indd 1 8/25/09 2:40:31 PM
This page intentionally left blank This page intentionally left blank
edited by
Young C Kim
California State University, Los Angeles, USA
HANDBOOK OF
COASTAL AND
OCEAN ENGINEERING
NE W J E RSE Y • L ONDON • SI NGAP ORE • BE I J I NG • SHANGHAI • HONG KONG • TAI P E I • CHE NNAI
World Scientifc
6914.tp.indd 2 8/25/09 2:40:35 PM
British Library CataloguinginPublication Data
A catalogue record for this book is available from the British Library.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center,
Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from
the publisher.
ISBN13 9789812819291
ISBN10 9812819290
Typeset by Stallion Press
Email: enquiries@stallionpress.com
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or
mechanical, including photocopying, recording or any information storage and retrieval system now known or to
be invented, without written permission from the Publisher.
Copyright © 2010 by World Scientific Publishing Co. Pte. Ltd.
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
Printed in Singapore.
HANDBOOK OF COASTAL AND OCEAN ENGINEERING
YHwa  Hdbk of Coastal & Ocean.pmd 10/21/2009, 6:30 PM 1
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Preface
Although coastal and ocean engineering is a very ancient ﬁeld with the construction
of Port Aur near the mouth of the Nile in 3,000 BC, signiﬁcant advances in this
ﬁeld have been made in the last several decades. The rise of interest in this ﬁeld can
be seen from the number of attendees by academics and practitioners in interna
tional conferences. The ﬁrst International Conference on Coastal Engineering was
held in Long Beach, California in 1950 with less than 100 people. When the same
conference was held in San Diego, California, in 2006, over 1000 delegates attended.
In the last several decades, the world has seen signiﬁcant coastal and ocean engi
neering projects, one of which is the Delta Project in the Netherlands. This project
was designed to shorten and strengthen the total length of coast and dykes washed
by the sea by closing oﬀ the sea arms in the Delta region. Other noteworthy coastal
engineering projects include the Kansai Airport Project in Japan and, in recent
years, the construction of mobile barriers at inlets to regulate tides in the Venice
Lagoon known as the Venice Project. Interest in coastal and ocean engineering
has arisen in recent years as humankind experiences coastal disasters that derive
from coastal storm, hurricane and coastal ﬂooding and seismic activities such as
tsunamis, and the impacts of climate change which result in sealevel rise. The
tsunami activity in Sumatra in December 2004 aﬀected countries throughout the
Indian Ocean and resulted in the loss of thousands of lives. Hurricane Katrina in
New Orleans also claimed many lives with property damage exceeding $63 billion.
Global warming and sealevel rise will aﬀect shoreline retreats, inundate low coastal
areas, damage coastal structures, and accelerate beach erosion. The need for better
understanding of our coastal and ocean environment has risen considerably in
recent years.
This handbook contains a comprehensive compilation of topics that are
the forefronts of many technical advances in ocean waves, coastal and ocean
engineering. It represents the most comprehensive reference available on coastal
and ocean engineering to date, and it also provides the most uptodate technical
advances and latest research ﬁndings on coastal and ocean engineering. More than
70 internationally recognized authorities in the ﬁeld of coastal and ocean engineering
contributed papers on their areas of expertise to this handbook. These interna
tional luminaries are from highly respected universities and renowned research and
consulting organizations from all over the world.
v
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
vi Preface
This handbook provides a comprehensive overview of shallowwater waves,
waterlevel ﬂuctuations, coastal and oﬀshore structures, ports and harbors, coastal
sediment processes, environmental problems, sustainable coastal development,
coastal hazards, physical modeling, and coastal engineering practice and education.
This book is an essential source of reference for professionals and researchers in
the areas of coastal engineering, ocean engineering, oceanography, meteorology, and
civil engineering, and as a text for graduate students in these ﬁelds. This handbook
will be of immediate, practical use to coastal, ocean, civil, geotechnical, and struc
tural engineers, and coastal planners and managers as well as marine biologists and
oceanographers. It will also be an excellent source book for educational and teaching
purposes, and would be a good reference book for any technical library.
I would like to express my indebtedness to those who guided me and supported
me as a mentor and a colleague throughout my professional life. They are:
Professor Robert L. Wiegel, University of California, Berkeley
Professor Joe W. Johnson, University of California Berkeley
Professor Robert G. Dean, University of Florida
Professor Fredric Raichlen, California Institute of Technology
Professor Raymond C. Binder, University of Southern California
Professor Shoshichiro Nagai, Osaka City University
Dr Basil Wilson, Science Engineering Associates
Dr Lars Skjelbreia, Science Engineering Associates
Dr Bernard LeMehaute, University of Miami
Professor Richard Silvester, University of Western Australia
Mr Orville T. Magoon, Coastal Zone Foundation
Professor Billy L. Edge, Texas A&M University
Professor Michael E. McCormick, US Naval Academy
Professor Yoshimi Goda, Yokohama National University and ECOH Corporation
Professor Philip L.F. Liu, Cornell University
Professor Forrest M. Holly, The University of Iowa
Dr Etienne Mansard, National Research Council, Canada
Professor J. Richard Weggel, Drexel University
Mr Ronald M. Noble, Noble Consultants, Inc.
I also wish to express my indebtedness to those who nurtured me from my early
teen years and changed my course of life. They are:
Dr Helen Miller Bailey, East Los Angeles College
Mr H. Karl Bouvier, Jet Propulsion Laboratory
I extend my gratitude to my wife, Janet, for her constant support, encouragement,
patience, and understanding while I was undertaking this task and to my daughter,
Susan Calix, for proofreading some of the materials.
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Preface vii
Finally, I wish to express my deep appreciation to Ms Kimberley Chua of World
Scientiﬁc Publishing Company who gave me invaluable support and encouragement
from the inception of this handbook to its realization.
Young C. Kim
Los Angeles, California
January 2008
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
This page intentionally left blank This page intentionally left blank
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contents
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Section 1: ShallowWater Waves
1. Wave Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
R. G. Dean and T. L. Walton
2. Wavemaker Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
R. T. Hudspeth and R. B. Guenther
3. Analyses by the Melnikov Method of Damped Parametrically
Excited Cross Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
R. B. Guenther and R. T. Hudspeth
4. Random Wave Breaking and Nonlinearity Evolution Across
the Surf Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Y. Goda
5. Aeration and Bubbles in the Surf Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
N. Mori, S. Kakuno and D. T. Cox
6. Freak Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
N. Mori
7. ShortTerm Wave Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
A. Kimura
Section 2: WaterLevel Fluctuations
8. Generation and Prediction of Seiches in Rotterdam Harbor Basins . . . . . 179
M. P. C. de Jong and J. A. Battjes
9. Seiches and Harbor Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
A. B. Rabinovich
ix
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
x Contents
10. Finite Diﬀerence Model for Practical Simulation of
Distant Tsunamis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
S. B. Yoon
Section 3: Coastal Structures
11. TsunamiInduced Forces on Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
I. Nistor, D. Palermo, Y. Nouri, T. Murty and M. Saatcioglu
12. Nonconventional Wave Damping Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
H. Oumeraci
13. Wave Interaction with Breakwaters Including Perforated Walls . . . . . . . . 317
K.D. Suh
14. Prediction of Overtopping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
J. van der Meer, T. Pullen, W. Allsop, T. Bruce,
H. Sch¨ uttrumpf and A. Kortenhaus
15. Wave RunUp and Wave Overtopping at Armored Rubble
Slopes and Mounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
H. Sch¨ uttrumpf, J. van der Meer, A. Kortenhaus,
T. Bruce and L. Franco
16. Wave Overtopping at Vertical and Steep Structures . . . . . . . . . . . . . . . . . . . 411
T. Bruce, J. van der Meer, T. Pullen and W. Allsop
17. Surf Parameters for the Design of Coastal Structures . . . . . . . . . . . . . . . . . . 441
D. H. Yoo
18. Development of Caisson Breakwater Design Based on
Failure Experiences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
S. Takahashi
19. Design of Alternative Revetments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
K. Pilarczyk
20. Remarks on Coastal Stabilization and Alternative Solutions . . . . . . . . . . . 521
K. Pilarczyk
21. Geotextile Sand Containers for Shore Protection. . . . . . . . . . . . . . . . . . . . . . . 553
H. Oumeraci and J. Recio
22. Low Crested Breakwaters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 601
A. Lamberti and B. Zanuttigh
23. Hydrodynamic Behavior of Net Cages in the Open Sea . . . . . . . . . . . . . . . . 633
Y.C Li
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contents xi
Section 4: Oﬀshore Structures
24. State of Oﬀshore Structure Development and Design Challenges . . . . . . . 667
S. Chakrabarti
Section 5: Ports and Harbors
25. Computer Modeling for Harbor Planning and Design . . . . . . . . . . . . . . . . . . 695
J.J. Lee and X. Xing
26. Prediction of Squat for Underkeel Clearance . . . . . . . . . . . . . . . . . . . . . . . . . . . 723
M. J. Briggs, M. Vantorre, K. Uliczka and P. Debaillon
Section 6: Coastal Sediment Processes
27. WaveInduced Resuspension of Fine Sediment . . . . . . . . . . . . . . . . . . . . . . . . . 775
M. Jain and A. J. Metha
28. Suspended Sand and Bedload Transport on Beaches . . . . . . . . . . . . . . . . . . . 807
N. Kobayashi, A. Payo and B. D. Johnson
29. HeadlandBay Beaches for Recreation and Shore Protection . . . . . . . . . . . 825
J. R.C. Hsu, M. M.J. Yu, F.C. Lee and R. Silvester
30. Beach Nourishment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843
R. G. Dean and J. D. Rosati
31. Engineering of Tidal Inlets and Morphologic Consequences . . . . . . . . . . . . 867
N. C. Kraus
Section 7: Environmental Problems
32. Water and Nutrients Flow in the Enclosed Bays . . . . . . . . . . . . . . . . . . . . . . . 901
Y. Koibuchi and M. Isobe
Section 8: Sustainable Coastal Development
33. Socioeconomic and Environmental Risk in Coastal
and Ocean Engineering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 923
M. A. Losada, A. Baquerizo, M. OrtegaS´ anchez,
J. M. Santiago and E. S´anchezBadorrey
34. Utilization of the Coastal Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 953
H.H. Hwung
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
xii Contents
Section 9: Coastal Hazards
35. Ocean Wave Climates: Trends and Variations Due to
Earth’s Changing Climate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 971
P. D. Komar, J. C. Allan and P. Ruggiero
36. Sea Level Rise: Major Implications to Coastal Engineering
and Coastal Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 997
L. Ewing
37. Sea Level Rise and Coastal Erosion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023
M. J. F. Stive, R. Ranasinghe and P. J. Cowell
38. Coastal Flooding: Analysis and Assessment of Risk. . . . . . . . . . . . . . . . . . . . 1039
P. Prinos and P. Galiatsatou
Section 10: Physical Modeling
39. Physical Modeling of Tsunami Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1073
M. J. Briggs, H. Yeh and D. T. Cox
40. Laboratory Simulation of Waves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1107
E. P. D. Mansard and M. D. Miles
Section 11: Coastal Engineering Practice and Education
41. Perspective on Coastal Engineering Practice and Education . . . . . . . . . . . 1135
J. W. Kamphuis
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contributors
Jonathan C. Allan
Coastal Field Oﬃce
Oregon Department of Geology and Mineral Industries
Newport, Oregon
jonathan.allan@dogami.state.or.us
William Allsop
Technical Director
HR Wallingford
Wallingford, UK
w.allsop@hrwallingford.co.uk
Elena Sanchez Badorrey
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
elenasb@ugr.es
Asuncion Baquerizo
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
Jurjen A. Battjes
Emeritus Professor
Environmental Fluid Mechanics Section
Delft University of Technology
Delft, The Netherlands
j.a.battjes@tudelft.nl
Michael J. Briggs
Research Hydraulic Engineer
Coastal and Hydraulics Laboratory
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi
michael.j.briggs@erdc.usace.army.mil
xiii
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
xiv Contributors
Tom Bruce
School of Engineering and Electronics
University of Edinburgh
Edinburgh, UK
tom.bruce@ed.ac.uk
Subrata Chakrabarti
Joint Professor, Civil and Mechanical Engineering
University of Illinois at Chicago
Chicago, Illinois
chakrab@aol.com
Peter J. Cowell
Associate Professor
School of Geosciences
Institute of Marine Science
University of Sydney
Sydney, Australia
Daniel T. Cox
Professor
School of Civil and Construction Engineering
Oregon State University
Corvallis, Oregon
dtc@oregonstate.edu
Robert G. Dean
Graduate Research Professor of Coastal Engineering, Emeritus
Department of Civil and Coastal Engineering
University of Florida
Gainesville, Florida
dean@coastal.uﬂ.edu
Pierre Debaillon
Research Hydraulic Engineer
Centre d’Etudes Techniques Maritimes Et Fluviales (CETMEF)
Compiegne, France
pierre.debaillon@equipement.gouv.fr
Martijn P. C. de Jong
Formerly at Environmental Fluid Mechanics Section
Delft University of Technology
Presently at Delft Hydraulics
Delft, The Netherlands
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contributors xv
Lesley Ewing
Senior Coastal Engineer
California Coastal Commission
San Francisco, California
lewing@coastal.ca.gov
Leopoldo Franco
Professor of Coastal Engineering
Department of Civil Engineering
University of Rome 3
Rome, Italy
leopoldo.franco@uniroma3.it
Panagiota Galiatsatou
Research Associate
Department of Civil Engineering
Aristotle University of Thessaloniki
Thessaloniki, Greece
pgaliats@civil.auth.gr
Yoshimi Goda
Professor Emeritus
Yokohama National University
Adviser to ECHO Corporation
Tokyo, Japan
goda@ecoh.co.jp
Ronald B. Guenther
Professor Emeritus
Department of Mathematics
Oregon State University
Corvallis, Oregon
guenther@math.orst.edu
John RongChung Hsu
Professor
Department of Marine Environment and Engineering
National Sun Yatsen University
Kaohsiung, Taiwan
Honorary Research Fellow
School of Civil and Resource Engineering
University of Western Australia
Nedland, Australia
jrchsu@mail.nsysu.edu.tw
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
xvi Contributors
Robert T. Hudspeth
Professor and Director, Emeritus
Coastal and Ocean Engineering Program
Oregon State University
Corvallis, Oregon
robert.hudspeth@orst.edu
HwungHweng Hwung
Professor of Hydraulic and Ocean Engineering
Director of Tainan Hydraulics Laboratory
Department of Hydraulic and Ocean Engineering
National Cheng Kung University
Tainan, Taiwan
hhhwung@mail.ncku.edu.tw
Masahiko Isobe
Professor and Special Adviser to the President
Department of Sociocultural Environmental Studies
Graduate School of Frontier Sciences
The University of Tokyo
Chiba, Japan
isobe@k.utokyo.ac.jp
Mamta Jain
Coastal Engineer
Halcrow Inc.
Tampa, Florida
mjain@halcrow.com
Bradley D. Johnson
Coastal and Hydraulics Laboratory
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi
Shohachi Kakuno
Professor and Vice President
Department of Civil Engineering
Osaka City University
Osaka, Japan
kakuno@ado.osakacu.ac.jp
J. William Kamphuis
Professor of Civil Engineering, Emeritus
Department of Civil Engineering
Queen’s University
Kingston, Ontario, Canada
kamphuis@civil.queensu.ca
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contributors xvii
Akira Kimura
Professor
Department of Social Systems Engineering
Tottori University
Tottori, Japan
kimura@sse.tottoriu.ac.jp
Nobuhisa Kobayashi
Professor and Director
Center for Applied Coastal Research
University of Delaware
Newark, Delaware
nk@coastal.udel.edu
Yukio Koibuchi
Assistant Professor
Department of Sociocultural Environmental Studies
Graduate School of Frontier Sciences
The University of Tokyo
Chiba, Japan
koi@k.utokyo.ac.jp
Paul D. Komar
Professor of Oceanography
College of Oceanic and Atmospheric Sciences
Oregon State University
Corvallis, Oregon
pkoma@coas.oregonstate.edu
Andreas Kortenhaus
LeichtweissInstitute for Hydraulics
Technical University of Braunschweig
Braunschweig, Germany
kortenhaus@tubs.de
Nicholas C. Kraus
Senior Scientist
Coastal and Hydraulics Laboratory
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi
nicholas.c.kraus@erdc.usace.army.mil
Alberto Lamberti
Professor
Department of Civil Engineering
University of Bologna
Bologna, Italy
alberto.lamberti@unibo.it
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
xviii Contributors
FangChun Lee
Department of Marine Environment and Engineering
National Sun Yatsen University
Kaohsiung, Taiwan
JiinJen Lee
Professor of Civil and Environmental Engineering
Sonny Astani Department of Civil and
Environmental Engineering
University of Southern California
Los Angeles, California
jjlee@usc.edu
YuCheng Li
Professor
School of Civil Engineering
Dalian University of Technology
Dalian, China
liyuch@dlut.edu.cn
Miguel A. Losada
Professor
Research Group on Environmental Flux Dynamics
CEAMA — Universidad de Granada
Granada, Spain
mlosada@ugr.es
Etienne P. D. Mansard
Executive Director
Canadian Hydraulics Centre
National Research Council Canada
Ottawa, Ontario, Canada
etienne.mansard@nrccnrc.gc.ca
Ashish J. Mehta
Professor of Coastal Engineering
Department of Civil and Coastal Engineering
University of Florida
Gainesville, Florida
mehta@coastal.uﬂ.edu
Michael D. Miles
Canadian Hydraulics Centre
National Research Council Canada
Ottawa, Ontario, Canada
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contributors xix
Nobuhito Mori
Associate Professor
Disaster Prevention Research Institute
Kyoto University
Kyoto, Japan
nobuhito.mori@hy5.ecs.kyotou.ac.jp
Tad S. Murty
Adjunct Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
tadmurty@gmail.com
Ioan Nistor
Assistant Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
inistor@uottawa.ca
Younes Nouri
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
Miquel Ortega
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
miguelos@ugr.es
Hocine Oumeraci
University Professor
LeichtweissInstitute for Hydraulic Engineering
and Water Resources
Technical University of Braunschweig
Braunschweig, Germany
h.oumeraci@tubs.de
Dan Palermo
Assistant Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
palermo@eng.uottawa.ca
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
xx Contributors
Andres Payo
Graduate School of Science and Technology
University of Kumamoto
Kumamoto, Japan
Krystian W. Pilarczyk
(Former) Manager, Research and Development
Hydraulic Engineering Institute
Rykswaterstaat
Delft, The Netherlands
HYDROpil Consultancy
Zoetermeer, The Netherlands
krystian.pilarczyk@gmail.com
Panayotis Prinos
Professor of Hydraulic Engineering
Department of Civil Engineering
Aristotle University of Thessaloniki
Thessaloniki, Greece
prinosp@civil.auth.gr
Tim Pullen
Senior Engineer
HR Wallingford
Wallingford, UK
tap@hrwallingford.co.uk
Alexander B. Rabinovich
P.P. Shirshov Institute of Oceanology
Russian Academy of Sciences
Moscow, Russia
Department of Fisheries and Oceans
Institute of Ocean Sciences
Sidney, B.C., Canada
abr@iki.rssi.ru.
Roshanka Ranasinghe
Associate Professor
UNESCOIHE/Delft University of Technology
Delft, The Netherlands
r.ranasinghe@unescoihe.org
Juan Recio
LeichweissInstitute for Hydraulic Engineering
and Water Resources
Technical University of Braunschweig
Braunschweig, Germany
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contributors xxi
Julie D. Rosati
Research Hydraulic Engineer
Coastal and Hydraulics Laboratory
U.S. Army Corps of Engineers
Mobile, Alabama
julie.d.rosati@erdc.usace.army.mil
Peter Ruggiero
Assistant Professor
Department of Geosciences
Oregon State University
Corvallis, Oregon
ruggierp@science.oregonstate.edu
Murat Saatcioglu
Professor
Department of Civil Engineering
University of Ottawa
Ottawa, Ontario, Canada
murat.saatcioglu@uottawa.ca
Juan M. Santiago
Associate Professor
CEAMA — Universidad de Granada
Granada, Spain
santi@ugr.es
Holger Schuttrumpf
Professor and Director
Institute of Hydraulic Engineering
and Water Resources Management
RWTH — Aachen University
Aaachen, Germany
schuettiumpf@iww.rwthaachen.de
Richard Silvester
Professor Emeritus
School of Civil and Resource Engineering
University of Western Australia
Nedland, Australia
rsilvest@bigpond.net.au
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
xxii Contributors
Marcel J. F. Stive
Professor and Director
Delft Water Research Centre
Department of Hydraulic Engineering
Delft University of Technology
Delft, The Netherlands
m.j.f.stive@tudelft.nl
KyungDuck Suh
Professor
Department of Civil and Environmental Engineering
Seoul National University
Seoul, Korea
kdsuh@snu.ac.kr
Shigeo Takahashi
Executive Researcher and Director
Tsunami Research Center
Port and Airport Research Institute
Yokosuka, Japan
takahashi s@pari.go.jp
Klemens Uliczka
Research Hydraulic Engineer
Federal Waterways Engineering and Research Institute (BAW)
Hamburg, Germany
klemens.uliczka@baw.de
Jentsje van der Meer
Principal
Van der Meer Consulting
Heerenveen, The Netherlands
jm@vandermeerconsulting.nl
Marc Vantorre
Professor
Division of Maritime Technology
Ghent University
Ghent, Belgium
marc.vantorre@ugent.be
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
Contributors xxiii
Todd L. Walton
Director
Beaches and Shore Resource Center
Florida State University
Tallahassee, Florida
twalton@mailer.fsu.edu
Xiuying Xing
Graduate Research Assistant
Sonny Astani Department of Civil and Environmental Engineering
University of Southern California
Los Angeles, California
Harry Yeh
Professor
School of Civil and Construction Engineering
Oregon State University
Corvallis, Oregon
harry@engr.orst.edu
Dong Hoon Yoo
Professor
Department of Civil Engineering
Ajou University
Suwon, Korea
dhyoo@ajou.ac.kr
Sung Bum Yoon
Professor
Department of Civil and Environmental Engineering
Hanyang University
Ansan, Korea
sbyoon@hanyang.ac.kr
Melissa MengJiuan Yu
Department of Marine Environment and Engineering
National Sun Yatsen University
Kaohsiung, Taiwan
Barbara Zanuttigh
Assistant Professor
Department of Civil Engineering
University of Bologna
Bologna, Italy
barbara.zanuttigh@mail.ing.unibo.it
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
This page intentionally left blank This page intentionally left blank
August 25, 2009 18:3 9.75in x 6.5in b684fm FA
The Editor
Young C. Kim, PhD, is currently a Professor of Civil Engineering, Emeritus
at California State University, Los Angeles. Other academic positions held by
him include a Visiting Scholar of Coastal Engineering at the University of Cali
fornia, Berkeley (1971); a NATO Senior Fellow in Science at the Delft University of
Technology in the Netherlands (1975); and a Visiting Scientist at the Osaka City
University for the National Science Foundations’ US–Japan Cooperative Science
Program (1976). For more than a decade, he served as Chair of the Department of
Civil Engineering and recently he was Associate Dean of the College of Engineering.
