases that
support spatial features, for example, MySQL and
PostGIS. There are several libraries, for example,
to convert between common spatial data formats,
or to change projections that are included in vari-
ous open source projects. GRASS has become
a well-known open source desktop GIS. For
online GIS applications, there are various open
source Web-mapping servers such as Deegree,
GeoServer, and the UMN MapServer. Web clients
such as Chameleon and ka-Map allow developers
to build compelling interfaces to work with maps
in your Web browser. Most of the architectures
1235

are built around OGC (open geospatial consor-
tium) standards such as the Web Feature Server
and Geography Markup Language specifications.
Online vector-based mapping used to be the
domain of proprietary plug-ins from leading GIS
vendors. As the XML-based scalable vector graph-
ics (SVG) file format gains popularity, vector-
based mapping is not anymore the monopoly
of the GIS industry and government agencies.
Many individual Web developers have started
to create compelling vector-based Web-mapping
applications.
With geographic data being transferred based
upon open standards and clear interfaces, it is easy
for developers to combine various data sources
and make interesting Web-mapping applications,
with no resources spent on buying software.
COLLABORATIVE MAPPING
Several projects are underway to collect geo-
graphic data for the creation of base maps. Amster-
dam RealTime (http://www.waag.nl/realtime/),
conceived as an art project, has been a trigger for
many people to realise the potential of collabora-
tive mapping. As part of an exhibition on historical
maps of Amsterdam in late 2002, visitors were
invited to participate and take a GPS device with
Figure 2. The iterative collaborative mapping process consists of four steps
1236
Collaborative Mapping and GIS
them for a week. During that time, their locations
were uploaded in real time to a Web server and
beamed onto a screen in one of the exhibition
rooms. Visitors of the exhibition saw a new map
of Amsterdam gradually appear as the individual
GPS tracks joined up and started to overlap.
These base maps typically consist of road
networks in urban areas that have been traced
moving around the area with a GPS device. Most
community mapping projects seem to follow
four steps: collection, editing, augmentation,
and distribution (Walsh, 2005). In the following
paragraphs, these steps of community mapping
projects are discussed. Although the paragraphs
are provided in this sequence, community map-
ping is an iterative process. The steps are regularly
iterated as new data becomes available, and these
steps may take place concurrently.
Collection of Base Map
Data and GPS Tracks
There are several ways to kick-start a collabora-
tive mapping project. One approach is to scan
and georeference out-of-copyright paper maps
or global satellite imagery, and incorporate these
as a background to provide a geographic context
for the project. The advantage of global satellite
imagery over scanned paper maps is the flexibility
Collaborative Mapping and GIS
in the projection. Opting for scanned paper maps,
there is not this flexibility, as the map image may
become distorted when changing the projection.
Another approach is to use readily available
vector-based geographic data, having the flex-
ibility of projecting the data while maintaining the
sharpness of vectors. For example, the TIGER/
Line data from the US Bureau of the Census pro-
vides a good source for geographic data to start a
community mapping project. As discussed previ-
ously, copyright restrictions may prevent projects
from incorporating these sources. For the Mumbai
Free Map project (http://crit.org.in/projects/gis),
the Collective Initiative Research Trust (CIRT)
builds the base map from the cumulative basis
of the urban research, design, and development
studies and interventions conducted previously
by CIRT. Thus, CIRT develops an open-access
spatial data infrastructure, and a set of simple tools
and applications localised in Indian languages,
for knowledge transfer and participatory urban
planning by communities and citizens in Mumbai.
The third approach to collect base map data is
the most labour intensive. This low-level approach
requires a large group of community members to
travel the area to be mapped with a GPS receiver,
and follow the tracks, paths, and streets to trace
the network. On their travels, members may add
additional waypoints for landmarks in the area.
Returning from these surveys, the community
members upload their GPS tracks over the Web
to a central server, or equally make their tracks
available online for others to harvest. Previously
collected GPS data created while engaging in a
game of geocaching or as a form of GPS-drawing,
can be equally made available for integration.
In the United Kingdom, GPSdrawing (http://
www.gpsdrawing.com/) uses a GPS receiver to
make pictorial drawings while moving around.
Conceived as an art project, base mapping is not
the objective here, but the accumulation of GPS
tracks makes a nice base map. Geowiki (http://
www.geowiki.com/) is a cartographic variation
of the interactive encyclopaedia Web site, Wiki-
pedia. GPS tracks are converted into PostScript
files that are merged, using a drawing package,
into a GIF file that is put on the Web site. Apart
from showing the road networks in some of the
larger towns in Britain, the Web site allows visi-
tors to enter information about pubs and shops,
with personal reviews.
Whereas Geowiki uses a rather onerous produc-
tion process to convert the GPS tracks into a base
map, recent initiatives build on the MapServer
Open Source WebGIS package to serve online
maps on the Web. The MapServer was set up in
compliance with standards from the OGC, par-
ticularly the Web feature server with transactional
extension (WFS-T) to support not only serving
maps on the Web, but also to edit the maps re-
motely. The London Free Map (http://uo.space.
frot.org/freemap/) initiative finds pedestrians,
bikers, and skaters in London volunteering to
collect GPS tracks. Of course, these data are dif-
ficult to integrate with the road networks for car
navigation, as these tracks are not representative
for cars. However, as personal navigation becomes
more important, it is useful to have navigation
systems that include cycle tracks and walk paths.
The London Free Map cooperates closely with
the Open Street Map (http://www.openstreetmap.
org) initiative. Steve Coast bikes around London
carrying a GPS device (Dodson, 2005). Coming
home, he downloads the GPS tracks onto his lap-
top to subsequently make them available online
to share the tracks with others. Street by street, a
copyright-free map of London is built.
Editing Waypoint Data
This step involves the actual creation of the base
map from the raw GPS tracks previously collected.
First, it has to be decided which tracks or individual
GPS points to consider for inclusion in this step.
Next, geometric and statistical algorithms have
to be applied to decide when several very close
lines are actually the same line and thus need to be
averaged together, or are actual parallel segments.
1237

Intersections have to be modelled into junctions at
which tracks have to be split in segments to ensure
topology. Other points may belong to traffic situ-
ations such as squares, roundabouts, and parks.
Quantitative mechanisms relying on geometric
algorithms, or qualitative mechanisms requiring
human intervention, may decide which GPS tracks
are relevant at this stage in the process. Geometric
algorithms may filter out GPS points that have
a significantly different accuracy from the rest.
This may be important in areas with bad satellite
reception such as urban canyons, or at the start of
a GPS track due to cold start: sometimes it takes
rather long for a GPS device to establish con-
nection with an appropriate number of satellites.
Equally, GPS points, tracks, or repositories can
be included or excluded from the GPS tracks that
are the input for the base map based on qualita-
tive mechanisms (Shorter, 2005). There may be
a hierarchy of users, with trusted users given a
higher moderation status based on the quality of
their contribution to the community, to review con-
tributions. To ensure community members remain
interested and help reviewing, RSS feeds of latest
updates and additional repositories of GPS data
can be made available to trigger people to help
moderate the GPS tracks. Moderators and com-
munity members can discuss among themselves
about the value of GPS repositories on discussion
forums that are provided at a central server, or that
can be left as comments on distributed repositories
and be amalgamated using RSS readers.
As a save approach, GPS data is not avail-
able for postproduction editing unless one com-
munity member has reviewed the contribution.
The strength of collaborative mapping is that
relationships within the community can be made
explicit. Therefore, members may only want to
include GPS tracks and landmarks in their base
map that were left by people within their personal
network of friends. Finally, previous moderations
of the GPS data remain available for revision of
the individual member, and they can roll back
moderations on request, just like a wiki.
1238
Collaborative Mapping and GIS
Edelkamp and Schrödl describe a “map refine-
ment process” to create a traversable road network
from GPS tracks (Edelkamp & Schrödl, 2003).
In the context of normalising and synthesising
GPS tracks, the road segment clustering stage
and the road centreline generation stage of the
map refinement process are relevant. The road
segment clustering stage identifies segments that
are common between different GPS tracks and
locates intersections of these segments, based on
the contiguity information and temporal order of
the GPS points. The road centreline generation
creates a geometric construct that captures the road
geometry. The centreline is generated from a set
of GPS points using a weighted least-squares fit.
The relative weight of each GPS point is derived
from the measurement error, which is usually
available from the GPS receiver.
Annotation of Geometry with
Real-World Semantics
This step ensures the geometry of the network
makes sense for people that use the network to
locate themselves and navigate around the area.
The driving restrictions at one-way streets and
turning restrictions at the intersection of segments
have to be added by ground truthing: surveying
the actual location to interpret the data correctly.
Surveyors go out with a printed base map to an-
notate and upload annotations at a later time, or
may well carry PDAs connected to a WFS-T map
server and add annotations on the fly.
The Downhill Map of Bristol (http://locate.ira-
tional.org/bristol_map/) is a map for skateboard-
ers, cyclists, and walkers. Artist Heath Bunting
and the Bristol Skate Survey group surveyed the
streets of Bristol, the road surface quality, and the
locations of fruit trees. Like Geowiki, the Downhill
Maps of Bristol incorporates landmarks along the
street network created. This initiative not only
simply records the longitude and latitude, but
also maps the attributes of the map object such
as height and road surface quality. From this data,
Collaborative Mapping and GIS
a functional online route planning service can be
created for cyclists, skateboarders, and skaters,
which no national mapping agency would be able
to offer (Albert, 2004). The Dutch GeoSkating
project (http://www.geoskating.com/) aims to cre-
ate a similar map of the outskirts of Amsterdam,
uploading the GPS tracks and notes on road surface
quality to a Web server in real time.
Not only navigation attributes such as road
type, turn restrictions, and one-way streets are col-
lected in this step, but also toponymy is important
here: the geographic names of roads, streets, and
landmarks. In the context of collaborative mapping
projects, geographic names are collected using
the notion of “consensus-based location,” a term
coined by Alasdair Turner at the PLAN meeting
in London, early 2005. The name of a location is
established through consensus among a group of
people. It is not determined a priori, but evolves
as people in a community add new annotations in
response to the needs of the moment.
For example, in the GeoNotes service (Espi-
noza et al., 2001), a “location tag” determines the
exact location of an annotation in the service in
case the accuracy of the positioning technology of
either the creator or the reader is too coarse. Thus,
other users can still verify the correct location of
the note without having to rely on the accuracy
of the positioning technology. Furthermore, the
positional accuracy of the location tag may evolve
with improving positioning technology, and de-
pending on what labels are defined.
The communal character is emphasised as
these location tags are shared in the community.
The set of location tags is stored in the service for
other people’s perusal. Upon tagging a note with
a location, other notes in the vicinity are scanned
for their location tags, and presented as a list of
options for location tags, sorted according to
popularity. The creator selects a tag from the list
or creates a new location tag. This service may be
bootstrapped with predefined labels taken from
gazetteers, the yellow pages, or tourist databases.
This strategy not only applies to point loca-
tions, but to streets segments and areas as well.
The Neighborhood Project (http://hood.theory.
org/) uses the TIGER/Line US Census data for
mapping city neighbourhoods based on the col-
lective opinions of Internet users. Addresses and
neighbourhood data collected from housing posts
on Craigslist are translated into latitude and lon-
gitude values and drawn on the map. The blobs
on the map thus give a notion of the spatial extent
of neighbourhoods derived from the collective
mental map of Internet users.
Distribution
Once the geometry of the street network and the
locations of landmarks have been created, they
have to be made available to a wide audience,
without copyright or financial restrictions. In the
previous steps, various projects were highlighted,
with the aim to build geographic databases along.
While they may coexist to make up a distributed
repository of GPS tracks, it may be advisable
to cooperate, to ensure GPS tracks from these
repositories can be easily incorporated into any
of the other repositories,and base maps can be
exchanged without any problem. Therefore, it
is advisable for these projects to work together,
but also to transcend boundaries and work with
other initiatives that may benefit from geographic
data, for example, civic initiatives. Already, the
London Free Map, Geowiki, and OpenStreetMap
are cooperating to build a street network for
London together.
There are various sets of geographic data cre-
ated as part of a collaborative mapping project. The
scanned and georeferenced base maps, the global
satellite imagery, incorporated as background,
can be made available, as well as the geographic
vector data. For example, the London Free Map
initiative hosts various BitTorrents to distribute
free geographic datasets in a cost effective way.
Also, the individual GPS tracks collected during
the project can be made available as GPX files, a
1239

lightweight XML grammar to express GPS data,
or encoded using the geography markup language
(GML) published online as a Web feature service
(WFS). Also, for applications that do not require
the original vectors, raster images of the street
network can be published as a Web coverage
service (WCS) that can be used as tiles to plot
other data on top of.
Challenges and Opportunities
Sawicki and Craig (1996) identify two contra-
dictory currents in information technology that
equally apply to collaborative mapping services.
On the one hand, there is a movement towards
more accessibility by more users: online mapping
has experienced a steady uptake, and more and
more people turn to Web maps to find their way in
the real world. Also, mobile phones and PDAs are
becoming new platforms for delivering geographic
information. Since the mid-1990s, more than 50%
of the citizens of the developed countries have
become owners of mobile phones, while mobile
communication has also been remarkably suc-
cessful in poorer countries (Ahas & Mark, 2005).
On the other hand, there is a movement towards
less accessibility. Although more users have ac-
cess to more data and information, the entry cost
of the education, training, capital cost, time, and
experience of the users to take advantage of in-
formation technology is becoming much greater
(Sawicki & Craig, 1996). As an example of this
movement, Ahas and Mark (2005) write that the
uptake of LBS has slowed down, as the services
proved too difficult for the public to obtain, and
users have difficulties imagining the potential of
spatial information. For collaborative mapping
initiatives to become successful, it is advisable
to start off with the lowest common technical
denominator, but to design and develop architec-
tures to cater for more advanced technology in the
future as location-aware devices become cheaper.
Already, projects such as the London Free Map
use Web-based open source GIS technology and
1240
Collaborative Mapping and GIS
cheap GPS-devices rather than location-aware
mobile technology.
Since the information value and pertinence
of collaborative mapping services is character-
ised not only by physical proximity, but also by
social proximity, participation becomes another
challenge when only a small group has access
to location-aware, mobile devices. Especially in
the case of applying collaborative mapping as a
form of community-based participatory planning
(Harris et al., 1996) or as a means to obtain data
for the social positioning method as input for
urban planning (Ahas & Mark, 2005), one has to
be aware that it is only based on the space-time
behaviour and mental map of only a small group
within society, with specific social identifica-
tions. The dependency of collaborative mapping
services on social proximity may well expect a
wider participation, as indicated by the popularity
of online social networking tools, thus possibly
pushing the uptake of LBS in general.
Current collaborative mapping initiatives rely
on GPS-devices to capture road network geometry
and open source GIS technology to serve maps
and edit the raw data. The ubiquity of location-
aware devices will increase over the next few
years. Not only will the pool of possible volunteers
to participate in collaborative mapping projects
therefore expand accordingly, but the collabora-
tive mapping process will also benefit from these
technological advances as more raw positional
data is being captured with increasing accuracy
and in real time (Ahas & Mark, 2005). Finally,
the distributed nature will be reinforced, as more
collaborative mapping services will derive spatial
annotations from more repositories, and will have
to take into account wider social networks when
establishing the pertinence of these annotations.
FUTURE TRENDS
Ben Russell (2003) suggests a “street server” as
a vehicle for distributed collaborative mapping.
Collaborative Mapping and GIS
Street servers can be thought of as a computer
network with both private and public functions
for a community of residents of a street. In the
context of collaborative mapping, these street
servers operate as a GPS tracks database that can
be linked with other street servers to form a map
covering a wider area, as far as GPS tracks match
up. Thus, mobile devices ask stationary street
servers about their position and how to travel
around the neighborhood. This network approach
contrasts with the current practice of serving GPS
tracks from a single, centralized Web server.
As geographic information is created collab-
oratively, people upload their annotations to a
central repository, or store their annotations and
tracks on their personal Web sites, thus creating
a distributed framework for sharing geographic
information. With increasing storage capacity of
mobile devices, people may even put their GPS
tracks on their location-aware mobile devices and
share the data with their peers in their social net-
work, just as digital music files are shared online
or made available as podcasts. Ed Parsons, CTO
of the Ordnance Survey, spoke in an interview
about the “napsterisation” of geographic informa-
tion, claiming that “the experience of the music
industry may also apply to geographic informa-
tion” (Westell, 2003).
Thus, as people roam around the city, their
mobile devices pick up blocks of geographic data
to locate themselves and navigate the area sur-
rounding them. That geographic data represents
the street network and landmarks of the city blocks
where they find themselves, as derived from col-
laboratively collected GPS tracks.
CONCLUSION
Collaborative mapping is the latest development
in the democratisation of geographic informa-
tion. The proliferation of positioning technology,
location-awareness, online communities, and open
source GIS technology have paved the way for
people being able to contribute geographic data
and access localised information. Various collab-
orative mapping projects are currently collecting
GPS tracks to derive street network data. The
framework of steps described in this chapter was
based on experiences from these projects, and will
stimulate the exchange of geographic information
among these projects. Hopefully, collaborative
mapping will become a complimentary input for
corporate location-based services.
REFERENCES
Ahas, R., & Mark, Ü. (2005). Location-based
services – New challenges for planning and
public administration? Futures, 37(6), 547–561.
doi:10.1016/j.futures.2004.10.012
Albert, S. (2004). Crossing the line between
mapping and mapmaking: Artistic approaches to
cartography. Society of Cartographers Bulletin,
38(1-2), 41–44.
Bleecker, J. (2005). A design approach for the geo-
spatial Web. Retrieved February 2006, from http://
www.oreillynet.com/pub/a/network/2005/06/07/
geospatialweb.html
Buchwalter, C. (2005). Integrated interactive
marketing:Quantifying the evolution of online
engagement. Retrieved from http://www.nielsen-
netratings.com/pr/wp_050401.pdf
Dodson, S. (2005, April 7). Get mapping. The
Guardian. Retrieved from http://www.guardian.
co.uk/online/story/0,3605,1453293,00.html
Edelkamp, S., & Schrödl, S. (2003). Route plan-
ning and map inference with global positioning
traces. In R. Klein, H. W. Six, & L. Wegner (Eds.),
Lecure notes in computer science 2598. Com-
puter science in perspective: Essays dedicated to
Thomas Ottmann (pp. 128-151). Berlin: Springer.
1241
 Collaborative Mapping and GIS

Espinoza, F., Persson, P., Sandin, A., & Nyström, Nielsen//NetRatings. (2002) Seven million Eu-
H. Cacciatore. E., & Bylund, M. (2001, September ropeans look to the Web for maps, guides &
30-October 2). GeoNotes: Social and navigational directions. Retrieved from http://www.nielsen-
aspects of location-based information systems. In netratings.com/pr/pr_020920_uk.pdf
Abowd, Brumitt, & Shafer (Eds.), Ubicomp 2001:
Plewe, B. (1997). GIS online: Information re-
Ubiquitous computing, International Conference,
trieval, mapping, and the Internet. Santa Fe:
Atlanta, Georgia (pp. 2-17). Berlin: Springer.
OnWord Press.
Harris, T. M., Weiner, D., Warner, T. A., &
Russell, B. (2003). Street servers – An outline of an
Levin, R. (1996). Pursuing social goals through
open source networked neigbourhood. Retrieved
particatory geographic information systems. Re-
from http://www.headmap.com/street-servers/
dressing South Africa’s historical political ecol-
street-servers.htm
ogy. In J. Pickles (Ed.), Ground truth. The social
implications of geographic information systems Sawicki, D. S., & Craig, W. J. (1996). The democ-
(pp. 196-222). New York: The Guildford Press. ratisation of data: Bridging the gap for community
groups. Journal of the American Planning As-
Mac Gillavry, E. (2004). Collaborative mapping:
sociation. American Planning Association, 62(4),
By the people, for the people. Society of Cartog-
512–523. doi:10.1080/01944369608975715
raphers Bulletin, 37(2), 43–45.
Shorter, C. (2005). Bicycle map builder white
McClellan, J. (2003, March 27). Get caught
paper. Retrieved from http://bikemap.openearth.
mapping. The Guardian. Retrieved from
com.au/docs/whitepaper/
http://technology.guardian.co.uk.online/sto-
ry/0,3605,922337,00.html Walsh, J. (2005). London free map. Retrieved from
http://uo.space.frot.org/?node=LondonFreeMap
McHaffie, P. H. (1995). Manufacturing metaphors:
Public cartography, the market, and democracy. Westell, S. (2003, July/August). The napsterisation
In Pickles, J. (Ed.), Ground truth. The social of geographic information. GI News.
implications of geographic information systems
(pp. 113–129). New York: The Guildford Press.

This work was previously published in Collaborative Geographic Information Systems, edited by S. Balram and S. Dragicevic,
pp. 103-120, copyright 2006 by IGI Publishing (an imprint of IGI Global).

1242
 1243

Chapter 75
Provision of Web 2.0 Services
by Interoperable GIS-
Powered Local Administration
Portal Systems
Anastasios Tsitsanis
National Technical University of Athens, Greece

Sotirios Koussouris
National Technical University of Athens, Greece

Rob Peters
University of Amsterdam, The Netherlands

ABSTRACT
eGovernment for municipalities, a rather mature field when talking about the traditional and plain service
delivery to citizens, is nowadays seeking ways to provide more quality based services to the citizens and
to become more direct in terms of communication and interaction. In this direction, the most modern
implementations are already coming up with the utilisation of various Web 2.0 services in an attempt to
become more attractive to the users and to gain a larger user base. However, the introduction of Web 2.0
application to eGovernment portals does not seem to be the solution on its sole; in contrast, when these
solutions are accompanied with GIS powered technology, participation and utilisation of those portals
seems to reach the anticipated figures, as interoperable GIS systems are able to offer added-value and
more “personalised” services to users.

INTRODUCTION to the actual end users and “clients” of public

During the last years, as a natural follow-up of
the evolution of back office local administration
information systems, significant progress has
been made towards electronic service delivery
DOI: 10.4018/978-1-4666-2038-4.ch075
administrations; the citizens. The modern era of
eBusiness and electronic service delivery, has
touched, apart from the private sector, also the
governmental services, not only at the levels of
central governments, but also at the levels of
regional and public administrations, as the major-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
ity of the by citizens demanded services, lies in
their jurisdiction. Therefore, all over this period,
eGovernment was a keyword in the ICT agenda,
as this new uptrend market was quite open and
populated with billions of potential customers (the
citizens), and huge investments were placed in
this domain, either by private firms or by govern-
ments (van Ark, Melka, Mulder, Timmer, & Ypma,
2003), in order to overcome the main obstacles
which had to do with system’s interoperability,
data harmonization and of course re-engineering
of the public sector’s processes in order to achieve,
along with the maximization of the use of ICT,
lower lead times and higher service rates.
Nowadays, with eGovernment being a rather
mature field, where the issues of the past have
been solved up to an extent, public administrations
and governments in general are seeking ways of
engaging their citizens more in their everyday
operation, in order to transform themselves
from a single service system, to a collaborative
process environment, where citizens are not only
seen as clients, but are also taking decisions or
can pinpoint various issues as a feedback. This
stream is strengthen with the use of various
Web 2.0 technologies, that are able to offer the
advantages of building consensus amongst the
citizens of the same region (Nations). This is far
more empowered with the use of Geographic
Information Systems (GIS), as the representation
of areas is a powerful way to attract citizens to
public administration platforms, as through the
photorealistic representation not only can they use
such platforms better, as they are offered with a
life-like user interface, but they also realize that
the area shown is a part of their everyday life and
therefore they tend to increase their interaction
with the various services offered.
In this context, this chapter presents how GIS
systems can interoperate with already established
public administration portals and with other exist-
ing and globally accepted technologies, in order
to offer to the citizens, but also to the public
officers, various mash-up services that extend
1244
from basic application filling (e.g. pinpoint road
malfunctions) to highly sophisticated services
as public debates, discussions and consultations
(about issues such as spatial planning, green IT,
advertisement, etc) which can take advantage of
GIS technology for presenting the as-is situation,
but also for presenting forecasting models based
on the decisions that will be drawn.
BACKGROUND
The rapid development in Information Technol-
ogy is nowadays opening new horizons regard-
ing the facilitation of everyday life of computer
users. As the Internet becomes a daily activity of
people’s life, more and more organizations tend
to offer internet based services, replacing their
traditional front-desk transactions. Following this
trend, governmental organizations, such as local
governments, local administration and various
public service offices are constantly launching
eGovernment portals that are not only offering
information to the public, but offer electronic
flavoured services as well, by promoting interop-
erability with the present underlying systems that
are currently serving the public (Jaeger, 2003).
Local Administration Entities, such as munici-
palities (Capgemini, 2006), are topping the list of
such organizations, as they have high figures of
everyday transactions numbers with citizens and
also possess a large number of “clients”, which
are naturally the inhabitants and the locally based
enterprises. The benefits that rise from such a
portal are the following (Charalabidis, Askounis,
Gionis, Lampathaki, & Metaxiotis, 2006):
• Alternative service channels for the citi-
zens and enterprises such as Internet,
Mobile phone access and also voice access
with the use of Interactive Voice Response
(IVR) systems.
• Optimization of the service levels as the
on-site presence becomes unnecessary.
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
• Exclusive services for facilitating different
groups such as disabled persons, the elder-
ly, the youth, etc.
• Workflow optimization and automation in
the highest possible degree by standardiz-
ing processes and documents
• Effective cooperation between different in-
house departments of local administration,
by intergrading and interconnecting servic-
es and processes that take part in common
transactions
Nowadays, municipal eGovernment Portals,
which will offer automated services and would
address the public, it is essential to comply with
international standards and system design tech-
niques that should guarantee the end system’s
functionality and feasibility. This section presents
the most important frameworks and standards
that should be followed when designing and
implementing eGovernment Portals (Pavlichev &
Garson, 2004). Various Frameworks are nowadays
present, offering the guidelines that should be fol-
lowed when designing systems and applications
seeking interoperability with underlying systems.
Those frameworks define in detail:
• Certification Frameworks for Public
Services web sites
• Interoperability structures for interconnect-
ing systems and developing applications
• Digital Authentication structures for the
end-users
• Standardization Meta-Data and XML
Schemas for data entities
The most known frameworks are the following:
• The UK Electronic Government
Interoperability Framework (Cabinet
Office, 2005)
• The German SAGA (Bundesministerium
des Innern, 2003)
• The European Interoperability Framework
(IDABC, 2008)
• The Greek eGovernment Interoperability
Framework (Information Society S.A.,
2008)
International Standards and state-of-the-art
Modeling Languages and technologies are used
in eGovernment portals as they preserve the fea-
sibility, the accessibility, the accessibility and the
security of the end product which is the portal.
Service provision can be dived into four differ-
ent levels, regarding the completion rate of each
service based on electronic means. Those levels,
as also adopted by the European Commission
(IDABC, 2006) are the following (Layne & Lee,
2001):
• Level 1-Information: This level contains
only information about the corresponding
service.
• Level 2-One-way interaction: A 2nd level
service provides to the end user download-
able material, such as.pdf forms or similar
documentations which have to be filled in
by the user and handed to the correspond-
ing office.
• Level 3-Two way interaction: This level
provides on line tools where a user can fill
in his request and initiate the process of the
transaction. For completing the transac-
tion, the end user must appear at the ser-
vice office and collect his receipt. Services
offered in level 3 require the authentication
of the user.
• Level 4-Transaction: Services in the 4th
level are fully automated and the end user
gets the service’s output in electronic for-
mat, after imitating the transaction. Such
services include steps as authentication,
decision, notification and delivery of
receipt.
1245
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
System Architecture
Most eGovernment portal architectures are based
on n-tier structures (data layer, application layers,
presentation layer). This particular approach al-
lows the scalability of modules according to the
portal needs, as the portal size and its functional
requirements will guide the total implementation.
Moreover, the sorting of operations in distinc-
tive levels in order to avoid unequal burdens of
particular sources, or of the whole system is also
guaranteed by the 3-tier architecture chosen and
last but not least this architecture offers the option
of selective expandability with no changes in the
system’s infrastructure.
Figure 1 depicts a conceptualisation of the
logical architecture model (Koussouris, Tsitsanis,
Gionis, & Psarras, 2008) of a system containing
parameterisable Common Off-the Self Com-
ponents (COTS), open source components and
Figure 1. Logical architecture of the system
1246
be-spoke components. The core platform is a
Content Management Platform System (CMS).
This system handles the presentation of the in-
formation and offers out-of-the-box tools for the
implementation of services belonging to levels 1
and 2. Other systems are GIS systems, Workflow
Engines, Citizen Relationship Management Sys-
tems and IVR Systems for enabling voice access.
The Content Management Platform the Citizen
Relationship Management System (CRM) and
the Workflow Engine all cooperate as the core
transactional components of the system. The CRM
serves the user authorization and identification
and tracks down all user activities, namely from
simple queries or questions asked, to the current
status of an online submitted application. This
way, the end-user is constantly aware of his opens
issues and on the other hand, the administration
authority is able to generate the end-users profile
in order to target the most needed services, an
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems

important issue in eGovernment portals (Medja- proach. Those interfaces are connected with the
hed, Rezgui, Bouguettaya, & Ouzzani, 2003). wrappers that facilitate the information flow to
Finally, the Workflow Engine offers the flex- and from the portal with the use of Web Services
ibility of adding, replacing and updating working (Soon, Vijayalakshmi, & Nabil, 2002). This ap-
processes, without requiring great code-writing proach enables the interconnection of the different
efforts. This way the process flow is constantly subsystems and guarantees the high performance
managed and the system guarantees the flow of as only the required interfaces are used. In order
documents to the appropriate users even at heavy to implement this architecture, the following steps
loads, surpassing the operation of manual systems are needed:
in Local Administration (Vassilakis, Laskaridis,
Lepoura, Rouvas, & Georgiadis, 2003). • Discover the inputs and the outputs of the
back-office systems
Interoperability Layer • Modeling the data that is transferred within
the system using XML
Speaking of interoperable platforms, mechanisms • Protocol and Communication channels de-
are implied that will facilitate the transformation, velopment. This refers to the wrappers and
mapping and transportation of information data the web services development by defining
packets between the various systems in order to the communication ways with the back-
enable their direct, errorless and flawless commu- office systems (.Net Calls, RPC Calls,
nication and collaboration. Therefore, concepts as Intermediate Tables, Direct DB Calls) and
interoperability layers are essential for eGovern- the portal (XML Schemas, Service Calls)
ment Portals, if services of level 3 and 4 are to be • Definition of workflow and application
offered to the public. Such a portal will operate calls. The application call can either be
as a front-end interface for the internet users, in triggered by the portal (in case of a request
order to be served by the Local Administration submission) where the portal is initiating
back-office systems and among them the GIS
services that are offered though them. As those
systems are of different technologies and of dif- Figure 2. Interoperability layer architecture
ferent generation, interoperability is the key which
adds value to the eGovernment portal by achieving
the interconnection and the cooperation between
heterogeneous systems (Koussouris, Charalabidis,
Gionis, Tsitsanis, & John Psarras, 2007).
An interoperability layer is designed in such a
way that future enhancements are possible and that
system and platform independence is preserved.
It contains “Encapsulation Software Components
(Wrappers)” that are responsible for the data
transportation between the Back Office systems
and the Portal, through specific interfaces.
As depicted in Figure 2, from each back-office
system only the required input and output inter-
faces, that are brought to life during a transaction,
are selected in a purely “follow the service” ap-

1247
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
the call and waits for a reply but can also
be trigger by the back-office system itself
(in case of a notification for a fee payment)
• Development of Security and
Authentication mechanisms
GIS POWERED WEB 2.0 SERVICES
FOR LOCAL ADMINISTRATION
In the recent year, the trend of Web 2.0 brought
many new applications in the surface, which at
first glance looked like fancy implementations
that targeted mostly the entertainment factor of
the internet. Those were web based GIS imple-
mentation like Google Maps or Live Maps, social
networks like Facebook, MSN, and various other
plain services or mash-ups that became popular
due to the ever-growing user bases of them. This
fact got the attention of many governmental or-
ganisations, such as public administrations, which
slowly started to adopt such solutions in the search
of establishing a more direct and straightforward
communication with the citizens and offering the
advanced tools that are able to increase their online
experience and satisfy their user needs.
More precisely, in the case of public admin-
istrations one can see that WEB 2.0 service are
constantly used as they can easily not only satisfy
the needs of the direct information flow and the
bi-directional communication between the citizens
and the administration, but are able to facilitate
the establishment of virtual online communities,
quite closely to the idea of a neighbourhood, where
citizens are granted the possibility to interact with
each other, express their opinions publicly and
in more general terms operate within the virtual
boundaries of these communities similarly to
the “agora” situations that existed in past direct-
democracy political environments. All the above
mentioned capabilities spring of course directly
of the well-know advantages of the modern In-
formation Technology and are also encountered
every day in various on-line initiatives, however
1248
the use of GIS systems and their interoperation
and co-existence with the various Web 2.0 services
adds the winning argument to those systems, as it
manages to literally lift degree of user satisfaction
and user popularity, e.g. having users satisfied
constantly using the system and not letting it die
(Smith, 2007). This winning point if no other than
the visual representation of the region a citizen
lives, a fact which not only makes the citizen more
familiar with the layout that is unfolded in front
of him, but also motivates him to use the system
as he can perceive at once that he is in a position
to call the shots on what is happening in his own
area, his “home”.
Having in mind all the above, public admin-
istrations are starting to pull various attempts
to utilise GIS-powered WEB 2.0 services and
mash-ups in order to satisfy the needs of their
citizens, to manage internal affairs and increase
their productivity and to open-up their “closed”
modus-operandi which is seen a one of the most
important factors contributing to the huge problem
of democratic deficit and of people not taking ac-
tively part in the decision making process, which
lead to misbelieve in politics and politicians as
well. In this context, a list of possibly services
powered by GIS is presented below:
• Interactive eGovernment platforms of-
fering public administration services:
These are systems that interact with citi-
zens for trivial issues such as data form
filling with the use of the exact location as
extracted from the GIS infrastructure but
also for more advanced issues that require
the fastest possible reaction of the munici-
pality’s officials like in cases of damages
in vital infrastructures of the city such as
water supply grid, vehicular grid, sewage
system as well as issues that deal with the
citizens’ security which maybe be in the ju-
risdiction of Municipal Police.
• Traffic Management System: Systems
using GIS and telematics infrastructures
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
in order to advise drivers about the fastest
and optimum routes throughout the border
of the city, targeting the highest possible
convenience of drivers and the protection
of the environment through the reduce
of CO2 emitted by the vehicles that pass
through the city.
• GIS services for internal general admin-
istration use: Those systems are able to
operate using multiple layers containing
geographical information of many kinds
such as: fire zones, earthquake zones, spa-
tial data, building data, digital background
of the area, satellite photos and legal data
connected to several points of the GIS map.
• Green Areas Management System:
These systems give the opportunity to the
city’s officials to manage in an efficient
way the green areas of the city and include
them in their spatial planning programs
(creation of new green areas (e.g. parks) in
the city, recreation and exploitation of un-
tended city locations, etc). These systems
are based on GIS solutions able to map in
detail the present situation in a public ad-
ministration and offers the pubic servants
an in-a-glance overview of how green ar-
eas are allocated and how they should be
managed. Moreover, these systems can be
accompanied by sophisticated artificial in-
telligence forecasting algorithms in order
to visualise what if scenarios based on the
data that correspond to each region.
• Crisis management systems: These refer
to systems utilized as high-level manage-
ment tools that are able to elaborate several
civil protection scenarios in cases emer-
gencies (fire, earthquake, flood, ground
frost and broiling weather) for achieving
the highest possible degree of protection of
the citizens in a fast and efficient way.
• eDemocracy GIS powered platforms:
Those are platforms targeting the enhance-
ment of citizens’ participation to the deci-
sion making process and make use of Web
2.0 technologies and tools (forums, social
networks, information provision, RSS
feeds, audio and video streaming tools,
legislation data, ePetitions, ePolls) in order
to give citizens the opportunity to actively
participate in the various steps of the local
decision making processes and make their
voice heard by the city’s officials.
• Entrepreneurship and general informa-
tion portals powered by GIS technolo-
gies: These systems refer to GIS supported
platforms where enterprises have the op-
portunity to pinpoint their presence to-
gether with their services and needs and in
such a way virtual marketplaces are creat-
ed that enable the citizens to have an online
yellow-page map of the business develop-
ments, the services and possibly the career
opportunities in their region.
The above mentioned list is of course end-
less, as modern technology is able to adopt GIS
solutions in almost any case, and the advantages
of such breeding are quite visible from the few
examples discussed before, as users are treated
with a familiar environment and this contributes
greatly to their willingness to utilise the systems
at a higher rate. Based on this discussion, the next
sections of the chapter is devoted to the show-
case of a real-life application of a GIS powered
portal that was implemented in the Netherlands
in order to help public administration to engage
stakeholder in the spatial planning process of the
region of Flevoland.
Using GIS-Powered Portals
for Conflict Resolutions and
Public Consultation
System Architecture and Features
Legal Atlas is a software application which in
the first steps of prototyping included navigation
1249
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
techniques that helped to navigate through four
layers of laws and regulations: European, national,
regional and municipal.
Early results from using and demonstrating
the system in the Netherlands (Wilson & Peters,
2004) showed not only high user satisfaction but
also gave the confidence that this type of solution
would be feasible in other Europe’s countries. As
a natural consequence, a system named Legal At-
las II was implemented. The Legal Atlas systems
should enable the ultimate transparency of local
regulations in the sense that the citizen could
ask “where can I do this?”, instead of: “can I do
this in place such and such?” (Wilson & Peters,
2004). Such an application consequently requires
a unique coding for all local and regional legal
functionalities affecting citizens and a mechanism
for semantically mapping (Boer, Hoekstra, Win-
kels, van Engers, & Willaert, 2002) between the
legal terms and legal procedures affecting those
required functionalities of citizens. In that way the
client of the service could view his case in the light
of aggregated data and monitor the consequences
of decisions and opposition. This brings a new
perspective on maintainability and consistency
of the legal system as a whole. These are of great
importance, since regulations will change over
time. The Legal Atlas makes use of the IMRO
(information model spatial planning) XML/GML
(Open Geospacial Consortium) coding scheme
to enable the user to query the underlying legal
system using very detailed coloured maps repre-
senting legal constraints and possibilities (supports
explanation and understanding by users).
The succeeding Legal Atlas II application was
based upon a (Winkels & Hupkes, 2007) open
and extensible architecture using semantic web
technologies and was designed using the expe-
riences from the first version. The codes were
translated to GML syntax rules which helps to
create a more flexible and maintainable codifica-
tion system. The main challenge was to extend
the retrieval possibilities of law from text to map,
in other words; by clicking on a regulation the
1250
user would be able to identify all objects or map
area’s where this regulation applies by means of
a visualisation mechanism.
While this may seem a small extension of the
first version, the reader should realise that there
is a big difference between the question: “what
will be allowed here? and: “where will this be
allowed?”. The second inverse question demands
a complete semantic definition of the world and
complete legal coverage (Boer, van Engers,
Peters, & Winkels, 2007) of that world as previ-
ously described. With additional legal reasoning
(Deontic) arithmetic, such a functionality (Boer,
van Engers, & Winkels, Using Ontologies for
Comparing and Harmonizing Legislation, 2003)
enabled questions like: return all contours on the
map where regulation “X” and “Y” are both ap-
plicable, and with exception of rule “Z”.
After some prototype evaluation actions, a
real-life prototype was implemented using IMRO
in combination with the GEMET (“GEneral Mul-
tilingual Environmental Thesaurus”). GEMET is
increasingly relevant because of INSPIRE. The
regional layers of data and most of the content
were provided by The Province of Flevoland
who maintains a number of databases (Thematic
Spatial Data Infrastructure) about the economy, the
environment, our cultural heritage and the public
infrastructure. Legal Atlas III is the evolution of
the above named systems and for the purpose of
European transferability Google maps technology
is used next to GIS technology to support “contour
editing” by visitors of the website. The visitor can
indicate an area by drawing a line on the map to
discuss and to support his case with documents,
research reports, legislation and plans that can be
uploaded, stored and downloaded for that drawn
area. The interface of the prototype shown above
is linked to the database of the regional authority
Flevoland. All map layers available can be pro-
jected on this Google map screen. This way the
conflict resolution system will maintain dynamic
input from legal constraints and map layer updates.
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
Regarding the conflict annotation engine, it
was mostly based on Semantic Web Technology
(Boer, van Engers, Peters, & Winkels, 2007).
The knowledge models about the Cross Border
Deliberation models and the domain knowledge
of the working scenarios are all described with
RDF/OWL (W3C, 2004). The reason for this
choice was that RDF/OWL is able to inference
and reasoning with these models. To publish this
information as a service the OpenRDF Sesame
server (OpenRDF.org) is used, which has a web
accessible SPARQL-endpoint which is an access
point to which SPARQL (W3C, 2008) queries
can be sent. The RDF that is stored within the
OpenRDF Sesame server is processed with
OWLIM (Ontotext AD, 2009), a high-performance
semantic repository which is packaged as a Stor-
age and Inference Layer (SAIL) for the Sesame
RDF database. It reasons about the RDF data and
propagates this by means of rule-entailment. The
SPARQL endpoint is used to fill up the Legal Atlas
III with information. The Legal Atlas III can be
seen as an interface for the OpenRDF Sesame
server, and the SPARQL-endpoint is the interface
between them.
The SPARQL queries are based on the sche-
mata of the RDF/OWL models. This means that
they are independent on the actual content. This
ensures that different content should be anno-
tated with the RDF/OWL models to makes sure
that the SPARQL queries are able to retrieve the
content. Here is a simple query that retrieves all
the concepts with which content can be tagged
with. The same concepts will also be used for
searching the OpenRDF Sesame server for the
appropriate content.
SELECT ?keyword ?label ?scheme
WHERE {
?keyword a skos:Concept.
OPTIONAL {
?keyword rdfs:label ?label.
FILTER (lang(?label) = ‘nl’)
} modelling and constraint analyses.
OPTIONAL {
?keyword skos:inScheme ?scheme.
}
}
The return is a gigantic list of all the concepts
that can be used for annotation. Such a huge list
might not be convenient in a user interface. There-
fore it is better to tone this list a bit down with a
practical list. Pruning this list down to a domain
is one way to limit the amount of concepts. The
following example shows the SPARQL query for
a specific domain, namely the IMRO2006 SKOS
(Geonovum, 2006) vocabulary.
SELECT ?keyword ?label
WHERE {
?keyword a skos:Concept.
OPTIONAL {
?keyword rdfs:label ?label.
FILTER (lang(?label) = ‘nl’)
}
?keyword skos:inScheme <http://
www.ravi.nl/imro2006/skos#imro2006>.
}
Figure 3 describes the architecture of the
interaction between the municipal portal and
the Annotation engine. The portal containing
Legal Atlas gets its information by sending the
appropriate queries to the SPARQL Endpoint.
The SPARQL endpoint is an access point to the
OpenRDF Sesame server. The OWLIM is an infer-
ence layer within the OpenRDF Sesame server.
This is filled with the Cross Border Deliberation
models and the domain models and the content
that is part of the public administration portal. The
Cross border deliberation models are schematic
representations of the decision process around a
theme like spatial planning or enforcement agree-
ments. The standardised schema’s are required to
facilitate cross border negotiations and to transfer
a formal decision process into (automated) legal
1251
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems

Figure 3. Architecture of the annotation engine of a map based conflict resolution portal

The portal also provides an interface for add-
ing data. Which adds RDF/OWL to the models.
These are then processed by the OWLIM module
and the result is propagated to the Portal by the
same SPARQL endpoint.
An important part of the interface is the geo-
graphical interface with Google Earth and Google
maps. The connection between the geographical
data and the metadata that describes them can
be seen in Figure 4. The Map has a number of
different regions which are of some type of land
use. These regions have a particular geographical
representations in the form of layers with polygons.
Problems with INSPIRE legal Mapping
for Dutch regional authorities
The approach of the example of using a GIS
powered portal in Flevoland is based upon stan-
dardized legal knowledge representations and
maps and allows the stakeholders involved to cope
with legal pluralism and the complexity caused
by the interaction between the building blocks
of European laws and local regulations. In this
context, this implementation helps users to have
access to the regulations relevant to the problem
at hand, in a concise and comprehensible way.
The map-based interface providing access to the
applicable regulations offers a powerful mean to
ensure usability and client satisfaction. Before

Figure 4. The interaction between the maps, its regions with particular landuse and the geographical
representation