For his dedicated teaching and outstanding professional activities, he was awarded
the universitywide Outstanding Professor Award in 1994.
Dr Kim was a consultant to the US Naval Civil Engineering Laboratory in Port
Hueneme and became a resident consultant to the Science Engineering Associates
where he investigated wave forces on the HowardDoris platform structure, now
being placed in Ninian Field, North Sea.
Dr Kim is the past Chair of the Executive Committee of the Waterway, Port,
Coastal and Ocean Division of the American Society of Civil Engineering (ASCE).
Recently, he served as Chair of the Nominating Committee of the International
Association of Hydraulic Engineering and Research (IAHR). Since 1998, he served
on the International Board of Directors of the Paciﬁc Congress on Marine Science
and Technology (PACON). He currently serves as the President of PACON. Dr
Kim has been involved in organizing 10 national and international conferences, has
authored three books, and has published 52 technical papers in various engineering
journals.
xxv
This page intentionally left blank This page intentionally left blank
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Chapter 1
Wave Setup
Robert G. Dean
Department of Civil and Coastal Engineering
University of Florida, Gainesville, FL, USA
dean@coastal.uﬂ.edu
Todd L. Walton
Beaches and Shores Resource Center
Florida State University, Tallahassee, FL, USA
twalton@fsu.edu
Wave setup is the increase of water level within the surf zone due to the transfer of
waverelated momentum to the water column during wavebreaking. Wave setup
has been investigated theoretically and under laboratory and ﬁeld conditions,
and it includes both static and dynamic components. Engineering applications
include a signiﬁcant ﬂooding component due to severe storms and oscillating water
levels that can increase hazards to recreational beach goers and can contribute to
undesirable oscillations of both constructed and natural systems including harbors
and moored ships. This chapter provides a review of the knowledge regarding wave
setup and presents preliminary recommendations for design. It will be shown that
wave setup is not adequately understood quantitatively for engineering design
purposes.
1.1. Introduction
Wave setup was brought to the attention of coastal engineers and scientists in the
1960s (i.e., see Ref. 1, p. 245) after the initial theoretic developments of Longuet
Higgins
2
and LonguetHiggins and Stewart
3,4
along with limited ﬁeld observations
and laboratory studies supported the existence of wave setup, the magnitude of
which was observed to be in the order of 10–20% of the incident wave height. It
was noted in early ﬁeld observations that water levels on the beach were higher
than those recorded by a tide gauge at the end of a pier suggesting a wave setup
physically forced by wind waves and swell.
1
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
2 R. G. Dean and T. L. Walton
Fig. 1.1. Deﬁnition sketch. Energy and momentum are transferred from winds to waves in the
generating area. The waves convey energy and momentum to the surf zone where the waves break.
Upon breaking, the energy is dissipated and the momentum is transferred to the water column
resulting in longshore and onshore forces exerted on the water column.
Wave setup is the additional water level that is due to the transfer of
waverelated momentum to the water column during the wavebreaking process.
As waves approach the shoreline, they convey both energy and momentum in the
wave direction. Upon breaking, the wave energy is dissipated, as is evident from the
turbulence generated; however, momentum is never dissipated but rather is trans
ferred to the water column resulting in a slope of the water surface to balance the
onshore component of the ﬂux of momentum (see Fig. 1.1). If waves are irregular,
in addition to a steady wave setup, the setup includes a dynamic component that
oscillates with the wave group period and there may be a weak resonance within
the nearshore amplifying this oscillating component. These have been termed infra
gravity waves and are more dominant for narrow banded spectra both in frequency
and in directional spreading. The oscillatory component is denoted “dynamic wave
setup” in this chapter.
This chapter discusses the signiﬁcance of wave setup to coastal engineering
design, provides a review of the classical linear wave theory of wave setup, reviews
results from laboratory and ﬁeld studies, summarizes results and recommends pre
liminary design approaches for the static component. To provide a “look ahead,”
we will see that the phenomenon of wave setup is not yet adequately understood
for satisfactory engineering calculations and that the eﬀects of proﬁle slope are very
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 3
signiﬁcant. The interested reader is also referred to an earlier review article on wave
setup by Holman.
5
1.2. Engineering Signiﬁcance of Wave Setup
Wave setup (both static and dynamic components) is relevant to a number of engi
neering applications. The contributions of wave setup under extreme storm events
can be substantial, adding several feet to the elevated water levels. The interaction
of wave setup with vegetation diﬀers from wind surge and thus it is important to
diﬀerentiate the two components, for example in ascertaining the beneﬁts of wet
lands in reducing wave setup. Finally, the oscillating component of wave setup is
relevant to beach safety in some locations and to many natural and constructed
coastal systems that have the capability to resonate including harbors and moored
ships.
1.3. Terminology and Related Considerations
Standard terminology deﬁnes the water level in the absence of wave eﬀects as “still
water level,” whereas wave setup will cause a departure from the still water level
and this water level including the eﬀects of the waves is the “mean water level.”
As implied, the mean water level is determined as the average of the ﬂuctuating
water level over a suitable time frame usually taken as a number of multiples of
the short wave period, say the spectral peak. In considering wave setup, often the
location of interest is that of the maximum wave setup at the shoreline. This raises
the question of whether wave setup is deﬁned at elevations above the maximum
rundown, say on the beach face where the water is present over only a portion of
the wave period. Since wave setup is deﬁned as the mean water level, over what
period should the water surface be averaged on the beach face which is “wetted”
over only a portion of the wave period? If the time average is over only the portion
of the period that water is present, in the upper limit, the maximum setup will be
the maximum runup. For purposes here, wave setup will usually be deﬁned only for
conditions where water is present over a full wave period.
When calculating wave runup on a structure such as a levee or revetment, the
question arises whether it is appropriate to ﬁrst calculate wave setup and then add
the wave runup which is usually empirically based on model results. In the more
recent empirical results (e.g., the TAW method, see Ref. 6), the runup is expressed
as a proportion of the signiﬁcant wave height at the base of the steeper slope (e.g.,
at a revetment or levee). The wave runup determined in the model on which the
method was based generally included some wave setup (or setdown) seaward of the
toe of the slope and included wave setup landward of the toe of the slope. Thus, in
the application of interest, the most appropriate approach is to calculate and include
wave setup at the toe of the slope; however, recognizing that the measured landward
runup includes setup, no additional setup should be added explicitly landward of
the toe of the slope.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
4 R. G. Dean and T. L. Walton
1.4. A Brief Review of Wave Setup Mechanics
1.4.1. Static wave setup for monochromatic waves
LonguetHiggins
2
and LonguetHiggins and Stewart
3,4
were the ﬁrst to formalize the
notion of wave momentum ﬂux and its relationship to wave setup. The momentum
ﬂux, S
ij
, is a secondorder tensor given by
S
xx
= E
n(cos
2
θ + 1) −
1
2
,
S
yy
= E
n(sin
2
θ + 1) −
1
2
,
S
xy
= S
yx
=
E
2
sin 2θ,
(1.1)
where E is the wave energy density, n is the ratio of wave group velocity to wave
celerity and θ is the angle between the wave direction and the xaxis. The term
S
xy
reads “the ﬂux per unit width, in the xdirection, of the ycomponent of
momentum,” etc.
The steadystate equations of motion obtained by time averaging over the short
wave period are, including the eﬀects of wind stress and bottom friction:
∂ (η
wind
+η
wave
)
∂x
= −
1
ρg(h + η)
∂S
xx
∂x
+
∂S
xy
∂y
−τ
sx
+τ
bx
and
∂ (η
wind
+η
wave
)
∂y
= −
1
ρg(h +η)
∂S
yy
∂y
+
∂S
yx
∂x
−τ
sy
+τ
by
.
(1.2)
In the above, η
wind
is the surge component due to the wind stress, η
wave
is the wave
setup, ρ is the mass density of water, g is the gravitational constant, h is the local
water depth, τ
sx
and τ
bx
are the surface and bottom shear stresses, respectively,
and similarly for the ydirection. The coordinate direction, x is oriented shoreward
and a righthanded coordinate system is considered.
The most simple solution is for waves propagating directly shoreward (S
xy
= 0)
in which the surface and bottom stresses are considered negligible, and all variables
are considered uniform in the ydirection. The resulting equation is
∂η
wave
∂x
= −
1
ρg(h +η)
∂S
xx
∂x
. (1.3)
To proceed, we need to determine a boundary condition for η
wave
,
a
at the seaward
end of the surf zone. LonguetHiggins
7
has shown that in the absence of energy
dissipation, the following general relationship for η applies
η = C −
1
2g
(u
2
−w
2
)
η=0
, (1.4)
a
For purposes of convenience, hereafter the subscript on η
wave
will be omitted such that the wave
setup is simply denoted as η.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 5
where u
2
and w
2
represent the time averages of the square of the ﬁrstorder hori
zontal and vertical wave velocities evaluated at the mean water surface, respectively.
Equation (1.4) is a type of a Bernoulli equation for unsteady ﬂows which, when eval
uated at the break point and considering no wave setup in deep water to evaluate
the constant, C = 0, the setup is negative (setdown) and given by
η
b
= −
H
2
b
k
b
8 sinh 2k
b
h
b
, (1.5)
where H
b
is the breaking wave height and k
b
is the wave number at breaking. For
shallow water conditions and depth limited breaking (H
b
= κ(h
b
+ η
b
)), Eq. (1.5)
yields
η
b
= −
κH
b
16
. (1.6)
As an example, for a κ value of 0.78, the wave setdown is approximately 5% of the
breaking wave height.
With the seaward boundary condition now established, for the case of shallow
water wavebreaking and the consideration of depth limited breaking across the surf
zone, the wave setup is
η = −
κH
b
16
+
3κ
2
/8
(1 + (3κ
2
/8))
(h
b
−h) . (1.7)
It is noted that in the above equation, the bottom shear stress has been taken as
zero and that a shoreward directed bottom shear stress on the water column as
would occur due to undertow would increase the wave setup. As examples, the ratio
of wave setup to breaking height at the still water line (h = 0) and at the location
of maximum wave setup (η = −h) for a κ value of 0.78 are
F
0

κ=0.78
≡
η(h = 0, κ = 0.78)
H
b
=
5κ
16(1 + (3κ
2
/8))
κ=0.78
= 0.198 (1.8)
and
F
max

κ=0.78
≡
η(h = −η, κ = 0.78)
H
b
= F
0
1 +
3κ
2
8
κ=0.78
=
5κ
16
κ=0.78
= 0.244 .
(1.9)
It is seen that the wave setup is strongly dependent on the value of the breaking ratio
κ which will be shown to decrease with decreasing beach slope. Figure 1.2 presents
the ratios, F
0
and F
max
versus κ. It is useful to relate κ in an approximate manner to
beach slope. Although there is not a onetoone correspondence, Fig. 1.3 is based on
the Dally et al.
8
wavebreaking model and provides an approximate correspondence
between uniform proﬁle slope and the associated κ value. It is evident that the
Dally et al. model provides reasonable κ values for smaller beach slopes (say less
than about 0.06), but the κ values are too large for steeper slopes.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
6 R. G. Dean and T. L. Walton
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.1
0.2
0.3
0.4
Kappa
Fo
Fmax
F
o
a
n
d
F
m
a
x
Fig. 1.2. Values of F
0
and F
max
versus wavebreaking index, κ (kappa).
0.00 0.02 0.04 0.06 0.08 0.10
Profile Slope
0.0
0.5
1.0
1.5
2.0
K
a
p
p
a
Fig. 1.3. Relationship between proﬁle slope and κ (kappa) value. Based on Dally et al.
8
wave
breaking model.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 7
1.4.2. Eﬀects of wave nonlinearity
Wave nonlinearity depends on the following parameters: H/L
0
and h/L
0
. The
nonlinearity is exhibited in the wave proﬁle by peaked crests and ﬂatter troughs
and increases with wave height and shallow water. Somewhat surprisingly, the
momentum ﬂuxes in shallow water are less for nonlinear waves than for linear waves
of the same height. This is primarily because the momentum ﬂuxes are proportional
to wave energy (Eq. (1.1)) and the wave energy is proportional to the rootmean
square of the water surface displacement that is less for nonlinear waves with long
troughs and peaked wave crests. Figure 1.4 presents the ratio of nonlinear to linear
momentum ﬂuxes as determined by Stream Function wave theory.
9–11
The reason
that the quantities for nonlinear waves are greater in deep water than for linear
waves is that the nonlinear calculations extend up to the actual free surface whereas
the linear quantities only extend up to the mean free surface.
1.4.3. Role of wave directionality
Equation (1.1) demonstrates that for a given wave height, the maximum shoreward
ﬂux of onshore momentum occurs for normally incident waves (θ = 0
◦
). Thus as
expected for directional waves, the S
xx
term is reduced. However, this reduction is
relatively small as can be demonstrated by considering a breaking wave direction
of 30
◦
relative to a beach normal (this represents a reasonably large wave obliquity
10
3.000
10
2.000
10
1.000
10
0.000
10
1.000 2 3 4 5 6 2 3 4 5 6 2 3 4 5 5 2 3 4 5 6
h/L
o
0.0
0.2
0.4
0.6
0.8
1.0
1.2
R
a
t
i
o
o
f
N
o
n
l
i
n
e
a
r
t
o
L
i
n
e
a
r
M
o
m
e
n
t
u
m
F
l
u
x
H
/
H
b
=
0
.
2
5
H
/
H
b
=
0
.
7
5
H
/
H
b
=
1
.
0
H
/
H
b
=
0
.
5
0
Fig. 1.4. Ratio of nonlinear to linear wave momentum ﬂux, S
xx
, for forty stream function wave
combinations.
12
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
8 R. G. Dean and T. L. Walton
at breaking). The reduction in S
xx
for a wave of given height and for shallow water
conditions is 16.7%.
1.4.4. Eﬀects of vegetation
The eﬀects of vegetation have been shown to result in a reduced setup and, in some
cases, may cause a setdown.
12
For linear waves, vegetation protruding through the
water surface experiences a net drag force (quadratically related to velocity) on the
vegetation in the direction of wave propagation and, of course, there must be an
equal and opposing force exerted on the water column. This opposing force acting
on the water column partially counteracts the force due to momentum transfer
and thus reduces the wave setup (similar to an oﬀshore directed wind stress). For
linear waves and vegetation which is submerged during the entire wave passage, no
net vegetationrelated force exists on the water column and thus there is no eﬀect
on the wave setup. However, due to the character of nonlinear waves with higher
and shorter shoreward velocities under the wave crests, even if the vegetation is
fully submerged during the passage of the wave, a net drag force is induced on the
vegetation in the wave propagation direction again resulting in a reduction in the
wave setup and, for some cases, a wave setdown.
1.4.5. Dynamic wave setup
It is noted that theoretical formulations of the dynamic wave setup must include
the time dependent terms in the counterparts of Eq. (1.2). The dynamic wave setup
or “surf beat” was ﬁrst identiﬁed through ﬁeld observations and measurements by
Munk
13
and Tucker.
14
A number of theoretical treatments of dynamic wave setup
based on various hypotheses have been developed with each focusing on a diﬀerent
mechanism. These include Symonds et al.
15
(timevarying breakpoint), Symonds
and Bowen
16
(trapping of long waves by longshore bars), Schaﬀer and Svendsen
17
(reinforcement of incoming and reﬂected long waves), etc. Kostense
18
conducted
laboratory experiments to investigate the dynamic setup component and found that
the results were in qualitative agreement with the theory of Symonds et al.
15
We can apply the results for monochromatic waves to investigate the approx
imate dynamic wave setup for a simple irregular wave case. Consider a bichromatic
wave system with wave heights H
1
and H
2
(H
1
> H
2
) and a small frequency dif
ference between the two components. If the resulting wave group varies so slowly
that static conditions occur within the surf zone, Eq. (1.7) applies and is written as
η = FH
b
, (1.10)
where H
b
is the breaking wave height and F is a proportionality factor depending
on whether the referenced setup is at the still water shoreline or the maximum wave
setup (see Eqs. (1.8) and (1.9)). The maximum and minimum wave setup values are:
η
max
= F(H
1
+ H
2
) ,
η
min
= F(H
1
− H
2
) .
(1.11)
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 9
Table 1.1. Static and dynamic wave setup characteristics for a
biharmonic wave system.
H
2
/H
1
η
max
/FH
1
η
avg
/FH
1
(η
max
−η
avg
)/η
avg
0.2 1.2 1.01 0.19
0.4 1.4 1.04 0.35
0.6 1.6 1.09 0.47
0.8 1.8 1.16 0.55
1.0 2.0 1.27 0.57
It can be shown that the average wave setup depends on the ratio H
2
/H
1
as shown
in Table 1.1. The fourth column presents the ratios of the maximum dynamic wave
set amplitude to the average wave setup component.
In the case with H
2
= H
1
, the dynamic wave setup displacement from the mean
setup equals 57% of the average wave setup (Table 1.1, Column 4).
In the above, we have examined the dynamic wave setup for the case of a simple
bichromatic wave system in which the diﬀerence in frequencies of the two com
ponents was fairly small. For the case of a wave spectrum, the situation is much
more complex with, for the case of a narrow spectrum, the group envelope varying
according to the Rayleigh distribution. For the case of a wide spectrum, the dynamic
component is reduced considerably.
1.5. Laboratory and Field Measurements of Wave Setup
Having reviewed the theory of wave setup and its relationship to various factors,
the two sources available for evaluation are laboratory and ﬁeld data.
1.5.1. Laboratory experiments on wave setup
Many laboratory investigations of static and dynamic wave setup have been con
ducted. The results of an early laboratory investigation with monochromatic waves
by Bowen et al.
19
are shown in Fig. 1.5. For this study, the ratio of maximum wave
setdown and wave setup on the beach face to breaking wave height are −0.035 and
+0.316, respectively, compared with −0.049 and +0.244 on the beach face for a
κ value of 0.78. The eﬀect of beach slope has been noted earlier and the relatively
large beach slope of 0.082 in these experiments is undoubtedly a contributor to
the large setup value. Later, laboratory investigations have included examination
of irregular waves including measurements of water particle velocities and pressures
which form the basis of the S
xx
momentum ﬂux component.
Battjes
20
conducted one of the earliest laboratory studies of wave setup due to
irregular waves. Setup was measured through bottom mounted manometers and it
was found that the wave setup was less than predicted. It was hypothesized that
this diﬀerence was possibly due to air in the water column of the manometers.
The entire setup was shifted landward relative to the theoretical and this delay was
later attributed to a “roller” that is transported along with the wave crest region and
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
10 R. G. Dean and T. L. Walton
Fig. 1.5. Measured wave setup and setdown in the laboratory.
19
conveys wave energy and momentum landward prior to transfer of the momentum
to the water column and the associated wave setup.
21
Later, Stive and Wind
22
con
ducted a very detailed laboratory investigation in which they demonstrated the role
of wave nonlinearity. In this study, the momentum ﬂux components (velocities and
pressures) were measured to the degree possible and it was found that the calcu
lated wave setup based on nonlinear wave theories was in much better agreement
with measured wave setup than calculations based on linear wave theories. In these
comparisons, it was not necessary to introduce the roller concept.
The two laboratory studies reviewed above have focused on static setup and
it has been noted that irregular waves also produce dynamic wave setup. Hedges
and Mase
23
have presented an interesting reanalysis of earlier runup laboratory
measurements by Mase
24
in which irregular waves provided the forcing.
b
The planar
slopes represented in the data were: 1:5, 1:10, 1:20, and 1:30. Figure 1.6 presents
an example of the form in which the data were plotted where the horizontal axis is
b
Walton
25
was the ﬁrst to analyze the Mase data to extract the static wave setup.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 11
Fig. 1.6. Variation of nondimensional runup with Iribarren number.
23
the Iribarren number, ξ
0
deﬁned as
ξ
0
=
tan α
H/L
(1.12)
in which tan α is the proﬁle slope. The interpretation of Fig. 1.6 is that for a zero
slope (zero Iribarren number), there would be no short wave runup; therefore, the
intercept represents the sum of the static and dynamic components of wave runup.
Equations of the following form were ﬁt to plots of the type of Fig. 1.6:
R
char
H
1/3
=
S
char
H
1/3
+c
char
ξ
0
, (1.13)
where the subscript “char” refers to the percent associated with the variable; for
example, the 2% runup is deﬁned as R
2%
. It was found that both S
char
and c
char
were
Rayleigh distributed with S
1/3
and c
1/3
equal to 0.27 and 1.04, respectively, where
only the ﬁrst term represents wave setup and is of interest here. The results for S
char
can be interpreted in terms of the static and dynamic wave setup components. As an
example, S
mean
= 0.17 and S
2%
= 0.37. Thus, the 2% value of the nondimensional
dynamic setup deﬁned here as ∆S
2%
is
∆S
2%
=
η
dyn,2%
H
1/3
= (S
2%
−S
mean
) = (0.37 −0.17) = 0.20 . (1.14)
Thus, the mean setup at the still waterline is 17% of the signiﬁcant wave height
measured at the toe of the slope and the 2% dynamic component at the still waterline
is 20% of the signiﬁcant wave height at the toe of the slope or slightly larger than
the mean wave setup. These results are interesting and of reasonable magnitudes;
however, there are two problems with recommending them for universal application.
First, we know that the mean setup depends on the slope (through the κ dependency
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
12 R. G. Dean and T. L. Walton
as discussed earlier), and the second is that the oscillating wave setup component
should depend on the width of the input spectrum. Referring to Fig. 1.6, which is
one of several similar plots presented in the Hedges and Mase paper, since each plot
may include a mix of beach slopes, the slope dependency is not resolved in the S
mean
results which of course are derived from the yintercept of these graphs. Secondly,
these experiments were not designed to evaluate the eﬀect of spectral width and the
spectral characteristics included in the experiments are not known. However, it is of
interest to identify the “representative” κ value and beach slope associated with a
S
mean
= η
avg
/H
1/3
of 0.17. Referring to Fig. 1.2, we see that the associated κ value
is approximately 0.63 for F
0
. Based on the Dally et al.
8
breaking wave model, the
associated beach slope from Fig. 1.3 is 1:29 compared to the beach slopes in the
Mase experiments ranging from 1:30 to 1:5.
1.5.2. Field experiments on wave setup
The paragraphs below describe several ﬁeld experiments and observations of wave
setup.
An early study of wave setup comprised a pair of observations at an exposed
coastal site (Narragansett Pier, RI) and a calmer water site (Newport, RI) where,
at the latter, wave action was assumed not a factor and was found to show an
approximate 3 foot water level diﬀerence during the peak of the 1938 hurricane
storm surge.
26
In a second early ﬁeld experiment on wave setup at Fernandina Beach, Florida,
Dorrestein
27
placed transparent plastic tubes with lightweight ﬂoats to track the
water surface on the beach in the zone of wave setup. To obtain the setup records,
16 mm movie ﬁlm recorded the tracked surface of the ﬂoats. A ﬂoat type tide gage
on the end of a ﬁshing pier provided oﬀshore water level records. Through analysis
of the tide gage records and the beach placed setup gages, Dorrestein
27
evaluated
the setup (with respect to the end of the pier) and compared observational results to
existing setup theory. He found the measured setup in four of ﬁve experiments to be
larger than the computed setup. One shortcoming of Dorrestein’s work is that the
water level records were only 72 s in length and thus subject to considerable scatter
and large standard deviation as later noted by Holman and Sallenger.
28
Although
rationale was provided by Dorrestein
27
for possible diﬀerences between measured
and computed setup in this early experiment, large discrepancies between measured
and analytically or numerically computed setup still exist today.
A North Sea ﬁeld wave setup experiment was conducted on the Island of Sylt
by Hansen.
29
Utilizing a combination of ultrasonic wave gages and pressure sensor
wave gages out to a distance of 1280 m from shore (10 m depth), Hansen
29
found
good correspondence of data to an empirical expression provided by:
η = 0.3H
os
= 0.42H
orms
. (1.15)
Hansen also noted the maximum wave setup to be approximately 50% of the signif
icant breaking wave height. It is not clear as to the methodology utilized to obtain
η
max
in this ﬁeld experiment.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 13
A wave setup ﬁeld experiment was conducted as part of the Nearshore Sed
iment Transport Study at Torrey Pines Beach, San Diego, California by Guza and
Thornton.
30
The Torrey Pines Beach face was gently sloping (beach slope ≈0.02)
and the beach material was a moderately sorted ﬁne grain sediment (≈0.1 mm).
A dual wire resistance runup meter was used for the recording and estimation of
the wave setup. It should be noted that the measurements of the wave setup were
considered to be the average runup determined from wires placed approximately
3 cm above the beach level rather than an actual water level at one location in
these experiments. Oﬀshore pressure sensors outside the surf zone at mean depths
of 7 to 10.5 m were used for estimating wave height with recording lengths of 4096 s.
Guza and Thornton
30
note speciﬁc problems in the data set, which are typical of
ﬁeld measurements, i.e., the diﬃculty in obtaining a common datum for the oﬀ
shore wave measurements and the beach wave setup measurements. Results of their
measurement program suggest an empirical relationship as follows:
η = 0.17H
os
= 0.24H
orms
(1.16)
with scatter that suggests η/H
os
ranging approximately from 0.05 to 0.50 for indi
vidual experiments.
Holman and Sallenger
28
conducted a ﬁeld experiment for measuring wave setup
as well as other surf zone parameters at the U.S. Army Corps of Engineers ﬁeld
research pier in Duck, NC, USA. Data on water level at the shoreline were col
lected using longshore looking time lapse photography from Super8 movie cameras
mounted on the research pier scaﬀolding. The beach at the experiment site had a
very steep foreshore (∼1 on 10) while the oﬀshore proﬁle slope is much milder (∼1
on 100). Beach material was bimodal in size with a median sand size of 0.25 mm and
a coarse fraction of 0.75 mm. Results of the experiments showed considerable scatter
and dependence on tide level. Regression lines were ﬁt to the data (segmented by
tide levels) with results as follows for high tide and midtide data:
η
Hs
= 0.35ξ
0
+ 0.14 (high tide) , (1.17)
η
Hs
= 0.46ξ
0
+ 0.06 (midtide) . (1.18)
As most of the data fell in a range of ξ
0
= 1 to 2, the maximum setup was noted
to be of the same order as the signiﬁcant wave height in many of the experiments,
much higher than theoretically suggested values. Note that in terms of H
orms
(based
on consideration of monochromatic theory results) the setup would be much higher
than most other studies show or suggest.
Although Holman and Sallenger
28
conclude from their experiments that the
setup is dependent on the Iribarren number, it is not entirely clear from their data,
especially for higher waves (i.e., see Fig. 1.4, Ref. 28). An additional problem that
must be considered when computing the Iribarren number for real beaches and
irregular waves is how to deﬁne beach slope. It should be noted that video camera
(visual) approaches estimate setup via the measurement of the water surface ele
vation on the beach (similar to the Guza and Thornton measurements) rather than
an actual vertically ﬂuctuating water level. The anomaly between dependence of
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
14 R. G. Dean and T. L. Walton
setup on Iribarren number as noted by Holman and Sallenger
28
is likely due to the
aforementioned relationship between the wavebreaking coeﬃcient, κ, and beach
slope.
Nielsen
31
and Davis and Nielsen
32
conducted a novel setup experiment on Dee
Why Beach, in New South Wales, Australia using a set of manometer tubes as shown
in Figs. 1.7 and 1.8 from Davis and Nielsen.
32
The tubes were deployed throughout
the beach face and surf zone. A total of 120 setup proﬁles were measured in 11 days.
Wave heights H
orms
ranged from 0.6 to 2.6 m in height and signiﬁcant wave periods
(T
s
) ranged from 5.8 to 12.1 seconds. A shoreline setup of about 40% of H
orms
was
found although Davis and Nielsen
32
point out that there is reason to believe that
the surf zone characteristics inﬂuence the relationship between wave height and
setup magnitude, and also note a problem of deﬁning beach slope via the Iribarren
number. Nielsen
31
and Davis and Nielsen
32
also observe that a major portion of the
setup occurred on the beach face as shown in Fig. 1.9.
Nielsen
31
points out that previous ﬁeld investigations have typically measured
the mean water level elevation on the beach as opposed to the average ﬂuctuating
mean water level in the vertical plane (i.e., the wave setup as usually deﬁned), and
that the two measurements are often diﬀerent in part due to the beach permeability,
which in turn is related to beach material size. The issue of extracting wave setup
from runup and rundown on the beach is illustrated in Fig. 1.10.
King et al.
33
collected wave setup data at Woolacombe Beach in North Devon,
U.K. which faces the North Atlantic Ocean. The beach face slope varied between
1 on 40 at high tide and 1 on 70 at midtide level with a tidal range of 3 m at
neap and as much as 9 m at springs. Beach face material consisted of ﬁne sand with
Fig. 1.7. Manometer setup of Davis and Nielsen
32
for measuring setup.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 15
Fig. 1.8. Schematic diagram of apparatus (from Ref. 32).
Fig. 1.9. Dimensionless setup versus total depth where much of the setup occurs on the beach face
(from Ref. 32). In this ﬁgure, B and D are equal to η and h as used in this chapter, respectively.
90% in the 0.125 mm to 0.25 mm size range. Pressure transducers were utilized to
collect wave and setup information at various stations across the beach and also
in a longshore direction to assess the spatial variability of the mean setup. Both
tripod mounted and buried pressure transducers were utilized. The buried pressure
transducers were 50 to 80 cm below the beach surface and were protected by a
porous cover. Instruments collected pressure data which were then transformed to
water level data over 4096 second intervals. Data did not include sampling in very
shallow water and the maximum wave setup was estimated by extrapolating the
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
16 R. G. Dean and T. L. Walton
Fig. 1.10. Illustration of diﬀerences between mean water level (MWL) shoreline and mean water
line on beach.
water surface from the most shoreward water stations. Wave setup estimated from
the data showed the wave setup to be roughly:
η = 0.10H
os
= 0.14H
orms
(1.19)
with most of the values of η/H
orms
between 0.11 and 0.15. The authors do not
speculate as to why such low values of setup (compared to analytical results) were
found in this measurement program.
Yanagishima and Katoh
34
discuss ﬁeld measurements of mean water level near
the shoreline on the Paciﬁc Coast of Japan as measured by an ultrasonic wave gage
mounted on a pier where the mean depth of water was ∼0.4 m. The setup was
determined via a multiple regression approach on 1305 sets of (20 minute records)
data taking into account astronomical tide, wind setup, and atmospheric pressure
head components of mean water level. Their data included 91 records in which the
oﬀshore wave height was above 3 m. Yanagishima and Katoh’s
34
regression analysis
suggested the following relationship:
η = 0.0520H
os
L
os
H
os
0.2
, (1.20)
which can be formulated in terms of Iribarren number for their beach slope (1 on
60) to the following:
η = 0.27H
os
(ξ
0
)
0.4
= 0.38H
orms
(ξ
0
)
0.4
. (1.21)
Yanagishima and Katoh
34
noted reasonable agreement with the theory of Goda
35
(to be discussed later). Even higher values of setup would be expected on the beach
face in accord with theory and ﬁndings of other researchers.
Greenwood and Osborne
36
conducted ﬁeld measurements on a Georgian Bay
Beach, in Lake Huron, Ontario, Canada. Lake Huron has no measurable tide and
the beach proﬁle at the site had a slope of 0.015 with a steeper sloped (0.031 to
0.047) inshore bar. Setup was measured using surface piercing resistance wire wave
staﬀs with the shoreward most gage being in approximately 0.4 m of water depth.
Measured setup values were found as follows:
η = 0.19H
os
= 0.27H
orms
. (1.22)
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 17
It is again noted that even higher values of setup would be expected on the beach
face in accord with theory and experience of other researchers.
Further work by Hanslow and Nielsen
37,38
utilized the manometer tube
deployment shown in Fig. 1.7 on three additional beaches (Seven Mile, Palm, and
Brunswick) in New South Wales, Australia. With beach face slopes ranging from
0.03 to 0.16 and mean grain sizes of swash zone beach material ranging from 0.18
to 0.5 mm, shoreline beach setup was measured using 20 minute record averages.
Using the data from these three beaches as well as earlier measurements at Dee
Why Beach (see Refs. 31 and 32), linear least square relationships were ﬁt to the
data as follows:
η = 0.27H
os
= 0.38H
orms
with R = 0.65 (1.23)
or
η = 0.040
H
os
L
0
= 0.048
H
orms
L
0
with R = 0.77 , (1.24)
where a somewhat higher value of explained regression was noted using wave height
and wave period. Data and regression lines for these two relationships are shown
in Figs. 1.11 and 1.12. The improvement in ﬁt due to inclusion of the deep water
wavelength is not evident visually.
A signiﬁcant ﬁnding of these studies was that a major portion of the setup
occurred on the beach face (see Fig. 1.9). Further measurements on wave setup at
two river entrances is also discussed in Hanslow and Nielsen
37
and Dunn et al.
39
with the result that the wave setup at river entrances was found to be (somewhat
surprisingly) negligible.
Lentz and Raubenheimer
40
report on a ﬁeld experiment at the U.S. Army Field
Research Pier in Duck, NC, USA where 11 pressure sensor gages and 10 sonar
altimeters extended across the surf zone from 2 to 8 m of water depth. Close
agreement with LonguetHiggins radiation stress theory for wave setup was noted
Fig. 1.11. Empirical relationship between setup and wave height (from Ref. 38).
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
18 R. G. Dean and T. L. Walton
Fig. 1.12. Empirical relationship between setup and wave parameters (from Ref. 38).
although the lack of setup measurements in shallow water (<2 m) did not allow
conclusions regarding the maximum setup that might be expected on the beach.
Raubenheimer et al.
41
report on a second ﬁeld experiment at the U.S. Army
Field Research Pier in Duck, NC, USA where 12 buried pressure sensor gages were
employed across the surf zone from the shoreline to 5 m of water depth. Again good
agreement with LonguetHiggins and Stewart
42
radiation stress theory was noted
by integration of the crossshore momentum equation to estimate the wave setup
for water depths greater than 1 m but the theory was found to underpredict wave
setup in shallow water (h < 1 m). The lack of setup measurements on the beach face
did not allow conclusions regarding the maximum setup that might be expected on
a beach although an empirical equation was provided to estimate wave setup at the
SWL line as follows:
η
SWL
H
os
= 0.019 + 0.003β
−1/3
f
, (1.25)
where β
f
is the average slope across the surf zone. Raubenheimer et al.
41
suggest
that theory underpredicts the setup by a factor of 2 for water depths less than 1 m.
Stockdon et al.
43
using video shoreline water level time series determined wave
setup and wave runup results during 10 diverse ﬁeld experiments (four from Duck,
NC, USA; four from West Coast beaches in California/Oregon, USA; and two from
Terschelling, The Netherlands). These wave setup results were analyzed to provide
empirical parameterizations for wave setup under many natural beach conditions
as follows:
η = 0.385β
f
H
0
L
0
, (1.26)
which, assuming that tanβ
f
≈ β
f
can be expressed in terms of the Iribarren
number as
η
H
0
= 0.385ξ
0
(1.27)
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 19
and for extremely dissipative beaches
η
H
0
=
0.043
H
0
/L
0
, (1.28)
where H
0
is the eﬀective deep water wave height, L
0
is the deep water wave length
associated with the peak spectral period and β
f
is the average slope over a depth
range deﬁned in terms of the standard deviation of the water surface displacement.
It should again be noted that the video camera (visual) approach estimates setup
via the mean of measurements of the water surface elevation on the beach rather
than the mean of ﬂuctuating water levels at one location.
Results from nine of the ﬁeld experiments presented here have been analyzed to
determine the average ratio of wave setup at the still water line to signiﬁcant wave
height and its associated standard deviation. The ratios at the still water line were
determined to be 0.191±0.100. Several caveats apply to these results. In cases where
the beach slope and/or the deep water wave steepness was incorporated into the
expression presented, these were taken as 0.01 and 0.04, respectively. Some of the
published expressions were in terms of the breaking wave height and some in terms
of the deep water wave height and no attempt was made to diﬀerentiate between
breaking and deep water wave heights. The Holman and Sallenger results were not
included in these results as they appeared to be anomalously high. Finally, the wave
setup ratio at the intersection of the mean water line intersection with the beach
proﬁle would be greater than the average ratio (0.191) above. Also, although not
examined in detail here, the dynamic wave setup which increases with energetic
narrow spectra, would also contribute to the total wave setup.
It is relevant to note that results from ﬁeld measurements are often not con
sistent, possibly due to:
(1) Limited measurement distances across the nearshore.
(2) Use of many diﬀerent approaches to measure/evaluate setup (i.e., videos,
pressure sensors, runup gages, manometers, etc.).
(3) Inherent diﬃculties in obtaining a consistent datum for nearshore measurements
and oﬀshore measurements.
(4) A clear deﬁnition of setup on the beach face is lacking due to the nature of the
permeable beach and the diﬃculty of subaerial setup measurements.
1.6. Published Guidance on Wave Setup for Engineering
Applications
Several sources of wave setup recommendations are available; two are reviewed
here. The U.S. Army Corps of Engineers 1984 Shore Protection Manual (SPM)
presents a graphical method to calculate wave setup at middepth of the surf zone.
This method, developed for irregular waves, is presented in Fig. 1.13 in which the
normalized setup has been multiplied by a factor of 2 to transfer approximately the
results to the still water shoreline. The eﬀect of beach slope and deep water wave
steepness in Fig. 1.13 are evident.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
20 R. G. Dean and T. L. Walton
Fig. 1.13. Nondimensional wave setup versus deep water wave steepness and proﬁle slope by
the 1984 Shore Protection Manual recommendations as incorporated in Appendix D of FEMA
44
Guidelines. Note that the normalized setup has been multiplied by a factor of 2 to transfer the
setup from the middepth of the surf zone as it appears in SPM to the approximate still water
level contour. Note: S in this ﬁgure is equal to η in this chapter.
Fig. 1.14. Nondimensional wave setup by Goda versus deep water wave steepness and relative
water depth within the surf zone. Proﬁle slope = 1:100.
Goda
35
has presented guidance for static and dynamic wave setups due to
irregular waves. The guidance for static setup and a proﬁle slope of 1:100 is shown
in Fig. 1.14. The eﬀects of various deep water wave steepness values are illustrated
in Fig. 1.14.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 21
Table 1.2. Comparison of nondimensional wave
setup by SPM and Goda methods for irregular
waves.
η/H
0
H
0
/L
0
SPM Goda
0.005 0.154 0.122
0.01 0.135 0.102
0.02 0.120 0.083
0.04 0.103 0.065
0.08 0.097 0.049
Note: Values in SPM method have been multi
plied by 2.0 to transfer from surf zone middepth
to still water line.
Table 1.2 presents a comparison of ratios of nondimensional wave setup values
at the still water shoreline as recommended by SPM and Goda. In examining the
results in Table 1.2, recall that an additional wave setup occurs from the still water
line to the location where the maximum setup intersects the beach proﬁle.
1.7. Summary and Recommendations
The reviews of theory, laboratory and ﬁeld data, and published guidance for engi
neering applications presented here have identiﬁed static and dynamic wave setup
components as contributing to the deviation from still water level in the surf zone
and their relevance to engineering design. Examination of the static wave setup
has reinforced the eﬀect of beach slope on wave setup. The theory presented here
does not account for the onshore bottom stress acting on the water column due to
undertow.
The available ﬁeld measurement results exhibit a wide range of wave setup to
wave height ratios. Some of this variability is undoubtedly due to the eﬀect of
proﬁle slope, which is not accounted for explicitly in some of the analyses and
part is due to the eﬀect of wavebreaking in depths greater than the shallow water
limit.
Design methodology should account for the static and dynamic wave setup com
ponents. In determining the wave setup to include in design, the characteristics of
the particular application of interest should be compared with those of the various
ﬁeld and laboratory experiments available including those referenced here. The dom
inant role of beach slope should be recognized. The preliminary results presented
here of η/H
s
= 0.191 ±0.100 may serve as a useful guide for the static wave setup
component.
It is hoped that further research with improved instrumentation, modern sur
veying techniques, and more diverse ﬁeld site studies will help to clarify both the
static and dynamic wave setup components for future design applications.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
22 R. G. Dean and T. L. Walton
References
1. A. T. Ippen, Estuary and Coastline Hydrodynamics (McGrawHill Book Company,
1966), 744 pp.
2. M. S. LonguetHiggins, Radiation stress and mass transport in gravity waves with
application to “surf beats”, J. Fluid Mech. 13(4), 481–504 (1962).
3. M. S. LonguetHiggins and R. W. Stewart, A note on wave setup, J. Marine Res. 21,
4–10 (1963).
4. M. S. LonguetHiggins and R. W. Stewart, Radiation stresses in water waves; a
physical discussion, with applications, DeepSea Res. 11, 529–562 (1964).
5. R. A. Holman, Wave setup, Handbook of Coastal and Ocean Engineering, Vol. 1,
ed. J. Herbich (Gulf Publishing, 1990), Chapter 11, pp. 635–646.
6. van der Meer, TAW Technical Report: Wave Runup and Wave Overtopping at Dikes,
Technical Advisory Committee for Flood Defence, Delft (2003).
7. M. S. LonguetHiggins, On the waveinduced diﬀerence in mean sea level between the
two sides of a submerged breakwater, J. Marine Res. 25(2), 148–153 (1967).
8. W. R. Dally, R. G. Dean and R. A. Dalrymple, Wave height variation across beaches
of arbitrary proﬁle, J. Geophys. Res. 90(C6), 11917–11927 (1985).
9. R. G. Dean, Stream function representation of nonlinear ocean waves, J. Geophys Res.
79(30), 4489–4504 (1965).
10. R. G. Dean, Evaluation and Development of Water Wave Theories for Engineering
Application; Volume I: Presentation of Research Results; Volume II: Tabulation of
Dimensionless Stream Function Variables, Special Report No. 1, Published by U.S.
Army Corps of Engineers, Coastal Engineering Research Center (1974).
11. R. A. Dalrymple, A ﬁnite amplitude wave on a linear shear current, J. Geophys. Res.
87(C1), 483–491 (1974).
12. R. G. Dean and C. J. Bender, Static wave setup with emphasis on damping eﬀects by
vegetation and bottom friction, Coast. Eng. 13, 149–156 (2006).
13. W. R. Munk, Surf beats, Trans. Am. Geophys. Union 30, 849–854 (1949).
14. M. J. Tucker, Surf beats: Sea waves of 1 to 5 minutes period, Proc. Roy. Soc. A 202,
565–573 (1950).
15. G. Symonds, D. A. Huntley and A. J. Bowen, Twodimensional surf beat: Long wave
generation by a timevarying breakpoint, J. Geophys. Res. 87(C1), 492–498 (1982).
16. G. Symonds and A. J. Bowen, Interaction of nearshore bars with wave groups,
J. Geophys. Res. 89(C2), 1953–1959 (1984).
17. H. A. Schaﬀer and I. A. Svendsen, Surf beat generation on a mildslope beach, Proc.
ASCE Int. Conf. Coastal Engineering (1988), pp. 1058–1072.
18. J. K. Kostense, Measurements of surf beat and setdown beneath wave groups, Proc.
ASCE Int. Conf. Coastal Engineering (1984), pp. 724–740.
19. A. J. Bowen, D. L. Inman and V. P. Simmons, Wave setdown and setup, J. Geophys.
Res. 73, 2569–2577 (1968).
20. J. A. Battjes, SetUp Due to Irregular Waves, Report No. 72–2, Communications on
Hydraulics, Delft University of Technology, Department of Civil Engineering (1972),
13 pp. (Paper also presented at the 13th Int. Conf. Coastal Engineering, Vancouver,
B.C.).
21. I. A. Svendsen, Wave heights and setup in a surf zone, Coast. Eng. 8, 302–329 (1984).
22. M. J. F. Stive and H. G. Wind, A Study of Radiation Stress and SetUp in the
Nearshore Zone, Publication No. 267, Waterlopkundig Laboratorium, Delft Hydraulics
Laboratory (1982), 25 pp.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
Wave Setup 23
23. T. S. Hedges and H. Mase, Modiﬁed Hunt’s equation incorporating wave setup,
J. Waterway Port Coast. Ocean Eng. 130(3), 109–113 (2004).
24. H. Mase, Random wave runup height on gentle slopes, J. Waterway Port Coast. Ocean
Div. 115(WW2), 649–661 (1989).