1252
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
explaining in more detail the various steps, it
should be stressed that although most of the tests
have been conducted in the domain of spatial
planning our approach is also applicable in other
legal domains. All laws have a jurisdiction and are
consequently location bound. While a legislative
maps approach has advantages in the domain of
spatial planning it is not limited to that domain.
Spatial planning is an area where many differ-
ent interests have to be weighed and balanced,
like many other area’s. When proposing realistic
solutions one should also take into account the
massive volumes of existing regulations putting
constraints on the possible alternative solutions
to the problem at hand.
INSPIRE (INSPIRE, 2007) is the European
wide standardisation effort to model the envi-
ronmental policies and vulnerable area’s in such
a way that monitoring and enforcement will
become much easier. The goal is to meet Kyoto
targets and the European parliament is serious
about it. The enforcement is likely to spark a
range of whole range law suits, since all NGO’s
will have immediate access to any building effort
or economic activity that threatens vulnerable
ecologies. It will be hard to change anything in
Spatial Europe without serious deliberations. Re-
garding INSPIRE (EUROSTAT & DGENV, K.U.
Leuven, 2003), the Dutch Provinces have carried
out an impact analysis, which mentions several
ambiguities with the new infrastructure, starting
with the definition of “Protected sites” according
to the International Union for the Conservation of
Nature: “A Protected Site is an area of land and/
or sea especially dedicated to the protection and
maintenance of biological diversity, and of natural
and associated cultural resources, and managed
through legal or other effective means”.
The scope of this definition does raise a dis-
cussion among policy advisors how to apply the
INSPIRE directive wisely. If too many “protected
sites” are defined it tends to weaken enforcement
possibilities and reporting to Europe may become
tedious. If too few area’s are being identified as
protected the Regional authority will lose those
“sites” it wants to protect. The latter occurs be-
cause of the hierarchical legal effect of European
directives. “Lesser” legal regimes become less
useful to oppose against economic interests. The
Ecological main infrastructure in the Netherlands,
that was meant to become the vehicle to ensure
less fragmentation of green area’s in an urban
country, may well become negligible in it’s effect
given the greater impact of Natura2000 area’s.
The Dutch provinces have several examples of
non-intended effects. The regulation of salt and
sweat water balance, for example, indicated that
one Province was obliged to subsidise farmers for
lack of sweat water while those farmers had one
of the best crops in Europe. The Dutch regions
do not want to lose this need for flexibility due
to inflexible standardisation. These findings are
relevant for our research on Legal Atlas.
It appears that many times the visualisation of
map layers for environmental purposes, like with
INSPIRE, creates a legal challenge; the feeding
time of Goose in the Netherlands is six months at
most. Then they fly back to the North. The map
showing area’s that are protected for the sake of
those goose should convey the temporal factor
which is unusual in legal constraints mapping.
The biggest problem with European standar-
disation efforts like INSPIRE for Dutch regions
are those that concern the harmonisation of cal-
culating methods behind the layers. The Dutch
regional authorities are not sure that all envi-
ronmental indicators are calculated in the same
way in Europe. This envisioned level of variety
requires more explanation, flexibility and enough
administrative room for local interpretation. Even
within the same country it takes time and effort to
understand the different ways to calculate sound
level measurements, visualisations and mappings.
Water quality differs with temperature. Ecological
complexity defies simple arithmetic’s. There is a
relationship between legislation and methods of
calculation; changes in calculation methods at
the European level may cause changes in Dutch
1253
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
and regional legislation. INSPIRE may well be
one of the directives with the highest impact on
regional authorities.
Conflict Resolution and the Use
Case INSPIRE in Flevoland
For the real life trial of Flevoland, it was chosen
to take the example of INPIRE annex III layer
11 and 18: areas with legal constraints, regulated
areas and habitat/Natura2000 areas. Figure 5 de-
picts the Natura2000 map layer as it is now part
of the information infrastructure of the region.
The map is represented by the vertical stripes
covering the entire Markermeer and Ijssel Lake.
In reality this is not the case. The Lake Marker-
meer requires numerous ecological measurements
to save the area from further deterioration. The
water management is suffering for slib that cre-
ates uninhabitable murky waters. Everyone agrees
something has to be done. At the same time the
already stressed ecological balance of the region
is even more threatened by the obligation to host
another 60.000 houses with accompanying rec-
reational and transport infrastructure. Calling the
region “locked under Natura2000” as indicated by
the map layer is a simplification that would bring
Figure 5. Example of Inspire type map layers
1254
people into the court room needlessly and too early.
The groups involved in protection of birds and
those involved in building the infrastructure have
been in court for the last 20 years. The chairman
of the bird protection stated that the courtroom
does not help “his” food chain to develop in a
natural way. Pressure groups require more insight
in each other’s negotiation space. Another reason
to have dynamic maps fuelled by dynamic con-
straint analyses is the fact that no protected area
is fixed in the way some maps tend to indicate.
INSPIRE has an explicit limitation to its scope
as support mechanism for transparency: it states
in article 13,1: “information services should not
be published by means of INSPIRE if this action
jeopardizes the very environment it tries to pro-
tect. We argue here that map-layers represented
in static ways powered by static legal- and policy
constraints do exactly that. They do not support
conflict resolution and negotiation but rather sug-
gest more inflexibility and legal rigidity.
Feed back on the tests with the Flevoland
regional development plan prove that questions
like: return all contours on the map that fulfils the
legal constraints “X” and “Y”, but not “Z” are
indeed answerable. How does one provide such
type of opportunity finding for the user in a mean-
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
ingful representation?. The answer may have been
developed already by Simcity game developers
who created manoeuvrability using a cursor and
“tiles” with fixed business rules in 1985 when
Simcity was still called “Micropolis” (Koster &
Wright, 2004). Simcity allowed the virtual
Mayor of a city to plan different types of objects
on a tiled map. The objects could be hotels, in-
dustry, infrastructure or housing, for example.
The “Mayor” was informed if his planned object
did fit with the designated area because those tiles
would lit up in red or green if the business rules
defining those tiles in the database would match
the object dragged by the Mayor ‘s cursor. A
housing block could not be set on a highway.
The research in 2009 indicated that the most
ergonomical lay-out for a use case would be, that
the user drags the cursor (Blue, for op sailing area)
Figure 6. Simcity approach to conflict resolution and opportunity finding
and the surrounding area “lights up” problem
areas (striped red, in this case a 5000 Hectare
swamp that is envisioned to be the ecological
motor for the lake) and areas up for negotiation
(orange). When the cursor is dragged further into
the protected area it turns red to indicate the legal
absolute negative. The dragging turned out to be
crucial since it resembles human negotiations
about proximity and the likelihood of damage
to the ecology in a user friendly way. Dragging
affords negotiations of NGO’s with conflicting
interests in line with Gibsons theory of usability
(Gibson, 1979).
Early Results
The upload and retrieval functionality has been
judged to be part of the Atlas behind the decision
1255
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
process. The Legal annotation engine is currently
being embedded behind the Spatial context (in
Dutch: “Omgevingsplan “) of Flevoland.
The interesting difference between a screen
showing map layers already there and those that
“turn up” while moving a qualified cursor (like
the one seeking space for open sailing area’s) is
the affordance of opportunity finding in contrast
with the annotation of an area that is “ locked up”.
The regional authorities call this functionality a
“seeking area”. They have created the unusual
legal term “seeking area” to obtain legal degrees
of freedom in development plans that do not occur
with fixed parameterisation.
The map layer turning up while approaching
with the qualified cursor simulates the “seeking
area” degree of freedom the authorities require to
state intentions with a certain level of flexibility,
like moving 2 km to the right or left if this helps
with the original ecological goals. In Figure 7
a bordered “seeking area” is indicating a new
island that is being planned. The Red Polygon
can be moved and changes colour based on the
legal business rules set. This is called the Simcity
approach of Legal Atlas, as it allows for playing
around until a solution is found fit for purposes
which is similar to the Simcity game. The Legal
Figure 7. The“seeking area” near Lelystad, Flevoland
1256
Atlas as described has been evaluated by 25
real life stakeholders in the redesign process of
Ijmeer/Markermeer. Early results indicate that
they regard the system as a positive contribution
to the decision making process. The critics state
that the systems is not complete and that transpar-
ency is not always desirable. Most participants in
the evaluation state that it would definitely help
stakeholders to inform their constituencies about
the interests at stake. In international context the
system is criticised for implying a naïve perception
of regulatory systems. Different cultures would
treat legal constrains in different ways.
CONCLUSION
As discussed in this chapter, the new emerging
trends of the Internet, like Web 2.0 services are
constantly gaining space in the eGovernment
agenda, as municipalities, and generally public
administration entities, have realised that these
new technologies are supported and followed
by a massive enthusiastic audience, which can,
if treated carefully, increase the degree of par-
ticipation of eGovenrment platforms. In parallel,
through the last years, the use of GIS systems and
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
more specifically GIS powered applications are
as well a very useful tool for visualising various
aspects from the everyday life and help users to
conceptualise in visual terms the data they are
provided with.
The most important result that comes out
from the use of GIS data and infrastructures in
cooperation with web 2.0 tools, is that users seem
to find especially fascinating the fact that they
can interact with the system, the administration
officers and with each other, not only for enter-
tainment reasons, but for real life events as well,
targeting the improvement of their everyday life,
their active participation in the decision making
processes, the most effective management of di-
sasters that can affect human life, the promotion
of environmental policies, the reinforcement of
entrepreneurship in a location based perspective
and last, but not least, the structured planning of
strategic actions for local public administrations
with the participation of elected representatives,
public administration officers, NGOs, citizens and
other stakeholders involved in the overall process.
As such, GIS powered portals are able to
provide information services but also to visualise
the application field of various issues, directives
and various decisions. To this fact, the interactive
character of new GIS applications plays a major
role, as the different layers are able to interoperate
with the underlying infrastructures of the public
administrations, and through mash up technologies
to present useful and interactive functionalities,
offering to the user a more user-friendly, familiar
and enhanced online experience. Summing up,
one can clearly see that the use of GIS systems
alongside public administration portals is no longer
a fairytale, but a reality and the advantages that
these hybrid solutions offer are highly respected
by users, as they are presented with the best
possible interface which targets exactly on their
living regions.
REFERENCES
W3C. (2004, 2 10). OWL Web Ontology Language.
Retrieved 5 18, 2009 from http://www.w3.org/
TR/owl-features/
W3C. (2008, 1 15). SPARQL Query Language
for RDF. Retrieved 5 12, 2009 from http://www.
w3.org/TR/rdf-sparql-query/
Boer, A., van Engers, T., Peters, R., & Winkels,
R. (2007). Separating Law from Geography in
GIS-based eGovernment Services. Artificial
Intelligence and Law, 15(1), 49–76. doi:10.1007/
s10506-007-9042-4
Boer, A., van Engers, T., & Winkels, R. (2003).
Using Ontologies for Comparing and Harmoniz-
ing Legislation. Proceedings of the International
Conference on Artificial Intelligence and Law
(ICAIL) (pp. 60-69). Edinburgh: ACM Press.
Boer, A., Hoekstra, R., Winkels, R., van Engers,
T., & Willaert, F. (2002). METAlex: Legislation in
XML. Legal Knowledge and Information Systems
(Jurix 2002), 1-10.
Bundesministerium des Innern. (2003). SAGA:
Standards and Architectures for eGovernment
Applications Version 2.0 (Vol. 59). KBSt Publi-
cation Series.
Cabinet Office. (2005, 3 18). e-GIF. Retrieved 5
12, 2009 from http://www.govtalk.gov.uk/sche-
masstandards/egif.asp
Capgemini. (2006, 6). Online Availability of Web
Public Services: How is Europe progressing? Web
based survey on Electronic Public Services. Report
of the Sixth Measurement. Retrieved 5 12, 2009
from http://ec.europa.eu/information_society/
eeurope/i2010/docs/benchmarking/online_avail-
ability_2006.pdf
1257
 Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems
Charalabidis, Y., Askounis, D., Gionis, G., Lam-
pathaki, F., & Metaxiotis, K. (2006). Organizing
Municipal eGovernment Systems: A Multi-facet
Taxonomy of e-Services for Citizens and Busi-
nesses. In M. Wimmer, H. Scholl, Å. Grönlund,
& K. Andersen (Ed.), Lecture Notes in Computer
Science, 5th International Conference of Elec-
tronic Government. (vol. 4084, pp. 195-206).
Krakow: Springer.
EUROSTAT & DGENV, K.U. Leuven. (2003).
EC-INSPIRE: Spatial Data Infrastructures in
Europe: State of play.
Geonovum. (2006). IMRO: Informatiemodel Rui-
mtelijke Ordening (Information model for spatial
planning). Retrieved 5 20, 2009 from http://www.
geonovum.nl/informatiemodellen/imro-2006/
Gibson, J. (1979). The Ecological Approach to
Visual Perception. Boston: Houghton Mifflin.
IDABC. (2006). eGovernment in the Member
States of the European Union.
IDABC. (2008, 11). European Interoperability
Framework for pan-European eGovernment ser-
vices. Retrieved 5 12, 2009 from http://ec.europa.
eu/idabc/en/document/2319
Information Society S.A. (2008, 10). Greek eGov-
ernment Interoperability Framework. Retrieved 5
12, 2009 from http://www.e-gif.gov.gr/
INSPIRE. (2007). INSPIRE DIRECTIVE. Re-
trieved from INSPIRE DIRECTIVE: http://
inspire.jrc.ec.europa.eu/
Jaeger, P. (2003). The endless wire: EGovernment
as global phenomenon. Government Informa-
tion Quarterly, 20(4), 323–331. doi:10.1016/j.
giq.2003.08.003
Koster, R., & Wright, W. (2004). A Theory of Fun
for Game Design. Paraglyph Press.
1258
Koussouris, S., Charalabidis, Y., Gionis, G., Ts-
itsanis, T., & John Psarras, J. (2007). Building a
Local Administration Services Portal for Citizens
and Businesses: Service Composition, Archi-
tecture and Back-Office Interoperability Issues.
Lecture Notes in Computer Science - Electronic
Government, 4656, 80-91.
Koussouris, S., Tsitsanis, A., Gionis, G., & Psarras,
J. (2008). Designing Generic Municipal Services
Process Models towards eGovernment Interoper-
ability Infrastructures. The Electronic Journal for
E-Commerce Tools & Applications, 2 (3).
Layne, K., & Lee, J. (2001). Developing fully func-
tional EGovernment: A four stage Model Govern-
ment. Government Information Quarterly, 18(2),
122–136. doi:10.1016/S0740-624X(01)00066-1
Medjahed, B., Rezgui, A., Bouguettaya, A., &
Ouzzani, M. (2003). Infrastructure for EGovern-
ment Web Services. IEEE Internet Computing, 7,
58–65. doi:10.1109/MIC.2003.1167340
Nations, U. UN eGovernment survey 2008: from
eGovernment to connected governance. United
Nations. Department of Economic and Social
Affairs. Division for Public Administration and
Development Management Administration.
Ontotext, A. D. (2009). OWLIM Semantic Re-
pository. Retrieved 5 12, 2009 from http://www.
ontotext.com/owlim/
Open Geospacial Consortium. (n.d.). Geography
Markup Language. Retrieved 5 12, 2009 from
http://www.opengeospatial.org/standards/gml
OpenRDF.org. (n.d.). OpenRDF.org. Retrieved 5
18, 2009 from http://www.openrdf.org/
Pavlichev, A., & Garson, G. (2004). Digital Gov-
ernment: Principles and Best Practices. Hershey,
PA: IDEA Group Publishing.
Provision of Web 2.0 Services by Interoperable GIS-Powered Local Administration Portal Systems

Smith, A. (2007). Achieving competitive advan-
tage through the use of geographic information
systems. Electronic Government, an International
Journal, 4 (3), 326-344.
Soon, A., Vijayalakshmi, A., & Nabil, A. (2002).
Dynamic composition of workflows for custom-
ized eGovernment service delivery. ACM Inter-
national Conference Proceeding Series, 129. Los
Angeles, California.
van Ark, B., Melka, J., Mulder, N., Timmer, M.,
& Ypma, G. (2003). ICT Investments and Growth
Accounts for the European Union. Groningen
Growth and Development Centre September.
Vassilakis, C., Laskaridis, G., Lepoura, G., Rou-
vas, S., & Georgiadis, P. (2003). A framework for
managing the lifecycle of transactional eGovern-
ment services. Telematics and Informatics, 20(4),
315–329. doi:10.1016/S0736-5853(03)00011-X
Wilson, F., & Peters, R. (2004). Mapping the law:
Knowledge support for business development
enquiry. Proceedings of the eChallenges 2004
Conference. Vienna.
Winkels, R., & Hupkes, E. (2007). Legal Atlas: Ac-
cess to Legal Sources through Maps. Proceedings
on the 11th International Conference on Artificial
Intelligence and Law (ICAIL 2007). Amsterdam:
AAIL, ACM.

This work was previously published in Interoperability in Digital Public Services and Administration: Bridging E-Government
and E-Business edited by Yannis Charalabidis, pp. 201-217, copyright 2011 by Information Science Reference (an imprint of
IGI Global).

1259
1260

Chapter 76
Sharing Environmental
Data through GEOSS
Gregory Giuliani Hy Dao
University of Geneva and UNEP, Switzerland University of Geneva and UNEP, Switzerland

Nicolas Ray Jaap Van Woerden

University of Geneva and UNEP, Switzerland UNEP, Switzerland

Stefan Schwarzer Ron Witt

UNEP, Switzerland UNEP, Switzerland

Andrea De Bono Martin Beniston

University of Geneva and UNEP, Switzerland University of Geneva, Switzerland

Pascal Peduzzi Anthony Lehmann

UNEP, Switzerland University of Geneva and UNEP, Switzerland

ABSTRACT
Understanding the complexity of earth-system processes is crucial to convey improved information on the
environment to decision-makers and the general public. Addressing this need by sharing environmental
data is challenging because it requires a common agreed framework that allows easy and seamless in-
tegration of data from different sources. In this regard, the Global Earth Observation System of Systems
(GEOSS) portends major benefits through various sharing mechanisms and by giving access to services
that could be linked together to process and generate new understandable knowledge and information.
Various United Nations projects could greatly benefit from the GEOSS approach.

INTRODUCTION have expanded to take on a remarkable place in

Today we are living in a globalized world with
rapidly evolving processes including climate
change, population growth or environmental
degradation. In parallel, means of communication
DOI: 10.4018/978-1-4666-2038-4.ch076
our society, allowing us to access an enormous
and continuous flow of information.
In the last 30 years, the availability of geospatial
data has grown dramatically following the evolu-
tion of communication technologies supported
by the rapid development of spatial data capture
means such as remote sensing imagery, sensors and

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Sharing Environmental Data through GEOSS
GPS (Philips, Williamson, & Ezigbalike, 1999).
One of the challenges we are facing today is to
make sense of this vast amount of data in order to
turn them into understandable knowledge (Gore,
1998). Concrete actions can be taken only on the
basis of knowledge and understanding, but often
we know too little about the state of our planet’s
environment to take informed and sound decisions
about how it should be managed.
Our planet is a multi-dimensional system made
of complex interactions highly interconnected
and continuously evolving at many spatial and
temporal scales (GEO secretariat, 2007b). This
means that to understand these interactions, we
need to gather and integrate different sets of data
about physical, chemical and biological systems.
Altogether, these sets of data constitute environ-
mental data sets or data related to the environment.
These data are often georeferenced, describing a
geographical location through a set of attributes
and thus could be understood as being part of geo-
spatial data. An environmental data set is seldom
interesting in itself, but rather displays its full
information potential when used in conjunction
with other data sets, allowing one to monitor and
assess the actual status of the global, regional or
local environments, to discover complex relation-
ships between them and to model future changes.
In 1998, the former vice-president of the United
States, Al Gore, presented his visionary concept of
a Digital Earth (Gore, 1998), a representation of
the Earth embedding a vast amount of geospatial
data and allowing to make better sense of it. To
achieve this vision, Gore highlighted the need for
a collaborative effort (from government, indus-
try, academia and citizens) and pointed out the
different technologies required: computational
power, mass storage, satellite imagery, broadband
network, interoperability and metadata.
Despite the fact that administrations and
governments are recognizing that geospatial data
are an important component of an information
infrastructure (such as e-governement) that needs
to be efficiently coordinated and managed for
the interest of all citizens (Ryttersgaard, 2001),
this huge amount of geospatial data is stored in
different places, by different organizations and
the vast majority of these data are not being used
as effectively as they should. In consequence,
a framework allowing one to discover, access,
publish, share, maintain and integrate geospatial
data appears to be essential. Such a framework is
commonly known as a Spatial Data Infrastructure
(SDI).
Different initiatives at the regional and global
levels are influencing and promoting the creation of
SDIs allowing data providers to share and publish
their data in an interoperable manner. These initia-
tives coordinate actions that promote awareness
and implementation of complementary policies,
common standards and effective mechanisms for
the development and availability of interoperable
geospatial data and technologies to support deci-
sion making at all scales for multiple purposes.
These initiatives are related to data access, harmo-
nization, standardization, interoperability, seam-
less integration and services. Such an initiative is
the Global Earth Observation System of Systems
(GEOSS) which is a worldwide voluntary effort,
coordinated by the Group on Earth Observation
secretariat, to connect already existing SDIs and
Earth Observation infrastructures. GEOSS is
foreseen to act as a gateway between producers
of geospatial data and end users, with the aim of
enhancing the relevance of Earth observations
for the global issues and offering public access to
comprehensive information and analyses on the
environment (GEO secretariat, 2005, 2007a). The
GEOSS Common Infrastructure (GCI) provides
core capabilities that allow users to search, access
and use data, information, tools and services, and
is made of five components: GEO portal (web
portal to access GEOSS and search registries),
GEOSS clearinghouse (connects the different
components), GEOSS components and services
registry (catalogue of services and components),
GEOSS standards and interoperability registry
(catalogue of standards to use allowing users to set
1261

up and configure an interoperable system), and a
best practices wiki (offers a single space to share,
discuss, propose and exchange ideas and best
practices within the community). These compo-
nents are dependent on the voluntary contributions
of members and participating organizations. To
support the nine defined Societal Benefit Areas
(SBAs) (disasters, health, energy, climate, water,
weather, ecosystems, agriculture, biodiversity),
the mechanisms for data sharing and dissemination
are presented in a 10-year Implementation Plan
Reference Document (GEO secretariat, 2005) pro-
viding data sharing principles that any volunteer
member must endorse. The key element o share
data through GEOSS is to agree on “interoper-
ability arrangements” (GEO secretariat, 2007a)
allowing different components of the system to
communicate with each other.
Turning data into understandable knowledge
requires that data coming from different sources
be easily and seamlessly integrated. With the
capabilities offered by standards like the one
proposed by the Open Geospatial Consortium
(OGC), geospatial community can not only dis-
cover, access and publish interoperable geospatial
data but also services that can be linked together,
in chains of services, to process data and gener-
ate new information. Moreover, by registering
services into GEOSS, these different resources
are now accessible in a standardized way and are
reusable for many different purposes.
The aim of this paper is to present experiences
gathered through different United Nations (UN)
and European research projects and to discuss
promises and challenges envisioned in partici-
pating to an initiative like GEOSS, both in term
of building chains of services and sharing data.
THE NEED FOR DATA SHARING
AND INTEGRATION
Until very recently, the different systems used to
acquire environmental data were mostly operating
1262
Sharing Environmental Data through GEOSS
in isolation, which made it difficult to easily dis-
cover, access and use the data content of these sys-
tems due to incompatibilities and inconsistencies
of formats and data models (Bernard & Craglia,
2005). In addition, there is typically insufficient
data exchange among different stakeholders,
which is partially due to differing data policies.
Other important impediments to the flow of data
are the delays in accessing data that prevent timely
use of information, duplication and redundancy of
data acquisition, potential high costs associated
with data creation and access, and unclear access
rights and licensing policies (GEO secretariat,
2005). Altogether, these difficulties lead to a
fragmentation of data sources, impeding their
effective and efficient use, requiring much more
time than necessary for data collection (Open
Geospatial Consortium, 2004). All the previous
considerations highlight the growing need to share
data in an interoperable way and to ensure that
data are easily accessible and discoverable, so that
they can be used as often and widely as possible
(Arzberger et al., 2004). Moreover, the adoption
of the Agenda 21 resolution, a United Nations
initiative proposing a set of actions to be taken at
different scales to promote a sustainable develop-
ment, fostered the importance of geospatial data to
support decision-making and management related
to degradation and threats affecting the environ-
ment (Nebert, 2005). Availability and access to
appropriate information, and the related develop-
ment of interoperable databases, are the necessary
conditions for creating the basis for supporting the
information management needs of implementing
and monitoring sustainable development policies
and goals, such as the United Nations Millen-
nium Development Goals (MDGs) (Henricksen,
2007). The MDGs are eight development objec-
tives (eradicate extreme poverty and hunger,
achieve universal primary education, promote
gender equality, reduce child mortality, improve
maternal health, combat different diseases, ensure
environmental sustainability, and develop and
Sharing Environmental Data through GEOSS
global partnership for development) that all UN
members have agreed to achieve by 2015.
Over the last twenty years, the emergence and
evolution of Geographic Information Systems
(GIS) technology and the advent of applications
such as Google Earth or OpenStreetMap (Craglia
et al., 2008), allowed for a clear change on how
geospatial data are handled and incorporated into
regular workflows of organizations and agencies
in the governmental, private and public sectors
(Booz, Allen, & Hamilton, 2005). Highlighting
these changes, Masser (2007) stated that to real-
ize the full potential and benefits of geospatial
data, access must be maximized with the help
of Spatial Data Infrastructures (SDIs), that al-
low users to share, discover, visualize, evaluate
and retrieve geospatial data. Moreover, the vast
amount of data needed to run a complex model
(e.g., in climatology or ecology), and the recogni-
tion that organizations and/or agencies need more
data than they can afford financially (Rajabifard
& Williamson, 2001), reinforce the concept that
once a particular set of geospatial data has been
created, it should be accessible to potential users in
both the public and private sectors (Ryttersgaard,
2001). This reinforces the need to store such data
in databases that are made widely accessible for
various purposes (Philips et al., 1999). As a con-
sequence, geospatial data can be seen as a shared
resource which is maintained continuously.
To remove the barriers that block and impede
a wide use of geospatial data and related infor-
mation, Masser (2005, 2007) identified different
needs such as eliminating or reducing restrictions
on data access and availability (but protecting
intellectual property rights), promoting interoper-
ability between different data sets and different
systems, and disseminating the information about
data (metadata). Altogether these objectives are
designed to create an environment that fosters
activities for using, managing, producing and
sharing geospatial data in which all stakeholders
can cooperate with each other and interact with
technology, to better achieve their objectives at
different political/institutional levels (Rajabifard
& Williamson, 2004). In this sense, interoperabil-
ity appears to be a key element enhancing data
sharing, communication and efficiency.
The great advantage of interoperability is
that it describes the ability of locally managed
and distributed heterogeneous systems (different
operating systems, different databases, different
data formats) to exchange data in real time to
provide a service (OGC, 2004). The shift towards
a processed-based infrastructure offering reusable
and standardized components responsive to user
needs and requests is supported by the Service
Oriented Architecture (SOA) concept. In a SOA,
services are the elementary components represent-
ing a set of operations that could be invoked by
users allowing them to access, in the case of the
geospatial community, distributed geospatial data
as well as geoprocessing services. To implement
and deploy geo-enabled services, the OGC pro-
poses a suite of standards that use services over
the Internet, so-called web services, giving access
to distributed data and services through Uniform
Resource Locators (URLs). This allows data
providers to publish standardized services inde-
pendently on how it is implemented and on which
platform it is executed. This emphasizes the full
potential of interoperability allowing an organiza-
tion to maximize the value and reusability of data
under its control and giving the ability to exchange
these data with other interoperable systems. Using
such OGC web services offers the possibility to
seamlessly couple and reuse them in a variety of
applications. By chaining together a series of web
services, users can perform a set of operations to
process data whereby new knowledge emerges
from relationships that were not envisioned before
(Open Geospatial Consortium, 2004). Granell et
al. (2009) define service chaining as a mechanism
for combining individual geospatial web services
to create customized web applications. Although
current SDIs mostly offer the abilities to search,
view and access data, with the support of interop-
erable services and SOA related concepts it is
1263

now possible to build new applications based on
distributed services (Friis-Christensen, et al., 2007;
Diaz et al., 2008). When services are organized
through a coherent chain, combined services can
achieve a larger task (Di, 2004). The International
Organization for Standardization (ISO) through
its ISO 19119 standard (ISO, 2005) defines three
types of chaining:
• Transparent (user-defined): the work-
flow is defined and managed by users.
• Translucent (workflow-managed): users
invoke a service that manages the chain.
Users are aware of atomic services that
constitute the chain.
• Opaque (aggregated): users invoke an ag-
gregated service that carries out the chain.
Users have no awareness of the atomic ser-
vices that constitute the chain.
In this paper, we will focus on the transparent
chaining either by hard coding or by using OGC
Web Processing Service (WPS) specification
(OGC, 2007).
Through its online catalogue of registered
services, GEOSS is an interesting and promising
entry-point to discover and access services that
could be integrated into service chaining process.
It offers a framework to share data, expose them
through interoperable services and allow the pro-
duction and dissemination of timely and accurate
data needed by decision makers and the public
(GEO secretariat, 2005).
SERVING DATA INTO GEOSS
In 1985, the United Nations Environment Pro-
gramme (UNEP)/Division of Early Warning and
Assessments/GRID-Europe was founded as one
of the first two centres of the Global Resource
Information Database (GRID) network to support
environmental decision-making within UNEP
and the UN system as a whole, by generating and
1264
Sharing Environmental Data through GEOSS
disseminating information about the state of the
world’s environment in a timely and understand-
able manner. To provide reliable environmental
assessments and early warnings, GRID-Europe
specialized in handling and analyzing spatial
and statistical data on environmental and natural
resource issues through computerized GIS and
remotely-sensed imagery. Over the years, GRID-
Europe has compiled an archive of global, Euro-
pean and other geospatial databases as part of its
information management function. The experience
and in-house capabilities of GRID-Europe offer
a great potential to make geospatial and tabular
databases compiled over the years available to a
large array of users. Since its foundation, the Ge-
neva office has received considerable support from
Swiss and local authorities as well. This supporting
was significantly reinforced, and GRID-Europe’s
institutional base broadened, with the signing of
a “Partnership Agreement” between UNEP, the
Federal Office for the Environment (FOEN) and
the University of Geneva in June 1998.
GRID-Europe closely monitors developments
in information technologies and examines their
utility for environmental monitoring and policy
formulation and thus is extending and developing
its field of activities using SDIs. Moreover, the
“Partnership Agreement” provides a major op-
portunity to work at different geographic scales
ranging from global, to regional (Europe) and
national (Swiss) and finally local (Geneva). Such
a specificity allows GRID-Europe to participate to
different applied research projects funded either by
the United Nations or the European Commission.
A common ground for these projects is to serve
and share data through the European Directive
on Infrastructure for Spatial Information in the
European Community (INSPIRE) (European
Commission, 2007), the United Nations Spatial
Data Infrastructure (UNSDI) (Henricksen, 2007),
as well as GEOSS.
Sharing Environmental Data through GEOSS
PREVIEW Global Risk Data Platform
The PREVIEW (Project of Risk Evaluation, Vul-
nerability, Information, and Early Warning) Global
Risk Data Platform (http://preview.grid.unep.ch)
is a collaborative effort of UNEP, United Na-
tions Development Programme (UNDP/BCPR),
United Nations International Strategy for Disaster
Reduction (UNISDR) and the World Bank to
share geospatial data on global risk from natural
hazards. Users can freely visualize, download
or extract data on past hazardous events, human
and economical hazard exposure and risk from
natural hazards. The platform covers nine types
of natural hazard: tropical cyclones and related
storm surges, drought, earthquakes, biomass fires,
floods, landslides, tsunamis and volcanic erup-
tions. The collection of data is made via a wide
range of partners. This geoportal was developed as
a support to the 2009 Global Assessment Report on
Disaster Risk Reduction (United Nations Interna-
tional Strategy for Disaster Reduction Secretariat,
2009), replacing the previous PREVIEW platform
initially designed by UNEP/GRID-Europe and
already available since 2000. The new PREVIEW
platform is fully compliant with the OGC Web
Services (OWS) to access data using Web Map
Service (WMS), Web Feature Service (WFS),
Web Coverage Service (WCS), geo-enabled Re-
ally Simple Syndication (GeoRSS) or Keyhole
Markup Language (KML) as well as metadata
using Catalogue Service for the Web (CS-W).
GEO Data Portal
The GEO Data Portal (http://geodata.grid.unep.
ch) is the authoritative source for data sets used by
UNEP and its partners in the Global Environment
Outlook (GEO) report and other integrated envi-
ronment assessments. Its online database holds
more than 550 different variables, as national,
sub-regional, regional and global statistics or as
geospatial data sets (maps), covering themes such
as Freshwater, Population, Forests, Emissions,
Climate, Disasters, Health and Gross Domestic
Product (GDP). The data can be displayed and
explored on-the-fly through maps, graphs, data
tables, downloaded in various popular formats,
or copied and pasted into word processors. All
information products in the GEO Data Portal can
be accessed and used as web services as well.
The retrieval of statistical and country-wide in-
formation has been enabled via a Simple Object
Access Protocol (SOAP) connection; data from
the database can be retrieved as maps via WMS
or WFS; graphs can be displayed via a direct
Uniform Resource Locator (URL) usage.
enviroGRIDS
EnviroGRIDS (http://www.envirogrids.net) is
a European research project that will last from
2009 until 2013 and is funded under the seventh
framework programme (FP7). The Black Sea
Catchment is largely following an ecologically
unsustainable pathway based on inadequate re-
source management that could lead to severe
environmental, social and economical problems,
especially in a changing climate (WWF, 2008).
The aim of the project is to build capacities in the
Black Sea region to use new international standards
to gather, store, distribute, analyze, visualize and
disseminate crucial information on past, present
and future states of this region, in order to assess
its sustainability and vulnerability. EnviroGRIDS
objective is to federate and strengthen existing
Observation Systems to address several GEOSS
Societal Benefit Areas within a changing climate
framework. The expected result will be a shared
information system that operates on the boundary
of scientific/technical partners, stakeholders and
the public. It will contain early warning systems
able to inform in advance decision-makers and the
public about risks to human health, biodiversity
and ecosystems integrity, agriculture production
or energy supply caused by climatic, demographic
and land cover changes on a 50-year time horizon.
To achieve and support the enviroGRIDS vision
1265

and objectives, a grid-enabled Spatial Data Infra-
structure (gSDI) is under construction. The aim
of the gSDI is to host and analyze the data for the
assessment of GEOSS Societal Benefit Areas, as
well as the data produced within the project. These
data must be gathered and stored in an organized
form and accessible in an interoperable way on
the grid infrastructure in order to provide a high
performance and reliable access through standard-
ized interfaces.
ACQWA
ACQWA (http://www.acqwa.ch) stands for As-
sessing Climate impacts on the Quantity and
quality of Water. It is also a FP7 European re-
search project lasting from 2008 until 2013. As
the evidence for human induced climate change
becomes clearer, so does the realization that its
effects will have impacts on natural environment
and socio-economic systems. Some regions are
more vulnerable than others, both to physical
changes and to the consequences for ways of life.
According to the description of work, the project
will assess the impacts of a changing climate on the
quantity and quality of water in mountain regions
which are particularly affected by rapidly rising
temperatures, prolonged droughts and extreme
precipitation. Modeling techniques will be used
to project the influence of climatic change on the
major determinants of river discharge at various
time and space scales. Regional climate models
will provide the essential information on shifting
precipitation and temperature patterns. Snow, ice,
and biosphere models will feed into hydrological
models in order to assess the changes in seasonal-
ity, amount, and incidence of extreme events in
various catchment areas. Environmental and socio-
economic responses to changes in hydrological
regimes will be analyzed in terms of hazards,
aquatic ecosystems, hydropower, tourism, agri-
culture, and the health implications of changing
water quality. Attention will also be devoted to the
interactions between land use/cover changes, and
1266
Sharing Environmental Data through GEOSS
changing or conflicting water resource demands.
Adaptation and policy options will be elaborated
on the basis of the results. The chain of processes
involved in climatic, cryospheric and hydrologic
models is complex because each process impacts
on different compartments of human and natural
systems. Different types of data covering various
geographical regions are therefore necessary to
build different sets of scenarios, which translates
into substantial amount of data.
TECHNICAL COMPARISON
AND COMMON GROUNDS
All these projects have in common that they already
share (or will share in a near future) their data and
metadata into the GCI. As a pre-requisite all the
registered services have to be interoperable using
mainly standards proposed by the OGC, but also
other protocols like the Simple Access Object
Protocol (SOAP). A short comparison of these
different projects is given in Table 1, indicating
which services are available, which software are
used to publish these services, what are the types
of models used to chain these layers, and finally
what are the challenges and difficulties raised
while integrating these services.
Most of these projects make use of free and
open source software (PostgreSQL/PostGIS,
MapServer, GeoServer, GeoNetwork and PyWPS)
because it can ease the portability and replicabil-
ity of tools developed. Indeed, many countries
with low to moderate incomes are often also af-
fected by natural hazards, environmental threats
or degradation, and these countries are espe-
cially interested to manage and share their geo-
spatial data using free and open sources software.
Having tools readily available to be deployed in
these countries is a strong incentive for capacity
building, knowledge transfer, and sharing of
expertise.
These projects are also strongly related to ca-
pacity building in order to enhance an “open and
Sharing Environmental Data through GEOSS
Table 1. Technical comparison of enviroGRIDS, ACQWA, GEO Data Portal and PREVIEW projects
Project name enviroGRIDS ACQWA
Services WMS, WFS, WCS, CS-W, WMS, WFS, WCS, CS-W,
KML, GeoRSS, WPS, WPS, KML, GeoRSS
grid services
Software GeoServer, ArcGIS GeoServer, GeoNetwork,
Server, PyWPS, GeoNet- PyWPS
work, gLite
Type of models - Hydrological models - Snow cover mapping
Challenges & - linking SDI and grid - capacity building
difficulties infrastructure - data integration
- capacity building
- authorization/ authenti-
cation
- portal integration
sharing spirit”. It is necessary to show and prove
the benefits of data sharing through appropriate
examples, to communicate best practices as much
as possible and to develop guidelines and poli-
cies. Altogether this will help to reach agreement
and endorsement on the use of new standards.
Such a participative approach will certainly
stimulate data providers to be more “open” and
in consequence to share their data. The different
projects presented before will organize different
workshops and develop various teaching mate-
rial allowing participants, ranging from students
to members of government, to learn how to use
the specific applications to share large amount of
data. Rajabifard and Williamson (2004) believe
that building capacities is an important challenge
for SDIs concepts to be accepted and adopted at
a large extent. For these authors, the best way to
reach this objective is to establish a long-term com-
mitment to education and research: otherwise the
SDI vision will remain unclear and unachievable.
Through these projects, the objective is to build
the capacity of scientists to share and document
their data in order to strengthen existing observa-
tion systems, the capacity of decision-makers to
use it, and the capacity of the general public to
understand the important environmental, social
and economic issues at stake.
GEO Data Portal PREVIEW
WMS, WFS, WCS, CS-W, WMS, WFS, WCS, CS-W,
SOAP KML, GeoRSS
GeoServer, GeoNetwork, GeoServer, GeoNetwork,
MapServer MapServer
- providing base layers - providing base layers
(socio-economic,...) (events, risk,...)
- data integration - data integration
- data/metadata harmoni- - data/metadata harmoni-
zation zation
- different standard imple- - capacity building
mentation
- capacity building
Through simple data integrating scenarios
(integration into other web portals or applications)
GEO Data Portal and PREVIEW have pointed out
different issues. Integrating some socio-economic
data sets coming from the GEO Data Portal with
natural hazards maps of the PREVIEW project to
compute, for example, economical exposition of a
country to a specific hazard, was impossible. This
problem comes from the different implementations
of OGC specifications between Mapserver (used
by the GEO Data Portal) and Geoserver (used by
PREVIEW). Indeed, it appears that Mapserver use
an argument “MAP” that is not standardized and
not recognized by all clients. This problem will
be solved by migrating to Geoserver all the data
services of the GEO Data Portal. Thus implemen-
tation of a same specification can differ from one
software to another and can impend a consistent
integration of services.
Another issue raised by data integration process
was raised by the United Nations High Commis-
sioner for Refugees (UNHCR) while trying to
integrate WMS data coming from the PREVIEW
project in order to identify area that are not suitable
to install refugees camps. It appears that the only
projection available was EPSG:4326 (Geographic)
whilst UNHCR geoportal makes use of Google
maps in EPSG:900913 (Spherical Mercator). This
1267

experience showed us that it is important, while
publishing data services, to support at least the
most frequent projection types. Geoserver sup-
ports natively all types of projections and it is
easy to reproject on-the-fly data stored in another
projection so that it can be integrated with data
with other projections. In addition, following the
size of the data set, an important processing over-
load has been observed caused by the on-the-fly
reprojection process. This can slow the service
chain and impend and efficient data integration.
In the ACQWA project, a specific constraint
is the important number of partners involved and
their different scientific backgrounds (climatolo-
gists, hydrologists, glaciologists, ecologists). It is
quite challenging to raise awareness on new tools
and way of gathering and exchanging data with-
out strongly influencing the way these different
communities are working with geospatial data.
For that reason, the aim is to concentrate on the
promotion of GEOSS as an interesting and useful
framework to handle and discover scientific data.
Obviously, a dedicated geoportal is under develop-
ment to register the main outputs of the ACQWA
project into GEOSS using OGC web services.
Nevertheless to show the benefits of working with
interoperable services, we are currently developing
a scenario to make estimation of snow cover from
remote sensing imagery using data coming from
the Moderate Resolution Imaging Spectrometer
Figure 1. Data sources and processing steps for a geoprocessing service estimating snow cover
1268
Sharing Environmental Data through GEOSS
(MODIS) and Shuttle Radar Topography Mis-
sion (SRTM). Project partners that are currently
working to produce such estimations are working
with PCI Geomatica, doing all the process chain
manually. Our objective is to help our partners
by publishing a WPS geoprocessing service that
allows them to automatize this analysis (Figure
1). Once retrieved by FTP, MODIS images are
saved on a server that store also SRTM tiles. All
data are in EPSG:4326 and will be available using
WCS standard published by Geoserver. Finally,
the WPS service, currently under development
using PyWPS, will implement the different steps
to process the data. A major difficulty encountered
until now is to “translate” the PCI functionalities
by finding the equivalent in Geographic Resources
Analysis Support System (GRASS) software.
Indeed, PyWPS does not process data by itself
and instead uses GRASS as a backend to access
all the geoprocessing functionalities.
Once the snow cover process is successfully
achieved, our hope is to convince other commu-
nities within the project to benefit from such an
approach and to develop other scenarios espe-
cially making use of climate data.
In the process of turning data into understand-
able information and knowledge by chaining
data services a new challenge has emerged. The
ever-increasing spatial and temporal resolution
of geospatial data are causing a tremendous in-
Sharing Environmental Data through GEOSS
crease in term of data volumes and the limits of
the processing capacities of traditional GIS and
SDI are being reached. With the advent of grid
computing and the progressive deployment of
large grid infrastructure projects (e.g., Enabling
Grids for E-sciencE) many scientific disciplines
now have access to sizable computing resources
and new opportunities are emerging. For Foster et
al. (2008) grid aims to federate resource sharing
in a dynamic and distributed environment across a
network allowing to access unused CPUs and stor-
age space to all participating computers. Currently,
SDIs are lacking processing power and should
therefore be made interoperable with grid infra-
structures, which are offering large storage and
computing capacities. Recent studies (Muresan et
al., 2008; Di, Chen, Yang, & Zhao, 2003) applied
a successful approach to extend grid computing
to the remote sensing community and to make
OGC web services grid-enabled. Both studies
considered that the grid has a great potential for
the geospatial disciplines. Padeberg and Greve
(2009) have identified several differences between
OGC-compliant SDIs and grid infrastructures
concerning service description, service interface,
service state and security. In particular, grid infra-
structures are based on SOAP messaging protocol
to invoke operations and Web Service Description
Language (WSDL) to describe services. OGC-
compliant does not support neither SOAP nor
WSDL, except WPS, and thus chaining geospatial
services with grid services could be problematic.
In addition, OGC standards do not provide any
security mechanisms (authentication, encrypted
communication between resources) which is a
major concern in grid infrastructures. Finally, Di
et al. (2003) showed that the current grid metadata
catalog system is not good enough to answer the
needs of the geospatial community, especially
the requirements of the ISO19115 standard. All
these differences must be overcome in order to
allow traditional SDIs to benefit from the power
of grid computing, and consequently to offer new
services to GEOSS.
The main scientific and technological chal-
lenge of the enviroGRIDS project will be to link
an SDI with a grid infrastructure to benefit from
the processing capacities offered by grids. Indeed,
WPS appears to be an adequate candidate to be
grid-enabled because, first, it supports SOAP
protocol and, second, geospatial community has a
growing processing need that current SDIs cannot
deliver. A grid-enabled SDI will allow users to
model high resolution hydrological models (e.g.,
Soil and Water Assessment Tool) of the Black Sea
catchment under various climate, land cover and
demographic scenarios. In order to develop such
a gSDI to support the development of Black Sea
portal functionalities, the different components
of the enviroGRIDS architecture are currently
being defined to highlight the main issues emerg-
ing from different conceptual and technological
solutions (Figure 2).
These issues concern the choices of software
components, data repositories, data management,
grid-oriented processing, grid portal, and interop-
erability between SDIs and grid infrastructures.
Although the use of grid-enabled web services to
access data sets stored in the SDI will also be
explored (Maué & Kiehle, 2009), bridging archi-
tectural gaps between grids and SDIs remains
very challenging (Padberg & Greve, 2009) with-
out extensions and customizations. For example,
an important question concerns the location of
geospatial data repositories: inside or outside the
grid? The answer is not trivial and will greatly
influence services and in particular chains of
services to process data. In the one hand, being
outside the grid, all OGC-compliant services
functionalities remain the same and grid services
are only used to process the data. On the other
hand, being inside the grid, all OGC web ser-
vices have to be modified to support grid environ-
ment, becoming grid-enabled. The latter would
allow benefiting from all the advantages of the
grid (security, replicability, scalability, storage
and processing capacities) but would obviously
require a lot of developments for adapting already
1269

Figure 2. EnviroGRIDS grid-enabled SDI components supporting Black Sea portal
existing SDIs. In consequence, an incremental
development and implementation strategy will be
developed taking into account different integration
scenarios aiming to hide the complexity of the
grid while preserving OGC interfaces.
CHALLENGES AND PROMISES
From the experience acquired, or being acquired,
through these different projects, it is obvious that
many challenges remain both tangible (e.g., tech-
nology) or less tangible (e.g., culture, behavior).
Nevertheless, it is critical to overcome them in
order to improve our knowledge, share our experi-
ence and attempt to strive towards a society that is
better informed. Achieving the goal of sustainable
1270
Sharing Environmental Data through GEOSS
development requires the integration of a large
number of different data from various sources.
Through agreed common standards and a clear
political will, these data can be integrated in an
interoperable way, leading to a new collaborative
approach to decision-making.
Having environmental data in digital form
allows easy storage and dissemination, facilitate
data exchange and sharing, faster and easier up-
date and corrections, ability to integrate data from
multiple source (see Figure 1), and customization
of products and services (Henricksen, 2007). In
this sense SDIs appear to be a good choice to
encompass the sources, systems, network link-
age standards and institutional issues involved
in delivering geospatial data from many data
sources to the widest possible group of potential
Sharing Environmental Data through GEOSS
users (Coleman, McLaughlin, & Nichols, 1997).
The fact that, during the last years, multiple
SDIs initiatives have been developed all around
the world, ranging from local to regional levels,
is a good sign. It appears that there is a grow-
ing recognition that geospatial data is a critical
element underpinning decision making in many
disciplines (Rajabifard & Williamson, 2001) and
as such needs to be effectively managed.
The SDI hierarchy model proposed by Rajabi-
fard (2002) is composed of inter-connected SDIs
developed at different levels (from local to global).
Each SDI of a higher level is formed by the integra-
tion of data developed and made available by the
lower level. Such a hierarchy can be approached
though two views: on one hand, it is an umbrella
in which the SDI at a higher level encompasses
all SDI components from lower levels. On the
other hand, it can be seen as the building blocks
supporting the access of data needed by SDIs at
higher levels. This hierarchy allows creating an
environment in which users working at any level
can rely on data from other levels and integrate data
from different sources (Mohammadi & Rajabifard,
2009). Such a hierarchy is clearly envisioned in
the concept of the system of systems on which
GEOSS relies, integrating systems together into
an information highway which both links together
environmental, socio-economic and institutional
databases and provides a movement of information
from local to global levels. For Masser (2006),
the SDI hierarchy poses the challenge of multi-
stakeholder participation in SDI implementation,
because the bottom-up approach differs a lot from
the top-down approach. The top-down vision,
common in the SDI literature, emphasizes the
need for standardization and uniformity while
the bottom-up view stresses the importance of
diversity and heterogeneity caused by the dif-
ferent needs of the various stakeholders. As a
consequence, it is necessary to find a consensus
ensuring sufficient standardization and uniformity
while recognizing the diversity and heterogeneity
of the different stakeholders acting at different
levels. In particular, building a system of sys-
tems like GEOSS is highly dependent on a clear
governance structure that is understandable and
acceptable by the volunteer participants in order
to develop a shared vision of the system and to
allow users to feel a common sense of ownership
(Masser, 2007). As it is reminded in the Strategic
Guidance document (GEO secretariat, 2007a), the
success of GEOSS will depend on interoperability
arrangements that data providers agree to endorse.
As a provider of environmental data, GRID-
Europe is continuously facing the challenge of
encouraging data providers to go “open” and to
share their data in an interoperable and OGC-
compliant way. At present, technology is no longer
a problem because solutions based on a variety
of software can be proposed and/or developed
depending on the requirements and the technical
capabilities available. The most difficult task is
to create an environment allowing wide agree-
ment on data sharing principles. In this particular
regard, the GEOSS “best practices wiki” could be
of great benefit to help people promote sharing
principles. A lesson learned from our experience
is that once users can discover data they need,
their most important preoccupation is to know
what is the quality of the data they are going to
access and whether they can trust this data. We
are convinced that sharing data is an efficient
way to eventually recognize whether this data is
of sufficient quality. By submitting/exposing the
data to the judgment of the broader community,
one can know if it is useful or not. Through data
sharing, one can also benefit from the interaction
with end users by receiving feedbacks and then
improve the data sets accordingly. Sharing data and
participating to GEOSS can therefore contribute
to the improvement of data, which in turn allows
better information and eventually better decisions.
In the current climate of economic constraints,
interoperability and standardization have never
been so important because a non-interoperable
system impedes the sharing of data, information
and resources, which increase the risk for a system
1271

to fail in delivering its expected benefits and to
remain unused (Open Geospatial Consortium,
2004). Geospatial data can be an expensive and
time consuming resource to produce, and for
this reason, it is of high importance to improve
accessibility and availability and promote its
reuse. Many decisions that organizations need to
make depend on good quality and consistent data,
readily available and accessible (Rajabifard &
Williamson, 2001). The process of reuse does not
only concern the data itself, but also encompasses
the capabilities, skills developed, invested effort
and capital. This process allows an organization
to share the costs of data, people, and technology,
which helps realize more rapid returns on invest-
ment. By reusing data, one can avoid duplication
of efforts and expenses and enable users to save
resources, time and effort when trying to acquire
or maintain data sets (Rajabifard & Williamson,
2001).
Percivall (2006) claimed that in a distributed
environment, the help of open standards such
as OGC can help scientists to rapidly find and
evaluate a lot of different data sets and processing
approaches, providing a flexible and coopera-
tive environment that foster collaboration in the
different scientific communities that work with
geospatial data. Thus, organizing the workflows
using standard-based web services could provide a
great benefit in term of productivity to address the
nine SBAs of GEOSS. OGC standards provide a
solid ground for interoperability between services
within distributed geoprocessing environment
offered by SDIs (Friis-Christensen et al., 2007).
In particular, the fact that these services can be
reused and chained within other applications is
a very useful aspect offering the opportunity to
solve specific problems in a more flexible way
than with stand-alone applications. Nevertheless,
some performance issues can appear with services
that need to access and move large amount of data.
This can negatively impact the execution time
of this service (e.g., huge overload in gathering
1272
Sharing Environmental Data through GEOSS
necessary data) especially if this service is chained
with other services.
Consequently, GEOSS represents a very
promising and potentially powerful framework to
share and expose data. In particular, the fact that
a good governance structure is already in place
allows a clear vision that can be easily shared
and endorsed by the participants. The fact that
participating to GEOSS is on a voluntary basis
could be seen either as a great opportunity or
as a risk. Indeed, the voluntary aspect poses the
threat that only a few data providers join such an
initiative and, as a consequence, the system could
miss its objectives. Nevertheless, the growing
number of components and services registered
through GEOSS is a good sign for optimism. In
particular, we think that international organizations
such as UNEP could play a major role by paving
the way toward a broader acceptance by similar
organizations. The fact that GEOSS is based on
distributed systems that can operate, evolve and
be managed in a relative independence appears
to be a good choice to find a consensus ensuring
sufficient standardization and uniformity, while
recognizing the diversity and heterogeneity of
the different stakeholders. Finally, GEOSS offers
a unique characteristic that justifies by itself its
existence, which is the possibility to see emergent
properties. For Béjar et al. (2009), this emergence
is the main objective of a system of systems, where
users perform functions that cannot be made with
any single component. This means that such a
system is more than the sum of its parts and offers
the possibility to better understand the complex
relationships between the different components
of the Earth system.
CONCLUSION
Geospatial data is a critical element underpin-
ning decision-making for many disciplines and
is indispensable to make sound decisions at all
levels, from global to local. Experiences from
Sharing Environmental Data through GEOSS
developed countries show that more than two-
thirds of human decision-making are affected by
spatially-referenced data (Ryttersgaard, 2001).
Even if the technology exists, organizations and
agencies around the world are still spending bil-
lions of dollars every year to produce, manage and
use geospatial data, but they still do not have the
information they need to answer the challenges our
world is facing (Rajabifard & Williamson, 2001).
The web service model proposed by the OGC
appears to be suitable to allow users to combine
different services to solve a specific problem in a
scalable and flexible way. Nevertheless, through
simple examples of services chaining, we have
highlighted different issues that could potentially
impede an easy integration: problems with dif-
ferent implementation of a same specification,
problems regarding different projections used
in different web applications, overload caused
by on-the-fly reprojection using large data sets.
Moreover, working with different communities
that are not necessarily aware of the possibilities
offered by OGC web services could limit the
diffusion of such approach outside the geospatial
community. These communities need to be con-
vinced, through simple examples, which working
with chained services can bring benefits in their
own working flows. Finally, grid computing ap-
pears to be a promising complement of traditional
SDIs capabilities to build WPS services for pro-
cessing large data sets. To achieve this objective,
implementation of SOAP protocol into OGC
specifications is a pre-requisite in order to allow
the two types of infrastructures to communicate
(interoperability) and to ease the combination of
OGC and grid services in efficient chains.
Ten years after, GEOSS could be seen as an
initial step to achieve Gore’s vision, because the
relevant technologies are available and there is
growing recognition that countries can benefit both
economically and environmentally from better
access to data. GEOSS has the potential to sup-
port the achievement of sustainable development
initiatives such as the UN Millennium Develop-
ment Goals and to offer a unique framework to
share data and collaborate for a better society. In
this sense, organizations such as UNEP can act
as a “catalyst”, contributing to GEOSS, building
capacities and ensuring that environmental data
are easily accessible. This is a necessary step to
ensure better-informed decision-making for the
more sustainable development of our planet.
ACKNOWLEDGMENT
The authors would like to acknowledge the
European Commission “seventh framework pro-
gramme” that funded the enviroGRIDS (Grant
Agreement n° 226740) and ACQWA (Grant
Agreement n°212250) projects, and UNEP for
its support. The views expressed in the paper are
those of the authors and do not necessarily reflect
the views of the institutions they belong to.
REFERENCES
Arzberger, P., Schroeder, P., Beaulieu, A., Bowker,
G., Casey, K., & Laaksonen, L. (2004). Promoting
Access to Public Research Data for Scientific,
Economic and Social Development. Data Science
Journal, 3, 17. doi:10.2481/dsj.3.135
Béjar, R., Latre, M. A., Nogueras-Iso, J., Muro-
Medrano, P. R., & Zarazaga-Soria, F. J. (2009).
Systems of Systems as a Conceptual Framework
for Spatial Data Infrastructures. International
Journal of Spatial Data Infrastructures Research,
4, 17.
Bernard, L., & Craglia, M. (2005). SDI - From
Spatial Data Infrastructure to Service Driven
Infrastructure. Paper presented at the 1st Research
Workshop on Cross-learning on Spatial Data Infra-
structures (SDI) and Information Infrastructures
(II), Enschende, The Netherlands.
1273