25. T. L. Walton, Interim guidance for prediction of wave runup on beaches, Ocean Eng.
19(2), 199–207 (1999).
26. T. Saville, Experimental determination of wave setup, Proc. 2nd Technical Conf.
Hurricanes, Miami Beach, FL, National Hurricane Research Project Report No. 50,
U.S. Department of Commerce Washington, D.C. (1961), pp. 242–252.
27. R. Dorrestein, Wave setup on a beach, Proc. 2nd Technical Conf. Hurricanes, Miami
Beach, FL, National Hurricane Research Project Report No. 50, U.S. Department of
Commerce, Washington, D.C. (1961), pp. 230–241.
28. R. A. Holman and A. H. Sallenger, Setup and swash on a natural beach, J. Geophys.
Res. 90(C1), 945–953 (1985).
29. U. A. Hansen, Wave setup and design water level, J. Waterway Port Coast. Ocean
Div. 104(WW2), 227–240 (1978).
30. R. T. Guza and E. B. Thornton, Wave setup on a natural beach, J. Geophys. Res.
96(C2), 4133–4137 (1981).
31. P. Nielsen, Wave setup: A ﬁeld study, J. Geophys. Res. 93(C12), 15643–15652 (1988).
32. G. A. Davis and P. Nielsen, Field measurement of wave setup, ASCE Int. Conf.
Coastal Engineering, Malaga, Spain (1988), Chapter 38, pp. 539–552.
33. B. A. King, M. W. L. Blackley, A. P. Carr and P. J. Hardcastle, Observations of wave
induced setup on a natural beach, J. Geophys. Res. 95(C12), 22289–22297 (1990).
34. S. Yanagishima and K. Katoh, Field observation on wave setup near the shoreline,
Proc. 22nd Int. Conf. Coastal Engineering, Vol. 1, ASCE, New York, N.Y. (1990),
Chapter 7, pp. 95–108.
35. Y. Goda, Random Seas and Design of Maritime Structures (World Scientiﬁc
Publishing Co., 2000), 443 pp.
36. B. Greenwood and P. D. Osborne, Vertical and horizontal structure in crossshore
ﬂows: An example of undertow and wave setup on a barred beach, Coast. Eng. 14,
543–580 (1990).
37. D. J. Hanslow and P. Nielsen, Wave setup on beaches and in river entrances, 23rd Int.
Conf. Coastal Engineering, Venice, Italy (1992), pp. 240–252.
38. D. J. Hanslow and P. Nielsen, Shoreline setup on natural beaches, J. Coast. Res.
SI15, 1–10 (1993).
39. S. L. Dunn, P. Nielsen, P. A. Madsen and P. Evans, Wave setup in river entrances,
Proc. 27th Int. Conf. Coastal Engineering, ASCE, New York, Sydney, Australia (2000),
pp. 3432–3445.
40. S. Lentz and B. Raubenheimer, Field observations of wave setup, J. Geophys. Res.
104(C11), 25867–25875 (1999).
41. B. Raubenheimer, R. T. Guza and S. Elgar, Field observations of wavedriven setdown
and setup, J. Geophys. Res. 106(C3), 4629–4638 (2001).
42. M. S. LonguetHiggins and R. W. Stewart, Radiation stresses in water waves: A
physical discussion with applications, Deep Sea Res. 11(4), 529–562 (1964).
43. H. F. Stockdon, R. A. Holman, P. A. Howd and A. H. Sallenger, Jr., Empirical param
eterization of setup, swash, and runup, Coast. Eng. 53, 573–588 (2006).
44. FEMA, Guidelines and Speciﬁcations for Flood Hazard Mapping Partners,
Appendix D: Guidance for Coastal Flooding Analysis and Mapping, Map Modern
ization Program, Washington, D.C. (2003). Also available at: www.fema.gov/fhm/
dl cgs.shtm.
July 31, 2009 8:18 9.75in x 6.5in b684ch01 FA
This page intentionally left blank This page intentionally left blank
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Chapter 2
Wavemaker Theories
Robert T. Hudspeth
School of Civil and Construction Engineering
Oregon State University, Corvallis, OR 97331, USA
robert.hudspeth@orst.edu
Ronald B. Guenther
Department of Mathematics
Oregon State University, Corvallis, OR 97331, USA
guenther@math.orst.edu
The fundamental solutions to the wavemaker boundary value problem (WMBVP)
are given for 2D channels, 3D basins, and circular basins. The solutions are given
in algebraic equations that replace integral and diﬀerential calculus. The solutions
are generic and apply to both full and partialdraft piston and hinged wave
makers; to doublearticulated wavemakers, and to directional wave basins. The
WMBVP is solved by conformal mapping and by domain mapping. The loads on
a wavemaker are connected to the radiation boundary value problem for semi
immersed bodies and demonstrate the connection of these loads to the added mass
and radiation damping coeﬃcients required to compute the dynamic response of
large Lagrangian solid bodies.
2.1. Introduction
Wavemaker theories play several important roles in coastal and ocean engineering.
The most important role is the application to laboratory wavemakers for both
wavemaker designs and wave experiments. A second role for wavemaker theories
is to compute a scalar radiated wave potential to compute the waveinduced loads
on large solid bodies applying potential wave theory. The displacements and rota
tions of a semiimmersed six degreesoffreedom large Lagrangian solid body are
related to the displacements and rotations of wavemakers. The boundary between a
planar wavemaker and an ideal ﬂuid requires special care because the ﬂuid motion
25
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
26 R. T. Hudspeth and R. B. Guenther
is an Eulerian ﬁeld with time and space as the independent variables, and the
planar wavemaker is a Lagrangian solid body with time and the wavemaker as the
independent variables. Consequently, the kinematic boundary condition will be dif
ferent from the free surface boundary that separates two Eulerian ﬂuid ﬁelds of air
and water. The boundary between the ﬂuid and wavemaker separates an Eulerian
ﬁeld (the ﬂuid) from a Lagrangian body (the wavemaker), and the wavemaker
kinematic boundary condition (WMKBC) must convert the Lagrangian wavemaker
motion to a Eulerian ﬁeld motion in order that the independent variables for both
dependent motion variables are equivalent. This may be accomplished by multi
plying the Lagrangian motion of the wavemaker by the unit normal to the boundary.
Because the unit normal is a function of space and the Lagrangian wavemaker
motion is a function of time, the product will produce a motion that is a function
of both space and time that are the independent variables of the Eulerian ﬂuid
ﬁeld. Although this fact is not central to the WMBVP, it is an important con
nection between the WMBVP and the radiation potential boundary value problem
for semiimmersed large Lagrangian solid bodies.
1
The formulae for computing the two fundamental ﬂuid unknowns for an incom
pressible ﬂuid of the velocity q(x, z, t) and the pressure p(x, z, t) from a scalar
velocity potential Φ(x, z, t) are given ﬁrst. The classical linear WMBVP for dimen
sionless 2D planar wavemaker is reviewed for two types of doublearticulated planar
wavemakers. The sway X
1
(t) displacement of a fulldraft piston wavemaker and the
roll Θ
5
(t) rotation of a hinged wavemaker are connected directly to the sway dis
placement and the roll rotation of a semiimmersed large Lagrangian solid body. In
this review, integral calculus formulae for computing the integrals that are required
to compute the coeﬃcients of the eigenseries for the ﬂuid motion, to compute the
loads on the wavemaker and the average power required to generate the propagating
waves are replaced by generic algebraic formulae. For example, an integral equation
that is required to compute the nth eigenseries coeﬃcient C
n
for the nth eigen
function Ψ
n
(K
n
, z/h) from a wavemaker shape function χ(z/h) may be computed
symbolically and expressed by a dimensionless algebraic formula I
n
(α, β, b, d, K
n
),
that is given by
C
n
= h
0
−1
χ(z/h)d(z/h)Ψ
n
(z/h)d(z/h) = I
n
(α, β, b, d, K
n
). (2.1)
The coeﬃcient in (2.1) may then be computed very eﬃciently by substitution into
algebraic formulae in all subsequent applications. Next a dimensionless theory for
both amplitudemodulated (AM) and phasemodulated (PM) circular wavemakers is
reviewed. Then, a dimensionless theory for doubleactuated wavemakers is reviewed.
Following that, a dimensionless directional wavemaker theory for large wave basins
based on a WKBJ approximation
1
is reviewed. Next, a theory for sloshing waves
due to transverse motions of a segmented wavemaker in a narrow wave channel is
reviewed. Then, 2D planar wavemakers are mapped to a unit circle by conformal
mapping and to a ﬁxed rectangular domain by domain mapping; and both the linear
and nonlinear wavemaker solutions are computed numerically.
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 27
2.2. Planar Wavemaker in a 2D Channel
Two generic planar wavemaker conﬁgurations are shown in Figs. 2.1(a) and 2.1(b).
The ﬂuid motion may be obtained from the negative gradient of a dimensional scalar
velocity potential Φ(x, z, t) according to
q(x, z, t) = u(x, z, t)e
x
+w(x, z, t)e
z
= −∇
2
Φ(x, z, t), (2.2a)
where the 2D gradient operator in (2.2a) is given by
∇
2
(•) =
∂(•)
∂x
e
x
+
∂(•)
∂z
e
z
.
Fig. 2.1(a). Deﬁnition sketch for a Type I planar wavemaker.
Fig. 2.1(b). Deﬁnition sketch for a Type II planar wavemaker.
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
28 R. T. Hudspeth and R. B. Guenther
The total pressure ﬁeld P(x, z, t) may be computed from the unsteady Bernoulli
equation according to
P(x, z, t) = p(x, z, t) +p
S
(z) = ρ
∂Φ(x, z, t)
∂t
−
1
2
∇Φ(x, z, t)
2
+Q(t)
−ρgz,
(2.2b)
where Q(t) = the Bernoulli constant; and the free surface elevation η(x, t) for zero
atmospheric pressure according to
η(x, t) =
1
g
∂Φ(x, η, t)
∂t
−
1
2
∇Φ(x, η, t)
2
+ Q(t)
; x ≥ ξ(η, t); z = η(x, t).
(2.2c)
The scalar velocity potential must be a solution to the Laplace equation
∇
2
2
Φ = 0; x ≥ ξ(z, t); −h ≤ z ≤ η(x, t), (2.3a)
with the following boundary conditions:
Kinematic Bottom Boundary Condition (KBBC):
∂Φ
∂z
= 0; x ≥ ξ(−h, t); z = −h. (2.3b)
Combined Kinematic and Dynamic Free Surface Boundary Condition (CKDFSBC):
∂
2
Φ
∂t
2
+g
∂Φ
∂z
−
¸
∂
∂t
−
1
2
∇Φ· ∇
∇Φ
2
+
dQ
dt
= 0; x ≥ ξ(η, t); z = η(x, t).
(2.3c)
Kinematic WaveMaker Boundary Condition (KWMBC):
A Stokes material surface for planar wavemaker is W(x, z, t) = x −ξ(z, t), and the
Stokes material derivative gives the KWMBC from
DW
Dt
=
∂Φ
∂x
+
∂ξ
∂t
−
∂Φ
∂z
∂ξ
∂z
= 0; x = ξ(z, t); −h ≤ z ≤ η(t). (2.3d)
Kinematic Radiation Boundary Condition (KRBC):
A KRBC is required as x →+∞for uniqueness to insure that propagating waves are
only right progressing or that evanescent eigenmodes are bounded. For a temporal
dependence proportional to exp±iωt, the KRBC may be expressed by
lim
x→+∞
∂
∂x
±iK
n
Φ(x, z, t) = 0. (2.3e)
A velocity potential ϕ(x, z) may be deﬁned by the real part of
Φ(x, z, t) = Re{ϕ(x, z) exp−i(ωt +ν)}, (2.4)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 29
where Re{•} means the real part of {•}; and ν = arbitrary phase angle. The
linearized WMBVP for kh = O(1) is
1
∇
2
2
ϕ(x, z) = 0; 0 ≤ x < +∞; −h ≤ z ≤ 0, (2.5a)
∂ϕ(x, z)
∂z
= 0; 0 ≤ x +∞; z = −h, (2.5b)
∂ϕ(x, z)
∂z
−k
0
ϕ(x, z) = 0; 0 ≤ x < +∞; z = 0, (2.5c)
lim
x→+∞
∂
∂x
−iK
n
ϕ(x, z) = 0, (2.5d)
∂ϕ(x, z)
∂x
exp −i(ωt +ν) = −
∂ξ(z, t)
∂t
; x = 0; −h ≤ z ≤ 0, (2.5e)
η(x, t) = Re
−iω
g
ϕ(x, 0) exp −i(ωt +ν)
; x ≥ 0; z = 0, (2.5f)
p(x, z, t) = ρ
∂Φ(x, z, t)
∂t
; 0 ≤ x < +∞; z = 0, (2.5g)
where k
0
= ω
2
/g.
Because the boundary conditions deﬁned by (2.5b)–(2.5e) are prescribed on
boundaries with constant values of the independent variables x and z, a solution by
the method of separation of (independent) variables may be computed.
1
The instan
taneous wavemaker displacement ξ(z, t) from its mean position x = 0 is assumed
to be strictly periodic in time with period T = 2π/ω, and may be expressed by
ξ(z/h, t) = Re
i
¸
S
(∆/h)
χ(z/h) exp−i(ωt +ν)
=
¸
S
(∆/h)
χ(z/h) sin(ωt +ν). (2.6)
The speciﬁed shape function χ(z/h) for the Type I wavemaker shown in Fig. 2.1(a)
is valid for either a doublearticulated piston or hinged wavemaker of variable draft
and is given by the following dimensionless equation for a straight line
2
:
χ(z/h) = [α(z/h) +β][U(z/h + 1 −d/h) −U(z/h +b/h)], (2.7a)
where α, β = dimensionless constants; U(•) = the Heaviside step function with two
boundary conditions given by
[S/(∆/h)]χ(z/h = −1 +d/h +∆
b
/h +∆/h) = S, (2.7b)
[S/(∆/h)]χ(z/h = −1 +d/h +∆
b
/h) = S
b
, (2.7c)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
30 R. T. Hudspeth and R. B. Guenther
that may be solved simultaneously for the dimensionless coeﬃcients α, β to obtain
α = (1 −S
b
/S); β = ∆/h +α(1 −d/h −∆
b
/h −∆/h). (2.7d,e)
The coeﬃcients α and β for the speciﬁed shape function χ(z/h) in (2.7a) may be
obtained for the Type II wavemaker shown in Fig. 2.1(b) by substituting
S =
¯
S +
ˆ
S; S
b
=
¯
S; ∆ = h −b −d
into the following boundary conditions
2
in (2.7b) and (2.7c):
¸
(
¯
S +
ˆ
S)
1 −b/h −d/h
¸
χ(z/h = −b/h) =
¯
S +
ˆ
S, (2.8a)
¸
(
¯
S +
ˆ
S)
1 −b/h −d/h
¸
χ(z/h = −1 +d/h) =
¯
S, (2.8b)
that may be solved simultaneously for the constant coeﬃcients α, β to obtain
α =
ˆ
S
¯
S +
ˆ
S
; β = 1 −
d
h
−
¯
S
¯
S +
ˆ
S
b
h
. (2.8c,d)
2.2.1. Eigenfunction solution to the WMBVP
Because all of the boundary conditions deﬁned by (2.5b)–(2.5e) are now prescribed
for constant values of the independent variables (x, z) and the dimensionless
parameter kh = O(1), a solution by separation of independent variables
1
is
suggested according to
ϕ(x, z) = X(x) • Z(z). (2.9)
The eigenseries solution may be written compactly as
1,3,4
Φ(x, z, t; K
n
) =
¸
n=1
C
n
coshK
n
(z +h) exp +i(K
n
x −ωt +ν), (2.10a)
where K
n
= k for n = 1 and K
n
= +iκ
n
for n ≥ 2 provided that
k
o
h −khtanhkh = k
o
h +κ
n
htan κ
n
h = 0; n > 2. (2.10b)
The eigenseries (2.10a) may be separated into a propagating Φ
p
(x, z, t; k) and
evanescent eigenmodes Φ
e
(x, z, t; κ
n
) or “local ” wave components
3
according to
Φ(x, z, t; K
n
) = Φ
p
(x, z, t; k) +Φ
e
(x, z, t; κ
n
)
=
C
1
coshk(z +h) +
¸
n=2
C
n
cos κ
n
(z +h)
¸
exp+i(K
n
x − ωt +ν).
(2.10c)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 31
The wave number k = 2π/λ where λ = wavelength. Because the numerical value of
kh must be computed from an eigenvalue problem in the vertical z coordinate, equiv
alence of the eigenvalue k to the wave number 2π/λ requires a pseudohorizontal
boundary condition of periodicity given by k = 2π/λ and ϕ(x + λ, z) = ϕ(x, z). It
is computationally eﬃcient to normalize the eigenseries in (2.10a) according to
Ψ
n
(K
n
, z/h) =
coshK
n
h(1 +z/h)
N
n
; n = 1, 2, . . . , (2.11a)
where the nondimensional normalizing constant N
n
is
N
2
n
=
0
−1
cosh
2
K
n
h(1 +z/h)d(z/h) =
2kh + sinh 2kh
4kh
; n = 1, (2.11b)
2κ
n
h + sin 2κ
n
h
4κ
n
h
; n ≥ 2. (2.11c)
The eigenseries in (2.10a) may be written as an orthonormal eigenseries by
Φ(x, z, t; K
n
) =
∞
¸
n=1
C
n
Ψ
n
(K
n
, z/h) expi(K
n
x −ωt −ν), (2.12)
where the orthonormal eigenfunction Ψ
n
(•,•) is dimensionless.
2.2.2. Evaluation of C
n
by WM vertical displacement χ(z/h)
The following dimensionless coeﬃcient computed from (2.5e) will replace integral
calculus with algebraic substitution for the coeﬃcients C
n
in the eigenseries (2.12):
I
n
(α, β, b, d, K
n
) =
−b/h
−1+d/h
[α(z/h) +β]Ψ
n
(K
n
, z/h)d(z/h)
=
α
(K
n
h)
2
N
n
K
n
h
1 −
d
h
sinh K
n
d −K
n
b sinhK
n
h
1 −
b
h
−coshK
n
h
1 −
b
h
+ coshK
n
d
+
β
(K
n
h)N
n
sinh K
n
h)
1 −
b
h
− sinhK
n
d
(2.13)
that is dimensionless when α and β are given by (2.7d) and (2.7e) or (2.8c) and
(2.8d). The coeﬃcients C
n
may be computed algebraically by (2.13) from the
KWMBC (2.5e) to obtain
C
n
= i
Sωh
K
n
∆
I
n
(α, β, b, d, K
n
), (2.14)
and the orthonormal eigenseries (2.12) is given by
Φ(x, z, t; K
n
)
=
∞
¸
n=1
iSωh
K
n
∆
I
n
(α, β, b, d, K
n
)Ψ
n
(K
n
, z/h) expi(K
n
x −ωt −ν). (2.15)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
32 R. T. Hudspeth and R. B. Guenther
2.2.3. Decay distance of evanescent eigenmodes n ≥ 2
Numerical solutions and experimental measurements of ocean and coastal designs
require that the KRBC (2.5d) be applied far enough away so that only the prop
agating eigenmode for n = 1 in (2.12) is measurable. The evanescent eigenseries
in (2.12) for n ≥ 2 will decay spatially at least as fast as the smallest evanescent
eigenvalue κ
2
. This eigenvalue must be κ
2
h > (n − 3/2)π = π/2. If the smallest
value for κ
2
h > π/2, then κ
2
> π/2h and ϕ(x, z) ∝ exp −(πx/2h). For the values
of the evanescent eigenseries to be less than 1% of their values at the wavemaker,
ϕ(x, z) ∝ exp −(πx
d
/2h) = 0.01 and πx
d
/(2h) = 4.6 ≈ 3π/2, and the minimum
decay distance is x
d
≥ 3h.
2.2.4. Transfer function for wave amplitude
from wavemaker stroke
The average rate of work or power done by a wavemaker of width B is
1
˙
W
τ
= P
τ
= B
τ+1
τ
h
0
−1
p(x, z, τ)u(x, z, τ)d(z/h)dτ, (2.16a)
where the temporal averaging operator is deﬁned by
•
τ
=
τ+1
τ
(•)dτ, (2.16b)
and
˙
W
τ
= P
τ
=
ρω
3
S
2
Bh
4
∆
2
2kh
I
2
1
(α, β, b, d, k), (2.16c)
so that all of the average power from a wavemaker is transferred to only the
propagating eigenmode. The average energy ﬂux in a linear wave is given by
1
˙
E
τ
=
ρgBA
2
2
C
G
, (2.16d)
where the group velocity C
G
is given by
1
C
G
=
C
2
¸
1 +
2kh
sinh 2kh
. (2.16e)
Equating (2.16c) to (2.16d) gives the following transfer function for a planar
wavemaker:
A
S
=
k
o
h
kh
Ψ
1
(k, 0)I
1
(α, β, b, d, k). (2.16f)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 33
2.2.5. Hydrodynamic pressure loads (added mass
and radiation damping)
The wave loads on a planar wavemaker may be estimated by integrating the total
pressure over the wetted surface of the wavemaker, i.e.,
F
M
=
0
S
P
n
r ×n
dS, (2.17a,b)
where the outward pointing unit normal n points from the wavemaker into the ﬂuid,
and the pseudounit normal n
for the rotational modes is given by
n
= r ×n = (z +h −d)n
x
e
y
= n
y
e
y
. (2.17c)
Force. For the Type I piston wavemaker of total width B, the horizontal component
of the pressure force on the ﬂuid side only may be computed from the real part of
F
1
(t) = Re
iρωBh
¸
n=1
C
n
−b/h
−1+d/h
Ψ
n
(K
n
, z/h)d(z/h) exp−i(ωt +ν)
¸
= −F
1
cos(ωt +ν −α
1
), (2.18a)
where the static component of the pressure force on the ﬂuid side only is
F
s
= −
ρgBh
2
2
[1 −2(d/h) + (d/h)
2
−(b/h)
2
]. (2.18b)
The hydrodynamic component of F
1
(t) may be separated linearly into a propagating
and an evanescent component that are related to the piston wavemaker translational
velocity and acceleration, respectively, from the real part of
F
1
(t) = −Re{[λ
11
(Sω) +µ
11
(−iSω
2
)] exp −i(ωt +ν)}
= −µ
11
(−Sω
2
sin(ωt +ν)) −λ
11
(Sω cos(ωt +ν)) (2.18c)
= Re{−µ
11
¨
X
1
(t) − λ
11
˙
X
1
(t)}, (2.18d)
where the added mass coeﬃcient µ
11
may be computed from the evanescent
eigenmodes only, and the radiation damping coeﬃcient λ
11
may be computed from
the propagating eigenmode only. The average power may be computed from
−F
1
˙
X
1
t
=
λ
11
(Sω)
2
2
. (2.19a)
Equating (2.19a) to (2.16d) yields
λ
11
=
A
1
S
1
2
ρBh
k
o
h
C
G
, (2.19b)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
34 R. T. Hudspeth and R. B. Guenther
that relates the radiation damping coeﬃcient to the square of the ratio of the
radiated wave amplitude to the amplitude of the wavemaker displacement.