Booz, A. (2005). Geospatial Interoperability
Return on Investment. Hamilton: NASA.
Coleman, D. J., McLaughlin, J. D., & Nichols,
S. (1997). Building a Spatial Data Infrastructure.
Paper presented at the 64th Permanent Congress
Meeting of the Fédération Internationale des
Géomètres (FIG), Singapore.
Craglia, M., Goodchild, M. F., Annoni, A., Ca-
mara, G., Gould, M., & Kuhn, W. (2008). Next-
Generation Digital Earth: A position paper from
the Vespucci Initiative for the Advancement of
Geographic Information Science. International
Journal of Spatial Data Infrastructures Research,
3, 22.
Di, L. (2004). Distributed Geospatial Information
Services – Architectures, Standards, and Research
Issues. In Proceedings of the International Ar-
chives of Photogrammetry, Remote Sensing, and
Spatial Information Sciences, Istanbul, Turkey
(Vol. 35, Part 2, Commission II).
Di, L., Chen, A., Yang, W., & Zhao, P. (2003).
The Integration of Grid Technology with OGC
Web Services (OWS) in NWGISS for NASA EOS
Data. Paper presented at the GGF8 & HPFC12,
Seattle, WA.
Diaz, L., Granell, C., & Gould, M. (2008). Case
Study: Geospatial Processing Services for Web-
based Hydrological Applications. Geospatial
Services and Applications for the Internet, 31-47.
European Commission. (2007). Proceedings of the
Directive 2007/2/EC of the European Parliament
and the Council of 14 March 2007 establishing an
Infrastructure for Spatial Information in the Euro-
pean Community (INSPIRE), Brussels, Belgium.
Foster, I., Yong, Z., Raicu, I., & Lu, S. (2008).
Cloud computing and grid computing 360-degree
compared. Paper presented at the 2008 Grid
Computing Environment Workshop, Austin, TX.
1274
Sharing Environmental Data through GEOSS
Friis-Christensen, A., Lutz, M., Ostländer, N., &
Bernard, L. (2007). Designing Service Architec-
ture for Distributed Geoprocessing: Challenges
and Future Directions. Transactions in GIS, 11(6),
799–818. doi:10.1111/j.1467-9671.2007.01075.x
GEO secretariat. (2005). Global Earth Observa-
tion System of Systems 10-Year Implementation
Plan Reference Document. Geneva, Switzerland:
GEO.
GEO secretariat. (2007a). Strategic Guidance for
Current and Potential Contributors to GEOSS.
Geneva, Switzerland: GEO.
GEO secretariat. (2007b). The Full Picture. Ge-
neva, Switzerland: GEO.
Gore, A. (1998). The Digital Earth: Understanding
our planet in the 21st Century (p. 4).
Gorgan, D., Bacu, V., Ray, N., & Maier, A. (2009).
EnviroGRIDS data storage guideline (p. 67).
Granell, C., Gould, M., & Esbri, M. A. (2009).
Geospatial Web Service Chaining. In Karimi, H. A.
(Ed.), Handbook of Research on Geoinformatics
(pp. 189–195). Hershey, PA: IGI Global.
Henricksen, B. (2007). UNSDI Compendium:
A UNSDI Vision, Implementation Strategy, and
Reference Architecture. UNGIWG.
ISO. (2005). ISO 19119: Geographic Information
– Services. Gland, Switzerland: ISO.
Masser, I. (2005). The Future of Spatial Data
Infrastructures. Paper presented at the ISPRS
Workshop on Service and Application of Spatial
Data Infrastructure, Hangzhou, China.
Masser, I. (2006). Multi-level Implementation of
SDIs. GIM International, 20, 4.
Masser, I. (2007). Building European Spatial
Data Infrastructure. Redlands, CA: ESRI Press.
Sharing Environmental Data through GEOSS
Maué, P., & Kiehle, C. (2009). Grid technologies
for geospatial applications: an overview. GIS
Science, 3, 65–67.
Mohammadi, H., & Rajabifard, A. (2009). Multi-
source Spatial Data Integration within the Context
of SDI Initiatives. International Journal of Spatial
Data Infrastructures Research, 4, 18.
Muresan, O., Pop, F., Gorgan, D., & Cristea, V.
(2008). Satellite Image Processing Applications
in MedioGRID. Paper presented at the 5th Inter-
national Symposium on Parallel and Distributed
Computing.
Nebert, D. D. (2005). Developing Spatial Data
Infrastructure: The SDI Cookbook.
(2004). Open Geospatial Consortium (p. 7). The
Havoc of Non-Interoperability.
Open Geospatial Consortium. (2007). OpenGIS
Web Processing Service (p. 87).
Padberg, A., & Greve, K. (2009). Gridification
of OGC web services: challenges and potential.
GIS Science, 3, 77–81.
Percivall, G. (2006). GEOSS to Benefit from
“Service Chaining” Based on OGC® Standards
(p. 5). Geoinformatics.
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 1,
edited by Donald Patrick Albert, pp. 1-17, copyright 2011 by IGI Publishing (an imprint of IGI Global).
Philips, A., Williamson, I., & Ezigbalike, C.
(1999). Spatial Data Infrastructure concepts. The
Australian Surveyor, 44(1), 8.
Rajabifard, A. (2002). Proceedings of Diffusion
of Regional Spatial Data Infrastructures: with
Particular Reference to Asia and the Pacific,
Melbourne, Australia.
Rajabifard, A., & Williamson, I. P. (2001). Spatial
Data Infrastructures: Concept, SDI Hierarchy
and Future Directions. Paper presented at the
Geomatics’80, Tehran, Iran.
Rajabifard, A., & Williamson, I. P. (2004). SDI
Development and Capacity Building. Paper pre-
sented at the GSDI-7, Bangalore, India.
Ryttersgaard, J. (2001). Spatial Data Infrastruc-
ture: Developing Trends and Challenges. Paper
presented at the International Conference on
Spatial Information for Sustainable Development,
Nairoibi.
United Nations International Strategy for Disaster
Reduction Secretariat (UNISDR). (2009). Global
Assessment Report on Disaster Risk Reduction
2009. Geneva, Switzerland: UN.
World Wide Fund. (2008). Greening the Black
Sea Synergy. Brussels, Belgium: WWF.
1275
1276

Chapter 77
Agriculture and Conservation
in the Natura 2000 Network:
A Sustainable Development
Approach of the European Union

Cristian Ioja Maria Patroescu

University of Bucharest, Romania University of Bucharest, Romania

Laurenţiu Rozylowicz Mihai Niţă

University of Bucharest, Romania University of Bucharest, Romania
Diana Onose
University of Bucharest, Romania

ABSTRACT
The Natura 2000 network represents a new approach in the sustainable spatial planning promoted at
the European Union level. Agricultural landscapes comprise 28.6% of the surface area of the Natura
2000 sites, many of which have significant conservation values. Plant and animal species, and ap-
proximately 30% of the natural habitats of community interest are directly influenced by the presence of
certain agricultural activities. This chapter presents a GIS analysis of the European Union Natura 2000
ecological network: spatial distribution of Natura 2000 sites in EU-27, dynamic agricultural surfaces
in Natura 2000 sites, and GIS tools in managing process. GIS techniques must represent the tool by
which the efficiency of this ecological network is monitored, as it must be permanently nourished with
important financial resources.

INTRODUCTION ing biological communities (Baur et al., 2006;

Ecologically, agriculture has a dual nature, being
considered the main risk affecting biodiversity at
a global level (Primack, Patroescu, Rozylowicz
& Ioja, 2008) but also the support for sustain-
DOI: 10.4018/978-1-4666-2038-4.ch077
Kuemmerle, Muller, Griffiths & Rusu, 2009;
Pykala, 2003; Ruprecht, Enyedi, Eckstein &
Donath, 2010). Globally, protected areas embody
the most frequently used instrument for limiting
threats upon representative samples of species
and natural ecosystems (Balmford et al., 2002),

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Agriculture and Conservation in the Natura 2000 Network
but also for preserving and valorising certain
traditional agricultural landscapes (Plieninger,
Hochtl & Spek, 2006).
The intensity of biodiversity threats induced
by agriculture is extremely high in EU-27, where
agricultural fields comprise 47.4% of the land
area (European Commission, 2009). Habitat
disturbance, overexploitation, pollution, invasive
alien species, and disease characterize the most
common threats encountered in the European
space in which agricultural activities affect bio-
diversity (Plieninger et al., 2006; Primack et al.,
2008; Schmitt & Rakosy, 2007).
Agricultural techniques are becoming increas-
ingly destructive, as the methods are devoted
mainly to an increase in productivity (Stoate et
al., 2009). This destruction can be minimized by
the new orientations of the Common Agricultural
Policy (CAP), promoting agri-environmental
schemes (EEA, 2005; Piorr et al., 2009). The CAP
will moderate the negative effects of agriculture
upon biodiversity (Henle et al., 2008), and reac-
tivate benefits generated by ecosystem services
(Hockings, 2003). The new schemes may also be
crucial in maintaining the efficiency of agricultural
activities (EEA, 2009; Piorr et al., 2009). Illustra-
tive of this issue is the promotion of incentives
for high nature value farmland, with the purpose
of increasing the conservation of avian species
dependant on agricultural habitats (EEA, 2004).
In this framework, the Natura 2000 network
represents an European Union goal, aimed at pro-
moting the conservation of natural habitats of flora
and fauna species, without excluding populated
communities and local economy (Mucher, Hen-
nekens, Bunce, Schaminee & Schaepman, 2009;
Pullin et al., 2009; Silva, 2009). The enforcement
of the Natura 2000 network appeared as a necessity
in fulfilling the objectives of reducing species and
habitats loss (Pullin et al., 2009). The European
Union, through the Convention on Biodiversity,
Habitats and Birds Directives, assumed these
objectives (Cogalniceanu & Cogalniceanu, 2010).
The projected benefits of the Natura 2000
network regarding risk control can confer long-
term protection to a greater number of species and
habitats, and a more efficient use of the available
natural resources (European Commission, 2009;
Ioja et al., 2010). Beside their inherent ecological
value (Gaston, Jackson, Nagy, Cantu-Salazar &
Johnson, 2008) in underdeveloped areas many
Natura 2000 sites are created for the poten-
tial to alleviate social and/or economic issues
(Dimitrakopoulos, Memtsas & Troumbis, 2004;
Mauerhofer, 2010; Rauschmayer, van den Hove
& Koetz, 2009).
The Natura 2000 network is perceived as
a “social network” fit for the European Union
landscape that intertwines the preservation of
nature with the maintenance of a sustainable tra-
ditional lifestyle for the local communities (e.g.,
supply of locally grown products/produce, local
employment opportunities, and/or eco-tourism)
(Bladt, Strange, Abildtrup, Svenning & Skov,
2009; Maiorano, Falcucci, Garton & Boitani,
2007; Young et al., 2007).
Integrating the socio-economic elements
among the management objectives of the Natura
2000 sites significantly complicates the manage-
ment methods (Pullin et al., 2009), the interest
in utilizing natural resources, and rendering
services being often of higher importance value
than conservation (Anthon, Garcia & Stenger,
2010; Berentsen, Hendriksen, Heijman & van
Vlokhoven, 2007).
To achieve the proposed conservation and
social goals, the Natura 2000 network requires
a considerable capital infusion of 5.58 billion €/
year for administration in EU developed coun-
tries only (Bladt et al., 2009). For comparison,
approximately the same amount is invested
yearly in conservation worldwide, with develop-
ing countries accounting for less than 12% of the
total (Bruner, Gullison & Balmford, 2004; James,
Gaston & Balmford, 2001).
The management process for the Natura 2000
sites is one aimed at transparency in decision mak-
1277
 Agriculture and Conservation in the Natura 2000 Network
ing and public disclosure (Maiorano et al., 2007;
Papp & Tóth, 2007), and mandates a significant
volume of spatial information for sustaining
restrictive decisions made in a region (Stolon,
2008). Information relating to the distribution
of protected species and habitats (Margules &
Pressey, 2000; Moilanen, Arponen, Stokland &
Cabeza, 2009; Wilson, Cabeza & Klein, 2009),
the manner in which risks are present (Hockings,
2003), the land use (Feranec, Jaffrain, Soukup &
Hazeu, 2010), or the habitat carrying capacity are
essential in avoiding the emergence of conflicts
among the protected areas managers and land
owners (Mauerhofer, 2010; Rauschmayer et al.,
2009; Young et al., 2007).
In this framework, GIS techniques represent
an important and integrated support for the pro-
cess of protected areas management, allowing a
systematic conservation planning (Margules &
Pressey, 2000; Wilson et al., 2009). For that reason,
the existence of spatial databases relating to the
distribution of species and habitats, the natural and
human risk characteristics, conflict or functional
areas in which specific rules must be adhered to,
are critical in fundament decisions formulated at
local, regional and global level in protected areas
(Bock et al., 2005; Boteva, Griffiths & Dimopou-
los, 2004; Mucher et al., 2009). GIS techniques
are essential in biodiversity management, as the
dimension of environmental changes are becoming
increasingly dynamic, and decisions must be made
by connecting diverse information (Feranec et al.,
2010). The failure of global and national politics
of biodiversity conservation is proportional to the
deficit of vital information (Gaston et al., 2008;
Ioja et al., 2010).
The objectives of this study are: a) assessing
the spatial distribution of the Natura 2000 network
by using GIS techniques; b) identifying methods
of analysing the relationship among agricultural
and conservation activities within Natura 2000,
and c) identification of digital tools for assessment
and management of conflict situations among
agricultural activities and conservation activities.
1278
This chapter, evidencing the dual nature of
agriculture, is structured into four sections. In the
introduction, the theoretical general framework
illustrates the relationship among Natura 2000
sites and agricultural activities, as presented at
an international level. Section one, Background,
presents a structured account of the Natura 2000
network, and correlates conservation activities
with agriculture generated issues in the Natura
2000 sites (abandonment of agriculture fields,
intensification of agricultural activities, conver-
sion of agricultural fields, development of inad-
equate utilization practices, and conflict with land
owners), at the same time listing the main data
sources used in assessing the relationship among
agriculture and Natura 2000. Section two identity
tools used in land-use analysis in the European
Union. The third section presents results obtained
by using GIS techniques concerning the spatial
distribution of Natura 2000 sites in EU-27 and the
dynamics of agricultural surfaces in the sites. The
last section of the article presents proposals and
future study areas for increasing the knowledge
level of the agricultural-conservation activities
relationship in Natura 2000 sites.
BACKGROUND
Globally, a large percentage of the population and
their agricultural activities, depend directly and/or
indirectly of biodiversity (Primack et al., 2008).
The accelerated loss of biodiversity impacts the
population’s ability to produce food in balance,
promotes poverty and associated problems, de-
creases the productivity of natural, semi-natural
and artificial ecosystems, broadens natural risks,
and commonly creates political instability (Balm-
ford et al., 2002; Cogalniceanu & Cogalniceanu,
2010; Mace & Baillie, 2007).
At the international level, the main instrument
used by countries is the Convention on Biological
Diversity (CBD), with the purpose of increasing
the efficiency of conservation measures at a local,
Agriculture and Conservation in the Natura 2000 Network
regional, and global level (Bock et al., 2005; Mace
& Baillie, 2007). CBD is aimed at identifying and
applying stimulants for preserving biodiversity,
promoting instruments and mechanisms of bio-
logical resources use, and the equitable distribu-
tion of the benefits of using biological resources
(Boisvert & Vivien, 2005).
What is Natura 2000 Network?
The European Union’s conservation policy is
directed to create a framework allowing member
states the organization of an ecological network
consisting of protected areas. Its purpose is to
optimize the efficiency of protection and manage-
ment actions for species and habitats of community
importance (Araujo, Lobo & Moreno, 2007; Pullin
et al., 2009; WG, 2002).
The Natura 2000 ecological network is the
cornerstone of conservation politics within the
European Union, being enacted through the
Habitats and Birds Directives (79/409/EEC, 1979;
92/43/EEC, 1992; 2006/105/EC, 2006). Through
the Natura 2000 network, conservation status
is offered to 198 habitats, 1288 plants species
(Chiarucci, Bacaro & Rocchini, 2008) and 1140
animal species, consisting of 986 vertebrates and
154 invertebrates (Fontaine et al., 2007; Ioja et
al., 2010). From the total habitats of community
interest listed in Annex I of the Habitats Directive,
16.16% are directly dependent on the continuation
of traditional agricultural practices, such as grass
cutting, pasturage, apiculture, etc., and others
(Henle et al., 2008). As an example, the lack of
pasturage activities can determine the progression
of areas from pastures to forests, resulting in the
loss of numerous species of plants and animals
adapted to those prior habitats (Baur et al., 2006;
Plieninger et al., 2006; Ruprecht et al., 2010).
Additionally, 65 natural habitats and 102 avian
species of community importance have an imperil
status due to agricultural practices (EEA, 2004).
The Natura 2000 network includes SPA (Spe-
cial Protected Areas, established under the Birds
Directive) and SCI (Sites of Community Impor-
tance, established under the Habitats Directive),
transforming after the letter of advice from the
European Commission in SAC (Special Areas
of Conservation) (European Commission, 2009;
Papp & Tóth, 2007). All Natura 2000 sites follow
a maintenance and/or restoration program in a
beneficial conservation manner consistent with
the state of natural habitat and/or populations of
species for which the site is designated (EEA,
2002, 2009).
Problems in Natura 2000 Network
Due to Agriculture Activities
The Natura 2000 is identified as a “social network”,
not only with regards to habitats and commu-
nity interest species conservation, but also to the
sustainable development of human settlements
(Cogalniceanu & Cogalniceanu, 2010). More to
the point, a series of economic problems, related
mainly to agricultural activities, are seen as sig-
nificant threats for species and natural habitats
within the Natura 2000 sites (EEA, 2005; Hartel,
Schweiger, Öllerer, Cogalniceanu & Arntzen,
2010; Rozylowicz & Dobre, 2010).
• The abandonment of agricultural fields
affects between 10% and 50% of the agri-
cultural land mass of EU-27 member states
(Henle et al., 2008), with an increased
propensity in countries such as the Czech
Republic, Ireland, Latvia, Romania, and
Bulgaria (Stoate et al., 2009). The aban-
donment of agricultural fields determines
the loss of traditional landscapes, increases
the risks of soil erosion, and deteriorates
habitats for numerous species (Baur et al.,
2006; Muller, 2002). EEA (2004) estimates
that between 5% and 65% of Important
Birds Areas are threatened by the abandon-
ment of agricultural fields, principally by
the transformation of pastures and grass-
lands into shrubs and forests.
1279
 Agriculture and Conservation in the Natura 2000 Network
• The intensification of agricultural ac-
tivities does not signify a normal reac-
tion to the integration of a territory in the
Natura 2000 ecological network, but still
is an important risk at the EU-27 level (es-
pecially in Ireland, Portugal, Spain and
the Baltic countries). Overutilization of a
region’s resources can produce a lessen-
ing in the natural and semi-natural ecosys-
tems’ capacity of support to biodiversity,
through an excessive infusion of pollutants
or through the depletion of nutritive sub-
stances (Kuemmerle et al., 2009; Schmitt
& Rakosy, 2007; Stoate et al., 2009).
• The conversion of agricultural fields into
dissimilar agricultural systems (as an ex-
ample, from perennial crops into annual)
or other categories (particularly forestry,
man-made permanent structures or surfac-
es, and similar categories) demonstrates a
serious risk to the biodiversity in EU-27,
as these changes suggest the destruction of
natural habitats (Stoate et al., 2009). This
demands sound recognition of the species’
distribution in the area and their key habitat
requirements, which can be realised by us-
ing GIS techniques (Feranec et al., 2010).
• The development of ineffective utiliza-
tion practices represent a risk difficult to
formulate without integrating prior experi-
ences concerning spatial databases, which
would then assist in the means for decision-
making. Ineffective practices involve any
manner of intervention affecting the stabil-
ity of natural ecosystems (as an example,
the premature mowing of grassland, mech-
anized intervention, etc.) (Kallimanis et
al., 2008; Muller, 2002; Piorr et al., 2009).
• The conflicts with farmers represent a
major issue amplified by the inclusion of
agricultural fields in the Natura 2000 net-
work. Interdictions and restrictions that
resulted from those inclusions amplified
territorial conflicts, particularly in regions
1280
where the resource utilization level was
relatively high (Henle et al., 2008).
Data for Natura 2000 Management
The management of Natura 2000 sites requires a
thorough knowledge of these threats, which must
be removed, minimized, or at worst controlled.
Towards this objective, GIS techniques present
a solid support, already being utilised in this
goal. Boteva et al. (2004), based on a multilevel-
criteria, analyse the significance of plant forma-
tions included in the Natura 2000 sites from the
island of Crete. Mucher et al. (2009) models the
spatial distribution of 27 Natura 2000 habitats in
Europe, identifying, at a continental scale, their
specific elements. Bock et al. (2005) uses the GIS
techniques in delimiting representative indicators
for monitoring biological diversity.
Amongst the notable data sources that may
be used in analysing the relationship agriculture-
Natura 2000 network, using GIS techniques, the
following are well-known:
1. EUNIS (http://eunis.eea.europa.eu): EUNIS
comprises a database created and maintained
by the European Topic Centre on Biological
Diversity for the European Environmental
Agency and European Environmental
Information Observation Network, whose
principal intent is to facilitate access to
information for environmental reports and
for the creation of the Natura 2000 network.
The database facilitates the extraction of data
such as the coherent description of species,
habitats and sites from the Natura 2000 net-
work, a description of species, habitats, sites
present in international conventions relevant
for biodiversity, and offers up-to-date data
about EEA activity in this field. EUNIS pro-
vides information for over 275,000 European
taxa. The degree of derived detail varies,
from taxonomic information to populations
and distribution. The database provides com-
Agriculture and Conservation in the Natura 2000 Network
prehensive information regarding various
habitats (natural or artificial, terrestrial or
aquatic) included in the EUNIS classifica-
tion and in the Annex I Habitats of the EU
Habitats Directive. EUNIS also provides
information regarding the protected areas
at the European Union level: Natura 2000
sites, nationally designated protected areas,
and internationally designated protected
areas – European Nature Diploma Areas
and Biogenetic Reserves.
2. Corine (Coordination of Information on
the Environment): Corine Land Cover
(CLC) is the first European wide compila-
tion of land cover information. The CLC
project is currently implemented in all
EU countries, and in Central and Eastern
European non-member countries (Belarus,
Ukraine, Moldova, etc.). The first compi-
lation reflects the situation as it existed in
1990 (CLC1990), with the next two decades
being updated and improved (CLC2000
and CLC2006). The compilations resulted
in several versions that eliminated the pro-
liferation of errors (geometric and thematic
mistakes). The scale (1:100,000), minimum
mapping unit (25 ha), and minimum width
of linear elements (100 meters) for CLC
mapping, represents parameters common
for CLC1990, CLC2000, and subsequently
CLC2006. The standard CLC nomenclature
includes 44 land cover classes, grouped into
three-levels, wherein the main level consists
of: 1) artificial surfaces, 2) agricultural
areas, 3) forests and semi-natural areas, 4)
wetlands, and 5) water bodies. CLC 2000
produced a new dataset whose objective
was to detect land cover changes (LCC) in
Europe occurring between 1990 and 2000,
and 2000 and 2006 compilations. The LCC
datasets aim to identify changes bigger than
5 ha.
3. Land and Ecosystem Accounting (LEAC):
is a means provided by the European Topic
Centre on Land Use and Spatial Information
that allows for the spatial analysis of land
cover change throughout the European
Union territory. This process describes
how, from a quantitative and qualitative
viewpoint, the ecosystems change over time.
To perform such an analysis, three online
databases are utilized:
◦◦ Land Cover Accounts (LEAC) based
on CORINE land cover changes da-
tabase. This database contains the
information on changes provided by
CLC2000 and CLC Changes 1990-
2000. This database is available for
Croatia, Bulgaria, Romania, Austria,
Belgium, Czech Republic, Denmark,
Estonia, France, Germany, Greece,
Hungary, Italy, Ireland, Latvia,
Lithuania, Luxembourg, Netherlands,
Poland, Portugal, Slovakia, Slovenia,
Spain, and the United Kingdom.
◦◦ PHARE 1975 – 1990 database. A
database that contains information
on changes identified using two time
horizons (late 1970’s – early 1990’s)
at the second level of CLC nomen-
clature, covering Bulgaria, Hungary,
Romania, Slovakia and the Czech
Republic.
◦◦ LaCoast 1975 - 1990 - Eastern Europe
Extension. LaCoast aims to quan-
tify the changes of land cover types
in a 10 km land strip inland from a
coastline. LaCoast uses CLC1990 as
its reference dataset and tracks dif-
ferences of land cover changes using
LANDSAT satellite images from the
mid-1970s. The dataset will be avail-
able for Belgium, Denmark, France,
Germany, Greece, Ireland, Italy,
Netherlands, Portugal, and portions
of Spain. An extension for Bulgaria,
Estonia, Latvia, Lithuania, Poland,
Slovenia, and Romania is avail-
1281
 Agriculture and Conservation in the Natura 2000 Network
able (LaCoast 1975 - 1990 - Eastern
Europe Extension).
4. EURURALIS (http://www.eururalis.eu):
EURURALIS is a study consisting of four
distinct world scenarios (Global Economy,
Global Cooperation, Continental Markets,
and Regional Communities). It has a time
horizon of three decades, in 10-year time
steps: 2000-2010, 2010-2020, 2020-2030,
and encompasses EU27 countries. The
projects have two objectives: to provide a
means for a structured strategic discussion
at a European scale, and to develop and/or
improve EURURALIS, by which a scientifi-
cally sound framework for integrated assess-
ment at the European level may be realized.
The results should enable meaningful nego-
tiations among policy makers, stakeholders
and scientists from various domains and/
or countries. The results of EURURALIS
should potentially summarize what may
occur in rural Europe towards 2030, based
on conditions that differ in nature, course,
duration, and/or place.
NATURA 2000: AGRICULTURE
RELATION USING GIS
Natura 2000 National Distribution
The current dimension of the Natura 2000 net-
work is extraordinary: 22,419 sites of community
importance covering 716,992 km2 (13.6% of the
continental surface), and 5242 special protected
areas, covering 574,819 km2 (11.1% of the con-
tinental surface) (European Commission, 2009).
The network is not distributed homogenously
in member states, their percentage of national ter-
ritory varying from the average value of 17.4%.
Thus, at a national level, the largest percentage
of fields managed in a regime of NATURA 2000
site appear in Slovenia (35.5%), Bulgaria (34%),
Slovakia (28.9%), and Spain (20.7%). The small-
1282
est percentages are in countries such as United
Kingdom (7.1%) and Denmark (8.9%).
The area of SCI’s (Sites of Community Impor-
tance) varies from 6.8% in the United Kingdom to
31.4% in Slovenia, the average value being 12.7%
(±6.02%). New EU members have percentages
of their national territory devoted to SCI of over
10%, with lowest values in the Czech Republic
(10%), Poland (11%), Latvia (11.3%), Cyprus
(11.5%), Slovakia (11.7%), with the highest in
Slovenia (31.4%), Bulgaria (29.6%), and Spain
(24.5%). In countries with a high economic level,
the percent of SCI’s is significantly lower, being
generally under 10% of their national territory.
Numerically, the largest number of SCI’s ap-
pears in Germany (4622) and Sweden (3983),
and the lowest in smaller countries (Cyprus,
Luxemburg, Malta). This represents an important
characteristic, particularly for understanding the
management effort for these protected areas at
the national level, as they have also a notable
conservative value (Gaston et al., 2008).
Some countries prefer having a greater number
of SCI’s, with smaller areas distributed on a larger
territory (e.g. Germany, the Czech Republic, Italy,
Sweden, Slovakia), while other countries choose
SCI’s with larger surfaces, in many cases concen-
trated in regions with lower population densities
(Portugal, Bulgaria, Greece, and Romania).
In Romania, the average surface area of an
SCI is 120.3 km2. The highest concentration of
these SCI’s is found in the Carpathian Mountains
and the Danube Delta areas, where the highest
preservation degree is recorded (Ioja, et al., 2010).
Thus, in a 0 m (delta area) and over 800 m areas
(Carpathian Mountain area), the protection of the
territory by the Natura 2000 sites exceeds 35%.
Lowest values (below 20%) occur at altitudes
below 600 m, where the suitability for human
activity is much higher (Rey, Groza, Ianoş &
Patroescu, 2007) (Figure 1).
The distribution of SPA’s (Special Protected
Areas) is dependent of the existence of avian
preferred habitats, such as wetlands, forestry areas,
Agriculture and Conservation in the Natura 2000 Network
Figure 1. Elevation distribution of NATURA 2000 sites in Romania
and/or widespread agricultural fields (Table 1).
The percent of SPA’s of the continental surface
of EU-27 is 11.9%, ranging from 2.9% of the
national territory in Ireland to 25.2% in Slovakia.
The smallest percentages (under 10%) are re-
corded in smaller countries (Malta, Cyprus,
Luxemburg) or those with a high percentage of
human altered habitats (Ireland, France). The
highest percentage occurs in countries that also
have large areas devoted to SCI’s, respectively
Slovakia (25.1%), Slovenia (23%), Bulgaria
(20.4%), and Spain (20.6%). Regarding the dis-
tribution of SPA’s average area at a national
level, it is observed that the lowest values occur
in Malta (1.2 km2), Luxemburg (11.2 km2), Bel-
gium (14 km2), with the highest in Poland (391
km2), Slovakia (322 km2), and Romania (292.7
km2) (Table 1).
Using GIS techniques, the sufficiency degree
in comprising natural habitats and species of
community interest within the Natura 2000 net-
work is revealed. Thus, at a national level, the
40%-60% rule (Papp & Tóth, 2007) is embraced
by five countries (Denmark, Netherland, Italia,
Belgium, and Finland). Therefore, values under
100% assume a new expansion of the Natura 2000
network at a national level, with impacts on land
use dynamics.
In Romania, with regards to species classes,
it was observed that most of the problems occur
within invertebrates and plants, where 40.9%
protected invertebrate species, respectively 17.9%
protected plant species do not benefit from protec-
tion inside a SCI (Ioja et al., 2010; Patroescu, Ioja,
Patroescu-Klotz & Necsuliu, 2006) (Figure 2).
This is of considerable importance in the network
expansion process or in its re-evaluation (European
Commission, 2009; Papp & Tóth, 2007).
With the use of GIS techniques, the distribution
of Natura 2000 sites may be observed in the nine
biogeographic regions of the European Union
(EEA, 2002), resulting in important information
in estimating their degree of conservation. Thus,
the best inclusion rate in Natura 2000 sites is
recorded in the Macaronesian and Black Sea
bioregions (100%), with the lowest inclusion rates
in the steppic, pannonian and continental biore-
gions (under 90%), the latter supporting the most
intense agricultural activities (Feranec et al., 2010;
Mucher et al., 2009).
Furthermore, to judge management effective-
ness it is important to understand the overlapping
levels of SCI’s, SPA’s, and the protected areas of
national interest. The overlapping of SCI’s and
SPA’s varies in EU-27 from 0% (Latvia) up to
52.8% (Slovenia). In many countries, overlaps
1283
 Agriculture and Conservation in the Natura 2000 Network

Table 1. NATURA 2000 sites distribution in relation to agriculture in EU-27

Average Average
% NATURA SCIs surfaces surface per % SCI SPA surfaces surfaces per % SPA
Country 2000 areas of NT km2 SCI km2 of NT km2 SPA km2 of NT
Austria 13,8 8978 53,4 10,7 9869 102,8 11,8
Belgium 12,7 3269 11,7 10,1 3282 14,0 9,7
Bulgaria 34 33430 146,6 29,6 23217 203,7 20,4
Cyprus 17,5 711 19,8 11,5 788 112,6 13,4
Czech 13,9 7854 7,3 10 9684 248,3 12,3
Deutschland 15,4 54342 11,8 9,7 59784 81,0 12,2
Denmark 8,9 19319 74,0 7,4 14718 130,2 5,9
Estonia 17,7 11321 21,3 16,7 12592 190,8 13,5
Spain 28,7 131434 90,8 24,5 105032 175,3 20,6
Finland 14,4 48552 28,3 12,7 30838 65,9 7,5
France 12,5 72418 53,0 8,5 76297 200,3 7,9
Greece 21 27798 116,3 16,3 16740 102,7 11,9
Hungary 21,1 13973 29,9 15 13512 245,7 14,5
Ireland 11,1 13558 32,0 10,7 2827 21,6 2,9
Italy 19,1 45309 19,8 14,3 43777 73,3 13,6
Lithuania 13,9 8822 31,6 13,2 6013 74,2 9
Luxemburg 17,8 399 8,3 15,4 145 11,2 5,6
Latvia 11,3 7856 24,2 11,3 6999 73,7 10
Malta 13 50 1,8 13,3 16 1,2 5,1
Netherland 13,8 14342 98,2 8,4 10125 131,5 12,6
Poland 19,4 38003 46,2 11 55228 391,7 15,6
Portugal 20,9 16788 174,9 17,4 10438 176,9 10,7
Romania 17,9 32833 120,3 13,2 31905 292,7 13,4
Sweden 13,8 64468 16,2 13,7 29874 56,3 6,2
Slovenia 35,5 6360 24,6 31,4 4656 172,4 23
Slovakia 28,9 5739 15,0 11,7 12236 322,0 25,1
United Kingdom 7,1 29066 46,7 6,8 16114 62,7 6,2
Total/ Average 17,6 716992 32,0 12,7 606706 115,7 11,9

of SCI’s and SPA’s occur on over 30% of their
total area (Belgium, Bulgaria, Slovenia, Romania,
United Kingdom, and Denmark), however, in
many circumstances, this is not total overlapping.
Thus, areas benefiting from several simultaneous
conservation status (protected areas of national
interest, SCI, SPA), are subject to difficulties in
their administration, for the management of hu-
man activities is inherently dissimilar in different
categories (Ioja et al., 2010).
Dynamics of Agriculture Areas
in Natura 2000 Sites in EU
Conservation activities are in direct competition
with activities for territorial resources utilization.
In Natura 2000 sites, all activities or actions that
may affect the conservation state of habitats and
species of community interest for which they were
designated are restricted or prohibited (Dimitra-
kopoulos et al., 2004; Mauerhofer, 2010).

1284
Agriculture and Conservation in the Natura 2000 Network
Figure 2. Representation degree of group species in the Romania SCIs
CORINE Land Cover and EUNIS databases,
with the aid of the ArcGis Desktop 9.3, were used
to determine the overlapping degrees of agricul-
tural fields in Natura 2000 sites. For identifying
the dynamics of agricultural fields between 2000
and 2006, CORINE Land Cover database for the
respective years was employed, observing the
evolution for the Natura 2000 sites.
Agricultural fields with high natural conser-
vation values represent 32% of the agricultural
areas of EU-27, with reduced percentages in
Baltic countries and Benelux, and with higher
percentages in Mediterranean countries, Austria,
Bulgaria, and Finland.
It is essential to understand the dynamics of
agricultural activities in order to administer Natura
2000 sites that comprise 9.2% of the agricultural
areas of EU-27 (from 0.9% in Finland to 22.3% in
Slovenia) that were being integrated into Natura
2000 sites (Table 2). The greatest percentage of
agricultural fields in Natura 2000 sites are located
in Mediterranean countries and in some countries
from the last enjoinment stages (Bulgaria, Slo-
venia, and Slovakia). Even if the purpose of the
NATURA 2000 network is not that of creating
wildlife sanctuaries, allowing economic activities
that do not convey perceivable disadvantages,
or even assisting preserved elements (such as
manual grasslands mowing, pasturing, extensive
plants culture, sustainable exploitation of natural
resources (wood, forest fruits, mushrooms, fish-
ing, or hunting) (Bladt et al., 2009; Gaston et al.,
2008; Phillips, 1994), any protected area imposes
narrow to broad restrictions in resources utilization
(Cogalniceanu & Cogalniceanu, 2010).
For that reason, Natura 2000 sites are areas in
which most of the technologies for intensive
agricultural utilization are incompatible with
conservation objectives (mechanization, the use
of fertilisers and pesticides, pest species control,
genetic modified organism) (Petrovan, Barrio,
Ward & Wheeler, 2010; Plieninger et al., 2006;
Pullin et al., 2009).
Within EU-27, 15.8% of arable land and 22.8%
of pastures and grasslands are utilized in an inten-
sive regimen, and therefore it is important to rec-
ognize their overlapping with currently designated
Natura 2000 sites. In the European Union, there
are three types of agricultural utilization systems
recognized as compatible with Natura 2000 sites
(EEA, 2004, 2005): type 1 – with a high percent-
age of semi-natural vegetation; type 2 – farms
dominated by lowered intensity agriculture or the
existence of natural-cultivate mosaics, and type
3 – farms containing rare species.
1285
 Agriculture and Conservation in the Natura 2000 Network

Table 2. Natura 2000 sites distribution in relation to agriculture in EU-27

Agricultural % agriculture Evolution of agriculture

% agriculture surfaces in land from land in NATURA 2000
land in % HVAF of NATURA 2000 NATURA 2000 sites between 2000 and
Country NATURA 2000 agricultural land km2 surfaces 2006
Austria 11,4 68 4079,6 35,4 No data
Belgia 7,3 19 1304,5 33,6 0,01
Bulgaria 22 37 14815,3 39,3 0,032
Cyprus 1,5 33 95,6 9,5 0,028
Czech 6,4 21 3168,6 28,8 0
Deutchland 9,1 15 19662,7 35,8 No data
Denmark 4,9 5 1688,6 44,0 -0,013
Estonia 5,5 22 932,7 11,6 +0.011
Spain 15,5 56 52760,3 36,5 No data
Finland 0,9 45 267,0 0,5 No data
France 8,2 22 28955,7 42,1 -0,011
Greece 14,2 59 12953,6 46,7 No data
Hungary 14,5 28 9893,2 50,3 +0.132
Ireland 3,1 20 1791,0 23,0 +0.130
Italy 10,3 33 18910,4 32,9 No data
Lithuania 3,8 15 1580,7 17,4 -0.033
Luxemburg 11 9 156,9 34,0 -0.013
Latvia 6,6 20 1883,4 25,9 0
Malta 5,8 no data no data no data 0.026
Netherland 4,6 14 1206,0 21,0 -0.18
Poland 10,7 24 21648,1 35,6 +0.003
Portugal 18,5 58 9316,4 48,4 +0.060
Romania 9,6 34 13856,6 32,5 +0.013
Sweden 4,3 24 2046,7 3,6 No data
Slovenia 22,3 78 1682,0 23,4 -0.064
Slovakia 16,3 22 4051,3 28,7 -0.353
United Kingdom 3 27 5810,5 33,6 No data
Total/Average 9,23 32 215619 28,6 -0.012

The importance of agricultural fields is ampli-
fied by the reality that 28.6% of the area of Natura
2000 sites is represented by agricultural fields,
with the lowest values being recorded in Nordic
countries (under 5%), and the highest in Mediter-
ranean countries or in those regions containing a
high percentage of plains and low hills (Hungary,
Romania, Denmark, and Poland) (Table 2). The
management of these areas is made difficult, with
agricultural sources framed as diffuse environ-
mental degradations and therefore, challenging
to control and monitor (Stoate et al., 2009).
Additionally, by analysing the prevalence of
Natura 2000 sites upon agricultural activities, a
decreasing tendency is observed in the 2000-2006
period, especially in countries in which the sites

1286
Agriculture and Conservation in the Natura 2000 Network
were established prior to 2006. Thus, in countries
such as Denmark, Netherlands, or France, reduc-
tion of agricultural areas within Natura 2000 sites
is already perceivable and forewarns of social
and economical incidences that may be deter-
mined by models such as land use, principally
those without economic support (Mauerhofer,
2010; Rauschmayer et al., 2009). In exchange,
countries in which the rehabilitation of traditional
agricultural landscapes present an advantage for
conservation activities from Natura 2000 sites
(Portugal) or countries demonstrating a higher
development of agriculture (Ireland, Hungary),
present expansion of agricultural areas (European
Commission, 2009) (Table 2).
Solutions and Recommendations
The European Union’s policy in the field of
biodiversity is in many aspects inordinate, af-
fecting the already fragile networks of protected
areas, principally in new member states (Ioja et
al., 2010). The orientation towards social and
economic aspects of the management methods
does nothing more than sidetrack attention from
the priority objective of protected areas, to that
of biodiversity conservation (Dimitrakopoulos et
al., 2004; Rauschmayer et al., 2009). More to the
point, the extensive development of the protected
area network does not contribute to alleviating
social and economic problems of disadvantaged
communities, generating at their level strong
regional tensions through imposed restriction
and the lack of financial compensation (Anthon
et al., 2010; Mauerhofer, 2010; Rauschmayer et
al., 2009). The mechanism that offers, just for
conservation, financial compensation, although
generous (200-450 €/ha/year) (Anthon et al., 2010;
Piorr et al., 2009), it is relatively difficult to ac-
cess, especially for disadvantaged communities;
where most of these problems converge (Young
et al., 2007). Thus, most financial compensations
are directed to countries with lowered biodiversity
but a strong economy (France, Germany, and Hol-
land), while countries with lowered development
do not portray the capacity of absorbing these
funds (Bladt et al., 2009).
The large numbers of species that typically
benefit from a strict preservation regime, coupled
with the tendency of covering large areas, predict-
ably mystify management objectives and the huge
funds necessary for management actions (James
et al., 2001). It is problematic to create a Natura
2000 network with projected low probability of
enforcement (Pullin et al., 2009) without a spatial
extend reassessment based on the principles of
spatial planning for conservation (Araujo et al.,
2007; Ioja et al., 2010; Wilson et al., 2009).
The efficiency of protected areas of European
regions should be linked to the design of protected
areas of controllable dimensions, which bestow
favourable habitats, principally for vulnerable,
endangered, and critically endangered plant and
animal species, and with mechanisms through by
which actions affecting natural ecosystems are
kept away from protected areas (Ioja et al., 2010).
Stimulating eco-labelling and the development
of a bio-agricultural product’s market is an ob-
jective that may be promoted in the Natura 2000
sites, although the economic efficiency of these
activities is uncertain (Bladt et al., 2009; EEA,
2004; Piorr et al., 2009; Plieninger et al., 2006;
Stoate et al., 2009). Member states should have
greater freedom in prioritizing protected areas and
in choosing the types of protected areas they can
efficiently manage (Rauschmayer et al., 2009;
Rozylowicz, L., Popescu, V. D., Patroescu, M.,
& Chisamera, G., 2011).
FUTURE RESEARCH DIRECTIONS
In order to benefit from the advantages of the Na-
tura 2000 network, and for establishing a proper
management organization, significant financial
resources are required, estimated at 6.1 billion €
each year (Bladt et al., 2009; WG, 2002), compa-
rable with those spent currently at a global level
1287
 Agriculture and Conservation in the Natura 2000 Network
(Balmford, Gaston, Blyth, James & Kapos, 2003;
James et al., 2001). These financial resources are
necessary for the management process, but prin-
cipally for offering compensations and financial
incentives to landowners or other entities for the
restrictions imposed on them (Anthon et al., 2010).
The assignment of any compensation and/or
stimulant-necessitates sound awareness of field
realities, respectively - the productivity of the
land, the presence of threatened habitats and spe-
cies, losses recorded by various resource users,
etc. In this objective, GIS methods should have
an important role, as they permit correlations
of multiple parameters needed in the process of
Natura 2000 sites management.
Spatial databases regarding the Natura 2000
network are in an evolutionary process, and the
diversity of future research is extremely high,
especially as a well-managed network based on
quality data. In this objective, the monitoring of
agricultural fields dynamics in Natura 2000 sites is
paramount, specifically regarding their conversion
and the agricultural methods used. These assess-
ments should be correlated with the state of natural
habitats and all species of community importance
for which a particular model of agricultural use
is applied. This is particularly important because
this conservation status may suggest acceleration
in the utilization of regional resources.
Special attention in the managing process
should be given to the monitoring of social and
economic benefits of communities from Natura
2000 sites, essential in maintaining this manner
of resource use.
CONCLUSION
The process of managing Natura 2000 sites is
at its beginning in EU-27. Scepticisms concern-
ing the success of this politics of biodiversity
social conservation are elevated, as the financial
resources necessary for the management process
are lacking. Additionally, restrictions imposed
1288
in promoting agricultural activities accentuate
much of the tensions present in local communi-
ties, where important conflicts currently subsist
(i.e. Germany). Natura 2000 is an objective of the
European Union, being defined as an ambitious
conservation program, a position that shapes
negative-national conservation politics.
Therefore, GIS techniques must represent the
tool by which the efficiency of this ecological
network is monitored, as it must be permanently
nourished with important financial resources.
ACKNOWLEDGMENT
We would like to thank Edward F. Rozylowicz
for proofreading and valuable comments and
anonymous reviewers for their suggestions.
REFERENCES
2006/105/EC. (2006). Council Directive 2006/105/
EC of 20 November 2006 adapting Directives
73/239/EEC, 74/557/EEC and 2002/83/EC in the
field of environment, by reason of the accession
of Bulgaria and Romania: Official Journal L 363,
20/12/2006.
79/409/EEC. (1979). Council Directive 79/409/
EEC of 2 April 1979 on the conservation of wild
birds (Vol. 79/409/EEC): Official Journal L 103,
25/04/1979.
92/43/EEC. (1992). Council Directive 92/43/EEC
of 21 May 1992 on the conservation of natural
habitats and of wild fauna and flora (Vol. 92/43/
EEC): Official Journal of E.U., L 206, 22.7.1992.
Anthon, S., Garcia, S., & Stenger, A. (2010).
Incentive contracts for Natura 2000 implementa-
tion in forest areas. Environmental and Resource
Economics, 46(3), 281–302..doi:10.1007/s10640-
009-9341-1
Agriculture and Conservation in the Natura 2000 Network
Araujo, M. B., Lobo, J. M., & Moreno, J. C. (2007).
The effectiveness of Iberian protected areas in
conserving terrestrial biodiversity. Conservation
Biology, 21(6), 1423–1432..doi:10.1111/j.1523-
1739.2007.00827.x
Balmford, A., Bruner, A., Cooper, P., Costanza,
R., Farber, S., & Green, R. E. (2002). Ecology
- economic reasons for conserving wild nature.
Science, 297(5583), 950–953..doi:10.1126/sci-
ence.1073947
Balmford, A., Gaston, K. J., Blyth, S., James, A.,
& Kapos, V. (2003). Global variation in terrestrial
conservation costs, conservation benefits, and
unmet conservation needs. Proceedings of the
National Academy of Sciences of the United States
of America, 100(3), 1046–1050..doi:10.1073/
pnas.0236945100
Baur, B., Cremene, C., Groza, G., Rakosy, L.,
Schileyko, A. A., & Baur, A. (2006). Effects of
abandonment of subalpine hay meadows on plant
and invertebrate diversity in Transylvania, Roma-
nia. Biological Conservation, 132(2), 261–273..
doi:10.1016/j.biocon.2006.04.018
Berentsen, P. B. M., Hendriksen, A., Heijman, W.
J. M., & van Vlokhoven, H. A. (2007). Costs and
benefits of on-farm nature conservation. Ecologi-
cal Economics, 62(3-4), 571–579..doi:10.1016/j.
ecolecon.2006.07.026
Bladt, J., Strange, N., Abildtrup, J., Svenning, J.
C., & Skov, F. (2009). Conservation efficiency of
geopolitical coordination in the EU. Journal for
Nature Conservation, 17(2), 72–86..doi:10.1016/j.
jnc.2008.12.003
Bock, M., Rossner, G., Wissen, M., Remm, K.,
Langanke, T., & Lang, S. (2005). Spatial indicators
for nature conservation from European to local
scale. Ecological Indicators, 5(4), 322–338..
doi:10.1016/j.ecolind.2005.03.018
Boisvert, V., & Vivien, F. D. (2005). The conven-
tion on biological diversity: A conventionalist ap-
proach. Ecological Economics, 53(4), 461–472..
doi:10.1016/j.ecolecon.2004.06.030
Boteva, D., Griffiths, G., & Dimopoulos, P. (2004).
Evaluation and mapping of the conservation sig-
nificance of habitats using GIS: An example from
Crete, Greece. Journal for Nature Conservation,
12(4), 237–250..doi:10.1016/j.jnc.2004.09.002
Bruner, A. G., Gullison, R. E., & Balmford, A.
(2004). Financial costs and shortfalls of managing
and expanding protected-area systems in develop-
ing countries. Bioscience, 54(12), 1119–1126..
doi:10.1641/0006-3568(2004)054[1119:FCAS
OM]2.0.CO;2
Chiarucci, A., Bacaro, G., & Rocchini, D. (2008).
Quantifying plant species diversity in a Natura
2000 network: Old ideas and new proposals.
Biological Conservation, 141(10), 2608–2618..
doi:10.1016/j.biocon.2008.07.024
Cogalniceanu, D., & Cogalniceanu, G. C. (2010).
An enlarged European Union challenges prior-
ity settings in conservation. Biodiversity and
Conservation, 19(5), 1471–1483..doi:10.1007/
s10531-010-9777-1
Dimitrakopoulos, P. G., Memtsas, D., & Troumbis,
A. Y. (2004). Questioning the effectiveness of the
Natura 2000 special areas of conservation strategy:
The case of Crete. Global Ecology and Bioge-
ography, 13(3), 199–207..doi:10.1111/j.1466-
822X.2004.00086.x
EEA. (2002). Europe’s biodiversity—biogeo-
graphical regions and seas (Vol. 1). Copenhagen,
Denmark: European Environmental Agency.
EEA. (2004). High nature value farmland - char-
acteristics, trends and policy challenges (Vol. 1).
Copenhagen, Denmark: European Environment
Agency.
1289
 Agriculture and Conservation in the Natura 2000 Network
EEA. (2005). Agriculture and environment in
EU-15—the IRENA indicator report (Vol. 6).
Copenhagen, Denmark: European Environmental
Agency.
EEA. (2009). Progress towards the European
2010 biodiversity target (Vol. 4). Copenhagen,
Denmark: European Environmental Agency.
European Commission. (2009). Natura 2000 ba-
rometer - December 2008. Natura 2000. European
Commission DG ENV Nature Newletter, 8-9.
Feranec, J., Jaffrain, G., Soukup, T., & Hazeu,
G. (2010). Determining changes and flows in
European landscapes 1990-2000 using CORINE
land cover data. Applied Geography (Seveno-
aks, England), 30(1), 19–35..doi:10.1016/j.ap-
geog.2009.07.003
Fontaine, B., Bouchet, P., Van Achterberg, K.,
Alonso-Zarazaga, M. A., Araujo, R., & Asche,
M. (2007). The European Union’s 2010 target:
Putting rare species in focus. Biological Con-
servation, 139(1-2), 167–185..doi:10.1016/j.
biocon.2007.06.012
Gaston, K. J., Jackson, S. E., Nagy, A., Cantu-
Salazar, L., & Johnson, M. (2008). Protected
areas in Europe - principle and practice. Annals
of the New York Academy of Sciences, 1134(1),
97–119..doi:10.1196/annals.1439.006
Hartel, T., Schweiger, O., Öllerer, K., Cogalnicea-
nu, D., & Arntzen, J. W. (2010). Amphibian distri-
bution in a traditionally managed rural landscape
of Eastern Europe: Probing the effect of landscape
composition. Biological Conservation, 143(5),
1118–1124..doi:10.1016/j.biocon.2010.02.006
Henle, K., Alard, D., Clitherow, J., Cobb, P.,
Firbank, L., & Kull, T. (2008). Identifying and
managing the conflicts between agriculture and
biodiversity conservation in Europe - a review.
Agriculture Ecosystems & Environment, 124(1-2),
60–71..doi:10.1016/j.agee.2007.09.005
1290
Hockings, M. (2003). Systems for assessing the
effectiveness of management in protected areas.
Bioscience, 53(9), 823–832..doi:10.1641/0006-
3568(2003)053[0823:SFATEO]2.0.CO;2
Ioja, C. I., Patroescu, M., Rozylowicz, L., Popescu,
D. V., Vergheleţ, M., & Zotta, M. I. (2010). The
efficacy of Romania’s protected areas network
in conserving biodiversity. Biological Con-
servation, 143(11), 2468–2476..doi:10.1016/j.
biocon.2010.06.013
James, A., Gaston, K. J., & Balmford, A.
(2001). Can we afford to conserve biodiversity?
Bioscience, 51(1), 43–52..doi:10.1641/0006-
3568(2001)051[0043:CWATCB]2.0.CO;2
Kallimanis, A. S., Tsiafouli, M. A., Pantis, J. D.,
Mazaris, A. D., Matsinos, Y., & Sgardelis, S. P.
(2008). Arable land and habitat diversity in Na-
tura 2000 sites in Greece. Journal of Biological
Research-Thessaloniki, 9, 55–66.
Kuemmerle, T., Muller, D., Griffiths, P., & Rusu,
M. (2009). Land use change in Southern Roma-
nia after the collapse of socialism. Regional
Environmental Change, 9(1), 1–12..doi:10.1007/
s10113-008-0050-z
Mace, G. M., & Baillie, J. E. M. (2007). The 2010
biodiversity indicators: Challenges for science and
policy. Conservation Biology, 21(6), 1406–1413..
doi:10.1111/j.1523-1739.2007.00830.x
Maiorano, L., Falcucci, A., Garton, E. O., &
Boitani, L. (2007). Contribution of the Natura
2000 network to biodiversity conservation in
Italy. Conservation Biology, 21(6), 1433–1444..
doi:10.1111/j.1523-1739.2007.00831.x
Margules, C. R., & Pressey, R. L. (2000). System-
atic conservation planning. Nature, 405(6783),
243–253. doi:10.1038/35012251
Agriculture and Conservation in the Natura 2000 Network