Moment. For the Type I wavemaker of width B, the dynamic pressure moment
on one side only of the wavemaker may be computed from the real part of
M
5
(t) = Re
¸
iρωBh
2
¸
n=1
C
n
−b/h
−1+d/h
¸
1 +
z
h
−
d
h
× Ψ
n
(K
n
, z/h)d
z
h
¸
exp −(ωt +ν)
¸
= −M
5
cos(ωt +ν −α
5
), (2.20a)
and the static component of the pressure moment on the ﬂuid side only is
M
s
=
ρgBh
3
6
1−
d
h
3
+ 2
b
h
3
+ 3
¸
d
h
2
−
b
h
2
¸
−3
d
h
¸
1−
b
h
2
¸¸
.
(2.20b)
The pressure moment M
5
(t) in (2.20a) may be separated linearly into a propagating
and an evanescent component that are related to the rotational velocity and accel
eration from the real part of
1
M
5
(t) = −Re
¸
µ
55
−i
Sω
2
∆(1 +∆
b
/∆)
+λ
55
−i
Sω
∆(1 +∆
b
/∆)
exp −i(ωt +ν)
,
M
5
(t) = −µ
55
¨
Θ
5
(t) −λ
55
˙
Θ
5
(t), (2.21a)
where
µ
55
= ρBh
4
¸
n=2
I
2
n
(α, β, b, d, κ
n
)
κ
n
h
, (2.21b)
λ
55
= ρωBh
4
I
2
1
(α, β, b, d, k)
kh
. (2.21c)
2.3. Circular Wavemaker
Havelock
5
applied Fourier integrals to develop a theory for surface gravity waves
forced by circular wavemakers in water of both inﬁnite and ﬁnite depth. The ﬂuid
motion may be obtained from the negative gradient of a scalar velocity potential
Φ(r, θ, z, t) according to
q(r, θ, z, t) = −∇Φ(r, θ, z, t), (2.22a)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 35
where the 3D gradient operator ∇(•) in polar coordinates is
∇(•) =
∂
∂r
e
r
+
1
r
∂
∂θ
e
θ
+
∂
∂z
e
3
(•). (2.22b)
The total pressure ﬁeld P(r, θ, z, t) may be computed from the unsteady Bernoulli
equation in polar coordinates according to
P(r, θ, z, t) = p(r, θ, z, t) +p
S
(z)
= ρ
∂Φ(r, θ, z, t)
∂t
−
1
2
∇Φ(r, θ, z, t)
2
+Q(t)
−ρgz, (2.22c)
where Q(t) = the Bernoulli constant, and the free surface elevation η(r, θ, t) for zero
atmospheric pressure according to
η(r, θ, t) =
1
g
∂Φ(r, θ, η, t)
∂t
−
1
2
∇Φ(r, θ, η, t)
2
+
Q(t)
ρ
;
r ≥ b +ξ(θ, η, t); z = η(r, θ, t). (2.22d)
The scalar velocity potential Φ(r, θ, z, t) must be a solution to the continuity
equation
∇
2
Φ =
1
r
∂
∂r
r
∂Φ
∂r
+
1
r
2
∂
2
Φ
∂θ
2
+
∂
2
Φ
∂z
2
= 0,
r ≥ b +ξ(θ, z, t); 0 ≤ θ ≤ 2π; −h ≤ z ≤ η(r, θ, t), (2.23a)
with the following boundary conditions:
Kinematic Bottom Boundary Condition (KBBC):
∂Φ
∂z
= 0; r ≥ b +ξ(θ, −h, t); 0 ≤ θ ≤ 2π; z = −h. (2.23b)
Combined Kinematic and Dynamic Free Surface Boundary Condition (CKDFSBC):
∂
2
Φ
∂t
2
+g
∂Φ
∂z
−
¸
∂
∂t
−
1
2
∇Φ· ∇
∇Φ
2
+
dQ
dt
= 0;
r ≥ b +ξ(θ, η, t); 0 ≤ θ ≤ 2π; z = η(r, θ, t). (2.23c)
Kinematic WaveMaker Boundary Condition (KWMBC):
∂Φ
∂r
+
∂ξ
∂t
−
1
r
2
∂Φ
∂θ
∂ξ
∂θ
−
∂Φ
∂z
∂ξ
∂z
= 0; r = ξ(θ, z, t); −h ≤ z ≤ η(b, θ, t). (2.23d)
Two types of circular cylindrical wavemaker displacements ξ(θ, z, t) may be ana
lyzed, viz., amplitudemodulated (AM) and phasemodulated (PM) wavemakers.
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
36 R. T. Hudspeth and R. B. Guenther
The distinction between these two types is in the azimuthal θ dependency of the
wavemaker displacement ξ(θ, z, t) from its mean position r = b, given by
ξ(θ, z, t) =
m
S
(∆/h)
χ(z/h)
cos mθ sin(ωt +ν)
sin(ωt +ν +mθ)
¸
.
(2.23e)
(2.23f)
Kinematic Radiation Boundary Condition (KRBC):
lim
√
K
n
r→+∞
∂
∂r
±iK
n
Φ(r, θ, z, t) = 0; r →∞. (2.23g)
Finally, physically realizable solutions to (2.23a) must be periodic in θ; i.e.,
Φ(r, θ, z, t) = Φ(r, θ + 2π, z, t). (2.23h)
The dimensional WMBVP may be scaled and linearized
1
by expanding the variables
in perturbation series with a dimensionless perturbation parameter ε = kA. A scalar
radiated velocity potential ϕ(r, θ, z) may be deﬁned by the real part of
Φ(r, θ, z, t) = Re{ϕ(r, θ, z) exp−i(ωt +ν)}. (2.24)
A linearized WMBVP may be obtained by setting the dimensionless parameter
kA = ε = 0 and by requiring that kh = O(1). This linearized WMBVP is
∇
2
ϕ(r, θ, z) = 0; b ≤ r < +∞; 0 ≤ θ ≤ 2π; −h ≤ z ≤ 0, (2.25a)
∂ϕ(r, θ, z)
∂z
= 0; b ≤ r < +∞; 0 ≤ θ ≤ 2π; z = −h, (2.25b)
∂ϕ(r, θ, z)
∂z
−k
o
ϕ(r, θ, z) = 0; b ≤ r < +∞; 0 ≤ θ ≤ 2π; z = 0, (2.25c)
lim
√
K
n
r→+∞
∂
∂r
−iK
n
ϕ(r, θ, z) = 0, (2.25d)
∂ϕ(r, θ, z)
∂r
exp−i(ωt +ν) = −
∂ξ(θ, z, t)
∂t
; r = b; 0 ≤ θ ≤ 2π; −h ≤ z ≤ 0,
(2.25e)
η(r, θ, t) = Re
−iωϕ(r, θ, z)
g
exp −i(ωt +ν)
; b ≤ r < ∞; 0 ≤ θ ≤ 2π; z = 0,
(2.25f)
P(r, θ, z, t) = {p(r, θ, z, t)} +p
s
(z)
= Re{−iωρϕ(r, θ, z) exp−i(ωt +ν)} −ρgz, (2.25g)
ϕ(r, θ, z) = ϕ(r +λ, θ, z); ϕ(r, θ, z) = ϕ(r, θ + 2π, z). (2.25h,i)
The speciﬁed wavemaker shape function χ(z/h) is valid for either a double
articulated piston or hinged circular AM or PM wavemaker of variable draft that
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 37
Fig. 2.2. Deﬁnition sketch for circular wavemaker.
is shown in Fig. 2.2 is identical to (2.7) for a 2D planar wavemaker with the
dimension b replaced with a and the stroke S replaced with the azimuthal stroke
m
S. The solution to the WMBVP (2.25) may be compactly expressed by the fol
lowing orthonormal eigenseries:
m
ϕ(r, θ, z) =
∞
¸
n=1
C
mn
Ψ
n
(K
n
, z/h)H
(1)
m
(K
n
r)M
A(P)
(mθ), (2.26a,b)
where the azimuthal mode function is
M
A(P)
(mθ) =
cos mθ
exp −imθ
¸
; m ≥ 0 and integer, (2.26c,d)
and where (2.26a) represents an AM wavemaker; (2.26b) represents a PM
wavemaker; Ψ
n
(K
n
, z/h) = the orthonormal eigenseries deﬁned in (2.11);
H
(1)
m
(K
n
r) = the Hankle function of the ﬁrst kind. When K
1
= k and K
n
= iκ
n
for
n ≥ 2 and integer,
H
(1)
m
(iκ
n
r) = J
m
(iκ
n
r) +iY
m
(iκ
n
r) =
2
π
i
−(m+1)
K
m
(κ
n
r), (2.26e)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
38 R. T. Hudspeth and R. B. Guenther
where K
m
(•) = the Modiﬁed Bessel (or Kelvin) function of the second kind of
order m. The coeﬃcients C
mn
may be computed by expanding the KWMBC in an
eigenseries following the procedure in (2.14) and obtaining
C
mn
= −
m
S
j
hω
K
n
∆
I
n
(α, β, a, d, K
n
)
L
n
(H
(1)
m
(K
n
b))
; n ≥ 1 and integer, (2.26f)
L
n
(Z
m
(ζ
n
)) =
dZ
m
(ζ
n
)
dζ
n
=
1
2
{Z
m−1
(ζ
n
) −Z
m+1
(ζ
n
)};
Z
m
(ζ
n
) = J
m
(ζ
n
), Y
m
(ζ
n
), K
m
(ζ
n
), H
(1)
m
(ζ
n
). (2.26g)
The solution to (2.25) is given by the real part of the following eigenseries expansion:
m
Φ
j
(r, θ, z, t)
[
m
S
j
hω]
=
m
Φ
pj
(r, θ, z, t) +
m
Φ
ej
(r, θ, z, t)
[
m
S
j
hω]
= −Re
I
1
(α, β, a, d, k)
k∆
Ψ
1
(k, z/h)H
(1)
m
(kr)
L
1
(H
(1)
m
(kb))
+
∞
¸
n=2
I
n
(α, β, a, d, κ
n
)
κ
n
∆
Ψ
n
(κ
n
, z/h)K
m
(κ
n
r)
L
n
(K
m
(κ
n
b))
¸
¸
¸
¸
¸
¸
×M
A(P)
(mθ)e
−i(ωt+ν)
. (2.27a,b)
Because the asymptotic behavior of the evanescent eigenseries K
m
(κ
n
r) depends
on the mode m
(1)
, it is not possible to specify a minimum distance from the wave
maker equilibrium boundary at r = b where the evanescent eigenvalues are less
than 1% of their value at the circular wavemaker boundary. The wave ﬁeld must be
computed far away from the wavemaker, and it is understood that far away must
be computed uniquely for each radial mode m for either an AM or PM circular
wavemaker. The evaluation of the power, forces, and moments, and added mass and
radiation damping coeﬃcients for both AM and PM circular wavemakers are given
by Hudspeth.
1
2.4. Directional Wavemakers
Directional wavemakers are vertically segmented wavemakers that undulate sinu
ously and, consequently, are also called snake wavemakers. Segmented directional
wavemakers may be driven either in the middle of each vertical segment or at the
joint between vertical segments. Because of these two methods of wave generation,
parasitic waves are formed along the wavemaker due to either the discontinuity of
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 39
the wavemaker surface (middle segment driven) or of the derivative of the wave
maker surface (joint driven). A dimensional scalar spatial velocity potential ϕ(x, y, z)
may be deﬁned by the real part of
Φ(x, y, z, t) = Re{ϕ(x, y, z) exp−i(ωt +ν)}. (2.28a)
The dimensional linear ﬂuid dynamic pressure ﬁeld p(x, y, z, t) and 3D ﬂuid velocity
vector ﬁeld may be computed from
p(x, y, z, t) = ρ
∂Φ(x, y, z, t)
∂t
, (2.28b)
q(x, y, z, t) = −∇
3
Φ(x, y, z, t), (2.28c)
∇
3
(•) =
∂(•)
∂x
e
x
+
∂(•)
∂y
e
y
+
∂(•)
∂z
e
z
. (2.28d)
The dimensional WMBVP for directional waves is given by
∇
2
3
ϕ(x, y, z) = 0; x ≥ 0; B ≤ y ≤ +B; −h(x, y) ≤ z ≤ 0, (2.29a)
∂ϕ(x, y, 0)
∂z
−k
o
ϕ = 0; x ≥ 0; −B ≤ y ≤ +B; z = 0, (2.29b)
∂ϕ
∂y
= 0; x ≥ 0; y = ±B; −h(x, y) ≤ z ≤ 0, (2.29c)
∂Φ(x, y, z, t)
∂x
= −
∂ξ(y, z, t)
∂t
;
x = 0,
−B ≤ y ≤ +B,
−h(0, y) ≤ z ≤ 0,
(2.29d)
lim
x→+∞
∂
∂x
−iK
n
ϕ(x, y, z) = 0, (2.29e)
∂ϕ(x, y, z)
∂z
= −∇
2
ϕ(x, y, z) · ∇
2
h(x, y); z = −h(x, y), (2.29f)
∇
2
2
(•) =
∂
2
∂x
2
,
∂
2
∂y
2
(•), (2.29g)
ξ(y, z, t) = Re
i
ˆ
U(y, z)
ω
[∆U(y, a)][∆U(z, b, d)] exp−i(ωt +ν)
¸
, (2.29h)
ˆ
U(y, z) =
Sω
∆/h
o
Γ(y)χ(z/h
o
), (2.29i)
∆U(y, a) = U(y +a
−
) −U(y −a
+
),
∆U(z, b, d) = U(z +h −d) −U(z +b),
(2.29j)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
40 R. T. Hudspeth and R. B. Guenther
Fig. 2.3. Deﬁnition sketch for rectangular directional wave basin.
where k
o
= ω
2
/g and where a
±
denotes the (possibly nonsymmetric) transverse ends
of the directional wavemaker in the transverse ydirection in Fig. 2.3. The solution
to the WMBVP in (2.29) is given by the following set of orthonormal eigenfunctions:
ϕ(x, y, z) = i
g
ω
¸
n=1
ζ
n
(x, y)Υ
n
(K
n
, z/h), (2.30a)
Υ
n
(K
n
, z/h) =
Ψ
1
(k, z/h)
Ψ
1
(k, 0)
; n = 1, (2.30b)
Ψ
1
(κ
n
, z/h); n ≥ 2, (2.30c)
k
o
h = K
n
htanhK
n
h = 0; n = 1, 2, 3, . . . , (2.30d)
where K
1
= k and K
n
= +iκ
n
for n ≥ 2 and Ψ
n
(•) is deﬁned in (2.11). The
orthonormal eigenfunctions (2.30b) and (2.30c) are applicable strictly only for con
stant depth wave basins; however, they may be applied to slowly varying depth wave
basins if (2.30b) and (2.30c) are considered to be evaluated only locally over rela
tively small horizontal length scales (e.g., several wavelengths λ), where the depth
may be considered to be locally equal to a constant by a Taylor series expansion of
the depth.
1
Substituting (2.30a) into (2.29a) yields the following 2D Helmholtz equation:
∇
2
2
ζ(x, y) +K
2
n
ζ(x, y) = 0; x ≥ 0; −B ≤ y ≤ +B. (2.31)
Alternatively, for wave basins with mildly sloping bottoms, the mild slope equation
may be applied according to
1
∇
2
• (CC
G
∇
2
ζ(x, y)) +K
2
n
CC
G
ζ(x, y) = 0, (2.32)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 41
where the wave group velocity C
G
is given by (2.16e). If the product CC
G
is a
constant, (2.32) reduces to the 2D Helmholtz equation (2.31). Applying the WKBJ
approximation
1
for the xdependent solution in the method of separation of variables
to (2.32) yields the following solution
1
:
Φ(x, y, z, t) = Re
i
g
ω
ζ(x, y)Υ
n
(K
n
, z/h) exp−i(ωt +ν)
¸
= Re
i
g
ω
M
¸
m=0
¸
n=1
A
mn
[C(x)C
G
(x)]
x=0
C(x)C
G
(x)
Ξ
m
(µ
m
, y/B)
×Υ
n
(K
n
, z/h) expi
x
Q
mn
dξ
exp−i(ωt +ν)
(2.33a)
Ξ
m
(µ
m
, y/B) =
cos µ
m
B(y/B −1)
M
m
; M
m
= 1 +δ
m0
; µ
m
=
mπ
2B
(2.33b–d)
Q
mn
=
K
2
n
−µ
2
m
; m > 0; K
1
= k > µ
2
m
. (2.33e)
The coeﬃcients A
m
may be computed from (2.29d) by expanding the wavemaker
shape function in orthonormal eigenfunctions
1
and are given by
A
mn
k
Q
mn
S
(∆/h
x0
)
= −Ψ
n
(K
n
, 0)I
n
(α, β, b, d, k)
×
+
a
+
/B
−
a
−
/B
Γ
j
(q
j
, y/B)Ξ
m
(µ
m
, y/B)d(y/B). (2.34)
2.4.1. Fulldraft piston wavemaker
The prescribed transverse ycomponent of the snake displacement of a fulldraft
(b = d = 0) piston (α = 0 and β = 1) wavemaker may be expressed as
Γ
j
(q
j
, y/B) =
+∞
¸
j=−∞
˜ c
j
exp i[q
j
B(y/B +ν
y
)], (2.35a)
where the coeﬃcients ˜ c
j
may be computed from the integral in (2.34) by
c
mj
=
+
a
+
/B
−
a
−
/B
Γ
j
(q
j
, y/B)Ξ
m
(µ
m
, y/B)d(y/B)
=
R
a
+
,a
−
+iI
a
+
,a
−
mB(q
2
j
−µ
2
m
)(1 +δ
m0
)
. (2.35b)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
42 R. T. Hudspeth and R. B. Guenther
If the fulldraft piston snake wavemaker spans the entire width of the wave basin
so that a
±
= ±B, then (2.35b) reduces to the integral in (2.34) and
c
mj
=
4q
j
B
sin[q
j
B(ν
j
−1)] + (−1)
m
sin[q
j
B(ν
j
+ 1)]
−i{cos[q
j
B(ν
j
−1)] + (−1)
m
cos[q
j
B(ν
j
+ 1)]}
¸
((q
j
B −(mπ)
2
)
2
)(1 +δ
m0
)
. (2.35c)
Values for c
mj
for (possibly nonsymmetric) values for a
±
are given by Hudspeth.
1
2.5. Sloshing Waves in a 2D Wave Channel
A long rectangular wave channel with a horizontal ﬂat bottom, two rigid vertical
sidewalls, and a wavemaker may generate either 2D, longcrested progressive waves
or two types of transverse waves, viz.,
(1) sloshing waves that are excited directly by transverse motion of the wavemaker
or
(2) cross waves that are excited parametrically by the progressive waves at a
subharmonic of the wavemaker frequency.
The WMBVP for 3D sloshing waves is identical to (2.5) for planar 2D wave
makers except for the KWMBC at x = 0 and an additional kinematic boundary
condition on the sidewalls of the 2D wave channel at y = ±B in Fig. 2.4. The
kinematic and dynamic wave ﬁelds may be computed from a dimensional 3D scalar
Fig. 2.4. Deﬁnition sketch for a sloshing wave channel.
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 43
velocity potential Φ(x, y, z, t). The ﬂuid velocity q(x, y, z, t) may be computed from
the negative directional derivative of a scalar velocity potential by
q(x, y, z, t) = −∇Φ(x, y, z, t). (2.36a)
The dimensional ﬂuid dynamic pressure ﬁeld p(x, y, z, t) may be computed from
p(x, y, z, t) = ρ
∂Φ(x, y, z, t)
∂t
. (2.36b)
A spatial velocity potential ϕ(x, y, z) may be deﬁned by the real part of
Φ(x, y, z, t) = Re{ϕ(x, y, z) exp−i(ωt +ν)}. (2.36c)
The WMBVP for sloshing waves is given by the following:
∇
2
ϕ(x, y, z) = 0; x ≥ 0; −B ≤ y ≤ +B; −h ≤ z ≤ 0, (2.37a)
∂ϕ(x, y, −h)
∂z
= 0; x ≥ 0; −B ≤ y ≤ +B; z = −h, (2.37b)
∂ϕ(x, y, 0)
∂z
−k
o
ϕ = 0; x ≥ 0; −B ≤ y ≤ +B; z = 0, (2.37c)
∂Φ(x, y, z, t)
∂x
= −
∂ξ(y, z, t)
∂t
; x = 0; −B ≤ y ≤ +B; −h ≤ z ≤ 0, (2.37d)
∂ϕ
∂y
= 0; x ≥ 0; y = ±B; −h ≤ z ≤ 0, (2.37e)
lim
x→+∞
∂
∂x
−iK
n
ϕ(x, y, z) = 0, (2.37f)
η(x, y, t) = Re
−iω
g
Φ(x, y, 0, t)
;
x ≥ 0,
−B ≤ y ≤ +B,
z = 0.