Mauerhofer, V. (2010). Missing links: How indi- Phillips, A. (1994). The IUCN action plan for
viduals can contribute to reserve policy enforce- protected areas in Europe. Conserving Europe’s
ment on the example of the European Union. Natural Heritage, 69-73.
Biodiversity and Conservation, 19(3), 601–618..
Piorr, A., Ungaro, F., Ciancaglini, A., Happe, K.,
doi:10.1007/s10531-009-9737-9
Sahrbacher, A., & Sattler, C. (2009). Integrated
Moilanen, A., Arponen, A., Stokland, J. N., & assessment of future CAP policies: Land use
Cabeza, M. (2009). Assessing replacement cost of changes, spatial patterns and targeting. Envi-
conservation areas: How does habitat loss influ- ronmental Science & Policy, 12(8), 1122–1136..
ence priorities? Biological Conservation, 142(3), doi:10.1016/j.envsci.2009.01.001
575–585..doi:10.1016/j.biocon.2008.11.011
Plieninger, T., Hochtl, F., & Spek, T. (2006).
Mucher, C. A., Hennekens, S. M., Bunce, R. G. Traditional land-use and nature conservation
H., Schaminee, J. H. J., & Schaepman, M. E. in European rural landscapes. Environmental
(2009). Modelling the spatial distribution of Na- Science & Policy, 9(4), 317–321..doi:10.1016/j.
tura 2000 habitats across Europe. Landscape and envsci.2006.03.001
Urban Planning, 92(2), 148–159..doi:10.1016/j.
Primack, R. B., Patroescu, M., Rozylowicz, L.,
landurbplan.2009.04.003
& Ioja, C. (2008). Fundamentele conservării
Muller, S. (2002). Appropriate agricultural man- diversităţii biologice. Bucureşti, Romania: AGIR.
agement practices required to ensure conservation
Pullin, A. S., Baldi, A., Can, O. E., Dieterich,
and biodiversity of environmentally sensitive
M., Kati, V., & Livoreil, B. (2009). Conserva-
grassland sites designated under Natura 2000.
tion focus on Europe: Major conservation policy
Agriculture Ecosystems & Environment, 89(3),
issues that need to be informed by conservation
261-266. doi: S0167-8809(01)00235-3
science. Conservation Biology, 23(4), 818–824..
Papp, D., & Tóth, C. (2007). Natura 2000 site doi:10.1111/j.1523-1739.2009.01283.x
designation process with a special focus on the
Pykala, J. (2003). Effects of restoration with
biogeographic seminars (2nd ed., p. 36). Budapest,
cattle grazing on plant species composition
Hungary: CEEWEB Office.
and richness of semi-natural grasslands. Biodi-
Patroescu, M., Ioja, I. C., Patroescu-Klotz, I., & versity and Conservation, 12(11), 2211–2226.
Necsuliu, R. (2006). Umweltqualitat in Rumanien. doi:10.1023/A:1024558617080
In Kahl, T., Metzeltin, M., & Ungureanu, R. (Eds.),
Rauschmayer, F., van den Hove, S., & Koetz, T.
Rumänien. Raum und Bevölkerung. Geschichte
(2009). Participation in EU biodiversity gover-
und Geschichtsbilder.Kultur. Gesellschaft und
nance: How far beyond rhetoric? Environment
Politik heute. Wirtschaft. Recht undVerfassung.
and Planning. C, Government & Policy, 27(1),
Historische Regionen (p. 984). Münster/Hamburg/
42–58..doi:10.1068/c0703j
Berlin/Wien, Germany; London, UK; Zürich,
Switzerland: LIT Verlag. Rey, V., Groza, O., Ianoş, I., & Patroescu, M.
(2007). Atlas de la Roumanie. Montpelier/Paris,
Petrovan, S. O., Barrio, I. C., Ward, A. I., &
France: Reclus.
Wheeler, P. M. (2010). Farming for pests? Local
and landscape-scale effects of grassland manage-
ment on rabbit densities. European Journal of
Wildlife Research, 57(1), 27–34..doi:10.1007/
s10344-010-0394-9

1291
 Agriculture and Conservation in the Natura 2000 Network
Rozylowicz, L., & Dobre, M. (2010). Assessing
the threatened status of Testudo hermanni boettgeri
Mojsisovics, 1889 (Reptilia: Testudines: Testudi-
nidae) population from Romania. North-Western
Journal of Zoology, 6(2), 190–202.
Rozylowicz, L., Popescu, V. D., Patroescu, M.,
& Chisamera, G. (2011). The potential of large
carnivores as conservation surrogates in the Roma-
nian Carpathians. Biodiversity and Conservation,
20(3), 561–579..doi:10.1007/s10531-010-9967-x
Ruprecht, E., Enyedi, M. Z., Eckstein, R. L., &
Donath, T. W. (2010). Restorative removal of plant
litter and vegetation 40 years after abandonment
enhances re-emergence of steppe grassland vegeta-
tion. Biological Conservation, 143(2), 449–456..
doi:10.1016/j.biocon.2009.11.012
Schmitt, T., & Rakosy, L. (2007). Changes of
traditional agrarian landscapes and their conser-
vation implications: A case study of butterflies in
Romania. Diversity & Distributions, 13, 855–862..
doi:10.1111/j.1472-4642.2007.00347.x
Silva, C. N. (2009). Protected areas and regional
development in Europe: Towards a new model
for the 21st Century. Tijdschrift voor Econo-
mische en Sociale Geografie, 100(1), 129–131..
doi:10.1111/j.1467-9663.2009.514_2.x
Stoate, C., Baldi, A., Beja, P., Boatman, N. D.,
Herzon, I., & van Doorn, A. (2009). Ecological
impacts of early 21st century agricultural change
in Europe - a review. Journal of Environmental
Management, 91(1), 22–46..doi:10.1016/j.jenv-
man.2009.07.005
Stolon, S. (Ed.). (2008). Assessment of manage-
ment effectiveness in European protected areas.
Sharing experiences and promoting good manage-
ment. Bonn, Germany: Bundesant für Naturschutz.
WG. (2002). Working Group on Article 8 of the
Habitats Directive. Final Report on Financing
Natura 2000. Retrieved from www.eeb.org/activi-
ties /biodiversity/Financing-Natura- 2000-WG-
final-report-art8.pdf
1292
Wilson, K. A., Cabeza, M., & Klein, C. J. (2009).
Fundamental concepts of spatial conservation
prioritization. In Moilanen, A., Wilson, K. A., &
Possingham, H. P. (Eds.), Spatial conservation
prioritization. Quantitative methods & computa-
tional tools (pp. 16–28). New York, NY: Oxford
University Press.
Young, J., Richards, C., Fischer, A., Halada, L.,
Kull, T., & Kuzniar, A. (2007). Conflicts between
biodiversity conservation and human activities
in the central and eastern European countries.
Ambio, 36(7), 545–550. doi:10.1579/0044-
7447(2007)36[545:CBBCAH]2.0.CO;2
ADDITIONAL READING
Baylis, K., Peplow, S., Rausser, G., & Simon, L.
(2008). Agri-environmental policies in the EU
and United States: A comparison. Ecological
Economics, 65(4), 753–764..doi:10.1016/j.eco-
lecon.2007.07.034
Brooke, C. (2008). Conservation and Adapta-
tion to Climate Change. Conservation Biol-
ogy, 22(6), 1471–1476..doi:10.1111/j.1523-
1739.2008.01031.x
Bunce, R. G. H., Metzger, M. J., Jongman, R. H.
G., Brandt, J., De Blust, G., & Elena-Rossello, R.
(2008). A standardized procedure for surveillance
and monitoring European habitats and provision of
spatial data. Landscape Ecology, 23(1), 11–25..
doi:10.1007/s10980-007-9173-8
Costanza, R., dArge, R., deGroot, R., Farber, S.,
Grasso, M., Hannon, B.,... vandenBelt, M. (1997).
The value of the world’s ecosystem services and
natural capital. Nature, 387(6630), 253–260.
doi:10.1038/387253a0
Engel, S., Pagiola, S., & Wunder, S. (2008).
Designing payments for environmental services
in theory and practice: An overview of the is-
sues. Ecological Economics, 65(4), 663–674..
doi:10.1016/j.ecolecon.2008.03.011
Agriculture and Conservation in the Natura 2000 Network
Garrod, G. (2009). Greening the CAP: how the
improved design and implementation of agri-
environment schemes can enhance the delivery of
environmental benefits. Journal of Environmental
Planning and Management, 52(5), 571–574..
doi:10.1080/09640560902958099
Gerard, F., Petit, S., Smith, G., Thomson, A.,
Brown, N., & Manchester, S. (2010). Land
cover change in Europe between 1950 and 2000
determined employing aerial photography. Prog-
ress in Physical Geography, 34(2), 183–205..
doi:10.1177/0309133309360141
Gibbs, J. P., Hunter, M. L. Jr, & Sterling, E. J.
(Eds.). (2008). Problem-Solving in Conserva-
tion Biology and Wildlife Management. Wiley-
Blackwell. doi:10.1002/9781444319576
Gohin, A. (2006). Assessing CAP reform: Sensi-
tivity of modelling decoupled policies. Journal
of Agricultural Economics, 57(3), 415–440.
doi:10.1111/j.1477-9552.2006.00058.x
Guzy, M. R., Smith, C. L., Bolte, J. P., Hulse, D.
W., & Gregory, S. V. (2008). Policy Research
Using Agent-Based Modeling to Assess Future
Impacts of Urban Expansion into Farmlands and
Forests. Ecology and Society, 13(1). Retrieved
from: http://www.ecologyandsociety.org /vol13/
iss1/art37/
Haberl, H., Gaube, V., Diaz-Delgado, R., Krauze,
K., Neuner, A., & Peterseil, J. (2009). Towards
an integrated model of socioeconomic biodiver-
sity drivers, pressures and impacts. A feasibil-
ity study based on three European long-term
socio-ecological research platforms. Ecological
Economics, 68(6), 1797–1812..doi:10.1016/j.
ecolecon.2008.11.013
Herzog, F. (2005). Agri-environment schemes
as landscape experiments - Preface. Agriculture
Ecosystems & Environment, 108(3), 175–177..
doi:10.1016/j.agee.2005.02.001
Hynes, S., & Garvey, E. (2009). Modelling
Farmers’ Participation in an Agri-environmental
Scheme using Panel Data: An Application to the
Rural Environment Protection Scheme in Ire-
land. Journal of Agricultural Economics, 60(3),
546–562..doi:10.1111/j.1477-9552.2009.00210.x
Keulartz, J. (2009). European Nature Conservation
and Restoration Policy-Problems and Perspec-
tives. Restoration Ecology, 17(4), 446–450..
doi:10.1111/j.1526-100X.2009.00566.x
Kleijn, D., Baquero, R. A., Clough, Y., Diaz, M.,
De Esteban, J., & Fernandez, F. (2006). Mixed
biodiversity benefits of agri-environment schemes
in five European countries. Ecology Letters, 9(3),
243–254..doi:10.1111/j.1461-0248.2005.00869.x
Knickel, K., Kroger, M., Bruckmeier, K., &
Engwall, Y. (2009). The Challenge of Evaluating
Policies for Promoting the Multifunctionality of
Agriculture: When ‘Good’ Questions Cannot
be Addressed Quantitatively and ‘Quantitative
Answers are not that Good’. Journal of Envi-
ronmental Policy and Planning, 11(4), 347–367..
doi:10.1080/15239080903033945
Ladle, R. J., & Whittaker, R. J. (Eds.). (2011).
Conservation biogeography. Oxford: Wiley-
Blackwell.
Lobianco, A., & Esposti, R. (2010). The Re-
gional Multi-Agent Simulator (RegMAS): An
open-source spatially explicit model to assess
the impact of agricultural policies. Computers
and Electronics in Agriculture, 72(1), 14–26..
doi:10.1016/j.compag.2010.02.006
Louette, G., Maes, D., Alkemade, J. R. M., Boitani,
L., de Knegt, B., & Eggers, J. (2010). BioScore-
Cost-effective assessment of policy impact on
biodiversity. Journal for Nature Conservation,
18(2), 142–148..doi:10.1016/j.jnc.2009.08.002
1293
 Agriculture and Conservation in the Natura 2000 Network
Macedo-Sousa, J. A., Soares, A., & Tarazona, J.
V. (2009). A conceptual model for assessing risks
in a Mediterranean Natura 2000 Network site.
The Science of the Total Environment, 407(3),
1224–1231..doi:10.1016/j.scitotenv.2008.09.052
Maguire, D., Batty, M., & Goodchild, M. (Eds.).
(2005). GIS, Spatial Analysis, and Modeling.
Esri Press.
Manly, B. F. J. (2008). Statistics for Environmen-
tal Science and Management, Second Edition
(Chapman & Hall/CRC Applied Environmental
Statistics): Chapman and Hall/CRC.
Margules, C. R., & Sarkar, S. (2007). Systematic
conservation planning. Cambridge: Cambridge
University Press.
Metzger, M. J., Bunce, R. G. H., Leemans, R.,
& Viner, D. (2008). Projected environmental
shifts under climate change: European trends and
regional impacts. Environmental Conservation,
35(1), 64–75..doi:10.1017/S0376892908004529
Mouro, C., & Castro, P. (2010). Local Commu-
nities Responding to Ecological Challenges-A
Psycho-social Approach to the Natura 2000 Net-
work. Journal of Community & Applied Social
Psychology, 20(2), 139–155. doi:.doi:10.1002/
casp.1025
Mucher, C. A., Klijn, J. A., Wascher, D. M., &
Schaminee, J. H. J. (2010). A new European Land-
scape Classification (LANMAP): A transparent,
flexible and user-oriented methodology to distin-
guish landscapes. Ecological Indicators, 10(1),
87–103..doi:10.1016/j.ecolind.2009.03.018
Opdam, P. F. M., Broekmeyer, M. E. A., &
Kistenkas, F. H. (2009). Identifying uncertain-
ties in judging the significance of human impacts
on Natura 2000 sites. Environmental Science
& Policy, 12(7), 912–921..doi:10.1016/j.en-
vsci.2009.04.006
1294
Pereira, J. M. C., & Duckstein, L. (1993). A
multiple criteria decision-making approach to gis-
based land suitability evaluation. International
Journal of Geographical Information Systems,
7(5), 407–424. doi:10.1080/02693799308901971
Piorr, A., Ungaro, F., Ciancaglini, A., Happe, K.,
Sahrbacher, A., & Sattler, C. (2009). Integrated as-
sessment of future CAP policies: land use changes,
spatial patterns and targeting. Environmental Sci-
ence & Policy, 12(8), 1122–1136..doi:10.1016/j.
envsci.2009.01.001
Primack, R. B. (2010). Essentials of Conservation
Biology (5th ed.). Sinauer Associates, Inc.
Primdahl, J., Vesterager, J. P., Finn, J. A., Vlahos,
G., Kristensen, L., & Vejre, H. (2010). Current use
of impact models for agri-environment schemes
and potential for improvements of policy design
and assessment. Journal of Environmental
Management, 91(6), 1245–1254..doi:10.1016/j.
jenvman.2009.12.012
Randolph, J. (2003). Environmental Land Use
Planning and Management. Island Press.
Reidsma, P., Tekelenburg, T., van den Berg, M., &
Alkemade, R. (2006). Impacts of land-use change
on biodiversity: An assessment of agricultural
biodiversity in the European Union. Agriculture
Ecosystems & Environment, 114(1), 86–102..
doi:10.1016/j.agee.2005.11.026
Scott, J. M., Davis, F., Csuti, B., Noss, R., But-
terfield, B., & Groves, C. (1993). Gap analysis - a
geographic approach to protection of biological
diversity. Wildlife Monographs, (123): 1–41.
Stephan, J. Goetz, F. B. (Ed.). (2009). New
Perspectives on Agri-environmental Policies:
A multidisciplinary and transatlantic approach
(Routledge Explorations in Environmental Eco-
nomics): Routledge.
Agriculture and Conservation in the Natura 2000 Network
Stokes, D. L., Hanson, M. F., Oaks, D. D., Straub,
J. E., & Ponio, A. V. (2010). Local Land-Use
Planning to Conserve Biodiversity: Planners’
Perspectives on What Works. Conservation
Biology, 24(2), 450–460..doi:10.1111/j.1523-
1739.2009.01356.x
Thiel, A. (2009). The use of ex-ante modelling
tools in European Impact Assessment: What role
does land use play? Land Use Policy, 26(4), 1138–
1148..doi:10.1016/j.landusepol.2009.02.005
Vandermeulen, V., Gellynck, X., Van Huylen-
broeck, G., Van Orshoven, J., & Bomans, K.
(2009). Farmland for tomorrow in densely popu-
lated areas. Land Use Policy, 26(4), 859–868..
doi:10.1016/j.landusepol.2008.10.014
Verboom, J., Alkemade, R., Klijn, J., Metzger, M.
J., & Reijnen, R. (2007). Combining biodiversity
modeling with political and economic develop-
ment scenarios for 25 EU countries. Ecological
Economics, 62(2), 267–276..doi:10.1016/j.eco-
lecon.2006.04.009
Verburg, P. H., Eickhout, B., & van Meijl, H.
(2008). A multi-scale, multi-model approach for
analyzing the future dynamics of European land
use. The Annals of Regional Science, 42(1),
57–77..doi:10.1007/s00168-007-0136-4
Verburg, P. H., van de Steeg, J., Veldkamp, A.,
& Willemen, L. (2009). From land cover change
to land function dynamics: A major challenge
to improve land characterization. Journal of
Environmental Management, 90(3), 1327–1335..
doi:10.1016/j.jenvman.2008.08.005
Vereijken, P. H., & Hermans, C. M. L. (2010). A
quick scan tool to assess the relative prospects of
European regions for sustainable agriculture in a
liberal market. Land Use Policy, 27(2), 440–448..
doi:10.1016/j.landusepol.2009.06.002
Westhoek, H. J., van den Berg, M., & Bakkes, J.
A. (2006). Scenario development to explore the
future of Europe’s rural areas. Agriculture Ecosys-
tems & Environment, 114(1), 7–20..doi:10.1016/j.
agee.2005.11.005
Yang, W. H., Bryan, B. A., MacDonald, D. H.,
Ward, J. R., Wells, G., Crossman, N. D., & Connor,
J. D. (2010). A conservation industry for sustaining
natural capital and ecosystem services in agricul-
tural landscapes. Ecological Economics, 69(4),
680–689..doi:10.1016/j.ecolecon.2009.11.028
KEY TERMS AND DEFINITIONS
Agriculture Land Use Change: process that
contribute to habitat degradation through which
one habitat-type is removed and replaced with
another habitat-type due to agriculture expansion.
In the process of land-use change, plants and ani-
mals which previously used the site are displaced
or destroyed, reducing biodiversity.
Agri-Environmental Schemes: term used to
describe communitarian, national or local schemes
that pay farmers to farm in an environmentally
sensitive way.
Biodiversity Conservation: how to protect
and restore biodiversity, or the diversity of life
on Earth. To preserve biodiversity, scientists must
answer three general questions. First, how is the
diversity of life distributed around the planet?
Second, what threats does this diversity face?
Third, what can people do to reduce or eliminate
these threats and, when possible, restore biological
diversity and ecosystem health?
Birds Directive: the EU Directive on the
conservation of wild birds (79/49/EEC) seeks to
protect all wild birds and the habitats of listed
species, in particular through the designation of
special protection areas (SPA).
1295
 Agriculture and Conservation in the Natura 2000 Network

Geographical Informational System: any biodiversity through the conservation of natural

system that captures, stores, analyzes, manages,
and presents data that are linked to location. The
term describes any information system that inte-
grate, store, edits, analyzes, shares, and displays
geographic information.
Habitats Directive: Council Directive 92/43/
EEC of 21 May 1992 on the conservation of natural
habitats and of wild fauna and flora. The aim of
this directive is to contribute towards ensuring
habitats and of wild fauna and flora in the Eu-
ropean territory of the Member States to which
the Convention on Biological Diversity applies.
NATURA 2000: the key instrument to protect
biodiversity in the European Union based on the
1979 Birds Directive and the 1992 Habitats Di-
rective. It is an ecological network of protected
areas, set up to ensure the survival of Europe’s
most valuable species and habitats.

This work was previously published in Agricultural and Environmental Informatics, Governance and Management: Emerg-
ing Research Applications, edited by Zacharoula Andreopoulou, Basil Manos, Nico Polman and Davide Viaggi, pp. 339-358,
copyright 2011 by Information Science Reference (an imprint of IGI Global).

1296
 1297

Chapter 78
A Neural Network for
Modeling Multicategorical
Parcel Use Change
Kang Shou Lu
Towson University, USA

John Morgan III

Towson University, USA

Jeffery Allen
Clemson University, USA

ABSTRACT
This paper presents an artificial neural network (ANN) for modeling multicategorical land use changes.
Compared to conventional statistical models and cellular automata models, ANNs have both the ar-
chitecture appropriate for addressing complex problems and the power for spatio-temporal prediction.
The model consists of two layers with multiple input and output units. Bayesian regularization was used
for network training in order to select an optimal model that avoids over-fitting problem. When trained
and applied to predict changes in parcel use in a coastal county from 1990 to 2008, the ANN model
performed well as measured by high prediction accuracy (82.0-98.5%) and high Kappa coefficient (81.4-
97.5%) with only slight variation across five different land use categories. ANN also outperformed the
benchmark multinomial logistic regression by average 17.5 percentage points in categorical accuracy
and by 9.2 percentage points in overall accuracy. The authors used the ANN model to predict future
parcel use change from 2007 to 2030.

INTRODUCTION economic development. Degraded environments,

Land use change has been a major cause of many
environmental problems. Changes are often nec-
essary to accommodate population growth and
DOI: 10.4018/978-1-4666-2038-4.ch078
however, threaten sustainable development. As
more states and local governments are willing to
create and implement smart growth strategies,
planners and policy makers need to know the
impacts of future land use change and related

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 A Neural Network for Modeling Multicategorical Parcel Use Change
regulatory decisions. Predictive modeling is not
only a learning tool to understand causal factors
and dynamic changes of a land use system, but also
an important mechanism to predict future changes
and simulate the potential effects under different
growth scenarios (Pijanowski et al., 2002).
Despite the emergence of numerous predictive
models over the last half-century, our ability to
accurately predict land use change remains lim-
ited. Recently, 18 scientists jointly conducted an
evaluation of the performance of nine land cover/
land use models applied in 13 regions (Pontius et
al., 2008). These models included conventional
logistic regression (McConnell, Sweeney, & Mul-
ley, 2004), cellular automata based SLEUTH
(Dietzel & Clarke, 2004; Silva & Clarke, 2002),
multiple agent based SAMMBA (Boissau &
Castella, 2003; Castella, Trung & Boissau, 2005),
multistage model CLUE and its variant CLUE-S
(Duan et al., 2004; Verburg et al., 2002; Veldkamp
& Fresco, 1996; Verburg & Veldkamp, 2004), and
multifunctional LTM (Pijanowski et al., 2005),
multiple objective GEOMOD (Pontius, Cornell
& Hall, 2001; Pontius & Malanson, 2005; Pontius
& Spencer, 2005). It was found that prediction
accuracy, as measured in percent of the correctly
predicted against the observed, falls within a range
of 1-73 percent, falling below 30 percent in seven
cases and above 50 percent only in three cases.
Some of the assessments were against the sample
datasets used for model calibration or training.
But judging on these results, how much can we
trust the predictions of the models?
Pontius et al. (2008) point out that something
must be wrong with the mechanics of these models.
We attribute the poor performances more to the
oversimplification of complex land use systems
in both semantics and syntax. Semantically, most
models, particularly cellular automata models,
tend to overemphasize “parsimony” and use
too few predictive variables to truly represent
complex reality. It may not be a coincident that
SLEUTH, constructed with only three effective
predictive variables, is the worst performer among
1298
all the models examined by Pontius et al. (2008).
Although there is no suggested threshold for how
many predictive variables are deemed appropri-
ate, a good model should include variables that
measure key drivers of changes in land use such
as population, economy, and technology (Turner
& Meyer, 1994). More importantly, the model
should be able to generate meaningful predictions,
and pass the reality test.
From the syntax perspective, most existing
models, particularly conventional statistical
models, lack the multilayered, hierarchical,
interconnected structure needed for modeling
complex land use systems as Batty and Torrens
(2005) have suggested. Multiple layers mimic
the processes of land use changes that involve
land transaction, speculation, and development;
whereas interconnections through these layers
capture the interrelationships and interactions
between dependent and independent variables,
whether they are observable, quantifiable or
comprehensible. The CLUE (-S) model is to some
degree structured in this way and thus performed
relatively well (Pontius et al., 2008). Batty and Tor-
rens (2005) did not provide specific suggestions
about what algorithms to use and how to derive
them in order to describe such interrelationships
and implement the framework. The structure of
the conceptual model, however, implicitly points
to an artificial neural network (ANN) approach
which does offer such capabilities.
This paper presents an artificial neural network
(ANN) for modeling multi-categorical parcel use
changes. We intended to improve the model by: 1)
incorporating more effective predictor variables
for a better semantic representation; 2) expand-
ing the model structure for a multiclass land use
system; 3) using a multilayer neural network
to enhance predictive power; and 4) applying a
Bayesian regularization (BR) training algorithm to
avoid the over-fit problem. We applied the model
in Beaufort County, South Carolina, to predict
and simulate future land use changes from 2000
to 2030 under different growth scenarios.
A Neural Network for Modeling Multicategorical Parcel Use Change
BACKGROUND
ANNs have been around since the early 1940s,
but it was not until the mid 1990s that they were
introduced into the geosciences for spatial interac-
tion, interregional telecommunication, resource
management, suitability assessment, and image
classification (Fischer & Gopal, 1994; Openshaw,
1993). Openshaw and Openshaw (1997) identified
four major benefits of neural networks for model-
ing: better performance, greater representational
flexibility and freedom from current model design
constrains, the opportunity to handle explicitly
noisy data, and the incorporation of spatial de-
pendency in the net representation (which is
currently ignored).
Researchers in land use modeling have been
a half-decade slower in realizing such benefits.
Pioneering studies mainly focus the use of ANNs
for simulating the change in urbanization patterns
(Li & Yeh, 2001; Pijanowski, Gage, & Long,
2000; Pijanowski et al., 2002; Yeh & Li, 2002).
Pijanowski et al. (2005) later incorporated an
ANN as one of the three components of the Land
Transformation Model (LTM), and applied it in
several other areas. Allen and Lu (2004, 2006)
found that ANN models not only attained fair to
high prediction accuracy but also outperformed
the logistic regression in several applications in
coastal regions. They also found that ANN mod-
els were capable of discerning isolated, sparsely
distributed features or events that other models
often fail to predict. Land use systems modeled
in these studies are all binary. It remains to be
tested whether ANNs can achieve similar success
in predicting multi-category land use change.
Figure 1 provides both a graphic illustration
and mathematic notations (MatLab) of a two-layer
feedforward neural network appropriate for land
use prediction. The network has R input units, S1
hidden units and S2 output units. The connectors
between the input units and the hidden units are
input weights, collectively denoted by IW1,1,
which is a R x S1 matrix. The superscripts rep-
resent the destination and source of a connector.
The connectors between hidden units and output
units (second layer) are layer weights, denoted by
LW2,1, which is a S1x S2 matrix. Each unit in the
hidden layer or output layer has a bias term, b1 or
b2. They together form a bias vector, b1 or b2. The
input vector is p, whereas a1 and a2 are the output
vectors for the hidden layer and output layer, re-
spectively. The vector a2 is also the output of the
whole network; f1and f2 are transfer functions for
the hidden layer and output layer, respectively. A
logistic function is often used as transfer function
for classification problems.
The prediction starts from left to right. First,
each hidden unit sums with its own bias b1 the
product of the input value and corresponding
weight lw1,1 from every input unit p. Then, a
transfer function, f1, is applied to the sum n1 before
its output a1 is sent to every unit of the output
layer. Each unit of the output layer repeats the
process, but uses the output of every hidden unit
as the input. The result is the output, a2, or predic-
tion of the network. This process is also called
feedforward prediction. For this approach, the
determination of the weights and biases for mul-
tilayer networks used to be a challenge until re-
cently.
A backpropagation training algorithm (Ru-
melhart et al., 1986) makes it possible to auto-
matically derive the weights and bias for a neural
network by learning from a sample. It starts with
random weights and bias and proceeds with the
feedforward prediction. The prediction error, the
difference between the output value and the ob-
served value (often called the target value), is then
calculated. The error is redistributed backward
to every hidden unit based on a gradient descent
rule. The information is then used for updating
the corresponding weight and bias for every unit
at a certain correction rate before starting another
cycle for the next input-target pair. The entire
process that goes through every pair in a sample
dataset is an epoch. It usually takes many epochs
before a network achieves a desirable goal and
1299
 A Neural Network for Modeling Multicategorical Parcel Use Change
Figure1. A two-layer feedforward neural network for modeling multi-categorical land uses
stops training. The backpropagation method has
revitalized neural network sciences and marks the
second generation of neural computation (Penny
et al., 1999).
Many new algorithms have been developed
for increasing the training speed, maximizing
approximation capability, or avoiding over-fitting
problems. One such innovative algorithm is the
Bayesian Regularization (BR) developed by
MacKay (1992). Instead of minimizing just sum
of squared error (SSE) as the training objective,
BR incorporates sum of squared weights (SSW)
into the performance function by using two ad-
ditional parameters (δ and γ) for regularizing
SSE and SSW, respectively. Built upon Bayes-
ian probability framework, BR is capable of 1)
determining all parameters based on the posterior
probabilities conveyed in the sample data, 2)
automatically selecting an optimal model that
avoids over-fitting problems for better generaliza-
tion, and 3) retaining great approximation power
once sufficiently trained. It is no longer necessary
to split a sample into three subsets for training,
1300
testing, and validation as an essential technique
to avoid over-fitting problems.
METHODS
Study Area
Beaufort County, South Carolina, was selected
as the study area. It is a well-known coastal
tourist destination. The county covers 923 mi2 in
area, with 63 percent land and 37 percent water.
Driven by tourism development, new immigration,
service sector expansion, and military demand,
Beaufort County has seen tremendous growth in
population, urban area, and impervious surface
over the past two decades. From 1990 to 2000,
its population increased by 39.93 percent, from
86,425 to 120,935. This growth rate triples the
national average (13 percent) and leads all coun-
ties in South Carolina. From 1990 to 2008, the
total number of built parcels increased by 136.4
percent from 23,446 to 56,929. Impervious sur-
face almost quadrupled (380 percent) in the fast
A Neural Network for Modeling Multicategorical Parcel Use Change

growing Bluffton area. The land conversion rate We have also assessed the effects of predictor
has been high, with an average of 2,575 parcels variables.
(1,500 acres) per year over the last decade. As one
of the top seven retiree communities in the US, the Data Preparation
county is expected to continue growing at a rapid
pace. These changes impose tremendous pressure Parcel use in 2008 was derived from the parcel
on natural environments, local infrastructure, and shape file (ESRI, Inc., Redlands, California)
public services. There is a tremendous demand and Tax Assessor’s Table provided by Beaufort
for predictions and impact assessments needed County. We grouped parcel uses into five general
for smart growth decisions. categories: 1) commercial, including all com-
mercial, industrial, institutional, and service/
Neural Network Specification utility uses; 2) urban residential, including all
developed accommodation use < 5 ac; 3) rural
We used a two-layer, feedforward neural network residential, with parcel size falling between 5-10
with 18 x 18 x 5 architecture to model a five-class ac; 4) recreational (only golf courses); and 5)
parcel use system. The five output units included undeveloped, including wetlands, open space,
commercial, urban residential, rural residential, forestlands, croplands, parks and reserves, and
recreation (golf course), and undeveloped, all of other preserved and protected lands. Of a total
which are binary (1 or 0). Each parcel is assigned 99,781 parcels, 33,250 (33.32 percent) were de-
to one type of use, and each parcel is encoded veloped or built between 1990 and 2008. We call
by a vector of five elements. For example, if this ‘new development’ or ‘net change’ for model
a parcel was developed for commercial use, it validation. Because of rich temporal information,
was coded as (1, 0, 0, 0, 0). Input units represent we were able to derive changes on an annual or
18 independent variables listed in Table 1. The decadal basis for different land uses, periods, and
number of hidden units was initially set to 18, but
may change, depending on the BR training result.
Table 1. Predictor variables
We used a logistic function, f = 1 / (1+ exp
(- x)), as the transfer function for all layers, and Group Independent Variables
a softmax function, yj = exp (- xj) / ∑ exp (- xk), Physical properties • land area
for land transition probability and output classi- • wetland ratio
• slope
fication. The former allows us to simulate parcel • distance to water front
changes under different scenarios. The BR algo- • forest wetland
rithm is available in the Neural Network Toolbox Road accessibility • cost distance to major road
• distance to local street
of Simulink 8 (MatLabtm). We wrote scripts to
Proximity to utilities • distance to waterline
enable data input, classification, and error assess- • and distance to sewer line
ment. Several ArcGIS-based submodules were
Neighborhood • cost distance to central business
developed for sample extraction, data processing, effects district
and spatial prediction. • developed parcels
• distance to built parcels
Performance of the ANN was assessed in mea- • agricultural use
sures of: 1) mean squared error (MSE), sum of Demographic • block population
squared error (SSE), and sum of squared weights features • block neighborhood population
(SSW); 2) prediction accuracy; and 3) the Kappa Land value • parcel land value
• total property values
coefficient. Multinomial logistic regression was
Land ownership • private ownership
used as a benchmark model for comparisons.

1301
 A Neural Network for Modeling Multicategorical Parcel Use Change
regions. We used the growth rates for three differ-
ent lengths of periods to simulate future changes:
1980-2008, 1990-2008, and 2000-2008.
We prepared two sets of data for all predictive
variables, 1990 for model training, and 2008 for
future prediction. National Wetland Inventory
data from 1989 were used to extract forested
wetland, land area, and wetland ratio. Slope was
derived from the 30 m DEM downloaded from
USGS’s National Map Seamless Server. Block
population data were from Census Data 1990 and
2000. Beaufort County provided the updated data
for roads and streets from which proximity and
coast distance variables were derived. Waterline
and sewerline layers were from South Carolina
Department of Commerce. Protected lands were
downloaded from the Natural Resources GIS Data
Clearinghouse of the South Carolina Department
of and updated by using the 2008 parcel data.
The tax assessor file was the source for building
counts, built year, parcel size, land price, land
value, and land ownership at the finest spatial
units. Since the coastal area is highly fragmented
by water bodies and wetlands, we constructed two
cost distance variables to measure the influences
of major roads and central business district over
a cost surface created by weighting these natural
barriers more as frictional factors.
All data were aggregated in parcel units and the
parcel shape file was joined with the tax assessor’s
table. Most distances were measured from the
centroid of a parcel, but distance to waterfront is
the zonal average of the grid’s Euclidian distances
to a parcel. Data for each independent variable
were scaled to a range between 0 and 1 by using
the same set of min-max values for both training
and prediction sets.
The training dataset was extracted by using
a pseudo-stratified systematic random sampling
method. The strata are five land use classes, but
subsample size for each stratum is not purely
proportional to the total number of parcels, but
proximately proportional to the total area of the
stratum in concern. The sample size (S) for the
1302
entire training set was estimated based on the
formula S=Nw / Et where Nw is the total number
of weights including biases, and Et is the targeted
error threshold. The final training sample includes
7,812 parcels, accounting for nearly 8 percent in
number and 17 percent in area of all parcels in the
county. We also prepared four more datasets for
model validation and testing. They included two
sample sets, a net change set, and the full dataset.
The first three sets accounted for 20 percent, 10
percent, and 33 percent of the total number of
parcels in the county respectively.
RESULTS
Model Performance
We performed multinomial regression analysis in
a preliminary study to screen predictive variables.
Eighteen of the variables included in the prelimi-
nary study were found statistically significant and
thus used to construct the ANN model. Based
on the 4:2:2 ratios, the training sample was split
into three subsets for training, validation, and
testing, respectively, during the training process.
The training process stopped within 100 epochs
in several runs. MSE is 0.018 for the training
subset and 0.038 for both testing and validation
subsets. Results of BR suggest that an optimal
model with 415 effective parameters (18 x 17 x
5) would be adequate.
The ANN model performed well against the
training dataset (Table 2). The prediction accu-
racy, overall or categorical, is quite impressive as
measured in terms of correct percent (82.0-98.5
percent) and Kappa coefficients (0.81-0.98), par-
ticularly for a change of 33.3 percent over newly
two decades. Kappa coefficient is considered a
conservative measure of agreement with chance
corrected, and a value above 0.80 suggests an
excellent prediction. Although both accuracy and
Kappa tend to increase with the total number of
A Neural Network for Modeling Multicategorical Parcel Use Change

Table 2. Prediction accuracy of the ANN model against the training sample

Predicted
Observed Row Total Accuracy (%) Kappa
I II III IV V
I 511 24 24 9 6 574 89.0 0.88
II 12 2815 30 6 14 2877 97.8 0.97
III 10 22 866 7 38 943 91.8 0.91
IV 4 5 5 218 34 266 82.0 0.81
V 16 11 18 2 3105 3152 98.5 0.98
Column total 553 2877 943 242 3197 7812 96.2
a
0.94b
Notes: I=Commercial, II=Urban residential, III=Rural residential, IV=Golf Course, and V=Other Use.
a. Overall prediction accuracy, b. Overall kappa coefficient.

parcels within the category, their variations across
different land uses are relatively small.
The benchmark MLR was also statistically
significant based on the likelihood ratio tests
(Chi-square = 11,100, df = 76, p < 0.001). It was,
however, outperformed by the ANN model in all
measurements (Tables 2 and 3). The ANN model
improved prediction accuracy by an average of
17.5 percentage points across the five uses and
by 9.2 percentage points in overall. The two
numbers are even more impressive for Kappa
coefficients (0.19 and 0.14). The greatest improve-
ment occurred where MLR failed to generate
satisfactory results, by 33.6 and 32.2 percentage
points for commercial and rural residential re-
spectively.
Results of model validation against four dif-
ferent datasets are summarized in Table 4. Over-
all accuracy is high in all cases (> 94.9 percent)
with only little variation (< 1.3 percentage points)
across different cases. For all but net change
validation, prediction accuracy dropped only by
average 2.4 percentage points in all 19 measure-
ments with a range of 0.8-2.6 percentage points.
The net change case is an exception because we
observed three extreme discrepancies: the largest
jump (+12.2), the smallest drop (-1.2), and the
largest drop (-11.1) in prediction accuracy from
those for the training set. ANN’s predictive abil-
ity did not show a consistent decay pattern which
was not expected.

Table 3. Prediction accuracy of the MLR model against the training sample

Predicted
Observed Row Total Accuracy (%) Kappa
I II III IV V
I 318 197 12 12 35 574 55.4 0.53
II 19 2659 50 3 146 2877 92.4 0.88
III 74 147 562 4 156 943 59.6 0.56
IV 13 24 5 177 47 266 66.5 0.66
V 5 52 8 4 3083 3152 97.8 0.96
Column total 429 3079 637 200 3467 7812 87.0
a
0.80b
Notes: I=Commercial, II=Urban residential, III=Rural residential, IV=Golf Course, and V=Other Use.
a. Overall prediction accuracy, b. Overall kappa coefficient.

1303
 A Neural Network for Modeling Multicategorical Parcel Use Change

Table 4. Prediction accuracy (%) of the ANN model in four validations

Parcel Use Validation Sets

Training Set
Category Use-based Area-based Net Change a Full Set
I 89.0 86.6 83.8 79.2 87.1
II 97.8 96.5 96.5 96.6 96.2
III 91.8 88.5 84.9 80.7 87.4
IV 82.0 79.8 77.6 94.2 80.6
V 98.5 97.7 97.6 97.7
Overall 96.2 96.4 96.1 94.9 96.2
Notes: I=Commercial, II=Urban residential, III=Rural residential, IV=Golf Course, and V=Other use.
a
Category V or Other Use is treated as the background category and is thus not applicable in this case.

Predicted Changes straints. Table 5 provides a quantitative summary

The predicted long-term trajectories for parcel
development are shown in Figure 2. They reflect
quantitative demands at three different growth
rates. We used 2008 data as the new input of the
trained ANN model to predict land transition
probabilities for five parcel uses. These two types
of information were used to map future change
through 2030. Figure 3 shows the distribution of
predicted parcel use changes for selected classes
and areas under selected growth rates and con-
of simulated regional development with and with-
out a growth boundary constraint.
DISCUSSION AND CONCLUSION
A neural network was applied to predict parcel
use change in Beaufort County, South Carolina.
The model is not only theoretically appropriate
for addressing complex problems, but also has
demonstrated improved predictive capability. The

Figure 2. Predicted trajectories of parcel development in Beaufort County at three different growth rates

1304
A Neural Network for Modeling Multicategorical Parcel Use Change

Figure 3. Predicted parcel use changes in Beaufort County from 2000 to 2030 under different growth
scenarios

Table 5. Potential impact of the growth boundary on regional development in Beaufort County

Growth Boundary Gain (or Loss)

Region Yes No Yes-No
N Area N Area N% Area%
St Henna Island 0 0 5132 4592 -100 -100
S Beaufort Islands 2726 2038 1664 875 64 133
Hilton Head 5239 3968 4884 3035 7 31
NE Corner 0 0 958 1341 -100 -100
Port Royal-City of Beaufort 10332 8148 9631 5945 7 37
Bluffton 9404 7318 8687 5677 8 29
Entire County 27701 21472 30956 21466 -11a <1a
Notes: Prediction is based on the 1990-2008 growth rate. Area units are acres.
a
Discrepancies are caused by variable parcel sizes due to the use of area-based demand.

1305
 A Neural Network for Modeling Multicategorical Parcel Use Change
accuracy ranges from fairly good to excellent in
all 29 measurements, regardless of variations in
land use category, sample set, parcel size, and
time span. The ANN outperformed both bench-
mark MLR and models evaluated by Pontius et
al. (2008). A critique of ANNs is the over-fitting
problem. But this problem, if present, might have
been kept to minimal in our case, due to the use of
BR and large training sample. Cross-validations
do not show any substantial drops in prediction
accuracy nor suggest any pattern of variations
across different sample sets and land uses. It
was found again that ANNs enable us to predict
rare, sparsely distributed events with the level of
accuracy other models often fail to reach (Allen
& Lu, 2006).
The use of parcels as objects to be modeled
and as units of analysis has brought both benefits
and issues. This approach adds more realism to
our modeling practice because of the use of real
world units for land transaction and develop-
ment (Landis, 1994). Many parcel attributes are
predictive variables with measurements that can
be readily incorporated into a model to improve
model semantics and prediction accuracy. Also,
many property variables are spatially well-defined
by parcel boundaries and no additional error is
induced by aggregation or disaggregation. For
the same reason, it is less difficult to interpret the
modeling results. Four issues we have encountered
are related to large parcel partitioning, vector
data processing, parcel use classification, and
data availability. How to model the large parcel
partitioning process remains to be a technical
challenge.
The predicted trajectories and growth patterns
have policy implications. The anticipated urban
growth ratio is relatively moderate (2:1-3:1) com-
pared to that (6:1) for the Charleston area (Allen
& Lu, 2003). Potential impacts of land develop-
ment can be relatively severe because Beaufort’s
physical environment is very fragmented and
ecological systems are quite vulnerable. While
new development is continuously spreading in the
1306
urban fringe areas, infill development is expected
to intensify in previously sprawled regions. Over
time, development will change the image of the
county as a well vegetated, properly landscaped
tourist destination. The simulation results suggest
caution in implementation of a growth boundary.
The policy may be able to preserve natural areas
and farm lands in North Beaufort, but it has a
potential risk of driving land conversion and en-
vironmental change in many peripheral islands in
South Beaufort too fast for coastal terrestrial and
aquatic life to adapt. Incentives should be in place
in order to regulate such developments.
ACKNOWLEDGMENT
We would like to thank South Carolina Sea Grant
and NOAA’s National Centers for Coastal Ocean
Science for funding this project.
REFERENCES
Allen, J. S., & Lu, K. S. (2006). Predicting trajec-
tories of urban growth in the coastal Southeast.
In G. S. Kleppel, M. R. Devoe & M. V. Rawson
(Eds.), Changing land use patterns in the coastal
zone: Managing environmental quality in rapidly
developing regions (pp. 47-67). New York, NY:
Springer.
Allen, J. S., & Lu, K. S. (2003). Modeling and
prediction of future urban growth in the Charleston
region of South Carolina: A GIS-based integrated
approach. Conservation Ecology, 8(2), 1–20.
Allen, J. S., & Lu, K. S. (2004). Artificial neu-
ral network vs. binary logistic regression: Two
alternative models for predicting urban growth
in the Myrtle Beach region. Paper presented at
the NOAA’s Land Use-Coastal Ecosystem Study
(LU-CES) Program.
A Neural Network for Modeling Multicategorical Parcel Use Change
Batty, M., & Torrens, P. M. (2005). Modeling and
prediction in a complex world. Futures, 37(7),
745–766. doi:10.1016/j.futures.2004.11.003
Boissau, S., & Castella, J. (2003). Constructing a
common representation of local institutions and
land use systems through simulation-gaming and
multi-agent modeling in rural areas of North-
ern Vietnam: The SAMBA-Week methodol-
ogy. Simulation & Gaming, 34(3), 342–347.
doi:10.1177/1046878103255789
Castella, J., Boissau, S., Trung, T. N., & Quang,
D. D. (2005). Agrarian transition and lowland-
upland interactions in mountain areas in northern
Vietnam: Application of a multi-agent simulation
model. Agricultural Systems, 86(3), 312–332.
doi:10.1016/j.agsy.2004.11.001
Dietzel, C. K., & Clarke, K. C. (2004). Spatial
differences in multi-resolution urban automata
modeling. Transactions in GIS, 8, 479–492.
doi:10.1111/j.1467-9671.2004.00197.x
Duan, Z., Verburg, P. H., Zhang, F., & Yu, Z.
(2004). Construction of a land-use change simu-
lation model and its application in Haidian Dis-
trict, Beijing. Acta Geographica Sinica, 59(6),
1037–1046.
Fischer, M. M., & Gopal, S. (1994). Artificial
neural networks: A new approach to model-
ing interregional telecommunication flows.
Journal of Regional Science, 34(4), 503–527.
doi:10.1111/j.1467-9787.1994.tb00880.x
Landis, J. (1994). The California urban futures
model: A new generation of metropolitan simu-
lation models. Environment and Planning. B,
Planning & Design, 21, 399–420. doi:10.1068/
b210399
Li, X., & Yeh, A. G. O. (2001). Calibration of
cellular automata by using neural networks for
the simulation of complex urban systems. En-
vironment & Planning A, 33(8), 1445–1462.
doi:10.1068/a33210
Lu, K. S., & Allen, J. S. (in press). Neural networks
for modeling urban growth patterns and potential
ecosystem impact in the coastal zone. In Brouwer,
F., & Goetz, S. J. (Eds.), The dynamics of land
use and ecosystem interactions: A transatlantic,
multidisciplinary and comparative approach. New
York, NY: Springer.
MacKay, D. J. C. (1992). Bayesian interpolation.
Neural Computation, 4(3), 415–447. doi:10.1162/
neco.1992.4.3.415
Openshaw, S. (1993). Modeling spatial interaction
using a neural net. In Fischer, M. M., & Nijkamp,
P. (Eds.), GIS spatial modeling and policy (pp.
147–164). Berlin, Germany: Springer-Verlag.
Openshaw, S., & Openshaw, C. (1997). Artificial
intelligence in geography. Chichester, UK: John
Wiley & Sons.
Penny, W. D., Husmeier, D., & Roberts, S. J.
(1999). Introduction. In Lisboa, P. J. G., Ifeachor,
E. C., & Szczepaniak, P. S. (Eds.), Artificial neu-
ral networks in biomedicine (pp. 1–9). Berlin,
Germany: Springer-Verlag.
Pijanowski, B. C., Brown, D. G., Manik, G., &
Shellito, B. A. (2002). Using neural nets and GIS
to forecast land use changes: A land transforma-
tion model. Computers, Environment and Urban
Systems, 26(6), 553–575. doi:10.1016/S0198-
9715(01)00015-1
Pijanowski, B. C., Pithadia, S., Shellito, B. A.,
& Alexandrids, K. (2005). Calibrating a neural
network-based urban change model for two metro-
politan areas of the Upper Midwest of the United
States. International Journal of Geographical
Information Science, 19, 197–215. doi:10.1080
/13658810410001713416
Pontius, R. G. Jr, Boersma, W., Castella, J., Clarke,
K., de Nijs, T., & Dietzel, C. (2008). Comparing
the input, output, and validation maps for several
models of land change. The Annals of Regional
Science, 24(1), 11–37. doi:10.1007/s00168-007-
0138-2
1307
 A Neural Network for Modeling Multicategorical Parcel Use Change
Pontius, R. G. Jr, Cornell, J., & Hall, C. (2001).
Modeling the spatial pattern of land-use change
with GEOMOD2: Application and validation for
Costa Rica. Agriculture Ecosystems & Environ-
ment, 85(1-3), 191–203. doi:10.1016/S0167-
8809(01)00183-9
Pontius, R. G. Jr, & Spencer, J. (2005). Uncertainty
in extrapolations of predictive land change mod-
els. Environment and Planning B, 32, 211–230.
doi:10.1068/b31152
Rumelhart, D. E., Hinton, G. E., & Williams, R.
J. (1986). Learning internal representations by
error propagation. In J. L. McClelland & D. E.
Rumelhart (Eds.), Parallel distributed processing:
Explorations in the microstructure of cognition:
Vol. 1. Foundations (pp. 318-362). Cambridge,
MA: MIT Press.
Silva, E., & Clarke, K. C. (2002). Calibration
of the SLEUTH urban growth model for Lisbon
and Porto, Portugal. Computers, Environment
and Urban Systems, 26, 525–552. doi:10.1016/
S0198-9715(01)00014-X
Turner, B. L., & Meyer, W. B. (1994). Global
land-se and land-cover change: An overview. In
Meyer, W. B., & Turner, B. L. II, (Eds.), Change
in land use and land cover: A global perspective
(pp. 1–10). Cambridge, UK: Cambridge Univer-
sity Press.
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue
3, edited by Donald Patrick Albert, pp. Donald Patrick Albert, copyright 2011 by IGI Publishing (an imprint of IGI Global).
1308
Veldkamp, T. A., & Fresco, L. (1996). CLUE-CR:
An integrated multi-scale model to simulate land
use change scenarios in Costa Rica. Ecological
Modelling, 91, 231–248. doi:10.1016/0304-
3800(95)00158-1
Verburg, P. H., Soepboer, W., Veldkamp, T. A.,
Limpiada, R., Espaldon, V., & Mastura, S. A. S.
(2002). Modeling the spatial dynamics of regional
land use: The CLUE-S model. Environmental
Management, 30(3), 391–405. doi:10.1007/
s00267-002-2630-x
Verburg, P. H., & Veldkamp, T. A. (2004).
Projecting land use transitions at forest fring-
es in the Philippines at two spatial scales.
Landscape Ecology, 19, 77–98. doi:10.1023/
B:LAND.0000018370.57457.58
Yeh, A. G. O., & Li, X. (2002). Urban simulation
using neural networks and cellular automata for
land use planning. In Richardson, D., & van Os-
terom, P. (Eds.), Advances in spatial data handling
(pp. 451–464). Berlin, Germany: Springer-Verlag.
 1309

Chapter 79
Historical GIS as a Platform for
Public Memory at Mammoth
Cave National Park
Katie Algeo
Western Kentucky University, USA

Ann Epperson
Western Kentucky University, USA

Matthew Brunt
Western Kentucky University, USA

ABSTRACT
The Mammoth Cave Historical GIS (MCHGIS) fosters new understandings of a national park landscape
as a historic farming community and offers a web-based platform for public memory of pre-park inhabit-
ants. It maps the 1920 manuscript census at the household level over a streaming topographic map and
georeferences Civilian Conservation Corps photographs of dwellings for visualization and analysis of
the area’s population on the eve of creation of Mammoth Cave National Park. A web interface to the
MCHGIS permits broader dissemination of archival holdings. Public participation GIS techniques are
adapted to initiate a virtual site of public memory to supplement the history presented by institutionally-
held materials with those donated from private holdings.

INTRODUCTION tools to explore them (Gregory & Healey, 2007).

GIS has proven a valuable tool for historical ge-
ographers in facilitating integration of data from
diverse sources, permitting visualization and
analysis of past places, and allowing dissemination
via the Internet of both digital databases and the
DOI: 10.4018/978-1-4666-2038-4.ch079
This article describes a historical GIS created to
document and enhance understanding of the his-
tory of the pre-park inhabitants of Mammoth Cave
National Park (Figure 1). The national park was
authorized by U.S. Congress in 1926, the same
year as Great Smoky Mountains and Shenandoah
National Parks, part of a wave of park-creation
intended to meet the recreational and psychologi-

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
cal needs for wilderness among the core of U.S.
population located east of the Mississippi River
(Ise, 1961). All three new parks had resident
populations, variables mixes of EuroAmericans,
African Americans and Native Americans. The
Mammoth Cave region had, by far, the highest
population density of the three new parks, yet it
has received the least scholarly treatment of its
displaced population.1 As a first step in addressing
this lacuna, the Mammoth Cave Historical GIS
(MCHGIS) provides a snapshot of the region’s
habitation on the eve of the national park’s cre-
ation. It combines 1920 manuscript census data
for individual households with photographs of
the dwellings where families lived, geolocating
both to known house sites. It thus provides not
only the basis for quantitative analysis of the pre-
park population, but a framework for qualitative
understanding of landscape and living conditions
in the region.
A number of GIS map populations at the
household level for small cities or portions of
larger urban areas (see, for example, DeBats,
2008; Schlichting & Tuckel, 2006). Historical
GIS at the individual or household level for rural
Figure 1. Mammoth Cave National Park is located in the karst region of central Kentucky and is easily
accessible from Interstate 65, a modern roadway that roughly follows the route of the old Dixie Highway
of the 1920s
1310
areas in the U.S. are rare because rural street ad-
dressing was not standardized until recently, and
many rural areas lack alternative data sets, such
as city directories, tax lists, or utility records, that
help locate urban populations. DeBat’s (2009)
GIS for rural Washington County, Oregon, used
federal plat maps created under the aegis of the
Donation Land Claim Act, a data set unique to
the Oregon Territory. Thomas and Ayers’ (2003)
Valley of the Shadow project, investigating the
impacts of slavery at the time of the Civil War,
created a detailed GIS of one Northern and one
Southern county from census and military service
records, letters, and newspaper articles. The web
site for this project, however, presents only static
maps.
The MCHGIS is methodologically innovative
in its use of qualitative techniques to map census
data for an entire rural community at the household
level and for its creation of an interactive web
interface that allows users to explore the GIS. A
detailed topographic map prepared in 1930 as part
of the land acquisition process in the Mammoth
Cave region showed house locations, but the Park
retained no information on who lived in each
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
house. The authors developed a three-prong ap-
proach to matching census households with known
house sites, comparing head-of-household house
ownership data from the census with national
park parcel purchase records, relying on local
knowledge of key informants, and interpolating
between existing matches based on the presumed
path of the census taker.
The MCHGIS is also innovative in its adoption
of techniques from web-based public participa-
tion GIS (PPGIS) to foster a collaborative site of
public memory for the pre-park population of the
Mammoth Cave region. Users may interact with
the MCHGIS through an easy-to-use web GIS
interface, viewing house and parcel layers overlaid
on a hill shaded topographic map. Clicking on a
house site brings up the 1920 census data for that
household where a match is known, as well as
any photographs of family members, the dwell-
ing or farm outbuildings in the GIS. Layers are
fully searchable to assist in the location of family
names. Members of the public are encouraged to
submit additional photographs and other historic
information for inclusion in the MCHGIS.
The goals of this paper are to demonstrate the
utility of historical GIS in achieving new under-
standings of a national park landscape, document
the progress to date of the MCHGIS in creating a
geospatial database of the pre-park population and
a platform for dissemination of pre-park history,
and describe the creation of a geospatial frame-
work for public memory through a participatory
collaboration with former park residents and their
descendents.
TOWARDS A NEW UNDERSTANDING
OF A NATIONAL PARK LANDSCAPE
Individual U.S. national parks have varied in the
policies applied to resident populations at the
time of park creation and in their subsequent ap-
proaches to interpretation of cultural landscapes
within their borders. Yellowstone National Park,
created in 1872 as the first U.S. National Park, es-
tablished the model of national park as uninhabited
wilderness, yet the delineation of its boundaries
antecedent to European American settlement
was the exception rather than the rule (Dilsaver
& Wyckoff, 2005). When Glacier National Park
was created in 1910, sixteen years before the
Eastern trio of Great Smokies, Shenandoah,
and Mammoth, private inholdings amounting
to 10,000 acres were allowed to remain inside
Glacier’s boundaries. Moreover, homesteaders
and private land claimants were guaranteed both
rights to their land and access corridors to use it
(Dilsaver & Wyckoff, 2005). In contrast, residents
inside the boundaries of the eastern trio of parks
authorized in 1926 faced displacement, through
condemnation, if necessary.
These three Eastern parks have adopted differ-
ent approaches through the course of the twentieth
century to interpretation of their cultural land-
scapes and displaced communities, differences
that affect the visitor experience in the parks today
as well as public understandings of their history.
Great Smoky Mountains National Park was the
first of the three to embrace interpretation of its
displaced community, through the (re-)construc-
tion of Cades Cove, owing largely to the interest
and initiative of one of the park’s rangers (Young,
2006). While initial approaches to interpretation
created a romanticized vision of an Appalachian
community in harmony with nature and embody-
ing a mythological primordial society, subsequent
refinements in interpretation were more culturally
sensitive and environmentally sustainable (Young,
2006). What is significant is that, from the 1940s,
this park embraced its settlement history as part
of the story told to visitors. Shenandoah National
Park has also seen contestation about interpretive
materials concerning its displaced population and
responded by altering displays most offensive
to descendents (Horning, 2001). Archaeological
surveys begun during the 1990s have furthered
interpretation of its pre-park communities, some
of which results are included on the park’s offi-
1311
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
cial web site in a section titled “The Displaced”
that explores cultural and economic differences
between various former communities within the
park and addresses the history of cultural resource
interpretation (Horning, 2001). In contrast to
these efforts at interpretation, similar efforts at
Mammoth Cave National Park are notably lack-
ing. A brand new visitor center opened in 2010,
featuring displays on the physical environment
and the development of tourism, but virtually
nothing on the pre-park population. The “History
and Culture” section of the park’s official web
site contains a database of individuals buried in
over 80 cemeteries located throughout the park,
a section on black history at Mammoth Cave that
recognizes the contributions of the many African
American cave guides, and a link to a selection of
digitized photographs taken by a Civilian Conser-
vation Corps photographer documenting the early
years of park administration. The “Brief History
of Mammoth Cave” that makes up the remainder
of this section discusses the national park idea,
the legislation behind the park’s creation, the cost
of land purchases, and the work of the CCC in
transforming the landscape, but the communities
and people who were displaced are not mentioned
at all. Thus, the official approach to interpretation
of pre-park communities at Mammoth Cave has
been resounding silence.
Scholars have begun to probe such silences in
park histories. As William O’Brien (2007) points
out in his work on segregation in Southern state
park systems, silences may initially occur because
these aspects of social history are painful or dis-
turbing to recall, but lack of discussion leads to a
loss of institutional memory that perpetuates the
silences. This is a historical trajectory that seems
to be operating at Mammoth Cave. Oral histories
by the Mammoth Cave’s first rangers document
a high level of bitterness among some of the
displaced, leading, in one instance, to a ranger
being shot (Lally, 1987). Such incidents might
understandably lead to avoidance of dealing with
the issue of displacement in park interpretation.
1312
As might be expected, however, the passing of
time has greatly tempered bitter feelings. The
former residents who are still alive were children
when they left the park and, in discussions with
the authors, stress that they had a child’s view
of the situation. They and their descendents,
while retaining feelings of attachment to former
homesteads, do not express ill will towards the
park, but an interest in preserving memories of
their ancestors and their communities. This seems
an appropriate time, then, to open a dialog about
appropriate and effective means of interpreting
the pre-park communities at Mammoth Cave and
foster a new understanding of this national park
landscape. The MCHGIS, an academic initiative
launched by the authors, contributes to this effort
by recovering and making accessible primary
materials that document the history of the people
and communities that were removed to make the
national park.
Most visitors to Mammoth Cave National Park
come to see a portion of the world’s longest known
cave system, the raison d’etre for the park, but
many also enjoy camping, hiking a forest trail,
or boating on the Green River. In the visitor’s
center or on a cave tour, they may learn about
the region’s distinctive karst geology, the delicate
cave ecosystem, or the history of cave exploration.
In short, the visitor’s experience is primarily an
encounter with nature writ large. While several
cemeteries along roadways are well preserved,
there are few other reminders of the population
that used to live here and little interpretation of
the park’s human history. It is safe to say that the
typical visitor remains unaware that the iconic
national park landscape that they encounter, one
dominated by forest and devoid of human habi-
tation, is a recent creation. Less than a century
ago the area was occupied by rural hamlets and
scattered farmsteads, with much of the land in
open pasture and cropland. Mammoth Cave was
a private tourism attraction with absentee owners
run by a local community of guides, hotel workers,
and other tourism support staff (Algeo, 2004).
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park

The MCHGIS reveals the extent to which
the national park landscape has been socially
constructed by mapping the pre-park popula-
tion, revealing a lightly, but extensively settled
landscape (Figure 2). Unlike the Great Smoky
Mountains, where vast tracts of unpopulated
land were held by timber companies (Campbell,
1960), the Mammoth Cave region was thoroughly
settled and farmed. Photographs taken during the
early 1930s by a Civilian Conservation Corps
photographer reveal extensive open pastures and
fields. Early park administrators followed a dual
policy of removing signs of human habitation and
re-creating a “natural” landscape via a massive
tree planting campaign to compose a landscape
that accorded with the institutional vision of
what a national park should be. Subsequent visi-
tors encountering the re-natured landscape, now
devoid of farm fields, pastures, and houses, have
reinforced for them the concept of a national park
as a wilderness devoid of human imprint. Thus,
the social construction of the Mammoth Cave
National Park landscape was both material and
ideational (Demeritt, 2002). Since interpretive
material presented at the park favors a discourse of
nature, science, and cave exploration, with human
elements being most prominent in the latter, inter-
subjective encounters between visitors and park
guides discursively reinforce notions of the park
as nature over the park as historic human habitat
(Berger & Luckmann, 1966). By contributing to
a greater understanding of the area’s settlement
history and highlighting the social as well as
environmental changes wrought by the national
park, the MCHGIS creates “new insights into the
geographies of the past,” thus fulfilling Gregory
and Healey’s (2007) criterion for “the real test for
historical GIS as a discipline” (p. 644).
Specific outcomes of the MCHGIS include
data integration, visualization, and analysis, which
Gregory et al. (2001) identify as the three advan-
tages of using GIS in historical research. Public

Figure 2. The Mammoth Cave Historical GIS with parcel and house layers shown over a modern
variable-resolution streaming topographic map. The MCHGIS reveals an extensive pre-park settlement
pattern (identified occupants from 1920 are shown as dark gray box with white outline in print edition)

1313
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park

records and archival photographs are correlated,
using location as an organizing framework (Fig-
ure 3). Linking the manuscript census record for
a household to the same location as a photograph
of a farmstead or the structure in which the fam-
ily lived adds a qualitative understanding of living
conditions that cannot be discerned from census
data alone, making possible new understandings
based on associations among data sets. Superim-
posing house locations on a topographic map
rendered with hill shading permits visualization
of settlement patterns and a better understanding
of how homesteads were situated in the landscape.
As more records and photographs are added to
the MCHGIS, it will become a data-rich explor-
atory environment that will assist genealogists
and family historians, as well as more academi-
cally-oriented analysis (Ruvane & Dobbs, 2008).
Preliminary demographic analysis of informa-
tion assembled in the MCHGIS suggests that a
new interpretation of forces behind the park’s
creation may be in order. Brunt (2009)2 paints a
portrait of a relatively stable agrarian society in
the Mammoth Cave region of 1920. Farming was
overwhelmingly the most common occupation
among the slightly more than 2100 individuals
living within what would become the park bound-
ary. Nearly two-thirds of families owned the homes
where they lived, in contrast to much higher
tenancy rates in other parts of the South. Thus,
the removal of the Mammoth Cave population
should not be confused with 1930s-era rural re-
settlement programs in which struggling farm
communities were relocated to model settlements.
Such analysis helps frame the creation of Mam-
moth Cave National Park as a political approach
to state-wide economic development rather than

Figure 3. The Mammoth Cave Historical GIS combines census, house site, and parcel data with CCC
photographs of dwellings that were razed to make way for the national park. Although the streaming
topographic map has resolution limitations at extremely large scales, it is useful for identifying landscape
features such as the African American cemetery on Jennie Mansfield’s land

1314
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
a humanitarian response to a failing agricultural
sector.
The example of Brunt (2009) places the MCH-
GIS within the emerging field of digital humani-
ties, establishing it as more than a digital archive,
but also a tool, study object, exploratory laboratory
and activist venue for engaging with and better
understanding our world (Svensson, 2010). While
early phases of the project focus on assembling
a digital library of materials and making them
publically accessible through innovative and ex-
perimental geospatial techniques, their potential as
an object of study and analysis is exemplified by
Brunt (2009). The activist element of the project is
embodied in the preservation and dissemination of
public memory about Mammoth Cave’s displaced
communities in the face of institutional silences.
The authors’ continued engagement with National
Park personnel to encourage interpretation of this
aspect of the park’s cultural history may likewise
be considered an activist endeavor.
MAPPING RURAL HOUSEHOLDS
FROM THE CENSUS
Mapping entire communities at the household
level rather than using aggregated data or sample
populations not only allows more precise measure-
ment and visualization, but also permits analysis
of within-group variation and individual-group
relationships to better reveal social dynamics
(DeBats, 2008). “In the historical world, how-
ever, the scarcity of places fully mapped at the
individual level remains a limiting feature for the
spatial analysis of past communities,” in large part
because the kind of spatial data about individuals
that we take for granted in contemporary society
is lacking (DeBats, 2008, pp. 17-18). Historical
GIS that have mapped individuals or households
are almost exclusively for urban areas. Urban
households are often possible to locate precisely
thanks to Sanborn fire insurance maps or plat maps
created for utility or tax purposes. Attempts to
create historical databases on individuals in rural
areas of the U.S. are few because of the greater
difficulty of attributing location to rural house-
holds. Household census records in many U.S.
cities included street addresses as early as 1880,
but rural areas lagged in systematic use of street
names and numbers (DeBats, 2008). As part of a
series of four case studies creating individual-level
GIS for different settlements in North America,
DeBats (2009) mapped 71% of inhabitants of
rural Washington County, Oregon, to a place of
residence for the 1850s. This was the only rural
U.S. case study among the four, and the high level
of mapping was made possible by a set of federal
land plats unique to the Oregon Territory.
The MCHGIS attempts to map all households
from the 1920 manuscript census for the area that
would become Mammoth Cave National Park.
U.S. Census Bureau policy mandates that data for
recent censuses be aggregated because of privacy
concerns, but individual data is released after 72
years. Thus, images of the original forms filled
out by census takers as they went from house to
house are available for both the 1920 and 1930
censuses. Of the two, the 1920 census provides
the better snapshot of the pre-park population for
Mammoth Cave. Although most land purchases
for the national park were made during the 1930s,
land speculators became active after Congress
authorized the park in 1926, and some families re-
located during the late 1920s in anticipation of the
park’s creation. The manuscript census contains a
wealth of socio-economic data about individuals,
including age, occupation, homeownership, and
household composition. Mapping the population
at the household level rather than aggregated by
census tract or other administrative unit both
permits fine-grained demographic analysis and
allows the household to serve as a container for
other types of data as the GIS expands.
A significant challenge in using the 1920 census
for rural Kentucky is that no location information
below the level of the census district is available.
Street names and house numbers were not then
1315
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
in use in the rural portions of the counties (Bar-
ren, Edmonson, and Hart) from which Mammoth
Cave National Park was carved. While manuscript
census forms for towns such as Brownsville, the
Edmonson County seat, have marginal notations
naming the street on which a house lies, forms
for rural areas simply number houses sequentially,
with no notation when the census worker left one
road for another or of which side of the road a
house occupied. Moreover, other forms of spatial
residential information, such as city directories,
fire insurance maps, plat maps, and utility records
are lacking for this area.
Two other data sets were useful for placing
households on the landscape – known house sites
and parcels. House sites were digitized by Park
Service personnel from a topographic map created
by a 1930 U.S. Department of the Interior survey
as part of the land acquisition process for the park.
This resulted in a point data layer identifying
664 former residences, but with no information
about who lived in the houses. These structures
were dismantled during the 1930s by the Civilian
Conservation Corps, which supplied manpower for
most of the landscape alterations needed to ready
the park for opening, including removing fences,
filling stock ponds, and a massive tree-planting
campaign. The challenge for the MCHGIS was to
match these former house sites with the families
who occupied them in 1920. The temporal offset
between the 1920 census and 1930 survey of house
sites must be kept in mind. A few houses extant in
1920 may not appear in the 1930 data set because
of their destruction, e.g. by fire, and any houses
constructed after 1920 will not have a matching
census household. The latter are likely to be few,
however, particularly after the 1926 authorization
of the national park when it was known that the
area was to be cleared of inhabitants.
Park Service personnel also digitized parcel
boundaries of the 532 tracts purchased from
private landholders to form the park. This layer
includes the name of the former land owner as an
attribute. Thirteen tracts were church or school
1316
properties, and thus, were not residential. When
overlaid with the house site layer, 152 tracts were
found to contain no houses, leaving 377 residential
tracts. Over a third of the residential parcels (142)
contained multiple dwellings. In this region, the
level of absentee ownership was relatively low,
but because extended family members or tenants
typically occupied additional houses on a family
farmstead, knowing a parcel’s owner was in many
cases not sufficient to pinpoint the house occupied.
A three-prong methodology was developed to
match households listed in the 1920 manuscript
census to house sites: 1) probable residency based
on the owner of a tract with a single house on it
being listed in the census as a head-of-household
and a home-owner; 2) local historical knowledge of
a key informant; and 3) interpolation of household
records along a road between two existing matches.
It is recognized that error may be introduced
during each phase of the methodology and that
error propagation may occur when interpolating
during the third phase from presumed matches
made during the first two phases. As Gregory and
Healy (2007) note, “Historical GIS databases are
rarely a simple digital facsimile of a single source.
Instead they take data from multiple sources,
integrate them in a manner that is sympathetic to
the sources’ limitations, and create metadata and
documentation to record the sources and standards
used” (p. 639). In order to track uncertainty, an
attribute was added to the MCHGIS house site
layer indicating the basis on which a match was
made with a census household and the level of
certainty of that match. This is what Gregory and
Healy (2007) term the “documentary” approach
to handling error in historical GIS databases (p.
641). As new sources of data are uncovered or
additional informants provide more information,
we will be able to reevaluate and revise matches
based on the best available information.
The first phase of the methodology compared
land tract owners with census heads-of-house-
holds. Of the 235 tracts containing a single house
site, 67 owners were located in the census as
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
heads-of-households who owned their own homes.
A simplifying assumption was made that these
homeowners likely lived on the tract they owned
and the census household was matched with the
house site. The second phase of the methodology
involved interviews with a key informant, Norman
Warnell, who has extensive local knowledge of the
Mammoth Cave region and of the families who
used to live there. Mr. Warnell taught science for
many years in the Edmonson County School Sys-
tem, authored a folk history of the Mammoth Cave
region (Warnell, 2006), served as an officer of the
Edmonson County Historical Society, and spends
much of his free time hiking off-trail through the
park, lending him an intimate familiarity with the
remnant roads and traces of old homesteads. Mr.
Warnell verified many of the household-house site
matches made during the first phase of the project,
and was able to supply, based on his knowledge
of the region, the names of residents for a number
of additional house sites. Some of his matches
were subsequently verified with former residents
and their descendants during a public exhibition
of the MCHGIS at a Homecoming event held at
the national park, giving the authors a high level
of confidence in the matches Mr. Warnell made.
The third phase of the methodology located
households within the manuscript census that were
already matched to house sites in close geographic
proximity and sought to interpolate the gaps
both in household records and house sites. That
is, if the manuscript census shows three records
between two census household records that are
already matched to house sites A and B, and the
house site layer shows three houses along the road
between houses A and B, then we assumed the
path of the census worker ran between houses A
and B and matched the intervening houses to the
intervening census records (Figure 4). Through
the combination of the three methods, 230 of 664
house sites (35%) were matched to 1920 census
household records (Brunt, 2009). Although this
match rate seems low, we are hopeful of being
able to make additional matches during the public
participation phase of the project, which will make
the MCHGIS accessible to many more potential
local informants and potentially uncover new
data sources.
DISSEMINATING PRE-PARK
HISTORY
A web-based interface was developed for the
MCHGIS to disseminate materials documenting
the historic settlement history of the Mammoth
Cave region. While the primary data used in the
MCHGIS is publicly accessible, there are practi-
cal barriers to access by the general public. For
instance, the 1920 manuscript census for Kentucky
is available on microfilm in a number of federal
depository libraries (or on subscription-based
services such as Ancestry.com), but one has had
to travel to the library to use the microfilm (or pay
a subscription fee for online access) and then one
simply has the raw census data. It is the integration
of census data in geographic context with topo-
graphic maps and photographs that fosters a richer
understanding of the pre-park communities. Given
the common availability of the internet, a web ver-
sion of the MCHGIS seems the best way to reach
the widest possible target audience. The visual
orientation of GIS aligns well with that of the web;
Map-based internet services, such as Mapquest,
YahooMaps, and GoogleEarth, are ubiquitous, as
are the concepts of zooming in and out to alter the
scale of a map, layers of information that can be
turned on and off, and keyword searches. Thus, the
skills needed to operate a simple web-based GIS
have become part of American popular culture.
Moreover, the cost-of-entry for the internet has
decreased dramatically, with cell phones as well
as personal computers providing access, and most
schools and public libraries providing computers
with internet access.
The web interface was created with ESRI
ArcGIS Server (version 9.3.1) and is hosted on a
university computer at http://161.6.109.206/mam-
1317
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park

Figure 4. Interpolation of census records between known matches allowed additional houses to be matched
to census households. In this figure, the residents of houses labeled “2” and “5” were determined by
key informants. The corresponding manuscript census records were located. Note that a generational
transfer is evident in the Ritter match, reflecting the temporal disjunction between the 1920 census and
1936 parcel data. When a slight misalignment of layers is taken into account (the Carwell house should
be on the Carwell parcel and the Hogan house on the Hogan parcel), it was possible to match three
additional houses with census records based on an inferred linear route by the census taker

moth_HGIS/. The internet mapping application
displays MCHGIS data including house, parcel
and roads layers on top of a streaming topographic
map provided by the ESRI Resource Center. Two
advantages of the streaming map are its seamless-
ness and its variable resolution, allowing the image
to remain clear at any zoom level. A high quality
topographic map facilitates user interpretation of
the MCHGIS, as many former residents and their
descendants associate their memories in refer-
ence to topographic features such as a streams,
ridges, and sinkholes. Layers may be turned on
and off, and standard map navigation tools al-
low the user to zoom and pan. An inset overview
map shows the extent of the main frame as the
user changes map scale. The “Search for Family
Name” tool allows users to enter part or all of a
family name and returns a sorted list of matching
records from the parcel and house layers (Figure
5). Users can right click on a result and zoom to
the matched feature. The search results include
key attributes for matched parcels, such as acre-
age, and for houses that have been matched to a
census household, including head-of-household
name, age, and occupation. A help page explains
all of these features.
Custom HTML code was added to the map
application to allow hyperlinking of photographs.
Fields in the parcel and house layers specify the
physical file location of a high resolution.jpg and
metadata documenting the photo’s source. The
hyperlink is displayed along with other attribute
information for a parcel or house, and when a user
follows the link, a new window displaying the

1318
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park

Figure 5. The parcel and house site layers are searchable by name, allowing users of the Mammoth
Cave Historical GIS to locate properties occupied by ancestors

photograph is opened. Captions written on the
verso of the CCC photographs, presumably by
the photographer or his assistant, in many cases
identified the parcel number of a farm or the name
of a house’s owner or resident, allowing photo-
graphs to be matched to parcel and house features.
The CCC photographs had been scanned as high
resolution.jpgs and the captions entered into a
database by Park Service personnel. Matching a
photograph to a house or parcel sometimes re-
quired reconciling spelling variations. Since the
authors used processed versions of the captions,
they could not distinguish between data entry
errors, cases when the photographer misrecorded
a name, or the local use of variant spellings. In
most cases, enough other information was present
to make a match. For instance, a photo captioned
“House to be razed on tract #281 T. M. Durham”
clearly is the house of Tom M. Denham on parcel
#281.
A pilot version of the MCHGIS was loaded
onto a laptop equipped with ESRI ArcMap soft-
ware and demonstrated at the 2009 Mammoth
Cave Homecoming, held on the Fourth of July
weekend. This is an annual event whose origins
date back over a hundred years as a combination
community celebration and tourism promotion
during the era of private ownership of Mammoth

1319
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
Cave. It has been revived by the National Park
Service in recent years for the same combination
of purposes. Musical performances, guided hikes
and other events of interest to the general public
are complemented by activities organized by and
for former residents of the park area and their
descendants, such as a genealogy room where
people gather to reminisce and share family his-
tory. Here people display extensive collections
of photographs, records, and genealogy charts
that document their family’s history. Reflecting
the diaspora of Mammoth Cave area residents,
some participants came from as far as Kansas and
Michigan. It was in this room that the MCHGIS
was demonstrated.
Feedback for the pilot system was overwhelm-
ingly positive from both Mammoth Cave area
descendants and national park personnel. Although
the system was in the early stages of develop-
ment, participants were receptive, even excited,
to use the interactive map and see family names
on parcels and houses and to see photographs ap-
pear upon clicking hyperlinks. Several attendees
offered historic photographs from their private
collections for scanning and inclusion in the
MCHGIS. Additional photographs were submit-
ted via email the following week after an article
describing the Homecoming and mentioning the
MCHGIS appeared in the Bowling Green Daily
News (Gaines, 2009).
PARTICIPATORY GIS AS
A FRAMEWORK FOR
PUBLIC MEMORY
The final phase planned for the MCHGIS is a col-
laborative site of public memory for the pre-park
communities. Public or collective memory resides
in consciously constructed representations of the
past, such as monuments and museums that both
memorialize the past and, at the same time, con-
tribute to its ongoing social construction (Foote,
1997). Public memory linked to places of social
1320
significance is a key means of fostering national
or group identity (Hoelscher & Alderman, 2004;
Leib et al., 2000), although separate groups may
not be in consensus about the meaning conveyed
by a particular commemorative structure (Forest
et al., 2004; Linantud, 2008).
The most obvious examples of public memory
are statues, museums and other elements of the
built environment, but non-material culture, such
as commemorative naming practices, can also be
key components of public memory (Alderman,
2000). The evolution of the web has led to virtual
forms of public memory, particularly web sites
that pay tribute to people and events (Hess, 2007).
Virtual sites of commemoration have several ad-
vantages over material ones. They are accessible
to a much larger potential audience, and access
is unconstrained by the exigencies of travel to a
physical site. Web sites can be developed rapidly,
while bricks-and-mortar monuments are often
subject to design review and political contesta-
tion. For instance, while the official memorial to
the victims of the September 11 terrorism attack
at the site of the World Trade Towers languished
over competing visions, dozens of web tributes
appeared. Although created by individuals, these
web sites opened “space for communal memo-
rializing” (Hess, 2007, p. 813) through means
such as digital guestbooks that allowed others to
post tributes, reflect and grieve online, resulting
in “collections of vernacular voices, often with
contradictory and controversial stances on the
interpretations of public memory” (Hess 2007,
p. 828). Such multivocality arguably fulfills the
“public” dimension of public memory better than
the elite visions of artists, government officials and
other experts who design and approve the typical
statue or museum. A memorial open to comment
and annotation, essentially allowing the public to
co-construct it, closes a notable gap in meaning
between production and consumption of public
memory, for a monument may be understood in
ways very different from those intended by its
designers, particularly when the passage of time
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
has blurred other memories of the person or event
(Johnston & Ripmeester, 2007).
These commemorative web sites are vernacular
forms of public memory, as opposed to the elite
forms embodied in monuments and museums
designed by experts and approved by a political
process. Traditionally, vernacular forms of public
memory have been less visible than elite forms,
by virtue of occupying less prominent positions
in the landscape (e.g. roadside crosses marking
the deaths of ordinary citizens versus generals
memorialized with a statue in the town square)
or because of use of ephemeral materials (e.g.
flower tributes to Princess Diana placed outside
the gates of Buckingham Palace), but the web,
with its ability to draw a virtual flash mob of at-
tention, through links and blogs, promises to be
something of an equalizer in terms of the promi-
nence of virtual real estate.
The methods of public participation GIS
(PPGIS) will extend the MCHGIS web site to
a virtual site of public memory for the former
residents of Mammoth Cave National Park and
their descendents (Epperson, 2010). PPGIS, also
called participatory GIS (PGIS) and community-
integrated GIS (Dunn, 2007), is well suited to
fostering a digital form of collaborative public
memory, although its use in this manner is rare
to date. The plethora of names for this form of
GIS signal a technology in its infancy for which
standards, including a name, have yet to be settled,
but they also indicate its perceived usefulness, as
many researchers are pursuing related forms of
development. What their efforts have in common
is the intent to democratize GIS, making it more
socially-inclusive by widening access to it among
a variety of social groups and by cultivating those
groups’ participation in creation of geospatial data
represented in GIS.
Participatory GIS has been most often used in
the areas of urban and environmental planning, to
solicit the perspectives and knowledge of residents
of a place whose lives will be impacted by decisions
being made about infrastructure and resources (e.g.
Craig et al., 2002; Elwood, 2002; Ghose, 2001).
Web-based GIS enhances public participation in
some of the same ways that web-based memori-
als do – by reaching a larger virtual audience
than could attend a physical meeting, by offering
asynchronous access, and by the possibility of
anonymity, which reduces psychological barriers
to participation (Wong & Chua, 2001). Although
technological barriers to participatory GIS still
exist in developing countries, among most of the
United States’ population, “GIS technologies are
no longer elitist” (Wong & Chua, 2001, p. 65),
a statement even more true a decade after Wong
and Chua wrote it.
Web sites such as the Mark Twain’s Missis-
sippi Valley (Vandecreek, 2007) and the Civil
War-era site, Valley of the Shadow (Thomas &
Ayers, 2003), foster public memory by making
available a huge variety of materials, text, images,
and video, for the public to explore. They use a
digital library model, presenting materials housed
at academic and historical institutions that have
been selected for preservation by experts and
digitized. This is essentially a top-down approach
to public memory. The MCHGIS mixes top-down
and bottom-up approaches, allowing integration
of institutional material selected by experts with
information from private holdings deemed sig-
nificant by members of the public. Most of the
material currently available to the public about the
pre-park population is institutional in origin, e.g.
the census and records collected by the nascent
national park administration, yet a much greater
volume of historical material remains in private
hands among former residents of the park area
and their descendants. Members of the public may
submit material for addition to the MCGIS by
identifying the relevant household and submitting
a digital file (e.g. a photograph or scanned docu-
ment) and a signed donation agreement. Ethical
concerns require that potential contributors to the
MCHGIS be aware that their digital donations will
be freely accessible and downloadable over the
internet and that they must be free of copyright.
1321
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
Donors will be asked to sign a contribution from
stipulating these conditions. They will be given a
choice of being noted as the donor or having their
contribution listed as anonymous. The privacy
insured by the latter choice may help encourage
participation by individuals who would hesitate to
be seen as critical of the National Park Service. It
is typical for web GIS to employ a “gatekeeper”
for new submissions (Dunn, 2007) in contrast to
the wiki model of community-created data sets
where anyone can write or edit a wiki page. Filter-
ing public submissions is desirable for data quality
control but also because the data structures behind
the web GIS are not directly modifiable by the
public. A GIS expert will need to add hyperlinks
for new photographs and incorporate other data
into existing structures.
Although participatory GIS is generally char-
acterized as inclusive and democratizing, actual
projects generally originate in technologically-
informed settings, especially academia and gov-
ernment. One danger, then, is that because the
projects are conceived by technocrats rather
than by members of the target communities, “‘a
web-based PPGIS may run the risk of providing
services that the public does not want” (Wong &
Chua, 2001, p. 626). Since the MCHGIS was de-
veloped by an academic group, we should explain
our perception of the need for it. First, the family
history activities at the Mammoth Cave Fourth
of July Homecomings demonstrate a significant
constituency for public remembrance of the pre-
park inhabitants. Second, interest in digital forms
of public remembrance and a willingness to share
privately-held historical materials are manifest
in the section of the web site MyFamily.com de-
voted to Mammoth Cave. MyFamily is a social
networking site run by Ancestry.com intended to
allow sharing of photographs, videos, and other
files, host discussions, and share an event calendar
among members of a “family,” with membership
defined by the person who creates a particular
family group. The Mammoth Cave “family” was
created by a descendant of former residents of the
1322
park area who is active in genealogy, and as of June
2010, the Mammoth cave “family” had over 900
members. Members use the site both for sharing
historical information and for keeping up with
the current lives of friends and family. Wong and
Chua (2001) note that web-based PPGIS present
barriers to use by some community groups, but
we anticipate that the Mammoth Cave MyFam-
ily group, which currently shares an abundance
of digital files, both image and text, and makes
sophisticated use of online chat and blogging
facilities, is a good target audience for a PPGIS
application because a large number of interested
individuals have demonstrated the kind of com-
puter skills necessary to use a simplified web GIS
interface. Hundreds of photographs have been
uploaded to the MyFamily Mammoth Cave web
site, but without a strong organizational frame-
work, images of a particular family or place can
be difficult to locate. The geospatial framework
of the MCHGIS offers a distinct advantage by
grouping materials related to a family, neighbor-
hood, and community. Moreover, as a publicly
accessible web site, anyone can use the MCHGIS,
not just those who have been invited to join, as
with MyFamily.com. In keeping with Dunn’s
(2007) view of community-integrated GIS, the
MCHGIS “acknowledges the ‘expert’ nature of
GIS as a technology but enhances citizen access
and participation and, hence, the democratic
potential” (p. 620).
REPRESENTATION AND
REPRESENTATIVENESS
One hope of the MCHGIS developers is to reach
former park area residents and descendants who
do not currently participate in Mammoth Cave
Homecomings or MyFamily.com and, thus, to
broaden the pre-park population represented in the
memory-making process. Alderman and Modlin
(2008) point out that commemorative web sites
can inadvertently create biased understandings by
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
focusing on the lives of a small number of atypical
individuals who might be taken as representative
of a larger group. Their recommended strategy
of documenting the wider community is at the
heart of the MCHGIS methodology. There are
substantial challenges to achieving this goal; how-
ever that is historical and institutional in nature.
Participants in the Mammoth Cave Homecom-
ings, the MyFamily web site, and local histori-
cal societies are the easiest to reach and invite
to become engaged with the MCHGIS project.
Participation in these venues is self-selected, to
a degree, but also influenced by existing social
networks and residential proximity. For instance,
although some Homecoming attendees traveled
from distant states, most of those the authors
met lived within a short drive of the park. As an
example of non-representative participation, the
authors did not observe any African Americans
in the genealogy room during the 2009 or 2010
homecomings, despite the fact that most of the
Mammoth Cave guides prior to park creation
were African American and most lived with their
families in the Flint Ridge area that is now part
of the park. Jerry Bransford, a descendent of the
storied Bransford family, many of whom were
guides during the nineteenth and early twentieth
century’s, continues his family’s guiding tradition
as a part-time ranger at the national park. Thus,
the dearth of African American participants in the
genealogy room activities does not stem from a
disinterest in their Mammoth Cave heritage, but
from other, unknown factors. No systematic study
has been done of where the displaced population
went after leaving the Mammoth Cave area. Anec-
dotal evidence is that some moved to nearby towns,
including Cave City, Glasgow, and Louisville, and
that some who had particularly strong ties to place
and the financial wherewithal to do so purchased
new homes just outside the park boundary. The
authors suspect that differences in socioeconomic
status, especially home and property ownership,
and race may have influenced migration decisions
following displacement. Testing of this hypothesis
must await a future study, but if true, differential
migration patterns are a historical factor that likely
affects participation in Homecoming events and
local historical societies.
FUTURE WORK
It is too early to know how successful a participa-
tory MCHGIS will be, but it holds the potential
for broadening representations of the history of
the Mammoth Cave region, letting the residents
and their descendants assert a presence that, in
many ways, has been officially minimized (e.g.
by the removal of most human artifacts from
the landscape and from the lack of interpretive
emphasis by park rangers). While it is no doubt
true that some segments of the former resident
population will benefit more from a historical
GIS than others, owing to differences in internet
use, interest in genealogy, and amount of family
history preserved, the MCHGIS offers the poten-
tial for participation to all with a minimum set
of technology resources which are now widely
available through internet-connected computers
in homes, public libraries, and schools.
To gauge who is using the web site and how
they are using it, a survey enables users to provide
anonymous feedback. The “Take a Survey” link
opens a new window with a ten-question survey
created in the Qualtrics online survey software.
The survey, which was approved by the authors’
university Human Subjects Review Board, queries
users about their background, prior experience
with web mapping applications, ease in using the
GIS application, and their perception of the site’s
usefulness. It also provides an opportunity to report
problems with the web site and leave additional
open-ended feedback. Analysis of survey results
after the web site has been online for a while will
allow the authors to assess their success in dis-
seminating knowledge of the site and tweak its
interface to resolve any usability issues.
1323
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
The work of adding historical materials to the
MCHGIS will be ongoing. User contributions will
be incorporated as they are emailed to project
staff. Based on a few initial contributions already
received, we anticipate that a certain amount of
email back-and-forth will be needed to best place
contributed photographs on the landscape or to
pinpoint which house or family a contribution
belongs with. Future plans include expanding
the multimedia capabilities of the MCHGIS to
include audio clips selected from oral history
interviews that have been recorded with former
residents over the years and that reside in the
Mammoth Cave National Park collection. Since
GIS-produced maps have the power to prompt
recollections, stories, and personal and family
narratives (Bodenhamer, 2007), we also plan to
explore using the MCHGIS as a prompt during
future oral history interviews, essentially to talk
with subjects about the materials available on
the site in hopes of evoking thick ethnographic
descriptions of communities and localities.
CONCLUSION
The applied value of the MCHGIS stems from
its preservation and dissemination of the human
history of Mammoth Cave National Park. It will
help the general public understand the landscape
transformation required to create the iconic “wil-
derness” they encounter when they visit the park
today, it will help rangers and other park personnel
interpret the area’s history, and it will help former
residents of the Mammoth Cave region and their
descendants retain a valued cultural identity, one
that has been marginalized through neglect in of-
ficial histories and interpretations. The MCHGIS
integrates and makes accessible data from multiple
archival sources, including the 1920 manuscript
census and a collection of photographs of many of
the pre-park dwellings taken by a CCC photogra-
pher. It also provides a platform for integrating in
the same geospatial framework some of the vast
1324
privately-held stores of historical materials, thus
increasing publicly accessible materials about the
region’s people and history.
We believe, with Bodenhamer (2007), that “a
GIS-facilitated landscape of memory… is a vision
worth pursuing” (p. 108) because of its potential for
fostering a more inclusive construction of history.
By welcoming multiple voices and privately-held
memories, the MCHGIS can reveal a much richer
past than can be read from institutional holdings
alone, a past that will inform expert interpreta-
tions, offer alternative visions, and fill some of the
silences surrounding a marginalized group. The
web-based public participation GIS will integrate
diverse forms of data in a geospatial framework,
serve as a repository of local knowledge, provide
persistent access to historical data, and serve as a
model for collaborative public history.
ACKNOWLEDGMENT
The authors would like to thank the generosity
of Charles Silbereisen and the Silbereisen Fund
in providing financial support that contributed to
this research project. They gratefully acknowledge
the efforts of Kevin Cary, GIS Lab Manager at
Western Kentucky University, for his tireless as-
sistance in keeping the GIS servers running and
his instruction in internet GIS. They appreciate
the support and assistance of Mammoth Cave Na-
tional Park personnel in locating archival material,
including Bob Ward, Joy Lyons, Terry Langford,
Chuck DeCroix, and Lillian Scoggins. And they
are deeply indebted to all those who shared their
memories, family histories, and photographs with
us, but especially Normal Warnell for the hours he
spent with us “talking Mammoth Cave.”
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
REFERENCES
Alderman, D. (2000). A street fit for a king: Naming
places and commemoration in the American South.
The Professional Geographer, 52(4), 672–684.
Alderman, D., & Modlin, E. A. Jr. (2008). (In)
visibility of the enslaved within online planta-
tion tourism marketing: A textual analysis of
North Carolina websites. Journal of Travel
& Tourism Marketing, 25(3-4), 265–281.
doi:10.1080/10548400802508333
Algeo, K. (2004). Mammoth Cave and the mak-
ing of place. Southeastern Geographer, 44(1),
27–47. doi:10.1353/sgo.2004.0002
Berger, P., & Luckmann, T. (1966). The social
construction of reality: A treatise in the sociol-
ogy of knowledge. Garden City, NY: Doubleday.
Bodenhamer, D. J. (2007). Creating a land-
scape of memory: The potential of humanities
GIS. International Journal of Humanities and
Arts Computing, 1(2), 97–110. doi:10.3366/
E1753854808000207
Brown, M. L. (2000). The wild East: A biography
of the Great Smoky Mountains. Gainesville, FL:
University Press of Florida.
Brunt, M. (2009). Analysis of Mammoth Cave
pre-park communities. Unpublished master’s
thesis, Western Kentucky University, Bowling
Green, KY.
Campbell, C. (1960). Birth of a national park
in the Great Smoky Mountains. Knoxville, TN:
University of Tennessee Press.
Craig, W. J., Harris, T. M., & Weiner, D. (2002).
Community participation and geographic infor-
mation systems. London, UK: Taylor & Francis.
DeBats, D. A. (2008). A tale of two cities: Using
tax records to develop GIS files for mapping and
understanding nineteenth-century U.S. cities.
Historical Methods, 41(1), 17–38. doi:10.3200/
HMTS.41.1.17-38
DeBats, D. A. (2009). Using GIS and individual-
level data for whole communities: A path toward
the reconciliation of political and social history.
Social Science Computer Review, 27(3), 313–330.
doi:10.1177/0894439308329759
Demeritt, D. (2002). What is the ‘social con-
struction of nature’? A typology and sympathetic
critique. Progress in Human Geography, 26(6),
767–790. doi:10.1191/0309132502ph402oa
Dilsaver, L. M., & Wyckoff, W. (2005). The
political geography of national parks. Pacific
Historical Review, 74(2), 237–266. doi:10.1525/
phr.2005.74.2.237
Dunn, C. E. (2007). Participatory GIS -- A people’s
GIS? Progress in Human Geography, 31(5),
616–637. doi:10.1177/0309132507081493
Dunn, D. (1988). Cades Cove: The life and death
of a southern Appalachian community, 1818-1937.
Knoxville, TN: University of Tennessee Press.
Elwood, S. A. (2002). GIS use in community
planning: A multidimensional analysis of em-
powerment. Environment & Planning A, 34(5),
905–922. doi:10.1068/a34117
Epperson, A. (2010). Internet GIS as a historic
place-making tool for Mammoth Cave National
Park. Unpublished master’s thesis, Western Ken-
tucky University, Bowling Green, KY.
Foote, K. (1997). Shadowed ground: America’s
landscapes of violence and tragedy. Austin, TX:
University of Texas Press.
Forest, B., Johnson, J., & Till, K. (2004). Post-total-
itarian national identity: Public memory in Germa-
ny and Russia. Social & Cultural Geography, 5(3),
357–379. doi:10.1080/1464936042000252778
1325
 Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
Frome, M. (1966). Strangers in high places: The
story of the Great Smoky Mountains. Knoxville,
TN: University of Tennessee Press.
Gaines, J. (2009, July 4). Mammoth homecom-
ing: National park hosts gathering for hundreds
of former residents of the area, once known as
Northtown. Bowling Green Daily News.
Ghose, R. (2001). Use of information technology
for community empowerment: Transforming
geographic information systems into community
information systems. Transactions in GIS, 5(2),
141–163. doi:10.1111/1467-9671.00073
Gregory, I. N., & Healey, R. G. (2007). Historical
GIS: Structuring, mapping, and analysing geogra-
phies of the past. Progress in Human Geography,
31(5), 638–653. doi:10.1177/0309132507081495
Gregory, I. N., Kemp, K., & Mostern, R. (2001).
Geographical information and historical research:
Current progress and future directions. His-
tory & Computing, 13(1), 7–21. doi:10.3366/
hac.2001.13.1.7
Hess, A. (2007). In digital remembrance: Ver-
nacular memory and the rhetorical construction
of web memorials. Media Culture & Society,
29(5), 812–830. doi:10.1177/0163443707080539
Hoelscher, S., & Alderman, D. H. (2004). Memory
and place: Geographies of a critical relationship.
Social & Cultural Geography, 5(3), 347–355.
doi:10.1080/1464936042000252769
Horning, A. J. (2000). Shenandoah’s secret history.
Archeology, 53(1), 44–52.
Horning, A. J. (2001). When past is present:
Archaeology of the displaced in Shenandoah
National Park. Retrieved from http://www.nps.
gov/shen/historyculture/displaced.htm
Ise, J. (1961). Our National Park policy: A critical
history. Baltimore, MD: Johns Hopkins Press.
1326
Johnston, R., & Ripmeester, M. (2007). A monu-
ment’s work is never done: The Watson monument,
memory, and forgetting in a small Canadian city.
International Journal of Heritage Studies, 13(2),
117–135. doi:10.1080/13527250601118951
Lally, K. (1987). Unpublished oral history
interview with Vernon Wells. Bowling Green,
KY: Mammoth Cave National Park Curatorial
Archives.
Lambert, D. (1989). The undying past of Shenan-
doah National Park. Lanham, MD: Roberts
Rinehart.
Leib, J. I. (2004). Robert E. Lee, ‘Race,’ repre-
sentation and redevelopment along Richmond,
Virginia’s Canal Walk. Southeastern Geographer,
44(2), 236–262. doi:10.1353/sgo.2004.0025
Leib, J. I., Webster, G. R., & Webster, R. H. (2000).
Rebel with a cause? Iconography and public
memory in the Southern United States. GeoJour-
nal, 52, 303–310. doi:10.1023/A:1014358204037
Linantub, J. L. (2008). War memorials and memo-
ries: Comparing the Philippines and South Korea.
International Journal of Heritage Studies, 14(4),
347–361. doi:10.1080/13527250802155869
O’Brien, W. (2007). The strange career of a Florida
state park: Uncovering a Jim Crow past. Histori-
cal Geography, 35(1), 160–184.
Pierce, D. S. (2000). The Great Smokies: From
natural habitat to national park. Knoxville, TN:
University of Tennessee Press.
Powell, K. M. (2007). The anguish of displace-
ment: The politics of literacy in the letters of
mountain families in Shenandoah National Park.
Charlottesville, VA: University of Virginia Press.
Reader, C., & Reader, J. (1978). Shenandoah
Heritage: The story of the people before the
park. Washington, DC: Potomac Appalachian
Trail Club.
Historical GIS as a Platform for Public Memory at Mammoth Cave National Park
Ruvane, M. B., & Dobbs, R. (2008). Genealogy,
historical geography, and GIS: Parcel mapping,
information synergies, and collaborative oppor-
tunities. In Timothy, D. J., & Guelke, J. K. (Eds.),
Geography and genealogy: Locating personal
pasts (pp. 43–62). Burlington, VT: Ashgate.
Schlichting, K., & Tuckel, P. (2006). Residential
segregation and the beginning of the Great Mi-
gration of African Americans to Hartford, Con-
necticut. Historical Methods, 39(3), 132–143.
doi:10.3200/HMTS.39.3.132-144
Svensson, P. (2010). The landscape of digital
humanities. Digital Humanities Quarterly, 4(1).
Thomas, W. G., & Ayers, E. L. (2003). An over-
view: The differences slavery made: A close
analysis of two American communities. The
American Historical Review, 108(5), 1299–1307.
doi:10.1086/529967
Vandecreek, D. E. (2007). Digital history and
the public: Envisioning the Mississippi Valley of
the nineteenth century. International Journal of
Humanities and Arts Computing, 1(2), 123–136.
doi:10.3366/E1753854808000220
Warnell, N. (2006). Mammoth Cave: Forgotten
stories of its people. Fort Washington, PA: Eastern
National.
This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 4,
edited by Donald Patrick Albert, pp. 19-37, copyright 2011 by IGI Publishing (an imprint of IGI Global).
Wong, S., & Chua, Y. L. (2001). Data intermedia-
tion and beyond: Issues for web-based PPGIS.
Cartographica, 38(3-4), 63–80.
Young, T. (2006). False, cheap and degraded:
When history, economy and environment collided
at Cades Cove, Great Smoky Mountains National
Park. Journal of Historical Geography, 32(1),
169–189. doi:10.1016/j.jhg.2004.06.001
ENDNOTES
1
For scholarly treatments of the Great Smoky
Mountains National Park and its displaced
population see Brown (2000), Dunn (1988),
Frome (1966), and Pierce 2000; for Shenan-
doah National Park, see Horning (2000),
Lambert (1989), Powell (2007), and Reader
and Reader (1978).
2
Brunt’s (2009) analysis is for the portion of
the national park that falls within Edmonson
County, which comprises the bulk of the
park. A very small portion of the national
park lies within Barren and Hart Counties.
1327
1328

Chapter 80
Applied Geospatial Perspectives
on the Rock Art of the
Lake of the Woods Region
of Ontario, Canada
John W. Norder
Michigan State University, USA

Jon W. Carroll
Michigan State University, USA

ABSTRACT
This study examines the role of rock art in the construction of Woodland Period (300 BC to AD 1700)
hunter-gatherer landscapes in the Lake of the Woods region of northwestern Ontario. The authors ex-
amine the distribution of documented pictograph sites relative to the locations of rock formations where
the geologic conditions would have favored the placement of pictographic rock art but are absent. Point
pattern analysis, Monte Carlo simulation, and least cost path analysis were used to analyze the findings.
The authors suggest that pictograph sites were placed at points on the landscape along water routes to
facilitate information exchange among highly mobile hunter-gatherers.