(2.37g)
A solution to (2.37) is given by the following eigenfunction expansions
1
:
Φ(x, y, z, t) = Re
¸
n=1
˜
ψ
n
(x, y)Ψ
n
(K
n
, z/h) exp−i(ωt +ν)
¸
, (2.38a)
ϕ(x, y, z) =
˜
ψ
n
(x, y)Ψ
n
(K
n
, z/h), (2.38b)
η(x, y, t) = Re
¸
n=1
ζ
n
(x, y) exp −i(ωt +ν)
¸
, (2.38c)
ζ
n
(x, y) = −i
ω
g
˜
ψ
n
(x, y)Ψ
n
(K
n
, 0), (2.38d)
˜
ψ
n
(x, y) = i
g
ω
ζ
n
(x, y)
Ψ
n
(K
n
, 0)
, (2.38e)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
44 R. T. Hudspeth and R. B. Guenther
where ζ
n
(x, y) is sometimes referred to as a displacement potential. The scalar
potential (2.38a) may be expressed from (2.38d) and (2.38e) as
Φ(x, y, z, t) = Re
¸
n=1
i
g
ω
ζ
n
(x, y)
Ψ
n
(K
n
, z/h)
Ψ
n
(K
n
, 0)
exp −i(ωt +ν)
¸
. (2.39)
The WMBVP may be expressed in terms of a displacement potential ζ
n
(x, y) by
∂
2
ζ
n
(x, y)
∂x
2
+
∂
2
ζ
n
(x, y)
∂y
2
+ K
2
n
ζ
n
(x, y) = 0;
x ≥ 0,
−B ≤ y ≤ +B,
−h ≤ z ≤ 0.
(2.40a)
¸
n=1
∂ζ
n
(x, y)
∂x
Ψ
n
(K
n
, z/h)
Ψ
n
(K
n
, 0)
= i
ω
g
U(y, z)
x = 0,
B ≤ y ≤ +B,
−h ≤ z ≤ 0.
(2.40b)
lim
x→+∞
∂
∂x
−iK
n
ζ
n
(x, y) = 0. (2.40c)
∂ζ
n
(x, y)
∂y
= 0; x ≥ 0; y = ±B; −h ≤ z ≤ 0, (2.40d)
where (2.40a) is the 2D Helmholtz equation.
9,10
Because the boundary conditions are prescribed on boundaries that are constant
values of (y,z), a solution to the WMBVP (2.40) may be computed by the method
of separation of variables and is given by
1
Φ(x, y, z, t)
= Re
i
g
ω
M
¸
m=0
C
m1
Υ
m
(y/B)
Ψ
1
(k, z/h)
Ψ
1
(k, 0)
expiP
m1
x
+
¸
m=M+1
C
m1
Υ
m
(y/B)
Ψ
1
(k, z/h)
Ψ
1
(k, 0)
exp
−Ξ
m1
x
+
¸
m=0
¸
n=2
C
mn
Υ
m
(y/B)
Ψ
n
(κ
n
, z/h)
Ψ
n
(κ
n
, 0)
exp
−Q
mn
x
exp−i(ωt +ν)
,
(2.41a)
where
Υ
m
(µ
m
, y/B) =
cos µ
m
B(y/B −1)
M
m
; M
2
m
= 1 +δ
m0
; (2.41b)
µ
m
=
mπ
2B
,
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 45
P
m1
=
K
2
1
−µ
2
m
=
k
2
− µ
2
m
; k > µ
m
; m ≤ M, (2.41c)
Ξ
m1
=
µ
2
m
−k
2
; k < µ
m
; m > M, (2.41d)
Q
2
mn
= µ
2
m
−K
2
n
; m ≥ 0; n ≥ 1, (2.41e)
n ≥ 1: K
1
= k > µ
m
: Q
m1
= i
k
2
−µ
2
m
= iP
m
; m ≤ M, (2.41f)
n = 1: k = µ
m
: Q
m1
= 0
k < µ
m
: Q
m1
=
µ
2
m
−k
2
= Ξ
m1
> 0; m > M, (2.41g)
n ≥ 2 : K
n
= iκ
n
: Q
mn
=
µ
2
m
+κ
2
n
> 0, (2.41h)
where M is the maximum integer value for m in order for µ
m
< k. Substituting
(2.41a) into (2.40b) yields the following solutions for C
mn
1
:
C
m1
=
ω
g
Ψ
1
(K
n
, 0)
P
m1
+1
−1
d
y
B
0
−1
d
z
h
U
y,
z
h
Ψ
1
K
n
,
z
h
Υ
m
µ
m
y
B
;
m ≤ M (2.42a)
C
m1
= −i
ω
g
Ψ
1
(K
n
, 0)
Ξ
m1
+1
−1
d
y
B
0
−1
d
z
h
U
y,
z
h
Ψ
1
K
n
,
z
h
Υ
m
µ
m
,
y
B
;
m > M + 1 (2.42b)
C
mn
= −i
ω
g
Ψ
n
(K
n
, 0)
Q
mn
+1
−1
d
y
B
0
−1
d
z
h
U
y,
z
h
Ψ
n
K
n
,
z
h
Υ
m
µ
m
,
y
B
;
m ≥ 0, n ≥ 2. (2.42c)
The ﬁrst three transverse eigenmodes are illustrated in Fig. 2.5.
2.6. Conformal Mappings
Conformal and domain mappings are applications of complex variables to solve 2D
boundary value problems. Conformal mapping is an angle preserving transformation
that will compute exact nonlinear solutions for surface gravity waves of constant
Fig. 2.5. First three transverse eigenmodes in a 2D wave channel.
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
46 R. T. Hudspeth and R. B. Guenther
form that may be treated as a steady ﬂow following a Galilean transformation from a
ﬁxed inertial coordinate system to a noninertial moving coordinate system. Domain
mapping is a transformation of the wavemaker geometry into a ﬁxed computational
domain where a solution may be computed eﬃciently.
2.6.1. Conformal mapping
1
Conformal mapping of the WMBVP provides a global solution that accurately
accounts for the singular behavior at all irregular points. The irregular points in
the physical wavemaker domain are transformed into both weak and strong singular
kernels in a Fredholm integral equation. The two irregular points on the WMBVP
boundary are located at (1) the intersection between the freesurface and the wave
maker boundary and (2) the intersection between the horizontal bottom and the
wavemaker boundary. These two irregular points exhibit integrable weakly singular
kernels. The farﬁeld radiation boundary exhibits a strongly singular kernel and sig
niﬁcantly aﬀects the solution. The irregular frequencies
3,4
are included in the global
solution by the Fredholm alternative. A theory for the planar WMBVP computes
a global solution by a conformal mapping of the physical wavemaker boundary
to a unit disk that includes the motion of an inviscid ﬂuid near irregular points
that are illustrated in Fig. 2.6. A numerical solution to Laplace’s equation in a
transformed unit disk may be computed from a Fredholm integral equation. The
WMBVP deﬁned by (2.5) is transformed into complexvalued analytical functions
where the complexvalued coordinates are deﬁned as z = x + iy. The coordinates
for the semiinﬁnite wave channel in Fig. 2.1(a) must be transformed to complex
valued coordinates z. The ﬂuid velocity q(x, y, t) and dynamic pressure p(x, y, t)
may be computed from a scalar velocity potential Φ(x, y, t) according to
q(x, y, t) = −∇Φ(x, y, t); p(x, y, t) = ρ
∂Φ(x, y, t)
∂t
. (2.43a,b)
The WMBVP and Type I wavemaker shape function are given by (2.5)–(2.7).
There are both Irregular (I) and Regular (R) points at the intersections between
the Smooth (S) and Critical (C) boundaries B
1
and B
2
in the WMBVP as illustrated
in Fig. 2.6 where these two boundary intersection points are identiﬁed as P
1
and P
2
.
The classiﬁcation of the boundary points P
1
and P
2
in Fig. 2.6 depends on (1) the
boundary conditions ϕ
i
(P
j
) and (2) the continuity of the boundaries B
m
and their
derivatives where i, j, and m = 1 or 2. A conformal mapping of the semiinﬁnite
wave channel strip in the physical plane will yield a Fredholm integral equation,
6,7
where these critical points may be transformed to singular points that are integrable
over a smooth continuous mapped boundary.
2.6.1.1. Conformal mapping to the unit disk 1
Two conformal mappings are: (i) the physical zplane to a semicircle in the Zplane
shown in Fig. 2.7; and (ii) a semicircle in the upper Zplane to a unit disk in the
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 47
Fig. 2.6. Combinations of Irregular (I) and Regular (R) boundary points P
1
and P
2
between
Smooth (S) and Critical (C) boundaries B
1
and B
2
intersections in the WMBVP.
8
Fig. 2.7. Mapping of the semiinﬁnite strip in the lower half x–yplane in the physical zplane to
the upper half X–Y plane in the Zplane.
8
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
48 R. T. Hudspeth and R. B. Guenther
Fig. 2.8. Mapping of the upper halfplane in the Zplane to the unit disk in the Qplane.
Qplane shown in Fig. 2.8. The Schwarz–Christoﬀel transformation
dz
dZ
=
C
1
√
Z + 1
√
Z −1
(2.44a)
may be integrated to obtain
z = x +iy
= −
h
π
Log[−Z −
Z
2
−1], (2.44b)
where the Log[•] denotes the principal value of Log[•] and the argument of the
Log[•] is −π ≤ arg < π. Inverting (2.44b) yields
Z = X +iY = −cosh(πz/h), (2.44c)
where
X = −cosh
πx
h
cos
πy
h
, (2.44d)
Y = −sinh
πx
h
sin
πy
h
. (2.44e)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 49
The radiation boundary in the zplane transforms to a semicircle in the Zplane by
R
2
= X
2
+Y
2
=
1
2
¸
cos
2πy
h
+ cosh
2πx
h
, (2.44f)
tan θ =
Y
X
= tanh
πx
h
tan
πy
h
. (2.44g)
Details of the transformation of the WMBVP are given by Hudspeth.
1
2.6.1.2. Mapping Zplane to a unit disk
The upperhalfplane of the Zplane may be mapped into a unit disk shown in
Fig. 2.8 by the following bilinear transformation:
Q = ξ +iζ = exp(iθ
0
)
¸
i −Z
i +Z
, (2.45a)
that may be inverted to obtain
Z = X +iY = i
¸
1 −Q
1 +Q
, (2.45b)
and the mapping function coordinates are
ξ =
1 −X
2
−Y
2
X
2
+ (Y + 1)
2
,
ζ =
2X
X
2
+ (Y = 1)
2
,
(2.45c,d)
that may be transformed into the cylindrical coordinates for the unit disk in
Fig. 2.8 by
r
2
= ξ
2
+ζ
2
=
(X
2
+Y
2
−1)
2
−4X
2
(X
2
+ (Y + 1)
2
)
2
; = arctan
ζ
ξ
= arctan
2X
1 −X
2
−Y
2
.
(2.45e,f)
Details of the transformation are given by Hudspeth.
1
A numerical solution to
the transformed WMBVP may be computed from the following Fredholm integral
equation
1
:
2π
π
ˆ
Φ(r, ) = −
+π
−π
¸
ˆ
Φ(ˆ r, ˆ )
∂G(r, ˆ r, , ˆ )
∂ˆ r
+G(r, ˆ r, , ˆ )
∂
ˆ
Φ(ˆ r, ˆ )
∂ˆ r
¸
ˆ rd ˆ ,
(2.46a)
where
G(r, ˆ r, , ˆ ) = −ln[ρ(r, ˆ r, , ˆ )]; ρ
2
(r, ˆ r, , ˆ ) = r
2
− 2rˆ r cos( ˆ −) + ˆ r
2
.
(2.46b,c)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
50 R. T. Hudspeth and R. B. Guenther
Fig. 2.9. Nodal points on the unit disk in the Qplane and the corresponding nodal points on the
wavemaker in the physical zplane.
8
Numerical solutions to (2.46a) may be computed by discretizing the unit disk
boundary as shown in Fig. 2.9. The numerical details regarding the evaluation
(2.46a) at the two weakly singular irregular points at B and D in the physical
zplane in Fig. 2.7 and the strongly singular point at ±π that is the vertical
radiation boundary A–E at +∞ in the physical zplane in Fig. 2.7 are tedious.
8
Global numerical solutions may be computed for both the linear and the nonlinear
WMBVPs.
2.6.1.3. Conformal mapping to the unit disk 2
The wavemaker geometry shown in Fig. 2.10 is mapped to the unit disk by two
transformations. The WMBVP is given by (2.5) and the WM shape function is
χ(y/h) = [α(y/h) +β][U(y/h + 1 −b
0
/h) −U(y/h +a
0
/h)]. (2.47)
In order to transform the wavemaker geometry to a Jacobian elliptic function,
it must be rotated and translated as shown in Fig. 2.11. The 90
◦
rotation to the
z
plane is given by
z
= x
+iy
= iz = −y +ix. (2.48a)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 51
Fig. 2.10. WMBVP
11
with the six critical boundary points at a–a
0
–b–b
0
–c–d.
Fig. 2.11. Rotation and translation of the physical wavemaker rectangular strip in the zplane to
the wplane.
11
The horizontal shift to the left in the z
plane is given by
z
= x
+iy
= w
−h/2 = −y −h/2 +ix. (2.48b)
In order to map the WM geometry in the zplane to the semicircle in the Zplane
shown in Fig. 2.12 as a Jacobian elliptic function, the rotated and translated strip
in the z
plane must have the dimensions of −K ≤ ξ ≤ +K and 0 ≤ ζ ≤ K
, where
K = h/2 and K
= 3h = 6K. This requires a coordinate ampliﬁcation given by
w =
2K
h
(x
+iy
)
=
2K
h
−y −
h
2
+ix
. (2.48c)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
52 R. T. Hudspeth and R. B. Guenther
Fig. 2.12. Mapping of the wavemaker semicircle in the Zplane to the unit disk in the Qplane.
11
The Schwarz–Christoﬀel transformation from the wplane to the Zplane is
dw =
ˆ
CkdZ
(a −Z)(b −Z)(c −Z)(d − Z)
. (2.48d)
The following change of variables:
Z = a
ˆ
Z; dZ = ad
ˆ
Z : κ = a/c;
ˆ
C = c,
modiﬁes (2.48d) to the following Jacobian elliptic function:
w =
Z
0
d
ˆ
Z
[(1 −
ˆ
Z
2
)(1 −κ
2 ˆ
Z
2
)]
1
2
¸
= sn
−1
[Z, κ], (2.48e)
where sn[Z, κ] = the Jacobian elliptic function of modulus κ or sine amplitude
function.
9
Deﬁne
ˆ
k = sn
−1
[1, κ], (2.48f)
and the mapping of the rectangle {x
1
, x
2
; y
1
, y
2
} = {0, 3h; 0, h} is given by
Z = X +iY
=
sn
−
2K
h
y +
h
2
, κ
dn
2Kx
h
,
ˆ
k
1 −dn
2
−
2K
h
y +
h
2
, κ
sn
2
2Kx
h
,
ˆ
k
+i
cn
−
2K
h
y +
h
2
, κ
dn
−
2K
h
y +
h
2
, κ
, sn
2Kx
h
,
ˆ
k
cn
2Kx
h
,
ˆ
k
1 −dn
2
−
2K
h
y +
h
2
, κ
sn
2
2Kx
h
,
ˆ
k
,
(2.48g)
where sn[•, •] in the copolar halfperiod trio in (2.48g) is deﬁned in (2.48e), and
cn[•, •] and dn[•, •] are deﬁned by cn
2
[•, •] = 1−sn
2
[•, •], dn
2
[•, •] = 1−κ
2
sn
2
[•, •].
The mapping from the semicircle in the Zplane to the unit disk in the Qplane
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 53
Fig. 2.13. Transformed boundary conditions mapped to arcs on the perimeter of the unit disk.
11
is shown in Fig. 2.12; and the mapping to the unit disk in the Qplane is shown in
Fig. 2.13. The mapping of the Zplane to the Qplane is given by
Q =
i −Z − ˆ α(i +Z)
i +Z − ˆ α(i −Z)
=
(1 − ˆ α)
2
−Z
2
(1 + ˆ α)
2
+ 2iZ(1 − ˆ α)
2
(1 − ˆ α
2
) +Z
2
(1 + ˆ α)
2
, (2.49a)
where −1 < ˆ α < +1. Changing variables to circular cylindrical coordinates by
R
2
(X, Y ) =
(1 − ˆ α
2
) −2Y (1 − ˆ α
2
) + (X
2
+Y
2
)(1 + ˆ α)
2
(1 − ˆ α)
2
−2Y (1 − ˆ α
2
) + (X
2
+Y
2
)(1 + ˆ α
2
)
, (2.49b)
θ(X, Y ) = arctan
¸
2X(1 − ˆ α
2
)
(1 − ˆ α)
2
−(X
2
+Y
2
)(1 + ˆ α)
2
, (2.49c)
the unit disk may be transformed into functions of the copolar trio of Jacobian
elliptic functions. The transformed WMBVP in circular cylindrical coordinates
is given by Hudspeth.
1
A general solution to the transformed WMBVP may be
written as
10
ϕ(R, θ) =
N
¸
n=0
R
n
¸
ˆ a
n
1 + δ
n0
cos nθ +
ˆ
b
n
sinnθ
, (2.50)
where δ
ij
is the Kronecker delta function. Substituting (2.50) into the generic
boundary conditions on each of the six arcs on the perimeter of the unit disk illus
trated in Fig. 2.13, multiplying each of these six boundary conditions by a member
of the set of the orthogonal series in (2.50), integrating over the interval of orthogo
nality −π ≤ θ ≤ +π yields the following matrix equation for each of the coeﬃcients
ˆ a
n
and
ˆ
b
n
AB = H. (2.51)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
54 R. T. Hudspeth and R. B. Guenther
Fig. 2.14. Physical ﬂuid domain.
12
2.7. Domain Mapping
Domain mapping of the WMBVP
12
follows the theory by Joseph.
13
The physical
ﬂuid domain shown in Fig. 2.14 for the fully nonlinear WMBVP is mapped to a
ﬁxed computational ﬂuid domain, and the discretized coupled freesurface boundary
conditions are computed by an implicit Crank–Nicholson (C–N) method.
14,15
At
each iteration of the C–N method, the potential ﬁeld is computed by the conjugate
gradient method.
15
The wavemaker motion Ξ(y/h, t) is assumed to be periodic with
period T = 2π/ω, and the WMBVP with the surface tension
ˆ
T is given by
∇
2
Φ(x, y, t) = ∆Φ(x, y, t) = 0;
0 ≤ y ≤ Γ(x, t),
Ξ(y/h, t) ≤ x ≤ L.
(2.52a)
∂Φ(x, y, t)
∂t
+
1
2
∇Φ(x, y, t)
2
−
ˆ
T
ρ
∂
2
Γ(x, t)
∂x
2
¸
1 +
∂Γ(x, t)
∂x
2
¸
3/2
+gΓ(x, t) = 0. (2.52b)
∂Φ(x, y, t)
∂y
−
∂Γ(x, t)
∂x
∂Φ(x, y, t)
∂x
+
∂Γ(x, t)
∂t
= 0;
Ξ(Γ(x, t), t) ≤ x ≤ L; y = Γ(x, t). (2.52c)
∂Φ(x, y, t)
∂y
= 0; Ξ(Γ(x, t), t) ≤ x ≤ L; y = 0. (2.52d)
∂Φ(x, y, t)
∂x
= 0; x = L; 0 ≤ y ≤ Γ(L, t). (2.52e)
∂Φ(x, y, t)
∂x
= −
∂Ξ(y/h, t)
∂t
+
∂Ξ(y/h, t)
∂y
∂Φ(x, y, t)
∂y
;
x = Ξ(y/h, t),
0 ≤ y ≤ Γ(0, t).
(2.52f)
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
Wavemaker Theories 55
The initial conditions for t = 0 are
Γ(x, 0) = H(x); (2.52g)
∂Γ(x, t) = 0;
Ξ(x, 0) ≤ x ≤ L. (2.52h)
Φ(x, Γ(x, 0), 0) = 0; (2.52i)
The physical ﬂuid domain shown in Fig. 2.14 may be mapped into a dimensionless
ﬁxed rectangle of dimensions 0 ≤ ξ ≤ 1 by 0 ≤ ζ ≤ 1 by the transforms
ξ =
x
L
; ζ =
y
Γ(x, t)
; τ = ωt; γ(ξ, τ) =
Γ(x, t)
h
, (2.53a–d)
and dimensionless variables by
q(ξ, ζ, τ) = −
∇Φ(x, y, t)
Aω
; p(ξ, ζ, τ) =
P(x, y, t)
ρA
2
ω
2
, (2.53e,f)
ϕ(ξ, ζ, τ) =
Φ(x, y, t)
Ahω
; w(ζ, τ) =
Ξ(y/h, t)
S
;
ˆ
T =
˜
T
ρALhω
2
. (2.53g–i)
Because ζ is a function of both x and y in (2.53b), transforming partial derivatives
with respect to x must be done with some care.
12
Details of these lengthy transfor
mations and the transformed WMBVP in the ﬁxed mapped domain are given by
Hudspeth.
1
References
1. R. T. Hudspeth, Waves and Wave Forces on Coastal and Ocean Structures (World
Scientiﬁc, Singapore, 2006).
2. R. T. Hudspeth, J. M. Grassa, J. R. Medina and J. Lozano, J. Hydraulic Res. 32, 387
(1994).
3. F. John, Commun. Pure Appl. Math. 2, 13 (1949).
4. F. John, Commun. Pure Appl. Math. 3, 45 (1950).
5. T. H. Havelock, Phil. Mag. 8, 569 (1929).
6. P. R. Garabedian, Partial Diﬀerential Equations (Wiley, Inc., New York, 1964).
7. P. M. Morse and H. Feshbach, Methods of Theoretical Physics (McGrawHill Book
Company, New York, 1953).
8. Y. Tanaka, Irregular points in wavemaker boundary value problem, PhD thesis,
Oregon State University (1988).
9. G. F. Carrier, M. Krook and C. E. Pearson, Functions of a Complex Variable, Theory
and Technique (McGrawHill Book Co. Inc., New York, 1966).
10. R. B. Guenther and J. Lee, Partial Diﬀerential Equations of Mathematical Physics
and Integral Equations (Dover Publications, Inc., New York, 1996).
11. P. J. Averbeck, The boundary value problem for the rectangular wavemaker,
MS thesis, Oregon State University (1993).
12. S. J. DeSilva, R. B. Guenther and R. T. Hudspeth, Appl. Ocean Res. 18, 293 (1996).
July 31, 2009 8:18 9.75in x 6.5in b684ch02 FA
56 R. T. Hudspeth and R. B. Guenther
13. D. D. Joseph, Arch. Rational Mech. Anal. 51, 295 (1973).
14. B. Carnahan, H. A. Luther and J. O. Wilkes, Applied Numerical Methods (John Wiley
and Sons, New York, 1965).
15. R. Glowinski, Numerical Methods for Nonlinear Variational Problems (Springer
Verlag, 1984).