INTRODUCTION sites as records of the ideologies, histories, poli-

The study of rock art has been an area of archaeo-
logical interest for well over a century. However,
it has only been within the past three decades that
this heritage resource has gained mainstream
disciplinary attention. Researchers have become
increasingly aware of the significance of these
DOI: 10.4018/978-1-4666-2038-4.ch080
tics, and religions of early societies ranging in
complexity from hunter-gatherers of the Upper
Paleolithic to the empires of Ancient Egypt and
Rome. In addition, rock art has been an area of
greater efforts in heritage management plans as
increased development, air pollution, and other
human-related factors have either destroyed,
degraded, or currently threaten these sites. In
this context, the study of rock art sites takes on

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
greater urgency, not only for archaeologists, but
for people with connected histories to these places
such as those in this study.
This study examines the role of rock art in the
construction of Woodland Period (300 BC to AD
1700) hunter-gatherer landscapes in the Lake of
the Woods region of northwestern Ontario (Fig-
ure 1). Across the broader Canadian Shield, in
which the study area is located, rock art has long
been suggested as a means by which ancestral
Algonquian Indian communities communicated
information. The majority of research has focused
on images at these sites and has been informed by
ethnohistoric research and ethnographic fieldwork
among Algonquian speaking peoples of the region
– the Cree, Ojibwe, and Salteaux First Nations
(Dewdney & Kidd, 1972; Conway, 1993; Moly-
neaux, 1980; Rajnovich, 1989, 1994; Steinbring,
1992; Vastokas & Vastokas, 1973). Interpretations
derived from this research have emphasized reli-
gious and ritual meanings of these sites in terms of
cosmological significance, ritual practice, and as
Figure 1. Example of a CSRAT pictograph site (photograph by J. Norder)
components of the spiritual landscape of historic
and ancestral Algonquian Indian communities.
Recent and current research has focused more
on this latter topic further linking these places to
descendant First Nations aboriginal communities
as part of their contemporary cultural landscapes
(Arsenault, 2004; Norder, 2007, in press-a).
This paper builds on work conducted by one
of the authors (Norder, 2003) that examines an
additional area of study by focusing on the po-
tential significance of pictographic rock art site
placement not just as part of a spiritual landscape,
but as part of the broader social landscape of the
makers. Fieldwork conducted in 2000 (Norder,
2003) and 2005 (Norder, in press-b) strongly
suggested that locations of pictographs, or paint-
ed rock art sites (Figure 2), were intentionally
selected and placed on the landscape to foster
forms of secular information exchange related to
population mobility, and seasonal aggregation.
Despite these initial findings, it has remained to
be determined more conclusively whether the
1329
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
spatial distribution on the landscape is a result of
random behavior or if there is some form of pat-
terning associated with their production. Using
point pattern analysis, Monte Carlo simulation,
and least cost path analysis, we propose that
pictograph sites were placed in strategic locations
along specific water routes, which served to fos-
ter information exchange and guide travel and
other forms of secular human activities and be-
haviors. Understanding this intentional placement
of pictographs speaks to how these pre-Contact
aboriginal communities actively used the land-
scape to facilitate information exchange predict-
ably across space and time in structured and so-
phisticated manners that have previously been
overlooked.
PICTOGRAPHIC ROCK ART IN
NORTHWESTERN ONTARIO
Rock art within the study area is considered to
be part of a larger regional expression of rock art
referred to as the Canadian Shield Rock Art Tradi-
tion (CSRAT). The CSRAT has a site distribution
that spreads from Quebec in the east through the
Figure 2. Location of study area
1330
upper Great Lakes region and westward up through
Manitoba into Saskatchewan. This geographic
range is roughly equivalent to the spread of mem-
bers of sub-Arctic Algonguian speaking peoples
that include, as noted above, groups of Ojibwe,
Cree, Saulteaux and a variety of other culturally
and linguistically related groups. As a result, and
with supporting archaeological evidence, CSRAT
sites are considered to be Algonquian in origin
and this recognition has subsequently been used
as justification for application of historic and
ethnohistoric accounts to the interpretations of
these locations.
Rock art in the CSRAT occurs in four forms.
The first, and the subject of the authors’ research
here, are pictographs (Figure 2). These also are
the dominant rock art form of the CSRAT. The
second are petroglyphs, which are images that
are carved, incised or pecked into a stone surface.
The third are petroforms, and specifically refer to
arrangements or alignments of boulders that either
depict recognizable human/animal figures or form
abstract geometric patterns. The fourth and last
are an unusual form of geoglyph, referred to in
the region as lichenographs, and are made by the
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
scraping away of surface molds and lichens from
a rock surface to create an image.
While the majority of pictographic rock art
in northwestern Ontario is likely pre-European
in origin the exact time span of its production is
not known with any certainty. Part of this stems
from the nature of the production of the art itself.
The primary pigment used in the production of
CSRAT pictographs, derived from the ground
mineral red ocher, is not dateable by any direct
means (Keyser & Klassen, 2001; Nieves, 2006).
However, it should be noted that based on eth-
nohistoric and ethnographic sources red ocher
was likely mixed with an organic binding agent,
usually some form of animal fat or fish oil that
would have been dateable. Nevertheless, direct
exposure to the extreme seasonal weathering
processes of the region would have dissolved
this organic component after a relatively short
period of time. Regardless, there is sufficient
circumstantial evidence based on multiple lines
of ethnographic, ethnohistoric, and archaeological
evidence to suggest an antiquity consistent with
the Woodland Period of the region (Rajnovich,
1994; Arsenault et al., 1995), during which the
most developed hunting and gathering societies
inhabited the region (Reid, 1977).
As with rock art globally, difficulty in direct
dating has focused most research on the pursuit of
the meaning of rock art imagery (Chippendale &
Nash, 2004; Chippendale & Taçon, 1998; Clotte
& Lewis-Williams, 1998; Coulson & Campbell,
2001; David, 2001; Diaz-Granados & Duncan,
2004; Domingo-Sanz, Fiore, & May, 2008;
Hays-Gilpin, 2003; Helskog, 2001; Keyser &
Klassen, 2001; Layton, 1992; Loendorf, Chip-
pendale, & Whitley 2005; Nash & Chippendale,
2002; Schaafsma, 1980). Within the study region
considerable effort has been made to interpret the
rock art imagery using methods of standard stylis-
tic analysis combined with ethnographic studies
of the Algonquian descendants of the makers
(Rajnovich, 1989, 1994; Conway, 1993; Conway
& Conway, 1990; Dewdney, 1975; Dewdney &
Kidd, 1967; Vastokas & Vastokas, 1973).
The results of these studies have shed a great
deal of insight into elements of cosmological and
ritual behaviors and meanings associated with
these sites, but as with dating, the precise mean-
ing of any given composition remains elusive.
As noted by many of the researchers (Rajnovich,
1994; Dewdney & Kidd, 1967), despite insights
provided by ethnographic and ethnohistoric
sources, any single image could represent any
number of real world phenomena or cultural
ideological concepts. As a result, without direct
transmission of a composition’s meaning from
maker to audience and transgenerationally from
that point onward the meaning of a rock art site
becomes altered or lost. As ethnographic field-
work conducted with contemporary Algonquian
descendants of the makers in 2005 indicated, this
was much the case (Norder, 2007, in press-a).
To date, systematic documentation and
research of the rock art of Lake of the Woods
has been limited, with the majority of surveys
conducted either by provincial government ar-
chaeologists in a patchwork fashion throughout
the region (Lambert, 1985; Pastershank, 1989;
Rusak, 1992) or as part of single or specific site
comparisons (Conway, 1989; Rajnovich, 1980a,
1980b; Steinbring & Simpson, 1983). The first
systematic survey of the region was begun by
Norder in 2000 and elaborated substantially in
2005. The goal of the initial survey in 2000 was
to research the potential for identifying functional
sites types based upon information encoded in a
given rock art site, a determination based on the
use of the Shannon Information Measure (Hartley,
1992; Hartley & Vawser, 1998; Shannon, 1949).
The measure determines the relative ease with
which an observer would be able to interpret a
given message based on the redundancy of the
content of the composition. Initial results sug-
gested that there were at least three different types
of sites that communicated messages of particular
significance to anyone with a basic familiarity with
1331
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
the regional iconography. In this context, most
northern Algonquian-speaking peoples would
have had this knowledge since they shared similar
cosmological, ritual, and iconographic systems
(Norder, 2003, in press).
Turning to the research presented here, there
were several questions that remained after analy-
sis of the results of the 2000 survey data that we
have since sought to address with the survey work
conducted in 2005. First and foremost was to
try and develop a more complete understanding
of the distribution of pictograph sites in a given
area. The work of prior researchers, including
that conducted by Norder in 2000 (Norder,
2003), remained limited and opportunistic in its
regional coverage. As a result, it was not pos-
sible to determine if the distribution of the sites
was the result of documentary bias or not. Much
of the analysis that follows addresses this issue,
but it also examines additional factors related to
issues of site placement. As we will demonstrate,
ancestral First Nations populations of the region
clearly had numerous choices available to them but
there was intentional selection of particular loca-
tions over others. The identification of intentional
spatial placement on the landscape and selection
of specific locations due to culturally significant
geomorphological features will hopefully facili-
tate interpretations regarding the broader social
significance of CSRAT rock art and potentially in
other comparable socio-environmental contexts.
SURVEY METHODOLOGY
One of the primary features of the survey meth-
odology employed in both 2000 and 2005 was a
100% shoreline survey of a selection of bodies
of water and navigable river systems in the study
region. While this methodology was a standard
feature of provincial archaeological projects done
during the preceding decades (Adams, 1983;
Halverson, 1988; Lambert, 1985; Pastershank,
1989; Rusak, 1989, 1992) where the focus was
1332
on locating pictographic and other types of rock
art sites, the methodology employed by Norder
(2003) was to locate not only pictographic rock
art sites but to also document locations where
pictographic art could have been placed, referred
to from this point forward as ‘sterile’ locations.
The identification of sterile locations was based
on two main criteria. First, they had to be clean
surfaces relatively free of the ubiquitous growths
of regional lichens and molds and, as a result,
would have been suitable ‘canvases’ for rock art
imagery. Second, they would need to be relatively
easy to see from the water, a common feature of
nearly all documented CSRAT pictograph sites.
As a result, the surface angle of the rock face to
the water surface had to fall within a range of
between 120 to 60 degrees with the ideal, based
on normative criteria of other CSRAT pictograph
sites, falling between 90 and 70 degrees. These
sterile locations were documented in order to
demonstrate the relative availability of ‘canvases’
well in excess of those actually selected (Norder,
2003, in press).
In addition, a series of geomorphological
variables were documented for both rock art sites
and sterile locations. Regional archaeologists have
suggested that certain geophysical properties of
the sites where pictographs and petroglyphs may
be found were also typically located in conjunc-
tion with certain surface features such as rock
overhangs, places to sit, the presence of cracks,
quartz or other crystalline mineral veins, orienta-
tion along cardinal or ordinal directions, angle to
water, etc. Such features were demonstrated to
have cultural significance to Algonquian com-
munities (Arsenault, 2004; Conway, 1993; Mo-
lyneaux, 1980; Steinbring, 1992). Basic statistical
analysis determined that there were several of
these variables, typically visual elements such
as the presence of quartz veins and crevices in
the rock surface, that were directly related to the
production of rock art sites, confirming elements
of these observations (Norder, 2003).
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region

The shoreline surveys conducted in 2000 and
2005 documented 51 pictograph sites (44 previ-
ously documented sites and seven new pictograph
sites identified in the 2000 survey) and 610
sterile locations (Figure 3). Because of the large
number of lakes within the study area, n>2000,
a systematic strategy was employed in 2005 that
focused on two of the major watersheds that empty
into Lake of the Woods (Figure 3). Within these
watersheds, comprised of slightly more than 300
navigable lakes and interconnecting waterways,
a two-tiered survey was undertaken. All lakes
with shorelines greater than 5 km in length were
surveyed in these two watersheds, while a 20
percent random sample of those with smaller cir-
cumferences were surveyed. Each watershed was
further divided into four quadrants for a total of
eight survey regions. Six of these were surveyed
using this methodology. The remaining two were
not due to time and access limitations. Two of the
lakes surveyed in 2000 fell into the 2005 survey
area and were resurveyed to check for consistency

Figure 3. Locations of pictograph sites, sterile locations, and 2005 survey areas

1333
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
of documentation procedures between the field
seasons. Overall, the combined shoreline surveyed
represented approximately 60 – 70% of the total
available shoreline of all bodies of water in the
surveyed quadrants.
In addition to the documentation procedures
noted above sites were also coded for particular
image types based on a hybrid classification
system of those previously developed for Cana-
dian Shield pictographic art (Dewdney & Kidd,
1967; Maurer & Whelan, 1977; Whelan, 1983)
and consisted of 10 major image classes and 57
sub-classes. This information was used in the
information calculation noted above. Point loca-
tions for both pictographs and sterile locations
were collected using a Garmin GPSMAP 60 and
then imported into ArcGIS 9.3. A Microsoft Ac-
cess personal database was created and populated
to manage survey data that was later tied to point
locations.
ANALYSIS
Our main object here is to determine the degree
to which rock art sites occur in discrete, non-
random spatial patterns on the landscape. Point
pattern analysis was selected first to determine
if these sites were arranged in an ordered man-
ner or if they were scattered randomly across the
landscape. The nearest neighbor analysis (Table
1) indicates clustering for both sites and sterile
formations at the 0.01 significance level. This
was not unexpected given that the distribution of
Table 1. Nearest neighbor analysis for both sites
and sterile formations
Sites Sterile Formations
N 51 610
Nearest Neighbor Ratio .67 .35
Z-score -4.38 -30.83
Critical Value -2.58 -2.58
1334
water bodies is relatively ubiquitous across the
survey area and the availability of surfaces with
both rock art and sterile locations on any given
body of water was generally only limited by the
size of the water body. In other words, every lake
surveyed had available and/or utilized surfaces.
As clustering was demonstrated for both location
types we decided to take another approach toward
understanding the patterning of both sites and
sterile formations.
Global and Local Moran’s I
The global Moran’s I statistic (Mitchell, 2005;
O’Sullivan & Unwin, 2003) was used because of
its ability to analyze point patterns and assess dis-
tributions across a region based on specific criteria.
In this case, the variable that we were concerned
with was the presence or absence of rock art. A
local Moran’s I (LISA) test (Anselin, 1995) was
subsequently run to identify any further cluster-
ing at a local scale (O’Sullivan & Unwin, 2003).
These calculations were both run to determine
the degree to which rock art sites (represented as
points in the GIS) were either clustered, dispersed,
or distributed randomly throughout the study
area (represented as polygons) (Conolly & Lake,
2006; O’Sullivan & Unwin, 2003; Wong & Lee,
2005). Both documented pictograph sites (n=51)
and sterile formations (n=610) were grouped and
included in the calculation, with the presence of
pictographs serving as the variable used to assess
spatial autocorrelation (Wheatley & Gillings,
2002). Our null hypothesis for these two tests was
that there was no significant statistical difference
in spatial autocorrelation between pictograph sites
and sterile formations.
The results of the global Moran’s I trial using
the values from the survey datasets resulted in a
Moran’s I index of 0.11 at the 0.05 significance
level with a z score of 2.01 and a critical value of
1.96. These scores indicate that rock art sites in
the survey area occur in a clustered pattern with
less than a 5% chance that the clustering is due
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
to random chance. Where the global Moran’s I is
able to characterize spatial autocorrelation across
the study area as a whole the LISA test is able to
pinpoint whether clustering is occurring at local
scales. The results indicated that 74.5% (n=38)
of the pictograph sites demonstrated clustering
at the 0.05 significance level, demonstrating a
high positive z score meaning that these sites
are clustered around other sites (Mitchell, 2005;
O’Sullivan & Unwin, 2003). There is a notice-
able fall-off in site placement as distance between
pictograph sites increase. Forty-nine percent
(n=25) of sites are located within 1 kilometer of
another site with a mean distance of 3.5 kilometers
between sites (Figure 4). One smaller subset of
pictograph sites (15.6%, n=8) were clustered in
an area where predominately sterile formations
exist. Low negative z scores associated with these
sites indicate that they are surrounded by locations
with dissimilar values (Mitchell, 2005; O’Sullivan
& Unwin, 2003).
The results of both spatial analyses are not
conclusive enough to support an interpretation
without further exploration at the global scale.
Taken as a static distribution of sites, there remains
the possibility that clustering is a result of random
Figure 4. Distance between nearest neighboring pictograph sites in meters
chance. To address this issue we thought it ap-
propriate to assess the likelihood that a global
Moran’s I calculation would return a similar result
if the site presence variable was ‘reshuffled’. For
this we chose to use a Monte Carlo approach to
test randomized datasets and to compare those
results to the original global Moran’s I analysis.
Monte Carlo Simulation
Monte Carlo simulation methods assess the likeli-
hood that a particular phenomenon does or does
not occur due to random chance (O’Sullivan &
Unwin, 2003; Steinberg & Steinberg, 2006). A
random number generator in Microsoft Excel 2007
was used to create 99 simulated datasets that were
then evaluated for spatial autocorrelation using
the global Moran’s I statistical analysis. Each
dataset contained the same number of sites (51)
and sterile formations (610), and the location for
all 661 points remained the same; however, the
designation for the presence or absence of a rock
art site at any particular location was randomly
varied in all 99 simulations.
The simulated patterning from these trials
ranged from random (n=89) to somewhat dispersed
1335
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region

(n=10). The mean value for the ninety-nine simu-
lated global Moran’s I datasets was 0.017 with a
standard deviation of 0.02. It is important to note
that none of the values for these simulated datasets
displayed a clustered distribution as in the case
of the actual survey dataset. The results of these
simulations and subsequent point pattern analysis
further support our assertion that pictographs were
deliberately placed at locations for reasons that
were not explainable by the earlier research of
others, leaving the question remaining of “why
these places?”
Site Choice
The original study done by Norder (2003) assessed
the relationships between several geomorpho-
logical variables including the following: panel
orientation, rock face angle, and outcrop surface
area. For these he found no statistically significant
relationship to pictograph sites suggesting they
were not variables considered in the choice of
pictograph creation. We further analyzed a subset
of the pictograph sites (n=26) and sterile locations
(n=139) compiled from the 2000 survey dataset
(Norder, 2003). One geomorphological factor
crucial to both image creation and visibility, the
angle of the rock face in relation to water surface,
was used for the trial. This variable is significant
for site selection because open-air pictograph sites
– versus sheltered, hidden, or cave sites – need to
be created in places where people will encounter
them if they are to facilitate communication. Rock
face angle (measured in degrees) for both sites and
sterile formations was compared for statistically
significant differences between the two groups
using an independent samples t test performed
in SPSS 16.0 (Sirkin, 1999). The results of the
t test (two- tailed significance = 0.192) indicate
that there is no significant difference at the 0.05
level (Norusis, 2008) between the pictograph sites
and sterile formations in terms of rock face angle
(Table 2), and this is consistent with Norder’s
(2003) analysis.
Least Cost Path Analysis
Norder (2003) observed that all documented
rock art sites are easily observable from passing
watercraft. This is especially important to con-
sider when understanding how hunter-gatherers
interact with and within this particular environ-
ment. Patterns of hunter-gatherer socio-spatial
organization differ across the world and the need
to move across the landscape varies depending on
physical geography, subsistence strategies, territo-
rial behaviors, and a host of other considerations
(Binford, 2001; Kelly, 1995; Wendrich & Barnard,

Table 2. Results of independent samples t test for rock face angle to water surface

Group Statistics
Site Pres- Std. Devia- Std. Error
ent N Mean tion Mean
No 139 88.5612 4.74894 .40280
Yes 26 90.0000 6.92820 1.35873
Assumptions=Equal variances assumed

Levene’s Test for Equality of

Variances t Test for Equality of Means
95% Confidence Interval of
the Difference
Sig. Mean Differ- Std. Error Dif-
F Sig. t df (2-tailed) ence ference Lower Upper
2.264 0.134 -1.309 163 0.192 -1.43885 1.09902 -3.60900 .73130

1336
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
2008). In the Lake of the Woods region travel by
water or alongside waterways was the primary
means of travel year-round for any sort of long
distance movement. Consequently, this had a clear
influence on how ancestral Algonquian peoples
viewed and interacted with the landscape in terms
of social practice and ideological belief (Norder,
2003). It is also no coincidence that water travel
is the easiest way to negotiate this rugged terrain.
A further critical step in developing our un-
derstanding of how pictograph sites function is to
analyze least cost paths and their relation to known
transportation routes (Lock, 2000; Wheatley &
Gillings, 2002). To this end, topographic and
hydrologic GIS data in both vector and raster for-
mats were obtained from the Geography Network
Canada (ESRI Canada, 2010) and values were
combined to form a cost weighted distance raster
(McCoy & Johnston, 2002) with a pictograph site
serving as the source for the cost calculations
(Figure 5). The cost raster was assigned a 10-meter
cell size to reflect the same contour intervals as
the vector data. It was not possible to create cost
weighted distance rasters for every pictograph
site due to the size of the study area and limita-
Figure 5. Location of pictographs relative to first least cost path trial
tions of computer resources to analyze the overall
dataset. Therefore, a subset of sites was selected
for testing the least cost paths (Figures 6 and 7).
The sites selected for the cost calculation were
selected from the area where the aforementioned
LISA test revealed a cluster of sites conspicuously
located in an area predominately surrounded by
sterile locations.
The Spatial Analyst tool in ArcGIS 9.3 was
used to calculate the least cost path for travel in
relation to the elevation of the surrounding terrain.
Figure 6 illustrates that if a traveler moves from
the source point in the east (designated 70105AD)
to the destination point (designated 62805Z) in
the west the least cost path for travel is by water.
Note the four pictograph sites located along the
shoreline adjacent to the least cost path. A second
least cost path was then calculated for a source
(designated T1) located 7.5 kilometers to the
northeast of the first source point (70105AD).
Figure 7 illustrates the location of the pictographs
relative to the least cost path for the second trial.
This statistical and geospatial analysis pre-
sented above suggests an answer to the question
of why pictographs appear in clustered patterns
1337
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region

Figure 6. Location of pictographs relative to second least cost path trial

Figure 7. Location of the pictographs relative to the least cost path for the second trial

1338
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
throughout the study area when they could be
produced in many locations. If the purpose of
these images is to convey information then they
have the requirement to be observable. Pictograph
sites are located where individuals could see them
from the water, the least cost mode of transporta-
tion through this rugged area. In the case of the
example presented in the figures above it is also
important to note that the least cost path matches
for much of its length a known travel route used by
the historic First Nations people as they traveled
through the region (C.S. Reid, personal commu-
nication, June 23, 2005), and it is also a modern
recreational canoe route (ESRI Canada, 2010).
ROCK ART AS MEDIUM TO
STRUCTURE MOBILITY
Our point pattern analysis indicates that pictograph
sites occur in spatially discrete patterns in the sur-
vey area. This finding supports previous research
interpretations by Norder (2003) that pictograph
site locations were deliberately selected. That is
to say, while the creators of rock art had many
options in terms of where they could place these
images, the results presented above indicate that
these locations were selected for reasons that were
not defined by just the availability of surfaces.
Some of these choices were defined by certain
aspects of surface geomorphology associated
with Algonquian religious and cosmological be-
liefs (Conway, 1993; Norder, 2003; Steinbring,
1992), but given the patterning demonstrated by
our analysis there is reason to suggest that other
social factors were also a consideration. In this
case, linear patterning and clustering of sites along
least cost paths, which, in some documented cases
follow known historic Indigenous travel routes,
indicates that Woodland period peoples may have
used these sites to structure population movements
at the regional scale, at least. In sum, the creation
and subsequent use of pictographic rock art in the
study area were the results of social processes
that required decisions regarding placement and
what type of information to communicate, which
could be either sacred, secular or, more likely, a
combination of the two.
In regards to the potential secular information
encoded in these pictograph sites Norder (2003,
in press-b) proposed that there were at least three
different classes of pictographs that communicated
different types of information, each ranging from
general to specialized in their content. Drawing
from both information theory and ethnohistorical
accounts, he suggested that these sites and their
active use facilitated the exchange of information
relating to factors including group identity, land
tenure, the locations of economically significant
resources, ritual practice and belief, and broader
Algonquian cosmology. Within a landscape of
abundant lacustrine and riverine features, the
primary mode of travel was via water either by
canoe from late spring to fall or by foot when they
were frozen over.
In this environmental context the placement
of pictographs throughout the region in specific
locations would have been, according to Norder
(2003), strategic in both facilitating population
movements and providing information to travel-
ers as they passed by specific site types. Given
the nature of Woodland period communities
inhabiting the region at large people were not
ignorant of territorial considerations and other
social obligations (Howey, 2007). Rather, the
people who created these images were hunting,
gathering and fishing specialists who needed to
plan and coordinate strategic movements across
the landscape in order to access the diverse and
seasonably abundant resource base. Examples
would include, but not be limited to, wild rice
fields, spring fish spawning areas (Reid, 1977),
and intercept locations for migratory caribou
herds (John O’Shea, personal communication,
January 6, 2010).
These strategic movements, known as residen-
tial and logistical population mobility, are prac-
ticed in diverse manners by small scale hunting and
1339
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
gathering communities across the globe (Barnard
& Wendrich, 2008; Binford, 1980; Kelly, 1995).
For our purposes we refer to population mobility
in the sense defined by Wendrich and Barnard
(2008) where the term refers to the requirement,
capability, and choice of humans to access cul-
turally defined resources across the landscape.
The social context of First Nations communities
was one where mobile Algonquian communities
practiced alternating patterns of aggregation and
dispersal, typically at the band and family level,
at different points during the year for social,
economic, and ideological purposes (Norder,
2003; Reid, 1977). In this sense, social interac-
tion between hunter-gather communities served
to buffer the uncertainty of resources and share
risk with other groups, ensuring that families and
communities survived during periods of resource
scarcity (Halstead & O’Shea, 1982).
In this setting, pictographic rock art in their
easily visible locations would have provided in-
formation not only in the absence of face-to-face
communication, the primary means of obtaining
the above information needed for scheduling and
exploitation of regional resources, but would have
indicated through their content, placement, and
active use where people could likely be found.
Active use could have been through the renewal
or addition of images at a pictograph site, but
the most expedient method of demonstrating use
would have been to leave offerings of some type
that would have been easily visible to those pass-
ing by. This latter practice was docmented in both
the ethnohistoric and ethnographic literature in the
region (Norder, 2003; Rajnovich, 1994) and is a
practice still done today by Algonquian communi-
ties living in the region (Norder, 2007, in press-a).
Overall, we would argue that pictograph sites
served as points on the landscape where people
could access information that was spatial, tempo-
ral, social, and spiritual in nature, which provided
guidance and minimized the potential for social
miscommunication in regards to how resources
1340
were both available and could be accessed in a
landscape where others had prior claim.
CONCLUSION
The purpose of this investigation has been to
develop a broader understanding for both the
placement and function of pictograph sites in the
Lake of the Woods region of Ontario. Point pattern
analysis, Monte Carlo simulation and least cost
path analysis indicate that pictograph sites were
placed throughout the study area in linear and non-
linear clusters that were located along likely travel
routes and locations were people were likely to
aggregate. These pictograph distributions afforded
ancestral Algonquian peoples the opportunity to
maximize information exchange while minimiz-
ing the cost of transportation by having clearly
marked routes of travel. This is an important
advantage for highly mobile, spatially dispersed
hunter-gatherers who needed to access resources
and maintain social ties across a relatively large
landscape, such as that of the study area.
The application of these results to contem-
porary research has important implications for
considering how rock art is perceived in the main-
stream literature. Less than a handful of studies
of rock art have applied geospatial methodologies
to any significant degree (Bradley, 1997; Hartley
&Vawser, 1998; McCall, 2010; Norder, 2003; Rob-
inson, 2010; Smith, 2001) despite the recognition
of landscape as a significant aspect of the overall
social context of these sites (Chippendale & Nash,
2004; David, 2001; Nash & Chippendale, 2002).
The work here contributes to an expanded view
of rock art as part of a more holistic conception
of cultural landscapes in mobile hunter-gatherer
societies, and recognizes the significance of these
sites in secular as well as sacred contexts.
As a final note, we want to reiterate the im-
portance of these sites not only to research but
as heritage resources important to contemporary
community and national histories and identities.
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
Within northwestern Ontario, because of human-
induced environmental change, these sites are
vulnerable as are rock art sites worldwide. For
example, since one of the one of the authors be-
gan visiting these sites in 1998 in northwestern
Ontario, the past 12 years has seen the degradation
of several of them at a pace that belies their great
antiquity. In other areas of northwestern Ontario,
humans have defaced these sites with graffiti or
removed portions of sites for private collections
or sale. These sites are valuable cultural resources
that need to be preserved for future generations.
Without them, we will be missing important in-
formation that helps us to understand how people
have interacted across space and time and the rich
heritage of First Nations communities.
REFERENCES
Adams, N. (1983). An archaeological survey of
the Nakina Lakes area near Geralton, Ontario.
Toronto, ON: Ontario Ministry of Citizenship
and Culture.
Anselin, L. (1995). Local indicators of spatial
association-LISA. Geographical Analysis, 27(2),
93–115. doi:10.1111/j.1538-4632.1995.tb00338.x
Arsenault, D. (2004). From natural settings to
spiritual places in the Algonkian sacred land-
scape: An archaeological, ethnohistorical and
ethnographic analysis of Canadian Shield rock-
art sites. In Chippendale, C., & Nash, G. (Eds.),
The figured landscapes of rock-art: Looking at
pictures in place (pp. 289–317). New York, NY:
Cambridge University Press.
Arsenault, D., Gagnon, G., Martijn, C., & Watch-
man, A. (1995). Le projet Nisula: recherché plu-
ridisciplinaire autour d’un site á pictogrammes
(DeEh-1) en Haute-Côte-Nord. Archaéologies
québécoises, série Paléo-Québec, 23, 17-57.
Balac, A. M., Chapdelaine, C., Clermont, N., &
Duguay, F. (Eds.). (1986). Archeologies Que-
becoises (pp. 17–57). Montreal, QC, Canada:
Recherches Amerindiennes au Quebec.
Barnard, H., & Wendrich, W. (Eds.). (2008). The
archaeology of mobility: Old World and New
World nomadism. Los Angeles, CA. Los Angeles:
University of California.
Binford, L. (1980). Willow smoke and dogs’ tails:
Hunter-gatherer settlement systems and archaeo-
logical site formation. American Antiquity, 45(1),
4–20. doi:10.2307/279653
Binford, L. (2001). Constructing frames of refer-
ence: An analytical method for archaeological
theory building using hunter-gatherer and envi-
ronmental data sets. Berkeley, CA: University of
California Press.
Bradley, R. (1997). Rock art and the prehistory
of Atlantic Europe: Signing the land. New York,
NY: Routledge Press.
Chippendale, C., & Nash, G. (Eds.). (2004).
The figured landscapes of Rock-Art: looking at
pictures in place. New York, NY: Cambridge
University Press.
Chippendale, C., & Taçon, P. (Eds.). (1998). The
archaeology of rock-art. New York, NY: Cam-
bridge University Press.
Clotte, J., & Lewis-Williams, D. (1998). Shamans
of prehistory. New York, NY: Harry N. Abrams
Publishing.
Conolly, J., & Lake, M. (2006). Geographical
information systems in archaeology. Cambridge,
UK: Cambridge University Press.
Conway, T. (1989). Scotia Lake pictograph site:
shamanic rock art in northeastern Ontario. Man
in the Northeast, 37, 1–23.
Conway, T. (1993). Painted dreams: Native Ameri-
can rock art. Minocqua, WI: Northword Press.
1341
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
Conway, T., & Conway, J. (1990). Spirits on stone:
The Agawa pictographs. Echo Bay, ON, Canada:
Heritage Discovery Books.
Coulson, D., & Campbell, A. (2001). African rock
art: paintings and engravings on stone. New York,
NY: Harry N. Abrams.
David, B. (2001). Landscapes, rock-art and the
dreaming: An archaeology of preunderstanding.
New York, NY: Continuum.
Dewdney, S. (1975). The sacred scrolls of the
southern Ojibway. Toronto, ON, Canada: Uni-
versity of Toronto Press.
Dewdney, S., & Kidd, K. (1967). Indian rock
paintings of the Great Lakes (2nd ed.). Toronto,
ON, Canada: University of Toronto Press.
Diaz-Granados, C., & Duncan, J. (Eds.). (2004).
The rock-art of Eastern North America: Capturing
images and insight. Tuscaloosa, AL: University
of Alabama Press.
Domingo-Sanz, I., Fiore, D., & May, S. (Eds.).
(2008). Archaeologies of art: Time, place, identity.
Walnut Creek, CA: Left Coast Press.
ESRI Canada. (2010). Geography Network
Canada. Retrieved from http://www.geogra-
phynetwork.ca
Halstead, P., & O’Shea, J. (1982). A friend in need
is a friend indeed: Social storage and the origins
of social ranking. In Renfrew, C., & Shennan, S.
(Eds.), Ranking, resource, and exchange: Aspects
of the archaeology of early European society (pp.
92–100). Cambridge, UK: Cambridge University
Press.
Halverson, C. (1988). An archaeological survey
of Stephen, Cameron, Flint and Kakagi Lakes
in Northwestern Ontario. Kenora, ON, Canada:
Ontario Ministry of Culture and Communications.
1342
Hartley, R. (1992). Rock art on the northern Colo-
rado Plateau: Variability in content and context.
Brookfield, VT: Avebury Press.
Hartley, R., & Vawser, A. (1998). Spatial behavior
and learning in the prehistoric environment of the
Colorado River drainage (southeastern Utah),
western North America. In Chippendale, C., &
Taçon, P. S. C. (Eds.), The archaeology of rock-
art (pp. 185–211). New York, NY: Cambridge
University Press.
Hayes-Gilpin, K. (2003). Ambiguous Images:
gender and rock art. Walnut Creek, CA: Alta
Mira Press.
Helskog, K. (Ed.). (2001). Theoretical perspec-
tives in rock art research. Oslo, Norway: The
Institute for Comparative Research in Human
Culture.
Howey, M. (2007). Using multi-criteria cost
surface analysis to explore past regional land-
scapes: A case study of ritual activity and social
interaction in Michigan, AD 1200-1600. Journal
of Archaeological Science, 34(11), 1830–1846.
doi:10.1016/j.jas.2007.01.002
Keiser, J., & Klassen, M. (2001). Plains Indian
rock art. Seattle, WA: University of Washington
Press.
Kelly, R. (1995). The foraging spectrum: Diversity
in hunter-gatherer lifeways. Washington, DC:
Smithsonian Institution Press.
Lambert, P. (1985). The Northwestern Ontario
rock art project: The 1984 results. Kenora, ON,
Canada: Ontario Ministry of Citizenship and
Culture.
Layton, R. (1992). Australian rock art: a new
synthesis. New York, NY: Cambridge University
Press.
Lock, G. (2000). Beyond the map: Archaeology
and spatial technologies. Oxford, UK: IOS Press.
Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region
Loendorf, L., Chippendale, C., & Whitley, D.
(Eds.). (2005). Discovering North American Rock
Art. Tucson, AZ: University of Arizona Press.
Maurer, J., & Whelan, J. (1977). A Canadian
Shield pictograph classification design. American
Antiquity, 42(2), 196–202. doi:10.2307/278980
McCall, G. (2010). Changing views of Drak-
ensberg San rock art: Examining landscape use,
ritual activity, and contact through multivariate
content-based spatial analysis. American Antiq-
uity, 75(4), 773–791.
McCoy, J., & Johnston, K. (2002). Using ArcGIS
spatial analyst. Redlands, CA: ESRI Press.
Mitchell, A. (2005). The ESRI guide to GIS analy-
sis: Vol. 2. Spatial measurements and statistics.
Redlands, CA: ESRI Press.
Molyneaux, B. (1980). Landscape images: rock
paintings in the Canadian Shield. Rotunda: The
Magazine of the Royal Ontario Museum, 13, 6–11.
Nash, G., & Chippendale, C. (Eds.). (2002). Eu-
ropean landscapes of rock art. New York, NY:
Routledge Press.
Nieves, A. (2006). Reconstructing ritual: Some
thoughts on the location of petroglyph groups
in the Nasca Valley, Peru. In Robertson, E. C.,
Seibert, J. D., Fernandez, D. C., & Zender, M. U.
(Eds.), Space and spatial analysis in archaeology
(pp. 217–226). Calgary, AB, Canada: University
of Calgary Press.
Norder, J. (2003). Marking place and creating
space in northern Algonquian landscapes: The
rock-art of the Lake of the Woods region, Ontario.
Unpublished doctoral dissertation, University of
Michigan, Ann Arbor, MI.
Norder, J. (2007). Iktomi in the land of the May-
maygwayshi: Understanding lived experience in
the practice of archaeology among American Indi-
ans/First Nations. Archaeologies, 3(3), 230–248.
doi:10.1007/s11759-007-9044-6
Norder, J. (in press). -a). The invention and endur-
ance of memory and place among First Nations
of northwestern Ontario, Canada. International
Journal of Historical Archaeology.
Norder, J. (in press-b). Landscapes of memory
and presence in the Canadian Shield. In L. Sund-
strom & W. DeBoer (Eds.), Enduring motives:
The archaeology of tradition and religion in
Native America. Tuscaloosa, AL: University of
Alabama Press.
Norusis, M. (2008). SPSS 16.0 statistical pro-
cedures companion. Upper Saddle River, NJ:
Prentice Hall.
O’Sullivan, D., & Unwin, D. (2003). Geographic
information analysis. Hoboken, NJ: Wiley.
Pastershank, G. (1989). An archaeological survey
of Sabaskong Bay, Lake of the Woods. Kenora,
ON, Canada: Ontario Ministry of Culture and
Communications.
Rajnovich, G. (1980a). Paired morphs at Cuttle
Lake in northwestern Ontario. Canadian Rock
Art Research Association Newsletter, 13, 46–50.
Rajnovich, G. (1980b). Two pictographs near
Ignace, northwestern Ontario. Canadian Rock
Art Research Association Newsletter, 14, 34–38.
Rajnovich, G. (1989). Visions in the quest for
medicine: an interpretation of the Indian picto-
graphs of the Canadian Shield. Midcontinental
Journal of Archaeology, 14(2), 179–225.
Rajnovich, G. (1994). Reading Rock Art: The
Indian Rock Paintings of the Canadian Shield.
Toronto, ON, Canada: Natural Heritage/Natural
History Press.
Reid, C. (1977). Environment, man, and Maymay-
gwayshi: The dimensions of past human cultural
impact in the Lake of the Woods area. Toronto,
ON, Canada: Ministry of Culture and Recreation.
1343
 Applied Geospatial Perspectives on the Rock Art of the Lake of the Woods Region

Robinson, D. (2010). Land use, land ideology:
an integrated geographic information systems
analysis of rock art within south-central California.
American Antiquity, 75(4), 792–818.
Rusak, J. (1989). An archaeological survey on
a portion of the Winnipeg River, Northwestern
Ontario. Kenora, ON, Canada: Ontario Ministry
of Culture and Communication.
Rusak, J. (1992). The White Otter Lake pictograph
project: The 1991 results. Kenora, ON, Canada:
Ontario Rock Art Conservation Association.
Schaafsma, P. (1980). Indian rock art of the
Southwest. Santa Fe, NM: School of American
Research.
Shannon, C. (1949). The mathematical theory
of communication. Urbana, IL: University of
Illinois Press.
Sirkin, R. (1999). Statistics for the social sciences.
Thousand Oaks, CA: Sage.
Smith, H. (2001). Rock art and the shape of land-
scape. In Szabo, J. (Ed.), Painters, patrons and
identity: Essays in Native American art to honor
J. J. Brody (pp. 211–240). Albuquerque, NM:
University of New Mexico Press.
Steinberg, S., & Steinberg, S. (2006). GIS: Geo-
graphic information systems for the social sci-
ences: Investigating space and place. Thousand
Oaks, CA: Sage.
Steinbring, J. (1992). Phenomenal attributes: sites
selection factors in rock art. American Indian
Rock Art, 17, 102–108.
Steinbring, J., & Simpson, D. (1983). The Mud
Portage site (DkKr-4): a preceramic context for
petroglyphs in Northwestern Ontario. American
Indian Rock Art, 10, 145–156.
Vastokas, J., & Vastokas, R. (1973). Sacred art
of the Algonkians: A study of the Peterborough
petroglyphs. Peterborough, ON: Mansard Press.
Wendrich, W., & Barnard, H. (2008). The ar-
chaeology of mobility: Definitions and research
approaches. In The archaeology of mobility: Old
World and New World nomadism (pp. 1–21).
Los Angeles, CA: University of California, Los
Angeles.
Wheatley, D., & Gillings, M. (2002). Spatial
technology and archaeology: The archaeological
applications of GIS. New York, NY: Taylor &
Francis. doi:10.4324/9780203302392
Whelan, J. (1983). Context and association in
Canadian Shield rock art systematic. Canadian
Journal of Archaeology, 7(1), 77–84.
Wong, D., & Lee, J. (2005). Statistical analysis
of geographic information with ArcView GIS and
ArcGIS. Hoboken, NJ: John Wiley & Sons, Inc.

This work was previously published in the International Journal of Applied Geospatial Research (IJAGR), Volume 2, Issue 4,
edited by Donald Patrick Albert, pp. 77-92, copyright 2011 by IGI Publishing (an imprint of IGI Global).

1344
 1345

Chapter 81
Cartography and Stability to
Enhance and Self Regulate
Proactive Routing in MANETs
Mohamed Amine Abid
University of Manouba, Tunisia

Abdelfettah Belghith
University of Manouba, Tunisia

ABSTRACT
In this chapter, the authors propose a novel proactive routing protocol where routing advertisement
frequency is self regulated locally at and by each node depending on its mobility level. The proposal
relays on an asynchronous distributed cartography gathering algorithm and a link stability criterion.
The cartography of the network is kept constantly valid in time and for all applied speeds. Each node
senses its own dynamics and locally chooses an appropriate routing advertisement period size. As such
stationary nodes generate little signaling traffic while fast moving nodes prefer small routing periods
to mitigate the effect of their mobility.

INTRODUCTION percent routing validity while that of OLSR can

The performances of our proposed routing pro-
tocol are evaluated and compared to the known
Optimized Link State Routing (OLSR) protocol
through extensive simulations. We first show that
the collected network cartography maintains over
time a validity ratio near one hundred percent
even for high node speeds. We then show that
our proposed routing protocol provides around 97
DOI: 10.4018/978-1-4666-2038-4.ch081
hardly deliver more than 60 percent at moderate to
high speeds and workloads. Finally, we show that
our proposed protocol provides better throughput
than OLSR reaching a 50% increase at moderate
to high speeds and workloads and far much less
end to end delays.
Routing in multi hop mobile ad hoc networks
(MANETs) plays a central and vital role, yet it
stands out as one of the most and fundamental
issues to attain adequate performances and viable
deployments. The crux of routing in such networks

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
lies in its ability to maintain and provide valid
routes without which traffic may wander inside
the network without being able to be delivered
to its ultimate destinations. Worse yet, on such
invalid routes, the underlying Media Access Con-
trol (MAC) protocol retransmits unacknowledged
unicast frames for several (by default seven)
times which consequently amounts to over using
the wireless communication medium and hence
yields poor performances. We may stipulate here
that the basic reason of the non proliferation of
such spontaneous networks, despite their practical
usefulness and need, is mainly due to their poor
performances caused by the lack of viable low
signaling and high validity routing algorithms.
In mobile ad hoc networks, nodes and links
can appear and disappear spontaneously as a
consequence of several facts such as the behavior
of users, the depletion of energy resources, but
more inherently and subtly the random mobility
of the different nodes. These aspects imply a
dynamically and randomly evolving topology in
both time and space making the routing function a
real difficult and challenging task. As a result, and
as the elapsed time since the start of the routing
period gets farther, the topological information
collected at the beginning of the routing period
becomes inaccurate and obsolete leading patently
to invalid paths.
Conventional protocols, whether reactive
- such as the Dynamic Source Routing (DSR)
protocol (Johnson, Maltz and Yih-Chun Hu,
2004) and the Ad-hoc On-demand Distance Vec-
tor (AODV) routing protocol (Perkins, Royer
and Chakeres, 2003) - or proactive - like OLSR
(Clausen and Jacquet, 2003) and the Destination
Sequenced Distance Vector (DSDV) routing
protocol (Perkins and Bhagwat, 1999) -, use in
general the hop-count metric to compute shortest
paths towards destinations. However, shortest
paths are not always reliable especially in case of
dynamic networks. Finding stable routes is rather
the main concern for dynamic multi-hop ad hoc
networks. Several works already established that
1346
choosing routes based on positions, battery level,
etc. of the nodes would make selected paths more
resilient to topological changes.
Numerous simulation studies were conducted
on different scenarios to evaluate the performance
of conventional protocols: for both proactive and
reactive routing protocols as those conducted by
Clausen, Jacquet and Viennot (2002), Novatnack,
Greenwald and Arora (2005), or Mbarushimana
Amine and Shahrabi (2002). Nevertheless, due to
the huge number of relevant and complex events
that can take place in mobile wireless ad hoc
networks and their impacts on the performance of
the protocols, results do not necessarily agree as
to which protocol yields satisfying performances
and lower control traffic and overhead. It is now
understood that proactive protocols in particular,
such as OLSR which is the main representative
of this family, cannot withstand high mobility in
an appropriate manner since the validity of the
routes decreases rapidly as we get farther from
the start of the routing period (Belghith and Abid,
2009-1, 2009-2).
In this paper, we propose a novel cartography-
based protocol that builds the network cartography
in an asynchronous distributed fashion, i.e., no
synchronization is required among the nodes in
the network. This network cartography provides
better vision of the network connectivity and al-
lows selecting viable links in a way that improves
the routing pertinence as a function of the network
dynamics. Yet, the signaling overhead required to
build the network cartography and to establish
the routing tables consumes less bandwidth than
that of the well known and adopted OLSR. Our
proposed routing protocol allows different nodes
to have different routing period sizes. The size
of the routing period adopted by a mobile node
is decided locally and dynamically. Conducted
simulations brought out the betterments achieved
by our routing proposal in terms of a much better
throughput, much less end to end delay and less
signaling traffic rate than those provided using
OLSR.
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
The paper is organized as follows:
1. In section II, we present some of the relevant
research work done in the field.
2. Section III presents a mechanism to col-
lect the cartography of the network in a
distributed way, and that can be integrated
to common proactive routing protocols.
3. In section IV, we present our routing protocol
and describe its functioning.
4. Section V is devoted to the conducted simu-
lations. We compare the results given by
our proposed protocol to that of the OLSR
(Clausen and Jacquet, 2003) routing proto-
col and we show the remarkable efficiency
achieved by our proposal.
5. Finally, in section VI, we present concluding
remarks and some future directions.
RELATED WORK
To withstand the network dynamics, stability-
based routing is an approach aiming essentially at
choosing routes which are more stable over time
and hence more resilient to dynamic changes in
the network topology.
In 2004, Chung proposed to classify links
based on the mobility behavior of their end point
nodes. Links between stationary or very slowly
moving nodes are considered as stationary links.
Links which exist only for a short period of time
are handled as transient links. Newly formed links
are also considered to be transient as they are more
likely to break down. Routing should favor the
use of stationary links whenever this is possible.
A ticket-based probing procedure is proposed to
find stable routes.
Classification of links is previously adopted
in Dube, Rais, Wang and Tripathi (1997) work.
Authors used the strength of the received signal
from each neighboring node to determine whether
the associated link is either weak or strong. Rout-
ing is then made through paths maximizing the
received signal strength.
Toh (1997) proposed the Associativity-Based
Routing (ABR) which uses a new metric called
associativity which defines the stability of a link
between any given two nodes. ABR considers that
the longer the two nodes have being neighbors,
the longer they would keep connected. To express
its associativity, each node broadcasts periodi-
cally a Beacon to indicate its presence. Upon the
reception of a Beacon, a counter associated to the
generating node is updated. The counter is reset to
zero if the associated node is no longer accessible.
The optimal route towards a destination is the one
maximizing its cumulative associativity metric.
Further stability-based routing techniques
can be found in (Beraldi, Querzoni and Baldoni,
2006), (Dubois-Ferriere, Grossglauser and Vet-
terli, 2003), (Roth and Wicker, 2003), (Yu, Li
and Gruber, 2003), (Camp, Boleng and Davies,
2002) and (Hui Zhang and Yu-Ning Dong, 2007).
In 2009, Belghith, Ben Mnaouer and Abid
presented the Probabilistic Source (or flow ori-
ented) Routing (PSR) which is a routing protocol
using localization information gathered at the start
of the routing period, along with a probabilistic
technique to compute the most stable paths. This
protocol focuses rather on the existence prob-
ability of complete routes. The stability of routes
may change as time progresses, and used paths
selection is made accordingly. The PSR protocol
requires, however, synchronization among all
nodes in the network.
Along the same line, other probabilistic tech-
niques were proposed to cope with the unpredict-
able behavior of ad hoc topologies. In (Tseng,
Li and Chang, 2003), a probabilistic technique
is proposed to estimate the residual lifetime of
routes. Routing is then made through the ones
with maximum residual lifetime. The lifetime of
a route is computed as a function of the existence
probability of each link which is derived from the
distribution governed by the underlying mobility
model.
1347
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
In contrast to the above protocols, our current
routing proposal does not require any synchroniza-
tion between nodes and defines a proactive routing
protocol based on the collected cartography of the
network. It allows each node to compute its own
routing period and routing period size in a local
and autonomic way as a function of the locally
perceived network dynamics.
NETWORK CARTOGRAPHY
Network cartography is the set of coordinates of
the different nodes within the network area and
as such it is much richer than the mere network
topology that just gives the connectivity within the
network. Our aim here is to design a distributed
algorithm that collects the cartography of the
network in such a way that it can be integrated in
any proactive routing protocol with no additional
signaling traffic apart from a very few additional
fields in routing announcement messages (i.e.,
Hello messages).
First of all, we intend to get a correct and valid
cartography and therefore we will not tolerate
delayed routing information. As such, we pro-
pose to distinguish between control (i.e., hello)
messages and data packets. Hello messages are
to be transmitted as soon as possible before any
other awaiting data packets. As such, received
or locally generated Hellos are put at the head
of the IP sending queue in front of any awaiting
data packet.
Secondly, we assume that each node is capable
of knowing its own geographical location. Such
an assumption can be justified by the recent avail-
ability of small and inexpensive low power GPS
receiver devices (Giordano and Stojmenovic,
2003). In case where these devices are not avail-
able or are not applicable (in-door networks case
for instance), many other approaches for inducing
relative coordinates still could be used. We can find
a large set of GPS-less or restrained techniques for
localization (Bulusu, Heidemann and Estrin, 2000;
1348
Capkun, Hamdi and Hubaux, 2001; Niculescu
and Nath, 2003; Nagpal, Shrobe, and Bachrach,
2003; Saad, Benslimane and Jean-Claude, 2008;
Boukerche, de Oliveira, Nakamura and Loureiro,
2007; and Takenaka, Mineno, Tokunaga, Miyauchi
and Mizuno, 2007).
In this current work, we assume that nodes are
location capable, relaying on GPS or one of the
already cited localization techniques. A forwarded
hello contains then both of its originating node’s
position and its forwarding node’s position. The
network cartography is built by every node through
the dissemination of the geographic information
integrated in the routing announcements (i.e.; the
Hello messages). What we need, for our current
purpose, is a collected cartography that reflects
adequately and accurately the actual real network
cartography when the routing table calculation
is needed. To evaluate the correctness of the
collected cartography, we compare it against the
real current cartography of the network at the
instant of the observation. Note that the actual
instantaneous cartography of the network can be
easily extracted from the simulator but it cannot
be known in practice. This is of no concern to
us here since all what we are looking for is to be
sure that the collected cartography does represent
adequately the real actual cartography. For now,
we will just define the validity of the Cartography;
its evaluation will be differed to when we propose
our routing protocol.
Consider a target mobile node “N”. When N
firstly advertised itself (i.e., sent its own generated
Hello), it was at position (x0; y0). In the Collected
Cartography of a node “A” that had already heard
N’s Hello, a new entry for N was created, show-
ing (x0; y0) as N’s coordinates. Since node N is
mobile, its position varies as a function of time,
and consequently it will be at position (xt; yt) at
instant “t” as time progresses. We say that N’s
position, as indicated by A’s current cartography, is
valid as long as the distance between the recorded
position (x0; y0) and the actual current position
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
(xt; yt) is less than a tolerated predefined value
denoted by “d”. That is:
(x −x0 ) + ( yt −y0 ) ≤ d
2 2
t at instant “NextUpdate”.
The validity of the cartography, as perceived by any
given node, say node A, represents the percentage
of nodes having valid positions among all nodes
in the network. Parameter d is a tuning parameter
whose value is relative to the transmission range
used, and is in general a very small fraction of
this range.
The Proposed Routing Protocol
The proposed routing protocol is an asynchronous
self adjustable routing protocol since it is based
on an asynchronous distributed cartography
algorithm and each node regulates locally and
dynamically its own routing period size according
to its own dynamics.
The Protocol Description
Given a tolerance value d, our aim is to collect a
valid cartography. That means, if node N’s speed,
or more exactly its remoteness from the position
where it was at its last announcement, is such that
this tolerance value d is exceeded (Position of
Node N may be no longer valid in all other nodes’
collected cartographies), N has to re-announce
itself to enforce the update of its position in the
collected cartographies at the other nodes. So,
each node has to compute the instant, at which
it has to re-announce itself based on its currently
observed dynamics, so that it will not exceed the
tolerance d and keeps the validity of collected
cartographies at a high acceptable level at all times.
The period size is then locally computed by each
node according to its own perceived dynamics. Let
“PSize” be the period size computed by a given
target node. Depending on its currently observed
dynamics, PSize may vary from the smallest de-
fined updating period denoted by “Tmin” to the
largest updating period denoted by “Tmax”. The
target node next announcement is then scheduled
Furthermore, the proposed protocol makes
each node track its remoteness, from the position
where it was at the instant of its last announcement,
every Tmin seconds (i.e., an observation instant
every Tmin seconds). This is needed by each node
to correctly and appropriately follow its dynamic
evolution in time. Let “Last_Update_t” be the
instant at which the target node has made its last
announcement. Let “∆d” be the remoteness of
the target node at instant (Last_Update_t + Tmin).
The target node instantaneous remoteness speed
is then Sd = ∆d/Tmin.
At this Speed, the cartography tolerance value
d will be exceeded after (d / Sd) seconds. The target
node routing period size is then updated as follows:
• PSize = Tmax : if Sd = 0 m/s
• PSize = min(Tmax; max(Tmin; d / Sd)) :
otherwise
This procedure is then repeated each Tmin
seconds. The target node will consequently re-
announce itself at:
NextUpdate = min(Last_Update_t + PSize ;
NextUpdate)
As previously pointed out, the minimum period
size that can be used is Tmin. Each node periodi-
cally tracks its remoteness each Tmin seconds, and
if the tolerance limit is exceeded, an immediate
announcement is performed.
As a result, each node will adapt its rout-
ing period according to its own sensed velocity
level. As such, its position is updated in all other
nodes’ collected cartographies only when needed.
Each node includes its own perceived position
in its locally generated Hello message, and then
broadcasts it to its neighbors. Upon the reception
1349
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
of this Hello, a node updates its perceived car-
tography, inserts its current position in the Hello
and forwards the message. As such, a forwarded
hello contains both its originating node’s posi-
tion as well as its forwarding node’s one to avoid
inconsistency problems already pointed out in
(Abid and Belghith, 2009-3).
The proposed cartography based routing pro-
tocol works as follow:
1. It is through Hello messages that a node
can maintain its Cartography up to date.
The Cartography includes entries to already
heard destinations. Each entry includes a
sequence Number (i.e., the period’s number),
an entry local start instant (recall that an
entry is deleted after a maximum duration
period (“Max_duration_period”), the cur-
rent perceived position of the destination
node and the destination node address.
2. Each node generates locally a Hello mes-
sage every computed period (at instant
NextUpdate). The Hello message contains
the following fields:
◦◦ Source address: the address of the
node that originated this Hello.
◦◦ Source sequence number: the se-
quence number that the originator of
the Hello message is currently using.
◦◦ Next hop Address: the Next hop to
reach the Source Address, this is just
the address of the node that sent this
Hello.
◦◦ Next hop sequence number: the se-
quence number that the node for-
warding the Hello message is cur-
rently using.
◦◦ Source position: the source node
coordinates at the Hello generation
instant.
◦◦ Next hop position: the sender node
current coordinates perceived at the
very last self announcement it made.
1350
3. Upon receiving a Hello message, a node
first consults the sender’s relative fields. If
no entry exists in the cartography pointing
out the position of the next hop address field
then an entry is added. Otherwise if the entry
is relative to a previous period (old sequence
Number compared to the ”Next hop sequence
number” field), then the entry is updated ac-
cordingly. A Hello is automatically created
to advertise this new entry to other nodes in
the network. Otherwise, the information is
silently ignored.
4. Then, the node checks whether it is the
originator of this Hello, in which case it
just discards the message. Otherwise, the
node tests the usefulness of the received
hello: if its Cartography already contains
an entry for the ”Source Address” (which
represents the destination to be reached),
the current sequence number of this entry
and the sequence number included in the
Hello are compared. The entry is updated
if it is older. A new entry is automatically
created in the Cartography, if the message
announces a destination not yet known.
5. If the received message is useful, the node
forwards it to its neighbors. It puts its ad-
dress in the ”Next hop Address” field, puts
its position in the corresponding field, puts
its currently used sequence Number in the
”Next hop sequence number”, and then
broadcasts it to its neighbors after a certain
delay as explained below.
6. We attempt to make the new information
spread in the network as fast as possible, to
ensure that all nodes in the network are using
the same collected cartography along time.
For that, and before forwarding the new heard
useful Hello, a node computes the distance
“D”, separating him from the ”Next hop
Address” node (the node bringing the new
information). The bigger this distance is (0
≤ D ≤ Transmission Range = R), the fastest
the forwarding must be accomplished: we
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
proposed that each node forwards the new
information after a delay time:
1 − min ( R , D ) / R
delayTime =  seconds conducted an extensive set of simulations. We
2
New information forwarding is distributed over
one half a second, accordingly to the distance
between the sender and the forwarder, to lower
the chance of collisions.
Once the cartography is collected, a node can
compute its routing table using any shortest path
algorithm executed on the corresponding cartog-
raphy. Note here that the cartography collection is
made in an asynchronous way since no synchro-
nization is required between nodes. Recall that in
ad hoc networks, a wireless link exists whenever
its two end point nodes are within the transmis-
sion range R. In the proposed routing protocol,
however, a node does not re-announce itself unless
the tolerance value d is exceeded. Furthermore, if
we consider the worst case where two neighbor
nodes can only move away linearly, then we must
reduce the neighborhood range by twice the value
of this tolerance. This will make our routes more
stable over time since a node can’t move farther
more than 2×d meters. In this work, we suppose
that this extreme case has low chance to happen,
and as such, we limit ourselves to an acceptable
threshold as it will be shown later, by applying
the chosen shortest path algorithm to our collected
cartography using an effective range between
nodes limited to R - d. Next, we investigate the
validity of the collected cartography which must
be kept at its highest levels at all applied speeds,
all loads and all times. The load effect, however,
can be overcome thanks to the adopted prioritiza-
tion IP model.
Simulation Set Up
To ascertain the validity of the cartography as a
function of the mobility, traffic load and time, we
have considered a simulation area of 1000m by
1000m, with 100 mobile nodes using a modified
Random Waypoint mobility model (Camp, Boleng
and Davies, 2002). The simulation surface was
divided into 9 areas, each of which contains from
11 to 12 nodes. The Random-Waypoint mobility
is then applied separately inside each area (Figure
1). This is basically to mitigate the known middle
concentration effect of the Random WayPoint
mobility model.
The Effectiveness of the Restricted
Random Waypoint Mobility Model
As proven in several previous works (Bettstetter,
Lassila, and Virtamo, 2006 ; Hyytia, Lassila and
Virtamo, 2006 ; and Yoon, Liu and Noble, 2003),
the Random Waypoint Mobility model suffers
from the problem of concentration of the mobile
nodes towards the middle of the simulation area.
This behavior would result in a reduced average
number of hops and a smaller signaling than in
actual networks.
Figure 1. Restricted random waypoint: Applied
random waypoint inside each area
1351
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
In the quest to mitigate such an undesirable
effect, we propose dividing the simulation area
into smaller zones, where the RWP mobility
model is applied independently. The concentration
problem will always be there, but instead inside
every sub-area. As such, we will guarantee the
presence of nodes at the edges of our area as time
progresses, and therefore the average number of
hops in the network will be kept enough high
during the whole simulation.
Before going any further, we need to better see
whether the problem is solved, or at least reduced,
using our proposed restricted random waypoint
mobility model. It is obvious that this concentra-
tion comes out deeper proportionally to the speed
of the nodes. Starting the simulation scenarios with
the initial distribution of nodes given by Figure
2, Figure 3 highlight the success of our model in
avoiding the concentration problem by showing
the nodes’ geographical distribution after 1500s
of simulation time.
As expected, we observe as the simulation
time runs that the nodes using our restricted mo-
bility model occupy the entire simulation area,
though some concentration within each sub area.
Figure 2. Initial geographical distribution of mobile nodes
1352
To better ascertain the odd behavior of the
random waypoint mobility model and its impact
on simulation results, we propose next to inspect
the average length of the shortest paths in the
network as a function of time. The statistics are
made by a randomly selected mobile node. For all
the applied scenarios, we’ve started by the same
initial network with a shortest paths’ mean length
equal to 3.8 hops. We observe through figure s
4 to 8 that the use of the non restricted random
waypoint mobility model leads to smaller average
path lengths than when we use our restricted model.
The mean length of the shortest paths from the
node making the statistics towards all the remain-
ing nodes in the network is smaller when using the
non restricted model; and this for all the applied
maximum speeds. Notice here that, as pointed out
in (Yoon, Liu and Noble, 2003), the hysteresis ef-
fect taking place when using high speeds is due to
the large simulation time needed by the random
waypoint model to reach the permanent regime,
as well as some of the initial parameters of this
model, such as the use of a non null pause time
and a null lower limit of the interval from which
the nodes’ speed is uniformly picked up.
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 3. The geographical distribution of the mobile nodes after simulation running time = 1500s

The Rest of the Simulation Set Up
We used a transmission range of 250 m, a toler-
ance d of 25 meters that is a tolerance equals
to one tenth of the used transmission range, a
network capacity of 54 Mbps and a maximum
MAC retransmission count equals to 3. We used a
priority IP module at the network layer to enforce
that Hellos are treated before any awaiting data
packet. The priority queue maximum size is set
to 100 packets where data packets are admitted
only if less than 70 packets are within the queue.

Figure 4. The mean length of the shortest paths as a function of time at a maximum speed of 1m/s

1353
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 5. The mean length of the shortest paths as a function of time at a maximum speed of 2m/s

Figure 6. The mean length of the shortest paths as a function of time at a maximum speed of 5m/s

The Maximum updating period size Tmax is Simulation Results

set to 24 seconds and the minimum, Tmin, is fixed
to 3 seconds. Finally, the Max_duration_period,
representing the life time of an entry in the rout-
ing table, is set to 30 seconds. All required
modifications are ported on the OMNeT++ net-
work simulator.
Figure 9 portrays the validity of the cartography as
a function of time and for different node speeds.
Recall that we are using a priority IP handling and
therefore the network load has a very little impact
on the validity of the cartography. For a null node
speed (no mobility) as well as a speed of 1 m/sec,
we get a validity of one hundred percent. As the
speed gets higher the validity of the cartography

1354
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 7. The mean length of the shortest paths as a function of time at a maximum speed of 10m/s

Figure 8. The mean length of the shortest paths as a function of time at a maximum speed of 20m/s

declines a little bit but still stays sufficiently high
around 97 percent even for a speed of 10 m/s.
SUITABILITY OF THE PROPOSED
ROUTING PROTOCOL
To experiment our new stability-based routing
proposal, we conducted a set of simulations using
the OMNeT++ simulator where we integrated our
proposed protocol, using the cartography collect-
ing algorithm and the Priority IP handling. We also
integrated OLSR under the same prioritization IP
model. Our simulation set up is as defined previ-
ously. The OLSR TC period is set to 8 seconds and
its Hello period to 2 seconds. The initial topology
of 100 nodes is chosen randomly, but in a way to
have 11 to 12 nodes in each area (in each area, we
place 11 nodes uniformly). We immobilized 10
source nodes at the left edge of the simulation area

1355
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
Figure 9. Validity of the cartography as a function of time and speed
(randomly distributed over the three sub-areas at
the left edge) and similarly 10 destination nodes
at the right edge. We also consider ten traffic
flows using these fixed source-destination pairs
(pairs are chosen randomly). As such, we insured
multi-hop routes (4 to 5 hops). Finally, the data
packet size used is 200 bytes.
Let us now investigate the correctness or the
validity of the routes given by each protocol
against the actual real network. The routing period
is divided into equally spaced observation points;
namely an observation point each second. At
each observation time point, we have then three
different views of the network: the actual real
view provided by the underlying mobility model
used in our OMNeT++ simulator (to which we
compare the routing protocols), the OLSR rout-
ing table and the routing table provided by our
proposed protocol. A route from a source node
to a destination node is termed valid if it does
exist in the real network regardless whether it is
optimal (a shortest route) or not.
To ascertain the efficiency and accuracy of
our proposed routing algorithm, we compute the
percentage of valid routes from a designated node
(node 1) to all the other 99 destinations given by
each one of the two protocols. Figure s 10, 11,
1356
12, 13, and 14 portray the validity of the routes
as a function of time within an observation Time
window of 25 seconds, for the two defined proto-
cols with a Null applied data load and respectively
for the speeds 0 m/sec, 1 m/sec, 2 m/sec, 5m/sec
and 10 m/sec.
We readily observe that for a null speed the
validity provided by both protocols is, as ex-
pected, equal to one hundred percent. As node
speed gets higher, the OLSR routing protocol
starts losing its efficiency by delivering the low-
est route validity. More interestingly, we notice
that our proposed protocol succeeds in keeping
this validity near the 100% over the time even for
high mobility levels.
Having such a result, we can expect that our
proposed routing protocol is going to completely
outperform OLSR. Figures 15, 16, 17, 18 and
19 represent the number of received packets per
flow (the network throughput) as a function of
the applied data load rate per flow and for dif-
ferent speeds. We clearly observe the sensitivity
of OLSR to the network dynamics. As the node
speed gets higher, our protocol delivers more
throughput than OLSR. While the superiority
of our protocol is clear over the entire range of
traffic loads, it gets larger as the traffic load gets
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 10. Validity of the routes as a function of time: our proposal vs OLSR, Speed=0 m/sec

Figure 11. Validity of the routes as a function of time: our proposal vs OLSR, Speed=1 m/sec

Figure 12. Validity of the routes as a function of time: our proposal vs OLSR, Speed=2 m/sec

1357
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 13. Validity of the routes as a function of time: our proposal vs OLSR, Speed=5 m/sec

Figure 14. Validity of the routes as a function of time: our proposal vs OLSR, Speed=10 m/sec

higher. Indeed, at moderate to high workloads,
OLSR is unable to cope with the loss of the rout-
ing validity as its routes validity gets deteriorated
as shown in the previous Figures (11 12, 13 and
14) and consequently more and more packets
wander through the network wasting resources
otherwise used to correctly forward data traffic.
For a sustained traffic load of 50 generated data
packets per second and a high mobility of 10 m/s
(Figure 19), our protocol yields more than 50%
additional throughput than that of OLSR.
Let us now, investigate the performance of
each protocol in terms of the end to end delay.
This metric measures the average end to end
delay in seconds spent by the received data pack-
ets over all defined flow. Figures 20, 21, 22, 23,
and 24 portray this average end to end delay as a
function of the traffic load per flow and for dif-
ferent speeds. We clearly observe, except for the
Null mobility case, that our routing protocol deliv-
ers packets to their ultimate destinations in much
less time than OLSR does. As the node speed gets
higher, OLSR amounts to very high average end
to end delay attaining hundreds of milliseconds
(Figure 24) for even moderate traffic loads. Our
protocol, at the opposite, sustains both high speeds
and workload and maintains a rather very adequate
average end to end delay under few tens millisec-
onds even for a traffic load of 30 packets per
second.

1358
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 15. Average number of received packets / flow as a function of the applied load: this proposal
vs OLSR, Speed=0 m/sec

Figure 16. Average number of received packets / flow as a function of the applied load: this proposal
vs OLSR, Speed=1 m/sec

Figure 17. Average number of received packets / flow as a function of the applied load: this proposal
vs OLSR, Speed=2 m/sec

1359
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 18. Average number of received packets / flow as a function of the applied load: this proposal
vs OLSR, Speed=5 m/sec

Figure 19. Average number of received packets / flow as a function of the applied load: this proposal
vs OLSR, Speed=10 m/sec

Figure 20. Average delay as a function of the applied load: this proposal vs OLSR, speed=0 m/sec

1360
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 21. Average delay as a function of the applied load: this proposal vs OLSR, speed=1 m/sec

Figure 22. Average delay as a function of the applied load: this proposal vs OLSR, speed=2 m/sec

Figure 23. Average delay as a function of the applied load: this proposal vs OLSR, speed=5 m/sec

1361
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Figure 24. Average delay as a function of the applied load: this proposal vs OLSR, speed=10 m/sec

When nodes are stationary, OLSR has more
chance to slightly outperform our protocol as can
be observed from Figure 20. Indeed our protocol
builds slightly longer routes than OLSR as can
be seen on Figure 25. Recall that for the sake of
stability, we choose to use shorter links (R - d).
As Shown in Figure 25, the average number
of hops that data packets have made before reach-
ing their destinations under our protocol is
slightly larger than the average length of the routes
built by OLSR. The latter provides exactly the
shortest routes computed on the real network at
the beginning of the routing period. Since, in this
case, the network is stationary; the shortest routes
remain valid until the end of the period.
Now, we turn to investigate the signaling cost
required for our proposal to work and compare it
to that of OLSR. Recall that our routing protocol
exchanges extended Hello messages to set up and
maintain an up to date cartography on which it
builds its routing decisions and period sizes. These
Hello messages are rather very short compared
to the huge TC messages that OLSR builds and
distributes among the different MPRs. As such, it
is paramount to investigate not only the average
number of signaling messages exchanged by each

Figure 25. Average length of used routes by received packets: this proposal Vs OLSR, with null speed

1362
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
node but instead the average number of signaling
bytes per second transmitted by each node within
the network. The later provides more insights as
to how much signaling capacity per second is
required from the underlying network.
Figure 26 portrays the average number of
signaling messages (Hello and TC messages
for OLSR and Hello messages for our proposal)
transmitted by each node during a simulation time
of 100 seconds, as a function of the node speed
and for a null traffic load. Recall that the signal-
ing traffic is virtually independent of the traffic
load as we have adopted an IP priority mechanism
for both implementations of OLSR and our rout-
ing protocol. As expected, OLSR transmits less
routing messages than our proposed protocol.
Furthermore, as the node speed gets higher, our
protocol requires more routing messages to be
transmitted as the routing periods get smaller. The
signaling cost of OLSR, on the other hand, remains
virtually insensitive to the node speed as both of
its Hello and TC periods are maintained constant.
We may also notice on Figure 26 that the curve
relative to OLSR is in fact slightly increasing with
the node speed. Indeed, the number of MPR nodes
in the network gets slightly higher as the speed
increases. Since the network topology changes
quickly at higher speeds, new nodes are selected
to be MPRs. But some of the older MPRs, despite
Figure 26. Average number of signaling messages as a function of the network maximum speed: this
proposal Vs OLSR, null load
the fact that they can no more fulfill their status,
they persist being MPRs for a period of 3×TC
seconds (that is 24 seconds in our case) as speci-
fied in the OLSR protocol.
Figure 26 shows that OLSR transmits less
routing messages than our proposed protocol, but
recall that each MPR accumulates all received
signaling messages (Hello and TC messages) and
aggregates them into one large TC packet, while
our algorithm sends each individual topological
change as a single message. This explains the low
validity of routes used by OLSR especially when
the velocity is high. In OLSR, a collided or lost
routing packet leads patently to the loss of a large
amount of topological information otherwise re-
quired to update routing tables. More interestingly,
OLSR utilizes much more capacity to transmit
its huge signaling messages to the opposite of
our proposed protocol. Figure 27 portrays the
average number of signaling bytes transmitted by
second by each node which represents the average
effective bandwidth consumed by the signaling
traffic transmitted by each node. We observe that
our proposed protocol outperforms OLSR even
at null speed. Indeed, in our proposed protocol,
each node adapts its routing period to the perceived
degree of its mobility. With a static network, our
proposal adopts an updating period of 24 seconds
(Tmax) for all the nodes, while OLSR maintains
1363
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
a Hello period of 2 seconds and a TC period of 8
seconds regardless of node speeds. This amounts,
as portrayed on Figure 27 to almost doubling the
signaling traffic for OLSR. As the node speed
increases, the signaling overhead gets bigger for
both protocols but the difference between the
two protocols gets smaller as our proposal adapts
the routing period sizes to the mobility degree.
For speeds of 5 m/s, respectively 10 m/s, OLSR
requires around 43%, respectively 19%, more
signaling traffic than our proposed protocol.
Recall here that our proposed routing protocol
generates control traffic proportionally to the
sensed node’s speed. A node never gets farther
than d meters from its last announced position
without spreading its new position. As a conse-
quence by willingly fixing the node’s speed, the
tolerance parameter d will directly affect the
signaling overhead. At a fixed speed, the com-
puted routing period size of a node increases as
the tolerance value d increases and vice versa.
The other side of the coin is, though, an undesir-
able increase in the number of hops of the se-
lected routes is obtained when the value of d gets
large. This leads also to a high increase of the
number of retransmissions needed to deliver data
packets to their ultimate destinations. As such,
our tolerance parameter becomes a tuning param-
eter that needs to be further investigated.
Figure 27. Average number of signaling bytes/sec as a function of the network maximum speed: this
proposal Vs OLSR, null load
1364
CONCLUSION
In this paper, we proposed a new asynchronous-
cartography-based-routing protocol that allows
each mobile node in the network to calibrate its
own routing period size according to its perceived
local dynamics. This calibration allows each node
to re-announce itself once it exceeds d meters from
its previously announced position. This tolerance
distance represents a real tuning parameter that
needs to be appropriately gauged. Increasing d
leads to reducing the control overhead, but simul-
taneously increases the number of hops composing
the computed routes.
To evaluate our local self regulating routing
and compare its performance to that of the seminal
OLSR, we conducted extensive simulations. As a
first step, we focused on the underling used mobil-
ity model: the random waypoint. This particular
model suffers from some known problems, mainly
the middle concentration problem. We proposed a
new restricted random waypoint model to precisely
solve this aforementioned problem. We showed
through the conducted simulations that the effect
of this problem is minimized using our proposed
mobility model.
We then moved on to evaluate our proposed
routing protocol. We showed in particular, through
different scenarios of node speeds and various
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
workloads that our routing proposal outperforms
OLSR in terms of a better routing validity, an
enhanced throughput, a much lower end to end
delay and a lower signaling overhead.
Currently, we are investigating the scalability
of our routing protocol and its performance under
different mobility and propagation models. We are
also investigating the tuning tolerance parameter
d to best improve the network performances.
REFERENCES
Abid, M. A., & Belghith, A. (2009-3). Functional
enhancements of proactive routing in mobile ad
hoc networks. Third IFIP International Conference
on New Technologies, Mobility and Security (IFIP
NTMS 2009), Cairo, Egypt, 20-23 December,
2009.
Belghith, A., & Abid, M. A. (2009-1). Autonomic
self tunable proactive routing in mobile ad hoc
networks. The 5th IEEE International Conference
on Wireless and Mobile Computing, Networking
and Communications (IEEE WiMob’09), Mar-
rakech, Morocco, October 12-14. 2009.
Belghith, A., & Abid, M. A. (2009-2). Dynami-
cally self adjustable proactive routing protocols
for mobile ad hoc networks. The 34th Annual
IEEE Conference on Local Area Networks (IEEE
LCN’09), Zurich, Switzerlan, October 20-23.
Belghith, A., Mnaouer, A. B., & Abid, M. A.
(2009). Suitability analysis of probabilistic rout-
ing for dynamic ad-hoc networks. The 5th IEEE
International Conference on Wireless and Mobile
Computing, Networking and Communications
(IEEE WiMob 2009), Marrakech, Morocco, Oc-
tober 12-14, 2009.
Beraldi, R., Querzoni, L., & Baldoni, R. (2006).
A hint-based probabilistic protocol for unicast
communications in MANETs. Ad Hoc Networks,
4, 547-566. Elsevier, 2006.
Bettstetter, C., Lassila, P., & Virtamo, J. (2006).
Spatial node distribution of the random waypoint
mobility model with applications. IEEE Trans-
actions on Mobile Computing, 5(6), 680–694.
Retrieved from http://ieeexplore.ieee.org/lp-
docs/epic03/ wrapper.htm?arnumber=1624340.
doi:10.1109/TMC.2006.86
Boukerche, A., de Oliveira, H. A. B. F., Nakamura,
E. F., & Loureiro, A. A. F. (2007). Towards an
integrated solution for node localization and data
routing in sensor networks. Proceedings 12th IEEE
Symposium on Computers and Communications
(ISCC ‘07), July 2007.
Camp, T., Boleng, J., & Davies, V. (2002). A
survey of mobility models for ad hoc network
research. Wireless Communication and Mobile
Computing (WCMC): Special ISSUE on Mobile
Ad Hoc Networking: Research. Trends and Ap-
plications, 2(5), 483–502.
Capkun, S., Hamdi, M., & Hubaux, J. P. (2001).
GPS-free positioning in mobile ad-hoc networks.
Proceedings of the 34th Hawaii International
Conference on System Sciences (HICSS), 2001.
Chung, W. H. (2004). Probabilistic analysis of
routes on mobile ad hoc networks. IEEE Com-
munications Letters, 8(8), 506–508. doi:10.1109/
LCOMM.2004.833841
Clausen, T., & Jacquet, P. (2003). Optimized link
state routing protocol (OLSR). Request for Com-
ment 3626. MANET Working Group.
Clausen, T., Jacquet, P., & Viennot, L. (2002).
Comparative study of routing protocols for mobile
ad hoc networks. The First Annual Mediterranean
Ad HOC Networking Workshop, September 2002.
Dube, R., Rais, C. D., Wang, K. Y., & Tripathi, S.
K. (1997). Signal stability-based adaptive routing
(SSA) for ad hoc mobile networks. IEEE Per-
sonal Communications Magazine, 4(1), 36–45.
doi:10.1109/98.575990
1365
 Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs
Dubois-Ferriere, H., Grossglauser, M., & Vetterli,
M. (2003). Age matters: Efficient route discovery
in mobile ad hoc networks using encounter ages.
Proc. of ACM MobiHoc03, 1-3 Annapolis, MD,
USA, 2003.
Giordano, S., & Stojmenovic, I. (2003). Position-
based ad hoc routes in ad hoc networks (pp.
287–300). Boca Raton, FL: CRC Press, Inc.
Hyytia, E., Lassila, P., & Virtamo, J. (2006). Spatial
node distribution of the random waypoint mobility
model with applications. IEEE Transactions on
Mobile Computing, 5(6), 680–694. doi:10.1109/
TMC.2006.86
Johnson, D. B., Maltz, D. A., & Hu, Y.-C. (2004).
The dynamic source routing protocol for mobile
ad hoc networks (DSR). Internet draft, IETF
MANET Working Group 2004.
Kleinrock, L. (1975). Queueing systems, volume
1: Theory. In Wiley Interscience (Ed.), Queueing
systems.
Mbarushimana Amine, C., & Shahrabi, A. (2007).
Comparative study of reactive and proactive
routing protools performance in mobile ad hoc
networks. The 21st International Conference on
Advanced Information Networking and Applica-
tions Workshop, AINA’07, IEEE 2007.
Nagpal, R., Shrobe, H., & Bachrach, J. (2003).
Organizing a global coordinate system from local
information on an ad hoc sensor network. Pro-
ceedings of the Second International Workshop
Information Processing in Sensor Networks
(IPSN), 2003.
Niculescu, D., & Nath, B. (2003). DV based
positioning in ad hoc networks. [Springer Neth-
erlands.]. Telecommunication Systems, 22(1-4),
267–280. doi:10.1023/A:1023403323460
1366
Novatnack, J., Greenwald, L., & Arora, H. (2005).
Evaluating ad hoc routing protocols with respect
to quality of service. Wireless and Mobile Com-
puting Networking and Communications WiMob,
05, 2005.
Perkins, C. E., & Bhagwat, P. (1999). Highly
dynamic destination-sequenced distance-vector
routing (DSDV) for mobile computers. ACM
SIGCOMM’94 Conference on Communications
Architectures, Protocols and Applications, 1999.
Perkins, C. E., Royer, E. M., & Chakeres, I. E.
(2003). Ad hoc on demand distance vector rout-
ing protocol. Internet draft, MANET Working
Group 2003.
Roth, M., & Wicker, S. (2003). Termite: Emergent
ad-hoc networking. Proceedings of the Second
Mediterranean Workshop on Ad-Hoc Networks,
Mehdia, Tunisia, 2003.
Saad, C., Benslimane, A., & Konig, J.-C. (2008).
AT-dist: A distributed method for localization with
high accuracy in sensor networks. International
Journal Studia Informatica Universalis, Special
Issue on Wireless Ad Hoc and Sensor Networks,
6(1).
Sridhar, K. N., & Chan, M. C. (2005). Stability and
hop-count based approach for route computation
in MANET. In Proceedings of ICCCN05.
Takenaka, T., Mineno, H., Tokunaga, Y., Miyauchi,
N., & Mizuno, T. (2007). Performance evaluation
of optimized link state routing-based localiza-
tion. 2007 IEEE International Performance,
Computing, and Communications Conference,
(pp. 150-159).
Toh, C.-K. (1997). Associativity-based routing for
ad hoc mobile networks. International Journal
on Wireless Personal Communications, 4(2),
103–139. doi:10.1023/A:1008812928561
Cartography and Stability to Enhance and Self Regulate Proactive Routing in MANETs

Tseng, Y., Li, Y. F., & Chang, Y. (2003). On route
life time in multi-hop mobile ad hoc networks.
IEEE TOC, 2(4).
Yoon, J., Liu, M., & Noble, B. (2003). Random
waypoint considered harmful. IEEE INFOCOM
2003 Twenty-Second Annual Joint Conference
of the IEEE Computer and Communications
Societies (pp. 1312-1321). IEEE. Retrieved from
http://ieeexplore.ieee.org/lpdocs/epic03/ wrapper.
htm?arnumber=1208967
Yu, D., Li, H., & Gruber, I. (2003). Path avail-
ability in ad hoc network. Proceedings of the 10th
International Conference on Telecommunications
(ICT03), volume 1, (pp. 383–387).
Yuan, X. (1999). On the extended Bellman-Ford
algorithm to solve two-constrained quality of
service routing problems. Proceedings of The
Eighth International Conference on Computer
Communications and Networks (IC3N’99), Bos-
ton, USA, 1999.
Zhang, H., & Dong, Y.-N. (2007). A novel path
stability computation model for wireless ad hoc
networks. IEEE Signal Processing Letters, 14(12).
doi:10.1109/LSP.2007.907994
Zygmunt, J., Haas, M. R., & Pearlman, P. S. (2002).
The interzone routing protocol (IERP) for ad hoc
networks. Internet Draft, work in progress.

This work was previously published in International Journal of Business Data Communications and Networking, Volume 6, Issue
4, edited by Debashis Saha and Varadharajan Sridhar, pp. 21-37 , copyright 2010 by IGI Publishing (an imprint of IGI Global).

1367
1368

Chapter 82
Learning Geospatial Concepts
as Part of a Non-Formal
Education Robotics Experience
Viacheslav Adamchuk Neal Grandgenett
McGill University, Canada University of Nebraska at Omaha, USA

Bradley Barker Megan Patent-Nygren

University of Nebraska-Lincoln, USA University of Nebraska-Lincoln, USA

Gwen Nugent Collin Lutz

University of Nebraska-Lincoln, USA Virginia Polytechnic Institute and State
University, USA
Kathy Morgan
University of Nebraska-Lincoln, USA

ABSTRACT
In the increasingly modern and technological world, it has become common to use global navigation
satellite system (GNSS), such as Global Positioning System (GPS), receivers, and Geographic Information
Systems (GIS) in everyday life. GPS-equipped mobile devices and various Web services help users
worldwide to determine their locations in real-time and to explore unfamiliar land areas using virtual
tools. From the beginning, geospatial technologies have been driven by the need to make efficient use
of natural resources. More recently, GPS-equipped autonomous vehicles and aircraft have been under
development to facilitate technological processes, such as agricultural operations, transportation, or
scouting, with limited or virtual human control. As outdoor robotics relies upon a number of principles
related to science, technology, engineering, and mathematics (STEM), using such an instructional context
for non-formal education has been promising. As a result, the Geospatial and Robotics Technologies
for the 21st Century program discussed in this chapter integrates educational robotics and GPS/GIS
technologies to provide educational experiences through summer camps, 4-H clubs, and afterschool
programs. The project’s impact was assessed in terms of: a) youth learning of computer programming,
mathematics, geospatial and engineering/robotics concepts as well as b) youth attitudes and motivation
towards STEM-related disciplines. An increase in robotics, GPS, and GIS learning questionnaire scores
and a stronger self-efficacy in relevant STEM areas have been found through a set of project-related
assessment instruments.