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Chapter 3
Analyses by the Melnikov Method of Damped
Parametrically Excited Cross Waves
Ronald B. Guenther
Department of Mathematics
Oregon State University, Corvallis, OR 97331, USA
guenther@math.orst.edu
Robert T. Hudspeth
School of Civil and Construction Engineering
Oregon State University, Corvallis, OR 97331, USA
robert.hudspeth@orst.edu
The Wiggins–Holmes extension of the generalized Melnikov method (GMM)
to higher dimensions and the extension of the Generalized Herglotz Algorithm
(GHA) to nonautonomous systems are applied to weakly damped parametri
cally excited cross waves with surface tension in order to demonstrate that cross
waves are chaotic. The GMM is a global perturbation analysis about a mani
fold of ﬁxed points that are connected by separatrices for higher dimensional
nonlinear dynamical systems. The Luke Lagrangian, density function for surface
gravity waves with surface tension and dissipation is expressed in three gen
eralized coordinates that are the timedependent components of three velocity
potentials that represent three standing waves. The Hamiltonian for these cross
waves is homomorphic to the Hamiltonian for a parametrically excited pendulum
that is an example of a Floquet oscillator that may be approximated by the
Mathieu equation. Neutral stability curves measured from wave tank data moti
vated the inclusion of dissipation in the Luke Lagrangian density function for
cross waves. An integral containing a generalized dissipation function that is pro
portional to the Stokes material derivative of the free surface is added to the Luke
Lagrangian integral so that dissipation is correctly incorporated into the dynamic
free surface boundary condition. The generalized momenta are computed from
the Lagrangian; and the Hamiltonian is computed from a Legendre transform of
the Lagrangian. This Hamiltonian contains nonautonomous components and is
transformed by three canonical transformations in order to obtain a suspended
system for the application of the Kolmogorov–Arnold–Moser (KAM) averaging
57
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
58 R. B. Guenther and R. T. Hudspeth
operation and the GMM. The system of nonlinear nonautonomous evolution equa
tions determined from Hamilton’s equations of motion of the second kind must
be averaged in order to obtain an autonomous system that may be analyzed by
the GMM. Hyperbolic saddle points that are connected by heteroclinic orbits
are computed from the unperturbed autonomous system. The nondissipative per
turbed Hamiltonian system with surface tension satisﬁes the KAM nondegeneracy
requirements; and the Melnikov integral is calculated to demonstrate that the
motion is chaotic. For the perturbed dissipative system with surface tension, the
only hyperbolic ﬁxed point that survives the averaged equations is a ﬁxed point
of weak chaos that is not connected by a homoclinic orbit; and, consequently, the
Melnikov integral is identically zero. The chaotic motion for the perturbed dis
sipative system with surface tension is demonstrated by numerical computation
of positive Liapunov characteristic exponents. A chaos diagram of the largest
Liapunov exponent demonstrates regions in the Floquet forcing parameter space
of possible chaotic motions; and the range of values of the Floquet parametric
forcing parameter ε and of the wavemaker dissipation parameter α in the α−ε
space where chaotic motions may exist.
1,3
3.1. Introduction
The cross waves shown in Fig. 3.1 are excited parametrically by progressive wave
maker waves at a subharmonic of the wavemaker frequency.
1
Parametrically excited
standing cross waves that oscillate in a direction transverse to the wavemaker
forcing with crests perpendicular to the wavemaker are analyzed by the gener
alized Melnikov method (GMM) and by the Generalized Herglotz Algorithm (GHA)
extended to nonautonomous systems. The cross wave wavelengths and wave modes
Fig. 3.1. Mode 2 cross wave in a 2D wave channel.
1–3
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 59
possible are determined by the channel width L
c
= 2(channel width)/n, where
n = the mode number and is equal to the number of halfwavelengths across the
channel. Floquet instabilities occur when the cross wave wavelengths also satisfy
the frequency dispersion equation for surface gravity waves. Energy is transferred
from the progressive wavemaker waves to the parametrically forced cross waves
through the spatial mean motion of the free surface and their growth is due to
the rate of working of the transverse stresses of the progressive wavemaker waves.
The cross wave instability is shown to be homomorphic to the Floquet oscil
lator instability
2
by constructing the neutral Floquet stability diagram shown in
Fig. 3.2. The simultaneous generation of a primary resonance (ω
p
: ω
c
= 2 : 1)
and a secondary resonance (ω
p
: ω
c
= 1 : 1) may be observed in the cross wave
data. The criteria for horseshoe maps and for homoclinic/heteroclinic orbits are
selected to test whether or not cross waves are a chaotic dynamical system. Specif
ically, the GMM provides local criterion for the transverse intersection of stable
and unstable manifolds of the perturbed system and for the resulting chaotic
motion near the unperturbed (undamped and unforced) homoclinic/heteroclinic
orbits. In order to apply the GMM to a suspended dynamical system that will
survive the KAM averaging operation, the nonautonomous Hamiltonian is trans
formed by three separate canonical transformations by applying the GHA for nonau
tonomous transformations. Three canonical transformations are applied in order
to (1) eliminate cross product terms by a rotation of axes; (2) to eliminate two
degrees of freedom, and (3) to suspend the nonautonomous terms in the wave
maker forcing component of the Hamiltonian by applying the Hamilton–Jacobian
transformation. The Liapunov characteristic exponents represent an alternative cri
terion for diagnosing the chaotic behavior of a dynamical system by measuring
the mean rate of exponential divergence of nearby trajectories, and is computed
numerically for the perturbed dissipative Hamiltonian when the GMM fails to
predict chaos.
Fig. 3.2. Neutral Floquet stability diagram for mode 2 cross waves (o = mode 2 cross waves and
= no cross waves).
1,2
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
60 R. B. Guenther and R. T. Hudspeth
Fig. 3.3. Deﬁnition sketch of ﬂuid domain.
3.2. Hamilton’s Principle for Cross Waves
The ﬂuid is assumed to be incompressible and inviscid and the ﬂow to be irrota
tional. The dimensional ﬂuid particle velocities u
and the dimensional total pressure
in the ﬂuid P
are computed from
u
= −
¯
∇
ϕ
;
P
ρ
= −g
z
+ϕ
t
−
1
2
[
¯
∇
ϕ
[
2
; (3.1a,b)
where ϕ
= a velocity potential; ρ
= the ﬂuid mass density; and g
= the gravita
tional acceleration. The ﬂuid domain is the 3D rectangular wave channel shown in
Fig. 3.3.
3.2.1. Variational principle
The Lagrangian L
with kinematic surface tension T
is given by
L
=
V
(t
)
¸
1
2
[
¯
∇
ϕ
[
2
−ϕ
t
+g
z
dV
+ T
S
η
(ζ
−1)
ζ
dS
η
; (3.2a)
where η
= the free surface proﬁle and where
dS
η
= ζ
dx
dy
; ζ
= [1 +[
¯
∇
ϕ
[
2
]
1/2
. (3.2b,c)
Generalized Hamilton’s principle with dissipation.
4
The ﬁrst variational of
(3.2) is
δ
t
2
t
1
L
dt
+
t
2
t
1
F
Dη
Dt
dt
= 0; (3.3a)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 61
where the generalized dissipation function per unit mass density F
(Dη
/Dt
) is
F
Dη
Dt
= −α
g
κ
S
η
Dη
Dt
δη
ζ
dS
η
; (3.3b)
where α = a dimensionless damping parameter and κ
= the cross wave
wavenumber. The dimensional boundaryvalueproblem may be obtained by
requiring that the variation of φ
and η
vanish at the arbitrary values t
1
and t
2
,
and is given by
δL
+
d
dt
(t
)
δϕ
dV
=
b
−b
χ
η
¸
1
2
[∇
ϕ
[
2
−ϕ
t
+g
η
−T
∇
(∇
η
/ζ
)
z
=η
δη
dx
dy
−
V
(t
)
(∇
2
ϕ
)δϕ
dV
+
S
(ϕ
n
+U
s
ˆ n
s
)
δϕ
dS
s
+T
S
η
η
n
δη
/ζ
dS
η
, (3.4)
where n
denotes a normal derivative. The Lagrangian in (3.2) reduces to
1
L
= −
1
2
V
ϕ
∇
2
ϕ
dV
+
S
1
2
ϕ
n
+U
s
ˆ n
s
ϕ
dS
+
1
2
b
−b
χ
η
g
η
2
dx
dy
+T
b
−b
χ
η
(ζ
−1)dx
dy
. (3.5)
The velocity potential ϕ
is the ﬁeld variable, and the free surface η
may be
expressed in terms of ϕ
by the free surface boundary conditions and the contact
line conditions at the vertical sidewalls.
1,3
The velocity potential ϕ
and the free
surface displacement η
are assumed to be linear combinations of the progressive
wavemaker wave and the parametrically excited cross wave given by the following:
ϕ
(x
, y
, z
, t
) = ϕ
p
(x
, z
, t
) +ϕ
c
(y
, z
, t
), (3.6a)
η
(x
, y
, t
) = η
p
(x
, t
) +η
c
(y
, t
). (3.6b)
The dimensional variables may be scaled by the following scales
1,5
:
x
=
x
k
; y
=
y
κ
; z
=
z
κ
; t
=
t
√
g
κ
;
h = κ
h
; ξ = k
; b = κ
b
;
(3.7a–g)
ω
2
p
= g
k
τ
λ
; τ
λ
= 1 + (τ/λ
4
r
); τ = T
κ
2
/g
;
τ
2
1
= 1 +τ; λ
2
r
= κ
/k
; ω
2
c
= g
κ
τ
2
1
;
(3.7h–m)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
62 R. B. Guenther and R. T. Hudspeth
ϕ
p
= ϕ
p
a
p
g
k
; ϕ
c
= ϕ
c
a
c
g
κ
; η
p
= a
p
η
p
;
η
c
= a
c
η
c
; χ
= a
w
χ; L
=
La
2
c
g
k
κ
.
(3.7n–s)
Scaling by (3.7) yields the following dimensionless scaling parameters
1,3
:
ε = κ
a
c
; β = ω
c
/ω
p
; γ = k
a
w
; Γ = a
p
/a
c
; (3.8a–d)
and the following dimensionless ﬁeld variables:
ϕ(x, y, z, t) = ϕ
c
(y, z, t) + Γλ
r
ϕ
p
(x, z, t);
η(x, y, t) = η
c
(y, t) + Γη
p
(x, t).
(3.9a,b)
The independent variation of ϕ
and η
vanish at arbitrary temporal values t
1
and t
2
in Hamilton’s principle (3.3) yielding the following scaled boundary value problem:
1
λ
4
r
ϕ
xx
+ϕ
yy
+ϕ
zz
= 0; γχ ≤ x ≤ ξ, y ≤ [b[, −h ≤ z ≤ εη, (3.10a)
ϕ
z
= −τ
1
η
t
+ε
1
λ
4
r
ϕ
x
η
x
+ϕ
y
η
y
; z = εη, γχ ≤ x ≤ ξ, y ≤ [b[,
(3.10b)
ε
2
1
λ
4
r
ϕ
2
x
+ϕ
2
y
+ϕ
2
z
−τ
1
ϕ
t
+η = τ
1
λ
4
r
η
xx
+η
yy
−αϕ
z
;
γχ ≤ x ≤ ξ,
y ≤ [b[, z = εη
(3.10c)
∇ϕ ˆ n = 0;
y = [b[, −h ≤ z ≤ εη, γχ ≤ x ≤ ξ
z = −h, y ≤ [b[, γχ ≤ x ≤ ξ,
(3.10d,e)
ϕ
x
= −
γ
ε
λ
4
r
τ
1
χ
t
+γλ
4
r
ϕ
z
χ
z
; x = γχ, y ≤ [b[, −h ≤ z ≤ εη, (3.10f)
plus an appropriate radiation condition at x = ξ. The free surface curvature requires
a dynamical constraint that is given either by the contact line condition η
n
= 0 or
by the edge constraint boundary condition η
= 0.
1,3
The wavemaker perturbation
forcing parameter γ is smaller than the Floquet parametric forcing parameter ε
because experiments demonstrate that the standing cross wave amplitude becomes
larger than the wavemaker forcing amplitude as t →∞. The parameter ordering is
0 < γ
2
< εγ < ε
2
< γ < ε < 1 or 0 <
γ
2
ε
< γ <
γ
ε
< 1. (3.11a,b)
The higher order terms that will be neglected are O(ε
2
), O(εγ), and O(γ
2
); but
terms O(γ/ε) will be retained. Neglecting these higher order terms, the nondi
mensional Lagrangian integrals may be approximated by the following linear
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 63
combination
1,3
:
L
χ
= −
γτ
1
2ε
b
−b
0
−h
[(ϕ
c
+ Γλ
r
ϕ
p
)χ
t
]
x=0
dzdy
+
b
−b
[εχ
t
(η
c
+ Γη
p
)(ϕ
c
+ Γλ
r
ϕ
p
)]
x,z=0
dy, (3.12a)
L
ξ
= −
Γ
2λ
3
r
b
−b
0
−h
[ϕ
px
(ϕ
c
+ Γλ
r
ϕ
p
)]
x=ξ
dzdy
+
b
−b
[εϕ
px
(η
c
+ Γη
p
)(ϕ
c
+ Γλ
r
ϕ
p
)]
x=ξ,z=0
dy, (3.12b)
L
η
=
1
2
b
−b
ξ
0
¦τ
1
(η
c
+ Γη
p
)
t
[(ϕ
c
+ Γλ
r
ϕ
p
)
+ε(η
c
+ Γη
p
)(ϕ
c
+ Γλ
r
ϕ
p
)
r
] + (η
c
+ Γη
p
)
2
¦
z=0
dxdy
−
γ
2
b
−b
¦χ[τ
1
(ϕ
c
+ Γλ
r
ϕ
p
)(η
c
+ Γη
p
)
t
+ (η
c
+ Γη
p
)
2
]¦
x,z=0
dy, (3.12c)
L
τ
=
τ
2
b
−b
ξ
0
¸
Γ
2
λ
4
r
η
2
p
x
+η
2
c
y
dxdy −
b
−b
¸
γχ
Γ
2
λ
4
r
η
2
p
x
+η
2
c
y
x,z=0
dy.
(3.12d)
The wavemaker forcing χ is speciﬁed to be
χ = f(z) sin
t
β
f(z) = 1 +
z
h
, for a full draft hinge,
f(z) = 1, for a full draft piston;
(3.13a,b)
and the deepwater cross wave and wavemaker wave are approximated by
ϕ
c
(y, z; t) = q(t) cos(y −b) exp[z], (3.14a)
ϕ
p
(x, z; t) = [Q
1
(t) cos x +Q
2
(t) sin x] exp[z/λ
2
r
], (3.14b)
where the variables q(t), Q
1
(t), and Q
2
(t) are the generalized coordinates. The free
surface displacement η is a solution to the linearized inhomogeneous boundary value
problem given by
L
η
η = [L
η
p
η
p
−f
p
] + [L
η
c
η
c
−f
c
] ≡ 0; (3.15a)
L
η
p
η
p
−f
p
= 0; L
η
c
η
c
−f
c
= 0; (3.15b,c)
η
p
x
= 0; x = 0, η
c
y
= 0; y = [b[, (3.15d,e)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
64 R. B. Guenther and R. T. Hudspeth
where (3.15d and 3.15e) are the contact line conditions and where
L
η
p
η
p
=
τ
λ
4
r
η
p
xx
−η
p
, f
p
= αλ
r
ϕ
p
z
−τ
1
λ
r
ϕ
p
t
; z = 0, (3.15f,g)
L
η
c
η
c
= τη
c
yy
−η
c
, f
c
= αϕ
c
z
−τ
1
ϕ
c
t
; z = 0. (3.15h,i)
Substituting (3.14a) and (3.15h,i) into (3.15c) yields
η
c
=
1
τ
2
1
(−qα +τ
1
˙ q) cos(y −b). (3.16)
Substituting (3.14b) and (3.15f,g) into (3.15b) yields
η
p
=
−Q
1
α +λ
2
r
τ
1
˙
Q
1
λ
r
τ
λ
cos x +
−Q
2
α +λ
2
r
τ
1
˙
Q
2
λ
r
τ
λ
sinx +
√
τ
λ
2
p
exp
¸
−
λ
2
p
√
τ
x
¸
. (3.17)
The Lagrangian may be decomposed into the following components:
L = L
0
+L
ε
+L
γ
+O(ε
2
, εγ, γ
2
); (3.18a)
where
L
0
=
−
bΓ
2
2
[ξ(Q
2
1
(t) +Q
2
2
(t)) +Q
1
(t)Q
2
(t)]
+
bτ
2
1
Γ
2
2λ
2
r
τ
2
λ
[ξλ
4
r
τ
λ
(
˙
Q
2
1
(t) +
˙
Q
2
2
(t)) + 4τ
˙
Q
1
(t)
˙
Q
2
(t)]
−
bξ
2
(q
2
(t) − ˙ q
2
(t))
; (3.18b)
L
ε
= −
εbΓτ
1
τ
2λ
3
r
τ
λ
q
2
˙
Q
2
−
εbΓ
2τ
1
λ
3
r
q ˙ qQ
2
; (3.18c)
L
γ
= −γb cos
t
β
f
1
Γλ
r
τ
1
Q
1
εβ
+
Γ
2
τ
2
1
βτ
λ
Q
1
(λ
2
r
˙
Q
1
+
√
τ
˙
Q
2
) +
q ˙ q
2β
+γb sin
t
β
Γ
2
Q
2
1
−
Γ
2
τ
2
1
λ
2
r
τ
2
λ
(λ
4
r
˙
Q
2
1
+λ
2
r
√
τ
˙
Q
1
˙
Q
2
+τ
˙
Q
2
2
) +
−˙ q
2
+q
2
2
,
(3.18d)
where f
1
is an integral of (3.13a or 3.13b), L
0
represents the free oscillations, L
ε
represents the Floquet parametric forcing of the cross wave by the progressive wave,
and L
γ
represents the completely nonautonomous perturbation. The analogy to the
nonautonomous Hamiltonian for a Floquet oscillator follows.
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 65
3.2.2. Legendre transform
The Legendre transform of the Lagrangian yields the following Hamiltonian:
H(q, p, t) = p ˙ q + P
1
˙
Q
1
+P
2
˙
Q
2
−L(q, ˙ q, t), (3.19)
where p = (p, P
1
, P
2
) are the conjugate momenta corresponding to the generalized
coordinates q = (q, Q
1
, Q
2
). The conjugate momenta are computed from
p
i
=
∂(L
0
+L
ε
)
∂ ˙ q
i
; i = 1, 2, 3, (3.20a)
p
1
= p = bξ ˙ q −
εbΓ
2τ
1
λ
3
r
qQ
2
; (3.20b)
p
2
= P
1
=
bτ
2
1
Γ
2
λ
2
r
τ
λ
ξ
˙
Q
1
+
2τ
λ
4
r
τ
λ
˙
Q
2
; (3.20c)
p
3
= P
2
= −
εbτ
1
Γτ
2λ
3
r
τ
λ
q
2
+
bτ
2
1
Γ
2
λ
2
r
τ
λ
ξ
˙
Q
2
+
2τ
λ
4
r
τ
λ
˙
Q
1
. (3.20d)
Inverting (3.20b–3.20d) yields
˙ q
1
=
p
bξ
+
εΓqQ
2
2τ
1
ξλ
3
r
; (3.21a)
˙
Q
1
=
λ
4
r
τ
λ
P
1
−
2τ
ξ
P
2
bτ
2
1
Γ
2
λ
6
r
ξ
−
ετ
2
q
2
Γλ
9
r
ξ
2
τ
1
τ
λ
; (3.21b)
˙
Q
2
=
λ
4
r
τ
λ
P
2
−
2τ
ξ
P
1
bξτ
2
1
Γ
2
λ
6
r
+
ετq
2
2Γξλ
5
r
τ
1
. (3.21c)
Substituting (3.18), (3.20), and (3.21) into (3.19) yields the following Hamiltonian:
H(q, p, t) = H
0
(q, p) +H
ε
(q, p) + H
γ
(q, p, t) +O(ε
2
, εγ, γ
2
), (3.22a)
where the free oscillation component is
H
0
=
bΓ
2
2
(ξ(Q
2
1
+Q
2
2
) +Q
1
Q
2
) +
1
2
bξq
2
+
p
2
2bξ
+
(P
2
1
+P
2
2
)
2bβ
2
Γ
2
ξ
−
2τP
1
P
2
bτ
2
1
Γ
2
λ
6
r
ξ
2
; (3.22b)
the Floquet parametric forcing component is
H
ε
=
εΓ
2τ
1
ξλ
3
r
qpQ
2
+
ετ
2τ
1
Γξλ
5
r
q
2
P
2
−
ετ
2
τ
1
Γξ
2
λ
9
r
τ
λ
q
2
P
1
; (3.22c)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
66 R. B. Guenther and R. T. Hudspeth
and the nonautonomous perturbed component from the wavemaker forcing is
H
γ
=
γ cos
t
β
¸
¸
¸
¸
f
1
bΓλ
r
τ
1
Q
1
εβ
+
Q
1
(λ
2
r
P
1
+
√
τP
2
)
βξλ
2
r
−
2τQ
1
(λ
2
r
P
2
+
√
τP
1
)
βξ
2
λ
6
r
τ
λ
+
pq
2βξ
¸
−γ sin
t
β
bΓ
2
Q
2
1
−
(λ
2
r
P
1
+
√
τP
2
)
2
bτ
2
1
Γ
2
ξ
2
λ
6
r
+
1
2
bq
2
−
p
2
2bξ
2
. (3.22d)
Although the nonautonomous Hamiltonian in (3.22d) is homomorphic to a nonau
tonomous Floquet Hamiltonian, a subtle but important distinction exists. The
nonautonomous component in a Floquet Hamiltonian is due to parametric forcing;
while, in contrast, the nonautonomous component in (3.22d) is due to external wave
maker boundary forcing that generates the progressive wave that then parametri
cally forces the cross waves. This sequence of external forcing to parametric forcing
distinguishes cross wave parametric forcing from the simple pendulum Floquet para
metric forcing.