DOI: 10.4018/978-1-4666-2038-4.ch082

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
INTRODUCTION
There is growing national concern that the United
States’ educational efforts are not producing
enough STEM (Science, Technology, Engineering
and Mathematics) professionals to meet the needs
of our increasingly sophisticated and technological
society, especially when compared to many other
countries around the world. National reports such
as the 2010 Rising Above the Gathering Storm
Revisited (National Academy of Science, 2010),
focus on the problem of U.S. competitiveness in
STEM areas as summarized across educational
reports and statistics. This diminishing STEM
competiveness directly threatens our economy. For
example, a sample from these business statistics
indicates that, of Walmart’s 6000 suppliers, at least
5000 now reside in China (Goodman & Pan, 2004).
Engineering statistics are particularly frightening.
As an example, during a recent period when two
high-rise buildings were built in Los Angeles, a
total of more than 5000 similar buildings were
built in Shanghai (Fineberg, 2006).
Educational statistics from national and
international reports further illustrate the need
for transformative national reforms in STEM
education, and the need to pursue more innova-
tive educational strategies. In 2010, the World
Economic Forum ranked the United States as
48th in the overall quality of mathematics and sci-
ence education (World Economic Forum, 2010)
and the ACT recently reported that 78 percent of
high school graduates did not meet the readiness
benchmark levels for one or more entry-level col-
lege courses in mathematics, science, reading and
English (American College Testing, 2008). The
United States government is becoming increas-
ingly concerned about this STEM educational
gap, and in response to these alarming national
statistics, President Obama, in his 2011 State of
the Union Address, identified this STEM education
concern as our nation’s new “Sputnik moment.”
Finally, while there is little doubt that the United
States will have to develop human potential in
the STEM fields to stay competitive in the global
economy, there is also the need for universal un-
derstanding of STEM concepts for everyone. For
example, the National Research Council (2011)
recently reported that:
a compelling case can also be made that under-
standing science and engineering, now more
than ever, is essential for every American citizen.
Science, engineering, and the technologies they
influence permeate every aspect of modern life.
Indeed, some knowledge of science and engineer-
ing is required to engage with the major public
policy issues of today as well as to make informed
everyday decisions, such as selecting among al-
ternative medical treatments or determining how
to invest public funds for water supply options
(National Research Council, 2011 p1-1).
What makes an effective context for STEM
education is steadily changing as computer tech-
nology influences the ways in which we can teach
and learn about these important disciplines (Heid,
2005; Hegedus & Kaput, 2004). Instructional
tools such as tablet computers, online simulations,
robots, graphing calculators, and various other
technologies are allowing schools and teachers to
more effectively provide the instructional depth to
STEM coursework and to teach core concepts in
the context of advanced technologies (Anderson,
et. al, 2011; Heid & Edwards, 2001; Whitemeyer,
Nicoletti & DePaor, 2010). Integrated technology
environments such as educational robotics may be
particularly engaging and motivating for students
in the instruction of STEM concepts (Gura & King,
2007). Students engaged in STEM lessons using
such platforms have the opportunity to become
involved in various interdisciplinary connections
between science, technology, engineering, and
mathematics that cover numerous achievement
levels (Avanzato, 2000; Heer, Traylor & Fiez,
2003; Thompson, Heer, Brown, Traylor & Fiez,
2004).
1369
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
An effective, relevant and motivating student
context for STEM education is indeed at the
foundational core of effective STEM education,
and represents a key feature of a desired national
response to this growing crisis (National Science
Board, 2009). Such a national response at the lo-
cal level would seem to suggest a collaborative
effort to make STEM learning more exciting and
effective for students, within a context of real life
applications and relevance. Formal educational
organizations, such as K12 school districts and
universities, must come together with informal
educational organizations, such as zoos and mu-
seums, to make STEM concepts come alive, and
to help students see the relevance and excitement
in what they study, and to learn STEM concepts
more effectively.
STEM learning can and does take place dur-
ing out-of-school hours (Eshach, 2007; Martin,
2004; Wellington, 1990), which is often referred
to as informal education (Fenichel & Schwein-
gruber, 2010; Martin, 2004; Wellington, 1990;).
Generally informal learning environments provide
experiences that are open ended, learner-centered,
voluntary, and nonlinear (Falk & Dierking, 2000;
Griffin, 1998; National Research Council, 2009).
Moreover, assessment in informal learning envi-
ronments is often comprised of immediate per-
formance feedback from the activity. According
to Fenichel & Schweingruber (2010), informal
Table 1. Types and characteristics of informal learning environments
Characteristics
Everyday learning
Social context Individual
Place Public or private spaces
Impetus of participation Self
Time-scale Short-term
Facilitation Learner
Curriculum None
1370
learning occurs in three distinct venues including
everyday learning, designed learning environ-
ments, and programs (Table 1).
Everyday informal learning can happen just
by watching a science program on TV, participat-
ing in on-line games, or reading a book or maga-
zine (Fenichel & Schweingruber, 2010). These
common learning experiences are usually chosen
and led by the learner, and are short in duration.
Moreover, informal learning can also take place
at institutions outside of the classroom including
museums, zoos, botanical gardens, libraries, and
planetariums. These institutions provide designed
environments that are characterized by the use of
artifacts, media, signage, and interpretation by
the educators to guide youth through the learning,
while at the same time these experiences are
short-term and completed in less than a day. While
more structured than the everyday learning set-
tings, the designed environments allow partici-
pants to chose their own learning experiences.
Finally, informal learning programs, which are
also referred to as non-formal learning environ-
ments, are usually adult led and supported by a
curriculum that can be sequential or non-sequen-
tial (Fenichel & Schweingruber, 2010). Programs
include experiences such as 4-H clubs, afterschool
programs, summer camps, museum programs and
are typically longer in duration than designed
learning environments.
Informal Learning Environments
Designed environments Non-formal
Usually group-based learning Group or individual
Museum, zoo, library, planetarium, Public or private space
with media and artifacts including designed spaces and
schools
Self, parent or teacher Self, parent or teacher
Short-term, sporadic Short-term to sustained long-
term
Facilitator–led then learner choice Facilitator led
Media and artifacts Formal curriculum
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
While formal learning environments are ar-
guably the most important centers of learning,
research shows that time spent in formal class-
rooms pales in comparison with the amount of
time spent outside of the classroom in the K-12
learning years. While it is hard to determine exactly
how much time students spend in the classroom,
estimates range from 9% to 18.5% (Fenichel &
Schweingruber, 2010; Sosniak, 2001). Medrich,
Roizen, Rubin and Buckley (1982) estimated that
individuals spend 11,000 hours in the classroom
and 65,000 hours outside of school during their
K-12 years or roughly 85% of their waking hours
outside of the classroom (Eshach, 2007).
Because youth spend so much time outside of
the classroom and because “schools cannot act
alone, and society must better understand and draw
on the full range of science learning experiences
to improve science education broadly” (Fenichel
& Schweingruber, 2010, p.12) informal science
programs play an increasingly vital role in STEM
learning. In the 2009 report, Learning Science
in Informal Environments: People, Places, and
Pursuits, the National Academy of Sciences states:
Learning science in informal environments has
the potential to bolster science education broadly
on a national scale. This is evident in reports from
national initiatives to improve education in sci-
ence, technology, engineering, and mathematics
(STEM) in the United States... The council’s report
cites informal education as one of three integral
pieces of the U.S. education system (the other
two being K-12 education and higher education)
needed to ensure “U.S. economic competitiveness,
particularly the future ability of the nation’s edu-
cation institutions to produce citizens literate in
STEM concepts and to produce future scientists,
engineers, mathematicians, and technologists”
(U.S. Department of Education, 2007, p. 5).
Federal interest in informal environments is also
reflected in the National Science Board’s report
on the critical needs in STEM education (National
Science Board, 2007). The National Science Board
report stresses the need for coherence in this
kind of learning and an adequate supply of well-
prepared and effective STEM teachers... Informal
education is described as an essential conduit to
increase public interest in and understanding and
appreciation of science, technology, engineering,
and mathematics. (National Research Council,
2010 p.13).
It seems clear that research supports the use
of STEM activities in the informal and the non-
formal learning environments to enhance, bolster
and support formal learning environments.
One promising approach to increasing STEM
attitudes, knowledge, skills and workforce capac-
ity in a non-formal learning environment is to blend
the use of robotics with geospatial technologies
as an instructional platform. The University of
Nebraska, with grant funding from the National
Science Foundation (NSF), has developed the
Geospatial and Robotics Program for the 21St
Century (GEAR-Tech-21) program to increase
science, technology, engineering and mathemat-
ics (STEM) achievement and create interest by
coalescing robotics and geospatial technolo-
gies. Through hands-on experimentation, such
technologies help youth to transform abstract
mathematics and science concepts into concrete,
real-world applications. Recent improvements
in cost and complexity make it possible for even
relatively young children to engage in hands-on
experimentation with robots, GPS receivers and
GIS software.
Robots have the potential to change and en-
hance the learning process in education (Cham-
bers & Carbonaro, 2003). It is most remarkable
that computer-based technologies can be used
as “mindtools” which involve students in using
modern technologies to solve problems (Jonas-
sen, 2000). Traditional education practices typi-
cally utilize a relatively direct form of instruction
whereby students learn from the “technology”
(e.g. drill and practice exercises), while in con-
1371
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
trast, when using “mindtools” such as robotics,
students learn with the “technology”.
When considering spatial technologies like
GPS and GIS, similar to robotics, there is grow-
ing potential for educational applications (Kerski,
2003). GPS/GIS have become important tools
in precision agriculture and natural resources
management, which are relevant to youth in rural
communities; these students often suffer directly
from poor performance in STEM-related achieve-
ment areas. For example, GIS maps allow users to
visualize geographic data to aid agriculture-related
decision-making processes. Data used in the GIS
maps can also be obtained using small handheld
GPS receivers, which are easy to use, even for
middle school aged youth. Combined, the GPS and
GIS technologies provide a powerful set of tools
to analyze and interpret spatial information and
to represent a useful context for applied learning
in the mindtools context.
The long-term goal of the program described
in this chapter is to improve STEM learning out-
comes and attitudes for youth who are 11-15 years
of age. Using robotics and GPS/GIS concepts,
the program seeks to: 1) promote their interest in
STEM fields, 2) introduce basic technology skills,
3) foster problem solving and inquiry skills, and
4) encourage teamwork.
Program Overview
The widespread accessibility of technologies such
as the LEGO NXT Mindstorm robotics kit, GPS
receivers, and GIS software like GoogleEarth
and ArcMap make it possible for youth (ages 11
to 15) to explore and practice STEM concepts
through the integration of these technologies.
The project started in 2006 as the Nebraska 4-H
Robotics and GPS/GIS in 4-H: Workforce Skills
for the 21st Century (NSF contract #ESI-0624591),
and was designed as a regional project for Ne-
braska. However, in 2008 the program received
additional funding (NSF contract #0833403) to
further scale-up the program from a regional to
1372
a national project. The purpose of the GEAR-
TECH-21 program is to improve STEM learning
outcomes and attitudes of youth aged 10 to 15 by
combining educational robotics and geospatial
(GPS/GIS) technologies. The combination of
these technologies is widely used in precision
agriculture and natural resources management and
has many implications for improved sustainable
farming methods and conservation. Information-
based management has allowed producers around
the world to optimize agriculture inputs, reduce
negative environmental impact and improve the
social status of relevant professions. The program
provides youth with lessons and experiences in
each of the fields culminating in experiences and
challenges that combine the technologies. The goal
is to spread the GEAR-Tech-21 program to 5000
youth in 450 clubs and camps over five years.
GEAR-Tech-21 is designed for delivery by
adult facilitators in non-formal learning environ-
ments such as 4-H clubs, summer camps, and
afterschool programs. In addition, GEAR-Tech-21
is an intensive two-year program that usually
begins with a 40-hour summer camp experience.
The camp activities include the building and pro-
gramming of robots, working with handheld GPS
receivers to explore and collect geo-referenced
information, and the development of GIS maps.
Youth then receive an additional 80-hours of
hands-on robotics and GPS/GIS training during
the school year in their afterschool programs
or in 4-H clubs. In year two, youth attend an
advanced summer camp followed by 80 more
hours of hands-on instruction during the school
year. In total, participating youth receive at least
240 hours of focused, hands-on experience over
two years. Many curriculum activities have been
published by the project and can be found at the
project website http://www.gt21.org.
In educational lexicon bringing an intervention
to scale usually involves increasing the number of
students, teachers, and locations that participate
with the goal of producing similar results (Baker,
2005; Coburn, 2003; Fishman, 2005; McDonald,
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
Keesler, Kauffman, & Schneider, 2006). More-
over, discussion of scale usually relates to formal
educational environments and focuses on effective
classroom interventions that can be brought to the
district level or interventions developed by outside
vendors brought in to be used at a district level
(Blumenfeld, Kempler, Krajcik, & Blumenfeld,
2006; Cohen & Ball, 2007a). Ultimately, the
goal of scaling an intervention is to successfully
replicate or adopt a program with more students,
in a variety of settings, and produce similar out-
comes as the original intervention (Baker, 2005;
Dede, 2006).
Research on programmatic scaling expands
on the definition and adds multiple dimensions
to the general notion of scale. According to scal-
ing is about the “adoption and diffusion of ideas
across contexts.” Much of the research devoted
to scale-up is concerned with increasing the num-
bers (Baker, 2005; Coburn, 2003; McDonald et
al., 2006). Constas and Brown, (2007) expand
the definition to include magnitude (how many
individuals and locations), breadth (what indi-
viduals and locations), and program complexity
(how many iterations of the intervention will
be supported). Coburn (2003) points out that in
some contexts the original intervention will be
replicated in different settings or the intervention
may be adopted and modified for use in a local
setting. Furthermore, Dede, (2006) separates the
scaling of an individual intervention from the idea
of systemic reform. Whereas a single intervention
can be scaled to multiple classrooms, a systemic
reform typically involves a suite of interventions
that potentially transforms an entire school district.
Research also suggests that the scale-up
process, when undertaken successfully, moves
through at least three different stages. The first
stage is to conduct individual studies or efficacy
trials to determine the effectiveness of the inter-
vention in a particular setting. The second stage
tests an intervention that has shown to be effec-
tive in the individual study in multiple settings
with multiple populations of students. The final
step, in a scale-up process, is ongoing evaluation
to determine efficacy and sustainability and to
find an institutional mechanism to accomplish
the scale-up (Horwitz, 2007; McDonald, et al.,
2006; Motiff, 2007).
In summary, traditional scale-up frameworks
focus primarily on expanding the number of youth
and sites. However, traditional scaling efforts have
not typically explored additional dimensions of
scale such as the nature of change envisioned for
the intervention, how it is sustained, and the role
of educators and sites in the scaling effort (Co-
burn, 2003). Moreover, (Dede, 2006) argues that
traditional scaling efforts need to avoid replicat-
ing a program intervention that worked locally
without taking into consideration the variations
and complexities of new locations. Addressing
the variation and complexities of a new target
population or setting, by carefully reflecting on
scale, can lead to a better understanding of what
is needed to make a difference in teaching and
learning for a wider or larger context.
Early in 2009, we began to apply the scaling
framework developed by Coburn (2003) and Dede
(2006) to the Geospatial and Robotics Technolo-
gies for the 21st Century (GEAR-Tech-21) project
to help our project grow and evolve into a na-
tional effort. The guiding principle of the scaling
framework is that scaling efforts need to avoid
replicating an intervention that worked locally
without taking into consideration the variations
and complexities of new locations. Addressing
the variation and complexities of a new target
population or setting, by carefully reflecting on
scale, can lead to a better understanding of what
is needed to make a difference in teaching and
learning for a wider or larger context. Overall the
scaling framework is successfully being applied to
the GEAR-Tech-21 project as the project moves
from a statewide initiative to a national program.
Clearly there are benefits and tradeoffs between
scaling in an informal environment as compared
to the formal education environment typically
occurring within schools and school districts.
1373
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
One apparent benefit is that there is no systemic
reform movement in the non-formal environment.
Often scaling up an intervention fails in the schools
because a) they fail to take root or b) classroom
teachers and other educators and staff find the
interventions inconsistent with their local practices
and norms (Blumenfeld, et al., 2006). Furthermore,
teachers are often resistant to changes in practice
that originate at the district or state level and are
filtered down to the classroom level (Blumenfeld,
et al., 2006). In addition, organizational rigidity in
the formal education system may leave little room
for innovation (Cohen & Ball, 2007b). Conversely,
in the non-formal learning environments there is
little rigidity in organization structure that would
prevent innovation. More importantly, non-formal
organizations, like 4-H, rely on volunteers to lead
education experiences. To support innovations,
new volunteers can be recruited and existing
volunteers retrained. Volunteers enjoy a degree
of flexibility not found in formal education. Un-
like in the formal education system, volunteers
lack the similarity in preparatory training and
past experiences. Many volunteers do not pos-
ses the typical background content knowledge
and educational skills of certified teachers while
some are professionals in STEM fields but lack
teaching experience. This breadth requires greater
and more diverse instructional resources are fre-
quently needed for the training and development
of volunteers. Finally, unlike formal learning
environments, where criterion-reference tests are
routine, data collection in a non-formal environ-
ment is an ongoing challenge to implement, due
to time, support and training constraints.
Overall the framework has been helpful in
scaling the GEAR-Tech-21 program. However, the
process of scaling the program from a regional to a
national program has encountered some challenges
including professional development and logistics.
To move the program to a national scale, the de-
livery approach had to be modified from GEAR-
Tech-21 staff leading the activities and camps to
a sustainable professional development program
1374
for leaders whereby they are trained to deliver the
program in their local community. For example
in 2009, training was held in seven states in more
than 19 different locations. In addition, not every
site that implements the program has the means
to easily acquire the robot kits and GPS units.
Therefore, equipment is shared amongst sites for
a small fee. Equipment sharing has proved to be
beneficial in getting new sites started. Moreover,
the program has had success in scaling to new
sites and additional audiences (Table 2).
Curriculum
Foundation-building activities in Year 1 allow
youth to develop fundamental skills and to learn
the vocabulary associated with each technology.
In Year 1, youth are introduced to the basics of
robotics, review underlying geography concepts,
gain experience with GPS-based navigation and
data collection and are introduced to using and
manipulating GIS maps. Year 2 foci include
more advanced robotics programming, GIS skill
development and integration of the technologies
while exploring their potential in the world of
agriculture and natural resource management.
In addition to the technology-focused concepts,
activities are also designed to develop 21st Century
Skills in critical thinking and problem solving,
communication and collaboration, and informa-
tion media and technology (Partnership for 21st
Century Skills, 2008; Pearson & Young, 2002).
Table 2. Project camps and club participation
Year Camps Camp Clubs Club
Participants Participants
2007 2 38 4 46
2008 6 147 13 200
2009 5 178 19 297
2010 22 450 46 732
2011* 40 800 80 1200
*Projected numbers of camps, clubs, and participants based on
2011 camp requests and projected club growth.
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
Designed for a non-formal learning envi-
ronment using Experiential (Kolb, 1984) and
Understanding by Design (Wiggins & McTighe,
2005) learning models, project activities offer
youth multiple experiences with each concept to
deepen their understanding and skill development.
The project uses experiential learning strate-
gies combined with on-line eLearning delivery
methods. The 4-H experiential learning model
provides learners with a concrete experience, a
chance to reflect and observe the phenomenon
being studied, and then to apply the knowledge
to a new situation (Woffinden & Packham, 2001).
Each activity consists of an introductory, a guided
large group exercise, a prescribed exercise, an
on-line exercise where youth research real-world
problems, industry applications and an additional
open-ended challenge applying the skills acquired
earlier in the program. At the conclusion, a series
of questions are provided for journaling, facilitated
group discussion or small group work to recap
the learning experience and help youth make the
connections to outside, real-world applications in
their own lives and career fields.
Such activities allow the program participants
to relate their findings to contemporary challenges
being solved using technology. Instead of focus-
ing upon specific pieces of knowledge related
to a given activity, our emphasis has been on
timing the introducion of technology to help in
the learning process and to determine what chal-
lenges outside the material of activity could be
addressed. Eventually, this motivates participants
to learn more related to the technologies and to
extend their perceptions of the environment in
which they live. This approach also strengthens
the connections youth participants have to career
possibilities in related technological fields.
GEAR-Tech-21 provides all youth with a basic
literacy in agriculture and natural resource systems.
Including applications in agriculture and natural
resource systems meets several curricular needs.
The agricultural and natural resources industries
have a particular appeal to 4-H’s traditional rural
audience where these industries are the primary
economic drivers. These industries are also of in-
terest to non-rural youth as they are allengaged as
consumers on a daily basis. The technologies used
in the curriculum are integrated in many formats
and are already in use by professionals working
in the field. Farmers utilize precision agriculture
technology and methods to manage production in
a way that allows for more efficient use of reduced
inputs, equipment automation and maximized crop
quality and quantity. Environmental professionals
use geospatial and robotics technologies to track
and manage populations and habitats. The food
processing industry uses robotics in processing
facilities and geospatial technologies in distribu-
tion and strategic planning.
The curriculum is published in a series of ac-
tivities using Adobe Captivate (Adobe Systems
Inc., San Jose, California) to produce flash-based
animations allowing for the inclusion of text,
imagery, videos and software animations. Direct
teaching of programming and mapping concepts
through click-through software animations allows
youth to see actual settings in the software (Figure.
1). This format permits participant interaction with
the media as well as for dissemination online, via
CD-ROM or in mobile formats. Adult facilitators
are supported with guides and worksheets avail-
able online to be downloaded in PDF format, on
CD-ROM and in print. The curriculum is divided
into single-skill, flash-based modules. To maxi-
mize usage the curriculum can be accessed on-line
or for sites with limited or no Internet connection,
the curriculum can be installed on the computer
and run locally. When running locally, learners
are not able to utilize the social networking tools
including the online discussion boards.
A series of integrated activities that combine
geospatial and robotics technologies make up the
core of the advanced year of the curriculum. For
example, in the Robot Tracking activity youth
program, the robot follows a planned course or
path while recording a track with a GPS sensor
or GPS data logger. Recorded data is then im-
1375
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience

Figure 1. A screenshot of a software animation from the web-based robot tracking activity

ported into a GIS or digital mapping application
such as ArcGIS (ESRI, Redlands, California) or
Google Earth so youth can see an aerial image
overlaid with their robot’s path. With the Robot
Sensor Exploration activity, youth use a remote
control program or run a programmed course
while collecting and storing sensor data for view-
ing in a GIS or in table format generated by a
spreadsheet application. When working on the
Boundary Mapping activity, youth rely upon the
robot’s sensors to identify a field’s natural bound-
aries while recording the track data with a GPS
sensor or data logger to import to GIS software
to build a base field file. The Surfacing Land
Measurements activity requires youth to use their
robot as a data collection device at predetermined
waypoints, measuring reflected light of foliage
(darker green colors indicate greater nitrogen
availability and absorption) and height of vegeta-
tion (more vegetative mass indicates greater nutri-
ent and moisture availability) with the light and
ultrasonic sensors, respectively (Figure 2). The
data is then imported into GIS software and in-
terpolated to create a field map showing vari-
ability of the measured characteristic (Figure 2).
In the Robotic Navigation activity, participating
youth identify a location using a handheld GPS
receiver and input the location into software de-

Figure 2. Summer camp activities: collecting data with the robot’s sensors and presenting findings

1376
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
veloped by an undergraduate student with the
GEAR-Tech-21 program. Using a GPS data log-
ger and Bluetooth, youth can then autonomously
direct their robot to the designated geographic
coordinates. Or, if youth have a GPS sensor avail-
able, the destination coordinates can be entered
into variable blocks within the LEGO NXT soft-
ware. The robot can then autonomously find the
waypoint location. The sensor allows for a larger
field to be used since Bluetooth is no longer a
limiting factor to do this exercise. A follow-up
challenge that is made possible with the sensor is
for the robot to truly geocache by using its other
sensors to find an object meeting given specifica-
tions within the range of error provided by the
GPS navigation. Finally, the curriculum includes
a series of open-ended challenges representing
real-world agricultural applications including
automated harvesting of row, vegetable and tree
crops, rounding-up wayward livestock, applying
fertilizers and pesticides to particular areas of a
field, field scouting, and assessing wildlife popu-
lations.
Hardware and Software
The LEGO Mindstorms NXT Education kit
(LEGO Systems, Inc., Enfield, Connecticut) has
been selected as the main robotics platform for
the curriculum. The unit has three output and four
input ports. Our activities use the output ports
to operate electrical motors (two for maneuver-
ing and one for a mechanical manipulator), and
Figure 3. Summer camp activities: digital map development and robot navigation
the input ports are used to equip the robot with
sound, ultrasonic distance, light reflectance and
mechanical switch sensors. LEGO Mindstorms
NXT software is used to program the robots and
to upload the code via either a USB, or a Blue-
tooth connection.
Garmin eTrex Legend and Rino 110 (Garmin
International, Inc., Olathe, Kansas) GPS receivers
have been used to log waypoints and track paths
as well as for navigation. GoogleEarth (Google,
Inc., Mountain View, California) software has been
used to introduce youth to the world of digital
geographic data. ArcGIS software has helped
to further develop digital mapping awareness
and provide freedom to customize geographic
data (Figure 3a). Using the DNRGarmin module
(Department of Natural Resources, St. Paul, Min-
nesota) allowed for a data interface between GPS
receivers and GIS software.
Through a set of sequential stand-alone ac-
tivities, youth gain perspective of the relationships
between angular measures of geographic latitude
and longitude and physical distances. From the
early part of the program they learn to define
travel speed and direction and to compute the
distance between two nearby locations with known
geographic coordinates. Later these primary skills
are used to gain an understanding of the main
principles of navigation.
Blending GPS and robotic technologies results
in a better experience for the camp participants.
When the two technologies are combined, program
participants have a greater opportunity to improve
1377
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
their perception of navigation principles that fre-
quently require both logic and creative thinking
skills to complete the prescribed challenges.
For the integrated robotics and GPS/GIS ac-
tivities an i-Blue747 GPS Logger (Transystem,
Inc., Hsinchu, Taiwan) has been used to record
geographic positions of the robot. To implement
robot navigation capability (Year 2 curriculum),
several stand-alone applications have been de-
veloped using LabVIEW (National Instruments
Corp., Austin, Texas). Our software allows simul-
taneous wireless communication between a laptop
computer and both the LEGO Mindstorms NXT
and i-Blue747 GPS Logger in order for robot
motion to be adjusted automatically based upon
its relative position to a targeted, georeferenced
marker (Figure 3b). The relatively short range
of Bluetooth communication (approximately
10 meters) and low GPS positioning accuracy
(approximately 3 meters) have been the main
limitations at this point in the project. However,
discovering and quantifying these limitations
have become a focal point of robot navigation
activities. Students have the ability to experiment
with different parameters of our navigation algo-
rithms to appreciate the importance of raw signal
averaging and relating expected outcomes with
the technical hardware capabilities. For example,
they can change parameters in relation to the speed
with which the robot reacts to changes in desired
travel direction and what proximity to the target
is considered acceptable.
A recently funded curriculum development
project by the National Science Foundation (NSF
#0733228) relates to the use of educational robot-
ics in formal educational settings and has added
university partners and broadened the scope of the
educational programming. The use of an educa-
tional cyber-infrastructure and a train-the-trainer
model to administer the curriculum are inherent
challenges being addressed by the expanding pro-
gram, but the project is indeed steadily growing
in scale and curriculum resources. In addition,
a robotics platform (CEENBoT) developed at
1378
the Peter Kiewit Institute (PKI, University of
Nebraska-Lincoln, Omaha, Nebraska) has been
considered as an alternative to the LEGO robotic
kit. The CEENBoT will be capable of operating
in a non-smooth surface environment (Figure 4).
An integrated GPS receiver for the LEGO kit is
another option under development.
Custom hardware, software, and firmware
have been developed to create an integrated GPS
solution for the LEGO Mindstorms NXT robot
(Figure 5). The integrated GPS receiver operates
seamlessly with the LEGO Mindstorms NXT
robot using one of its standard input ports. The
NXT brick can simultaneously take input from
up to four sensors. Each sensor port uses a 6-pin
modular connector that offers both analog and
digital interfaces. The integrated GPS receiver
takes advantage of a digital protocol to commu-
nicate with the NXT brick.
The integrated GPS sensor is based on the
USGlobalSat EM-406A GPS receiver and the
popular Arduino microcontroller platform. The
Arduino platform offers two advantages. First,
the hardware platform facilitates rapid prototyp-
ing for sensor development. Second, the Arduino
is a very popular platform that simplifies micro-
controller programming significantly. Thus, the
Arduino by itself could potentially be used as a
teaching tool in future extensions of the program
curriculum.
Figure 4. New educational materials: one of
CEENBot prototypes
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
Figure 5. New educational materials: add-on
LEGO Mindstorm NXT GPS module
The USGlobalSat EM-406A GPS receiver
features high tracking sensitivity, the ability to
track up to 20 satellites at one time, and respect-
able accuracy for an inexpensive receiver. GPS
information can be accessed within NXT-G,
which is the graphical programming environment
bundled with the LEGO Mindstorms NXT kit. A
graphical programming “block” is used to request
GPS data from the sensor and to integrate into the
overall program.
Using a directly connected GPS sensor removes
many of the shortcomings associated with using
the Bluetooth-based i-Blue747. The Bluetooth
range restrictions are effectively eliminated, al-
lowing the robot to explore areas well outside of
a 10-meter radius. Furthermore, all processing
can be done locally on the robot as opposed to
a laptop. Custom data processing and filtering
algorithms can also be utilized to improve the
accuracy of the raw GPS data stream.
Learning Impact Research
Various assessment instruments have been used
to document participant learning and attitudinal
impact during summer camps, afterschool club
gatherings and other relevant activities such as
short-term interventions, virtual competitions,
etc. (Barker, Grandgenett, Nugent & Adamchuk,
2010; Nugent et al., 2010). For example, in the
summer of 2010, 20 GEAR-Tech-21 camps were
held throughout the United States. A total of 371
youth participated. The overall mean age for youth
participating in the camps was 11.4, with 76%
male and 24% female. 74% were Caucasian or
Asian; 26% were from minority groups.
To measure STEM learning, a 41-item,
paper-and-pencil, multiple-choice assessment,
covering a variety of topics including computer
programming, mathematics, geospatial concepts,
engineering/robotics and engineering design was
developed. The STEM concept test was admin-
istered on the first day of the camp prior to the
start of program activities. The same assessment
was also administered on the last day of camp as
a post-test.
There was a significant overall increase in
learning from the pretest to the post-test (Table
2). There was also a significant increase in scores
on the questions dealing with geospatial concepts.
These results suggest that the 4-H robotics and
geospatial summer camp program is a promising
approach for supporting robotics and geospatial
learning in a non-formal education setting.
To capture changes in attitudes of the 4-H Ro-
botics and GPS/GIS youth participants, an assess-
ment instrument was developed, which included
two scales dealing with GPS/GIS. Specifically, the
instrument measured a youth’s perceived value of
GPS/GIS technologies and their self-efficacy in
performing GPS/GIS tasks. A sample task value
question was “I like learning new technologies
like GIS.” A sample self-efficacy question was: “I
am confident that I can use GPS technologies to
get to places that I have never been before.” Youth
responded using a 5-point Likert scale from 1 =
strongly disagree to 5 = strongly agree. By focusing
on performance tasks, these scales complemented
the multiple-choice content test, which assessed
general comprehension and knowledge. Partici-
pants scored higher on the post-test than on the
pretest for both of these GPS/GIS scales, although
the task value scale was not significant (Table 3).
1379
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
Table 3. Student increases in STEM learning and attitudes
Pre Mean
STEM Learning
Overall 15.62
GPS/GIS 2.65
STEM attitudes
GIS/GPS task value 3.90
GIS/GPS self-efficacy 3.90
Future Developments
The GEAR-Tech-21 program has experienced
significant and steady growth since the first pilot
camps and clubs in 2007. Camps have grown from
two sessions with 38 participants in 2007 to a pro-
jected 40 camps with 800 participants in the summer
of 2011. Similar growth patterns exist with clubs,
beginning with four clubs with 46 participants in
2007 to 46 clubs with 732 participants in the 2010-
2011 school year. The program’s goals are to scale
up to serve at least 5000 youth in 450 clubs and
camps cumulatively over five years.
Shifts in the delivery methods of the GEAR-
Tech-21 program have facilitated the growth
from 2007 to 2011 and beyond. For the first three
years, project staff directly planned, coordinated,
and taught the camp programs. In 2010, GEAR-
Tech-21 staff trained local camp coordinators to
lead the activities with youth using a camp-specific
curriculum. As part of the camp training sessions,
GEAR-Tech-21 staff openly invited all interested
educators in the area. With this method, many
more potential GEAR-Tech-21 volunteers were
trained than if only the camp coordinators were
trained. Camp training sessions were held at 15
locations across the country and reached a total
of 180 attendees.
The increase in adult volunteers and staff
trained to lead GEAR-Tech-21 camp programs
allowed multiple sessions to occur at the same
times in different locations. Over four times as
many camp sessions (22 sessions with 450 par-
ticipants) were held in 2010 as compared to the
previous years (5 sessions with 178 participants).
1380
Post Mean t(df) P
19.00 12.70 (306) p <.001
3.05 4.19 (306) p <.001
3.96 1.59(289) p =.057
4.30 8.50(279) p <.001
In 2011, GEAR-Tech-21 expects 40 camp
sessions serving about 800 youth. The continued
growth demands additional training opportuni-
ties for camp coordinators. This need will be
met through the implementation of a train-the-
trainer model. The first group of GEAR-Tech-21
“Regional Trainers” includes six 4-H staff and
volunteers who will provide training to camp co-
ordinators and other interested educators in three
states. Regional Trainers were offered a stipend
and the equipment required for the program. They
are requested to train a minimum of 40 people per
state. Regional Trainers will provide training to
at least five camps in three states while GEAR-
Tech-21 staff will travel as well as use distance
education technologies to reach the remaining
camps. The project will continue to add three or
more new Regional Trainers each year to assist
with training all interested GEAR-Tech-21 camp
coordinators and club leaders.
The primary barriers to creating new clubs and
camps are the availability of confident, trained
adult leaders and the equipment required. While
the need to educate adults is well addressed through
the training system, assisting groups in getting
equipment is more challenging to scale. GEAR-
Tech-21 currently lends nearly 100 robotics kits
out to clubs and camps each year for a small fee.
This resource helps the local GEAR-Tech-21
program get started. After the initial interest is
generated, equipment resources or funding are
best acquired locally. GEAR-Tech-21 professional
development sessions address the challenge of
finding equipment by providing ideas for com-
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
munity connections as well as brainstorming for
additional local resources among the attendees.
Although the National Science Foundation has
made GEAR-Tech-21’s curriculum development
and growth possible, project leaders intend to
make the project sustainable beyond additional
grant funding. Initial steps toward sustaining the
project include designing curriculum components
for volunteers and participants to purchase (printed
leaders’ guides, youth workbooks, curriculum
DVDs, supporting camp materials, t-shirts, etc.).
GEAR-Tech-21 also expects to continue to de-
velop the integrated GPS sensor for the LEGO
Mindstorms NXT kit into a commercially avail-
able product. In addition, GEAR-Tech-21 staff
will continue to seek grant funding to assist with
sustaining the project.
SUMMARY
Improving STEM education is a steadily growing
and increasingly alarming national challenge, and
the Robotics and GPS/GIS in 4-H: Workforce Skills
for the 21st Century project has been initiated to
help meet that challenge and to provide a non-
formal education experience for 11-15 year old
youth to encourage their consideration of STEM
areas for their future academic and post-graduate
careers. The use of robotics, GPS, and GIS tech-
nologies under the theme of precision agriculture
and natural resource management appears to ap-
peal to many groups of youth when participating
in summer camps. Students participating in the
program showed increased short-term STEM
learning and attitudes.
REFERENCES
American College Testing. (2008). ACT college
readiness report news release. Retrieved from
http://www.act.org/newsroom/releases/view.ph
p?year=archive&p=314&lang=english
Anderson, W., Banerjee, U., Drennan, C. L., El-
gin, S., Epstein, I. R., & Handelsman, J. (2011).
Changing the culture of science education at re-
search universities. Science, 331(6014), 152–153.
doi:10.1126/science.1198280
Avanzato, R. (2000). Mobile robotics for fresh-
man design, research, and high-school outreach.
Proceedings of the IEEE, International Confer-
ence Systems, Man, and Cybernetics, 1, 736–738.
Baker, E. (2005). Principles for scaling up:
Choosing, measuring effects, and promoting the
widespread use of educational innovation. In
Schneider, B., & McDonald, S. (Eds.), Scale-up
in education: Ideas in principle (pp. 37–54).
Lanham, MD: Rowman & Littlefield.
Barker, B., Grandgenett, N., Nugent, G., & Ad-
amchuk, V. (2010). Pairing educational robotics
with geospatial technologies in informal learning
environments. Journal of Youth Development,
5(2), 48–56.
Blumenfeld, P. C., Kempler, T. M., Krajcik, J. S.,
& Blumenfeld, P. (2006). Motivation and cogni-
tive engagement in learning environments. In
Sawyer, R. K. (Ed.), The Cambridge handbook of
the learning sciences. New York, NY. Cambridge.
Chambers, J., & Carbonaro, M. (2003). Design-
ing, developing, and implementing a course on
LEGO robotics for technology teachers education.
Journal of Technology and Teacher Education,
11, 209–241.
Coburn, C. E. (2003). Rethinking scale: Mov-
ing beyond numbers to deep and lasting
change. Educational Researcher, 32(6), 3–12.
doi:10.3102/0013189X032006003
Cohen, D., & Ball, D. (2007a). Conceptualizing
scale-up: An educational perspective. In Sch-
neider, B., & McDonald, S. (Eds.), Scaling-up in
education. Lanham, MD: Rowman & Littlefield.
1381
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
Cohen, D., & Ball, D. (2007b). Educational in-
novation and the problem of scale. In Schneider,
B., & McDonald, S. (Eds.), Scale-up in education.
Lanham, MD: Rowman & Littlefield Publishers
Inc.
Constas, M., & Brown, K. (2007). Toward a pro-
gram of research on scale-up: Analytical require-
ments and theoretical possibilities. In Schneider,
B., & McDonald, S. (Eds.), Scale-up in education:
Ideas in principle (pp. 247–258). Lanham, MD:
Rowman & Littlefield.
Dede, C. (2006). Scaling up. In Sawyer, R. K.
(Ed.), The Cambridge handbook of the learning
sciences (pp. 551–566). New York, NY: Cam-
bridge University Press.
Eshach, H. (2007). Bridging in-school and out-
of-school learning: Formal, non-formal, and in-
formal education. Journal of Science Education
and Technology, 16(2), 171–190. doi:10.1007/
s10956-006-9027-1
Fenichel, M., & Schweingruber, H. (2010). Sur-
rounded by science: Learning science in informal
environments. Board on Science Education, Center
for Education, Division of Behavioral and Social
Sciences and Education. Washington, DC: The
National Academies Press.
Fineberg, H. (2006). Address to The Academy of
Medicine, Engineering & Science of Texas Annual
Meeting. January 5, 2006.
Fishman, B. (2005). Adapting innovations to par-
ticular contexts of use: A collaborative framework.
In Dede, C., Honan, J., & Peters, L. (Eds.), Scaling
up success: Lessons learned from technology-
based educational improvement. San Francisco,
CA: Jossey-Bass.
Goodman, P., & Pan, P. (2004, February 8).
Chinese workers pay for Wal-Mart’s low prices:
Retailer squeezes its Asian suppliers to cut costs.
The Washington Post.
1382
Gura, M. M., & King, K. P. (2007). Classroom
robotics: Case stories of 21st century instruction
for millennial students. Charlotte, NC: Informa-
tion Age Publishing.
Heer, R. L., Traylor, T. T., & Fiez, T. S. (2003).
Enhancing the freshman and sophomore ECE
student experience using a platform for learn-
ing. IEEE Transactions on Education, 46(4).
doi:10.1109/TE.2003.818752
Hegedus, S. J., & Kaput, J. J. (2004). An introduc-
tion to the profound potential of connected algebra
activities: Issues of representation, engagement
and pedagogy. Proceedings of the 28th Conference
of the International Group for the Psychology of
Mathematics Education, 3(1), 129-136.
Heid, M. K. (2005). Technology in mathematics
education: Tapping into visions of the future. In
Masalski, W. J., & Elliott, P. C. (Eds.), Technology-
supported mathematics learning environments, the
sixty-seventh yearbook of the National Council of
Teachers of Mathematics. Reston, VA: National
Council of Teachers of Mathematics.
Heid, M. K., & Edwards, M. T. (2001). Computer
algebra systems: Revolution or retrofit for today’s
classrooms? Theory into Practice, 40(2), 128–136.
doi:10.1207/s15430421tip4002_7
Horwitz, P. (2007). Successful scale-up in three
stages. In Schneider, B., & McDonald, S.-K.
(Eds.), Scale-up in education: Ideas in principle
(pp. 259–268). Lanham, MD: Rowman and
Littlefield.
Jonassen, D. H. (2000). Computers as mindtools
for schools: Engaging critical thinking (2nd ed.).
Upper Saddle River, NJ: Merrill.
Kerski, J. J. (2003). The implementation and ef-
fectiveness of geographic information systems
technology and methods in secondary educa-
tion. The Journal of Geography, 102, 128–137.
doi:10.1080/00221340308978534
Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience
Kolb, D. A. (1984). Experiential learning: Experi-
ence as the source of learning and development.
Englewood Cliffs, New Jersey: Prentice-Hall.
Martin, D. W. (2004). Doing psychology experi-
ments (6th ed.). Wadsworth.
McDonald, S.-K., Keesler, V. A., Kauffman, N.
J., & Schneider, B. (2006). Scaling-up exemplary
interventions. Educational Researcher, 35(3),
15–24. doi:10.3102/0013189X035003015
Medrich, E. A., Roizen, J., Rubin, V., & Buckley,
S. (1982). The serious business of growing up.
Berkeley, CA: University of California Press.
Motiff, R. (2007). Forecasting the effects of
scaling up social programs. In Schneider, B.,
& McDonald, S. (Eds.), Scale-up in education:
Ideas in principle (pp. 173–186). Lanham, MD:
Rowman & Littlefield.
National Academy of Sciences. (2010). Rising
above the gathering storm, revisited: Rapidly
approaching category 5. 2005 “Rising Above the
Gathering Storm” Committee. Washington, DC:
National Academies Press. Retrieved from http://
www.nap.edu/catalog/12999.html.
National Research Council. (2009). Learning sci-
ence in informal environments: People, places, and
pursuits (P. Bell, B. Lewenstein, A. W. Shouse,
& M. A. Feder, Eds). Committee on Learning
Science in Informal Environments. Board on Sci-
ence Education, Center for Education. Division
of Behavioral and Social Sciences and Education.
Washington, DC: The National Academies Press.
National Research Council. (2011). A framework
for K-12 science education: Practices, crosscut-
ting concepts, and core ideas. Committee on a
Conceptual Framework for New K-12 Science
Education Standards. Board on Science Educa-
tion, Division of Behavioral and Social Sciences
and Education. Washington, DC: The National
Academies Press.
National Science Board. (2007). Science, tech-
nology, engineering, and mathematics (STEM)
education issues and legislative options. In Nata,
R. (Ed.), Progress in education (Vol. 14, pp.
161–189). Washington, DC.
National Science Board. (2009). National Science
Board STEM education recommendations for the
President–Elect Obama administration. NSB-
0901. Retrieved on March 3, 2011, from http://
www.nsf.gov/nsb/publications/2009/01_10_
stem_rec_obama.pdf.
Nugent, G., Barker, B., Grandgenett, N., & Adam-
chuk, V. (2010). Impact of robotics and geospatial
technology interventions on youth STEM learning
and attitudes. Journal of Research on Technology
in Education, 42(4), 391–408.
Organization for Economic Cooperation and
Development. (2010). Pisa 2009 results: What
students know and can do - Student performance
in reading, matchematics and science. Retrieved
on August 11, 2011, from http://www.oecd.org/
edu/pisa/2009
Partnership for 21st Century Skills. (2008). 21st
century skills education and competitiveness
guide. Retrieved from http://www.p21.org/
documents/21st_century_skills_education_and_
competitiveness_guide.pdf
Person, G., & Young, A. (2002). Technically
speaking, why all Americans need to know more
about technology. Washington, DC: National
Academy Press. Retrieved from http://www.nap.
edu/openbook.php?isbn=0309082625
Sosniak, L. (2001). The 9% challenge: Education
in school and society. Teachers College Record,
103, 15.
Thompson, T., Heer, D., Brown, S., Traylor, R., &
Fiez, T. S. (2004). Educational design, evaluation,
and development of platforms for learning. Paper
presented at the ASEE/IEEE Frontiers in Educa-
tion Conference, Savanna, Georgia, October, 2004.
1383
 Learning Geospatial Concepts as Part of a Non-Formal Education Robotics Experience

U.S. Department of Education. (2007). Report of
the Academic Competitiveness Council. Washing-
ton, DC: Author.
Wellington, J. (1990). Formal and informal learn-
ing in science: The role of interactive science
centers. Physics Education, 25(5), 247–252.
doi:10.1088/0031-9120/25/5/307
Whitmeyer, S., Nicoletti, J., & De Paor, D.
(2010). The digital revolution in geologic map-
ping and presentation. GSA Today, 20(4/5), 4–10.
doi:10.1130/GSATG70A.1
Wiggins, G., & McTighe, J. (2005). Understand-
ing by design (2nd ed.). Upper Saddle River, NJ:
Merrill.
Woffinden, S., & Packham, J. (2001). Experiential
learning, just do it! The Agriculture Education
Magazine, 73(6), 8–9.
World Economic Forum. (2010). The global com-
petitiveness report 2009-2010. Retrieved from
http://www.weforum.org/pdf/GCR09/Report/
Countries/United%20States.pdf

This work was previously published in Robots in K-12 Education: A New Technology for Learning, edited by Bradley S. Barker,
Gwen Nugent, Neal Grandgenett and Viacheslav I. Adamchuk, pp. 284-300, copyright 2012 by Information Science Reference
(an imprint of IGI Global).

1384
 1385

Chapter 83
A Reflection on the Ph.D.
Program in Spatially
Integrated Social Science at
the University of Toledo
Bhuiyan Monwar Alam
The University of Toledo, USA

Jeanette Eckert
The University of Toledo, USA

Peter S. Lindquist
The University of Toledo, USA

ABSTRACT
The use of spatial analysis tools is on the rise in many academic fields and practical applications. These
tools enhance the ability to examine data from spatial perspectives. Though the study of place and space
has traditionally been the domain of the field of geography, growing numbers of researchers are turning
to these tools in the social sciences and beyond. The University of Toledo has established a unique Ph.D.
granting program to encompass the theories, tools, and applications of spatially integrated social sci-
ence. In the first couple of years of its inception the program has attracted students from different places
and diverse backgrounds. It is expected that the program will continue to thrive in attracting diverse
students, securing external grants, and positively impacting on the economy of Northwest Ohio. This
paper is a personal reflection of the views of the authors on the Ph.D. program in Spatially Integrated
Social Science at the University of Toledo two years after its inception in fall 2009. The views, by no
means, are of the University of Toledo, its SISS program, or any of the participating departments and
faculty members.

DOI: 10.4018/978-1-4666-2038-4.ch083

Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 A Reflection on the Ph.D. Program in Spatially Integrated Social Science at the University of Toledo
INTRODUCTION
The use of a variety of spatial analysis tools is
increasing in social science research. While social
problems and study areas span multiple disciplines,
the question of location has historically been the
domain of geographers. The prevalence of cutting-
edge technologies like Geographic Information
Systems (GIS) and Global Positioning Systems
(GPS) in the social sciences is facilitating the
study of the role of place in the society. Thus,
geography can be a unifying field for the social
sciences, wherein traditional data analysis can be
conducted within the context of place.
While each discipline has historically pos-
sessed its own set of analytic tools, the increasing
ability to measure processes in a spatiotemporal
context facilitates sharing of these new tools, a
general development that can then be applied to
local and unique conditions (Goodchild et al.,
2000). Human demographic data, for example, has
historically been presented in a series of tables.
However, Weeks (2004) argues that such data is
spatial in nature because it deals with character-
istics of human populations in specific regions. It
varies based on time, space, and scale. The recent
shift towards locational information in data sets
and displaying data in interactive maps in addition
to traditional tables and charts has brought to light
the importance and usefulness of spatial ways of
looking at data in the social sciences.
As GIS tools have become more powerful and
more widely available, they have been increasingly
used in the application side of the social sciences.
More and more it is being realized that these tools
are equally powerful for theoretical queries and
social research. As Sui (2004) asks, GIS is the
answer, but what is the question? Sui (2004) goes
on to argue that advances in geocomputation will
continue to bring social science disciplines into the
trend of spatializing previously non-spatial data.
Spatial analysis is both an inductive and de-
ductive tool, as it can reveal unforeseen patterns
and test existing theories about expected patterns
1386
(Goodchild & Janelle, 2004). Anselin (2006)
categorizes spatial analysis techniques into three
general groups: exploratory spatial data analysis
(searching for patterns), visualization (methods
of showing the patterns), and spatial modeling
(methods of explaining and predicting patterns).
Recently, the development of the GeoDa software
has created a program capable of both advanced
spatial statistics and mapping (Anselin et al.,
2006).
WHY SPATIALLY INTEGRATED
SOCIAL SCIENCE?
As much as GIS has revolutionized many disci-
plines and practices, including the social sciences,
GIS itself does not fully account for or measure
the complexities and relationships inherent in
spatial data (Páez & Scott, 2004). Páez and Scott
(2004) argue that current GIS software, even with
the recent inclusion of spatial modeling, do not
adequately account for heterogeneity, interdepen-
dence or spatial association, and thus are not as
accurate or useful as it could be if more advanced
spatial statistics were used. Logan et al. (2010)
argue that GIS gives way to more complex spatial
analysis tools when patterns on a map lead to ad-
ditional questions that simple visualization cannot
answer. For example, a traditional chloropleth
map made in GIS does not take into account the
outliers or an uneven population. The longstanding
method of collecting data in tables and translating
it to maps only scratches the surface of displaying
and analyzing spatial data.
The last decade has seen the development of
a cohesive shift towards spatial ways of thinking
about social data and a demand for tools that facili-
tate this approach (Voss, 2007). This area of study
is increasingly referred to as Spatially Integrated
Social Science (SISS). A few key programs have
progressed SISS techniques, including the Uni-
versity of California at Santa Barbara (UCSB),
which is home to many centers and programs
A Reflection on the Ph.D. Program in Spatially Integrated Social Science at the University of Toledo
related to the field, perhaps most notably the
Center for Spatially Integrated Social Sciences
(CSISS, 2011). CSISS focuses on integrating the
social sciences via the context of place, and also
holds workshops for students, researchers, and
instructors (CSISS, 2011). Also at UCSB is the
National Center for Geographic Information and
Analysis and the UCSB Spatial Studies Center,
making the University a vital hub of spatial infor-
mation research. CSISS reports that the majority
of participants in their offered workshops, for
which they receive far more applicants than they
can accept, are from the field of geography, though
the ratio is decreasing, indicating greater interest
from other disciplines (CSISS, 2003). CSISS also
produces sample syllabi to be used to teach spatial
analysis in a variety of disciplines.
The Pennsylvania State University (PSU)
has a strong GIS component as well, hosting the
multidisciplinary Geospatial Information Systems
Council. PSU’s Population Research Institute
(PRI) focuses on demography in the context of
a variety of disciplines, and conducts statistical
and GIS analyses of population data. The PRI
also offers an interdisciplinary dual Ph.D. de-
gree in demography (http://www.pop.psu.edu/).
PSU’s partnership with CSISS has encouraged
participation in workshops on spatial analysis
by demography students and faculty, helping to
further the integration of the social sciences, GIS,
and spatial statistics.
The University of Illinois at Urbana-Cham-
paign has established the University Consortium
of Geographic Information Science, another
prominent GIS program to promote the interdis-
ciplinary use of geospatial techniques. The GeoDa
Center for Spatial Analysis and Computation at
Arizona State University is the current home of
the GeoDa software (http://geodacenter.asu.edu/).
Another spatial analysis software tool is CrimeStat,
developed at the Inter-University Consortium for
Political and Social Research at the University
of Michigan. CrimeStat is in use for practical
application by police departments as well as by
criminal justice and social science researchers
(Inter-University Consortium for Political and
Social Research at the University of Michigan,
2011). This is a short sampling of some of the
centers that have created spatial analysis tools
and are focusing on an interdisciplinary approach
to furthering spatial research. Many universities
across the United States offer classes, certificates,
and degrees in GIS. The SISS program at the
University of Toledo establishes a Ph.D. degree
that focuses on the theories, tools, and approaches
of advanced spatial analysis.
THE SISS PROGRAM AT THE
UNIVERSITY OF TOLEDO
The SISS program is the only Ph.D. granting pro-
gram of this nature in the Midwest. The Master’s
Certificate Program in GIS at the University of
Toledo is housed within the Department of Ge-
ography and Planning, however students from
several disciplines, ranging from public health
to civil engineering, continuously enroll in GIS
classes, demonstrating the wide application of
this technology. Contrarily, the SISS program is
an interdisciplinary program incorporating the de-
partments of Geography and Planning, Economics,
Political Science and Public Administration, and
Sociology and Anthropology (Lindquist, 2009).
The SISS program at the University of Toledo
incorporates many of the approaches outlined by
CSISS. It is centered on spatial analysis tools, theo-
ries, and problems as well as spatial information
processing technologies, including GIS, remote
sensing, and digital image analysis (SISS Ph.D.
Graduate Program Handbook, 2009).
The Department of Geography and Planning at
the University of Toledo was already home to the
Center for Geographic Information Sciences and
Applied Geographics (GISAG), a well-equipped
research center which conducts much of the
University of Toledo’s GIS work (Czajkowski et
al., 2003). Through the GISAG, there have been
1387
 A Reflection on the Ph.D. Program in Spatially Integrated Social Science at the University of Toledo
many joint projects between the departments of
Geography and Planning, Economics, Political
Science, Environmental Sciences, Business Ad-
ministration, and Civil Engineering. Thus, the
Ph.D. program in SISS is a natural extension of
this collaboration and the increasing applicability
of spatial analysis (SISS, 2011). It is expected that
this degree program will solidify such research
focus, attracting faculty and students of diverse
background, and strengthen the university’s role
as a research institution.
The original proposal for the SISS program
was submitted to the Ohio Board of Regents
in 2007, and the program was approved by the
Board of Regents and the University of Toledo’s
Board of Trustees in September 2008 (Lindquist,
2009). The first and second classes of four and
five students entered into the program in fall 2009
and fall 2010. Although the application process
for admission is on the move, it is anticipated that
the number of students who would enroll into the
program in fall 2011 will be stable compared to
last two years. As such one can be optimistic about
the sustainability of the program. The admitted
students come from a diverse background: geog-
raphy, planning, health education, engineering,
sociology, economics, and such. The applicants
have to meet three pre-requisites for entering
into the program. The first of those is that they
have to have a master’s or equivalent in a Social
Science discipline. The second and third pre-
requites include prior successful completion of two
courses in GIS and one course in graduate-level
multivariate statistics. A typical student would
successfully complete six core courses (18 credit
hours), three advanced seminar courses (nine
credit hours) and three elective courses (nine
credit hours) to fulfill the requirement of 36 credit
hours of coursework. The core courses consist of
Spatial Statistics, Geographical Information Sci-
ence in SISS, Foundations of SISS, SISS Theory,
Advanced Spatial Data Analysis, and Research
Design. The advanced seminar courses, on the
other hand, include Geo-Computation, Advanced
1388
Qualitative Analysis in SISS, Policy Evaluation
and SISS, Space and Society: Critical Theory in
SISS, Discrete Choice Spatial Process Modeling,
Advanced Modeling Methods and Techniques in
SISS, Spatial Perspectives on the Environment,
Spatial Transport Modeling and Planning, Directed
Readings in SISS, Seminar in Special Topics, and
Doctoral Dissertation. The elective courses can be
taken from any of the participating departments. In
addition, she/he has to complete 24 credit hours of
dissertation research and successfully defend the
research as the requirements to receive the Ph.D.
degree (Lindquist, 2009; SISS, 2011).
Currently the core instructors are comprised
of nine faculty members representing the par-
ticipating departments. Because the core faculty
members have diverse background and expertise,
and that the admitted students come with similar
diverse backgrounds, the students have wide open
options to expertise in different subfields of SISS:
transportation, housing, environment, economics,
public health, demography, and such. The list of
core faculty members and their expertise are avail-
able from the SISS program website (SISS, 2011).
EXPECTED CONTRIBUTION TO
NORTHWEST OHIO’S ECONOMY
Northwest Ohio is, like many older industrial
areas, in a period of self-reinvention. The City of
Toledo has a strong legacy in manufacturing and
supplying glass and parts to Detroit auto industry.
The region continues to shift away from traditional
manufacturing due to decline in that industry, but
is hopeful to make use of the skilled labor pool
to focus on high tech manufacturing and alterna-
tive energies. The University of Toledo is very
involved in this transition, and plays a vital role
in the regional economy. As such, the creation
of the Ph.D. program in SISS at the University
of Toledo is expected to contribute to this by at-
tracting a unique group of students, faculty, and
collaborators. The SISS program is also expected
A Reflection on the Ph.D. Program in Spatially Integrated Social Science at the University of Toledo

to contribute to the local economy by working as
a research center assisting local organizations,
governments, and businesses.
One specific area where the SISS program ex-
pects to have a significant impact is in the area of
transportation research. The Greater Toledo region
taking several unique variables into account, as
well as to be able to measure and understand the
relationships and connections of different places
and scales. There are several reasons, therefore,
to be optimistic that the Ph.D. program in SISS
at the University of Toledo will