Damping forces. The scaled Hamilton’s principle (3.3) is given by
1,3
δ
t
2
t
1
Ldt =
t
2
t
1
3
¸
i=1
D
i
δq
i
dt = −
t
2
t
1
b
−b
ξ
0
(αϕ
z
δη)dxdydt, (3.23a)
where D = (D
1
, D
2
, D
3
) is a set of generalized components of the damping force
corresponding to the set of generalized coordinates q = (q, Q
1
, Q
2
) and where
δη =
3
¸
i=1
∂η
∂q
i
δq
i
+
∂η
∂ ˙ q
i
δ ˙ q
i
. (3.23b)
Integrating by parts, the last term
1,3
reduces D to the canonical variables q, p
D
i
= α
b
−b
ξ
0
¸
−ϕ
z
∂η
∂q
i
+
∂
∂t
ϕ
z
∂η
∂ ˙ q
i
dxdy; i = 1, 2, 3, (3.24a)
D
1
=
α
τ
1
p +
αεbΓ
2τ
2
1
λ
3
r
qQ
2
+
α
2
bξ
τ
2
1
q, (3.24b)
D
2
=
α
λ
2
r
τ
1
P
1
−
2ατ
ξλ
6
r
τ
1
τ
λ
P
2
−
αεbΓτ
2
λ
9
r
ξτ
λ
q
2
+
α
2
bξΓ
2
λ
2
r
τ
λ
Q
1
, (3.24c)
D
3
=
α
τ
1
λ
2
r
P
2
+
αεbΓτ
2λ
5
r
τ
λ
q
2
+
2α
2
bτΓ
2
λ
6
r
τ
2
λ
Q
1
+
α
2
bξΓ
2
λ
2
r
τ
λ
Q
2
. (3.24d)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 67
3.3. Hamilton’s Dynamic Equations and Canonical
Transformations
Hamilton’s equations of motion are given by
˙ q
i
=
∂H
∂p
i
; ˙ p
i
= −
∂H
∂q
i
−D
i
; i = 1, 2, . . . , N. (3.25a,b)
If the canonical transformation Q = Q(q, p, t), P = P(q, p, t) is computed from a
generating function F(u, U, t) computed from the GHAType I (Appendix A), then
the Hamiltonian K for the new set of variables is given by
K(Q, P, t) = H[q(Q, P, t), p(Q, P, t), t] +
∂F(u, U, t)
∂t
; (3.26a)
and the transformed Hamilton’s equations of motion are
˙
Q
i
=
∂K
∂P
i
+
N
¸
j=1
˜
D
j
∂q
j
∂P
i
; i = 1, 2, . . . , N, (3.26b)
˙
P
i
= −
∂K
∂Q
i
−
N
¸
j=1
˜
D
j
∂q
j
∂Q
i
; i = 1, 2, . . . , N, (3.26c)
˜
D
i
(Q, P, t) = D
i
[q(Q, P, t), p(Q, P, t)]; i = 1, 2, . . . , N. (3.26d)
3.3.1. Three canonical transformations and analyses by the GMM
In order to apply the GMM to the dynamical system in (3.26), canonical transfor
mations are required ﬁrst to simplify terms in H
0
; second to simplify terms in H
ε
;
and third to suspend nonautonomous terms in H
γ
. The GHA will be applied to
two of the following three canonical transformations, because the rotation of axes
transformation is well known so that the GHA will not be applied for that trans
formation even though it is still applicable.
(1) Rotation of axes. The cross product terms in (3.22b) may be eliminated by
q = ˜ q, p = ˜ p, (3.27a,b)
Q
1
=
˜
Q
1
cos θ +
˜
Q
2
sin θ, P
1
=
˜
P
1
cos θ +
˜
P
2
sinθ, (3.27c,d)
Q
2
=
˜
Q
2
cos θ −
˜
Q
1
sin θ, P
2
=
˜
P
2
cos θ −
˜
P
1
sinθ, (3.27e,f)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
68 R. B. Guenther and R. T. Hudspeth
that satisfy the Poisson bracket requirements in Appendix A. The rotation angle
θ = π/4 eliminates the cross product terms, and the transformed Hamiltonian is
given in component form by
H →
˜
H =
˜
H
0
+
˜
H
ε
+
˜
H
γ
(t) +O(ε
2
, εγ, γ
2
), (3.28a)
where
˜
H
0
=
¸
bξ
2
˜ q
2
+
˜ p
2
2bξ
+
¸
bΓ
2
ξ
2
˜
Q
2
1
+
˜
P
2
1
2bβ
2
Γ
2
ξ
¸
+
¸
bΓ
2
ξ
2
˜
Q
2
2
+
˜
P
2
2
2bβ
2
Γ
2
ξ
¸
+
¸
bΓ
2
4
(
˜
Q
2
2
−
˜
Q
2
1
) +
τ(
˜
P
2
1
−
˜
P
2
2
)
bτ
2
1
Γ
2
λ
6
r
ξ
2
¸
, (3.28b)
˜
H
ε
=
εΓ
2
√
2τ
1
ξλ
3
r
˜ q˜ p(
˜
Q
2
−
˜
Q
1
) −
ετ(2τ +ξλ
4
r
τ
λ
)
2
√
2τ
1
Γξ
2
λ
9
r
τ
λ
˜ q
2
˜
P
1
+
ετ(−2τ +ξλ
4
r
τ
λ
)
2
√
2τ
1
Γξ
2
λ
9
r
τ
λ
˜ q
2
˜
P
2
, (3.28c)
˜
H
γ
=
γ cos
t
β
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
f
1
bΓλ
r
τ
1
(
˜
Q
1
+
˜
Q
2
)
√
2εβ
+
(
˜
Q
1
+
˜
Q
2
)((λ
2
r
−
√
τ)
˜
P
1
+ (λ
2
r
+
√
τ)
˜
P
2
)
2βξλ
2
r
−
τ(
˜
Q
1
+
˜
Q
2
)((λ
2
r
+
√
τ)
˜
P
2
+ (−λ
2
r
+
√
τ)
˜
P
1
)
βξ
2
λ
6
r
τ
λ
+
˜ p˜ q
2βξ
¸
−γ sin
t
β
¸
¸
¸
bΓ
2
2
(
˜
Q
1
+
˜
Q
2
)
2
−
((λ
2
r
−
√
τ)
˜
P
1
+ (λ
2
r
+
√
τ)
˜
P
2
)
2
2bτ
2
1
Γ
2
ξ
2
λ
6
r
+
1
2
b˜ q
2
−
˜ p
2
2bξ
2
¸
.
(3.28d)
The ﬁrst three energy square brackets [•] in (3.28b) are the action variables in the
next canonical transformation.
(2) Action/Angle Transformation. The three new canonical variables are
given by
ˆ p =
bξ
2
˜ q
2
+
˜ p
2
2bξ
,
ˆ
P
1
=
bΓ
2
ξ
2
˜
Q
2
1
+
˜
P
2
1
2bβ
2
Γ
2
ξ
,
ˆ
P
2
=
bΓ
2
ξ
2
˜
Q
2
2
+
˜
P
2
2
2bβ
2
Γ
2
ξ
,
(3.29a–c)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 69
that satisfy the Poisson brackets. The Herglotz auxiliary functions X
i
are the ratios
X
1
= ˜ q/˜ p; X
2
=
˜
Q
1
/
˜
P
1
; X
3
=
˜
Q
2
/
˜
P
2
. (3.30a–c)
The angle variables are
ˆ q
i
=
3
¸
j=1
˜ q
j
∂˜ p
j
∂ˆ p
i
+
∂F
∂ˆ p
i
; i = 1, 2, 3, (3.31a)
where
ˆ q = tan
−1
(bξX
1
);
ˆ
Q
1
= β tan
−1
(bβξΓ
2
X
2
);
ˆ
Q
2
= β tan
−1
(bβξΓ
2
X
3
).
(3.31b–d)
The canonical transformation to action/angle variables is given by the following:
˜ q =
2ˆ p
bξ
sin ˆ q, ˜ p =
2bξˆ pcos ˆ q, (3.32a,b)
˜
Q
1
=
2
ˆ
P
1
bξΓ
2
sin
ˆ
Q
1
β
,
˜
P
1
= Γβ
2bξ
ˆ
P
1
cos
ˆ
Q
1
β
, (3.32c,d)
˜
Q
2
=
2
ˆ
P
2
bξΓ
2
sin
ˆ
Q
2
β
,
˜
P
2
= Γβ
2bξ
ˆ
P
2
cos
ˆ
Q
2
β
, (3.32e,f)
that satisfy the Poisson brackets.
1,3
The transformed Hamiltonian is
˜
H ⇒
ˆ
H =
ˆ
H
0
+
ˆ
H
ε
+
ˆ
H
γ
(t) +O(ε
2
, εγ, γ
2
), (3.33a)
ˆ
H
0
= [ ˆ p +
ˆ
P
1
+
ˆ
P
2
] +
2β
2
τ
λ
6
r
ξτ
2
1
ˆ
P
1
cos
2
ˆ
Q
1
β
−
2β
2
τ
λ
6
r
ξτ
2
1
ˆ
P
2
cos
2
ˆ
Q
2
β
−
ˆ
P
1
2ξ
sin
2
ˆ
Q
1
β
+
ˆ
P
2
2ξ
sin
2
ˆ
Q
2
β
, (3.33b)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
70 R. B. Guenther and R. T. Hudspeth
ˆ
H
ε
=
εˆ p
4
bξ
3
λ
3
r
τ
1
ˆ
P
2
¸
cos
2ˆ q −
ˆ
Q
2
β
−cos
2ˆ q +
ˆ
Q
2
β
¸
−
ˆ
P
1
¸
cos
2ˆ q −
ˆ
Q
1
β
−cos
2ˆ q +
ˆ
Q
1
β
¸
−
εβτ ˆ p
2
bξ
5
λ
9
r
τ
1
τ
λ
(2τ +ξλ
4
r
τ
λ
)
ˆ
P
1
¸
cos
ˆ
Q
1
β
−
1
2
cos
2ˆ q +
ˆ
Q
1
β
−
1
2
cos
2ˆ q −
ˆ
Q
1
β
¸
+ (2τ −ξλ
4
r
τ
λ
)
ˆ
P
2
¸
cos
ˆ
Q
2
β
−
1
2
cos
2ˆ q +
ˆ
Q
2
β
−
1
2
cos
2ˆ q −
ˆ
Q
2
β
¸
,
(3.33c)
ˆ
H
γ
=
γ cos
t
β
ˆ
P
1
sin
ˆ
Q
1
β
+
ˆ
P
2
sin
ˆ
Q
2
β
√
bλ
r
f
1
τ
1
βε
√
ξ
+
(λ
2
r
+
√
τ)(−2τ +ξλ
4
r
τ
λ
)
λ
6
r
ξ
2
τ
λ
ˆ
P
1
cos
ˆ
Q
1
β
+
(λ
2
r
+
√
τ)(−2τ +ξλ
4
r
τ
λ
)
λ
6
r
ξ
2
τ
λ
ˆ
P
2
cos
ˆ
Q
2
β
+
ˆ p sin(2ˆ q)
2βξ
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
−
γ
ξ
sin
t
β
−β
2
λ
6
r
τ
2
1
¸
(λ
2
r
−
√
τ)
ˆ
P
1
cos
ˆ
Q
1
β
+(λ
2
r
+
√
τ)
ˆ
P
2
cos
ˆ
Q
2
β
¸
2
+
¸
ˆ
P
1
sin
ˆ
Q
1
β
+
ˆ
P
2
sin
ˆ
Q
2
β
¸
2
+ ˆ p(sin
2
ˆ q −cos
2
ˆ q)
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
.
(3.33d)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 71
The nonautonomous perturbed component in (3.33d) will not survive the KAM
averaging theorem, and some of the nonautonomous terms may be suspended
by applying the following Hamilton–Jacobian nonautonomous canonical transfor
mation.
(3) Hamilton–Jacobian transformation. The following nonautonomous trans
formations:
ˆ q =
t
2β
+q +Q
1
,
ˆ
Q
1
= t + 2βQ
1
,
ˆ
Q
2
= 3t + 2β(q +Q
1
+Q
2
) (3.34a–c)
satisfy the Poisson brackets. The Herglotz auxiliary functions X
i
= X
i
(p, q) satisfy
the nonzero Jacobian condition given by
∂(ˆ q, X)
∂(p, q)
=
4β
2
qQ
1
Q
2
= 0. (3.35)
Solving (3.35) for q
i
= q
i
(p, X), substituting q
i
(p, X) into (3.34), and inverting
yields
p =
X
1
2β
(2βˆ q −
ˆ
Q
1
); P
1
=
−X
2
2β
(t −
ˆ
Q
1
);
P
2
=
−X
3
2β
(2t + 2βˆ q −
ˆ
Q
2
).
(3.36a–c)
The generating function F(•, •, •) for the nonautonomous canonical transformation
may be computed from the following indeﬁnite integral:
F(ˆ q, p(ˆ q, X, t), t)
= −
3
¸
i=1
X
i
¸
q
∂p
∂X
i
+Q
1
∂P
1
∂X
i
+Q
2
∂P
2
∂X
i
dX
i
=
−1
4β
2
[X
1
(2βˆ q −
ˆ
Q
1
)
2
+X
2
(t −
ˆ
Q
1
)
2
+X
3
(2t + 2βˆ q −
ˆ
Q
2
)
2
]. (3.37)
Computing the three remaining conjugate momenta from
ˆ p
i
= −
3
¸
j=1
q
j
∂p
j
∂ˆ q
i
−
∂F
∂ˆ q
i
; i = 1, 2, 3 (3.38a)
gives
ˆ p = p −P
2
,
ˆ
P
1
=
1
2β
(P
1
−p),
ˆ
P
2
=
1
2β
P
2
. (3.38b–d)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
72 R. B. Guenther and R. T. Hudspeth
The transformations in (3.34) and (3.38) satisfy the Poisson brackets, and the ﬁrst
square bracket term [•] in (3.33b) may be transformed to
[ ˆ p +
ˆ
P
1
+
ˆ
P
2
] =
1 −
1
2β
(p −P
2
) +
P
1
2β
. (3.39)
The primary Floquet resonance condition is, approximately, ω
c
−ω
p
≈ O(ε) so that
a detuning parameter Ω may be deﬁned by
1 −
ω
p
2ω
c
=
1 −
1
2β
= εΩ; (3.40)
and the primary parametric Floquet resonance is ω
p
: ω
c
= 2:1. A generating
function is now given by
F(ˆ q, p, t) = p
−ˆ q +
1
2β
ˆ
Q
1
+
1
2β
P
1
(t −
ˆ
Q
1
) +
1
2β
P
2
(2t + 2βˆ q −
ˆ
Q
2
); (3.41)
and the transformed Hamiltonian H(p, q, t) with autonomous and nonautonomous
terms by
ˆ
H −
∂F
∂t
⇒H = H
0
+H
0
(t) +H
ε
+H
ε
(t)
+H
γ
+H
γ
(t) +O(ε
2
, εγ, γ
2
), (3.42a)
where
H
0
=
−
1
8βξ
+
βτ
2ξλ
6
r
τ
2
1
(P
1
−p) +
−
1
β
+
1
8βξ
−
βτ
2ξλ
6
r
τ
2
1
P
2
, (3.42b)
H
ε
= εΩ(p −P
2
) +
ε(p −P
2
)
√
P
1
−p
4
2bβξ
5
λ
9
r
τ
1
τ
λ
[(2βτ
2
−ξλ
4
r
τ
λ
(λ
2
r
−βτ)) cos(2q)], (3.42c)
H
γ
=
γ
√
bf
1
λ
r
τ
1
2ε
2ξβ
3
P
1
−p sin(2Q
1
) +γ
2β −1
4βξ
(P
2
−p) sin 2(q +Q
1
), (3.42d)
and where the nonautonomous terms are given by Fadel.
5
(4) Transformation to original variable (q
orig
, p
orig
) ⇒ (q, p). In order to
compute the generalized damping forces in (3.24), the original canonical variables
(q
orig
, p
orig
) must be expressed as functions of the ﬁnal transformed canonical
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 73
variables (q, p) by
q
orig
=
2
bξ
p −P
2
sin
q +Q
1
+
t
2β
, (3.43a)
p
orig
=
2bξ
p −P
2
cos
q +Q
1
+
t
2β
, (3.43b)
Q
1
orig
=
1
Γ
√
2bβξ
¸
P
1
−p sin
2Q
1
+
t
β
+
P
2
sin
2(q +Q
1
+Q
2
) +
3t
β
, (3.43c)
P
1
orig
= Γ
bβξ
2
¸
P
1
−p cos
2Q
1
+
t
β
+
P
2
cos
2(q +Q
1
+Q
2
) +
3t
β
, (3.43d)
Q
2
orig
=
1
Γ
√
2bβξ
¸
−
P
1
−p sin
2Q
1
+
t
β
+
P
2
sin
2(q +Q
1
+Q
2
) +
3t
β
, (3.43e)
P
2
orig
= Γ
bβξ
2
¸
−
P
1
−p cos
2Q
1
+
t
β
+
P
2
cos
2(q +Q
1
+Q
2
) +
3t
β
. (3.43f)
3.3.2. Transformed damping forces
The transformed damping forces are
1,3
D
1
=
√
2bβξα
τ
1
p −P
2
cos
q +Q
1
+
t
2β
+
√
2bξα
2
τ
2
1
p −P
2
sin
q +Q
1
+
t
2β
+
εα
2λ
3
r
ξ
√
βτ
2
1
p −P
2
sin
q +Q
1
+
t
2β
P
2
sin
2(q +Q
1
+Q
2
) +
3t
β
−
P
1
−p sin
2Q
1
+
t
β
¸
¸
¸
¸
¸
, (3.44a)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
74 R. B. Guenther and R. T. Hudspeth
D
2
=
√
bβαΓ(2τ +λ
4
r
ξτ
λ
)
τ
1
τ
λ
λ
6
r
√
2ξ
P
1
−p cos
2Q
1
+
t
β
+
√
bβαΓ(−2τ +λ
4
r
ξτ
λ
)
τ
1
τ
λ
λ
6
r
√
2ξ
P
2
cos
2(q +Q
1
+Q
2
) +
3t
β
−
2Γεατ
2
λ
9
r
ξ
2
τ
2
λ
(p −P
2
) sin
2(q +Q
1
) +
t
2β
+
α
2
Γ
√
bξ
λ
2
r
τ
λ
√
2β
P
1
−p sin
2Q
1
+
t
β
+
α
2
Γ
√
bξ
λ
2
r
τ
λ
√
2β
P
2
sin
2(q +Q
1
+Q
2
) +
3t
β
,
(3.44b)
D
3
=
αΓ
√
bβξ
√
2λ
2
r
τ
1
−
P
1
−p cos
2Q
1
+
t
β
+
P
2
cos
2(q +Q
1
+Q
2
) +
3t
β
¸
¸
¸
¸
¸
+
εαΓτ
λ
5
r
ξτ
λ
(p −P
2
) sin
2(q +Q
1
) +
t
2β
+
α
2
Γ
√
b(2τ −λ
4
r
ξτ
λ
)
λ
6
r
τ
2
λ
√
2βξ
P
1
−p sin
2Q
1
+
t
β
+
α
2
Γ
√
b(2τ +λ
4
r
ξτ
λ
)
λ
6
r
τ
2
λ
√
2βξ
P
2
sin
2(q +Q
1
+Q
2
) +
3t
β
,
(3.44c)
that are completely autonomous and will survive the KAM averaging operation.
Averaging H(q, p, t) over the dimensionless cross wave period 2π yields
'H` = H
0
(p, P
1
, P
2
) +H
ε
(q, p, P
1
, P
2
) +H
γ
(q, p, P
1
, P
2
, Q
1
)
=
−
1
8βξ
+
βτ
2ξλ
6
r
τ
2
1
(P
1
−p) +
−
1
β
+
1
8βξ
−
βτ
2ξλ
6
r
τ
2
1
P
2
+ε
¸
Ω(p −P
2
) +
(p −P
2
)
√
P
1
−p
4
2bβξ
5
λ
9
r
τ
1
τ
λ
[(2βτ
2
−ξλ
4
r
τ
λ
(λ
2
r
−βτ)) cos(2q)]
¸
+γ
¸
√
bf
1
λ
r
τ
1
2ε
2ξβ
3
P
1
−p sin(2Q
1
) +
2β −1
4βξ
(P
2
−p) sin2(q +Q
1
)
¸
.
(3.44d)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 75
The averaged system of autonomous equations is
1,3
˙ q =
−a
1
+εΩ +εa
3
2P
1
+ P
2
−3p
√
P
1
−p
cos[2q]
−γc
1
sin[2Q
1
]
ε
√
P
1
−p
−γc
2
sin[2(q +Q
1
)] +α
2
d
1
−αε
¸
d
2
(P
1
−p) +d
3
(P
2
−p)
√
P
1
−p
sin[2q]
, (3.45a)
˙ p =
4εα
3
P
1
−p(p −P
2
) sin[2q] −2γc
2
(P
2
−p) cos[2(q +Q
1
)]
+αd
4
P
2
−
α
τ
1
p + 2αεd
2
P
1
−p(p −P
2
) cos[2q]
, (3.45b)
˙
P
1
=
−
4γ
ε
c
1
P
1
−p cos[2Q
1
] −2γc
2
(P
2
−p) cos[2(q +Q
1
)]
+α
(4λ
2
r
−1)(P
2
−p) −P
1
2λ
2
r
τ
1
−αεd
5
P
1
−p(p − P
2
) cos[2q]
, (3.45c)
˙
P
2
= αP
2
τ −λ
4
r
ξτ
λ
λ
6
r
ξτ
1
τ
λ
, (3.45d)
˙
Q
1
=
a
1
+εa
3
(p −P
2
)
√
P
1
−p
cos[2q] +γc
1
sin[2Q
1
]
ε
√
P
1
− p
+α
2
d
6
+αεd
3
P
2
−p
√
P
1
−p
sin[2q]
, (3.45e)
˙
Q
2
=
a
2
−εΩ −2εa
3
P
1
−p cos[2q] +γc
2
sin[2(q +Q
1
)]
+α
2
d
7
+αεd
2
P
1
−p sin[2q]
, (3.45f)
where a
i
, c
i
, and d
i
are tabulated in Appendix B.
3.4. Application of the GMM and Largest Liapunov
Characteristic Exponent
The GMM determines the existence of transverse homoclinic/heteroclinic points
that are transverse intersections between the stable and unstable manifolds to any
invariant sets of the perturbed system when a homoclinic/heteroclinic orbit exists
to a hyperbolic invariant manifold in the unperturbed (undamped α = 0 and unper
turbed γ = 0) system. The unperturbed vector ﬁeld may be computed by setting the
perturbation (wavemaker forcing) parameter γ = 0 and the dissipation parameter
α = 0 in (3.45), and are
˙ q = −a
1
+εΩ +εa
3
2P
1
+P
2
−3p
√
P
1
−p
cos(2q), (3.46a)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
76 R. B. Guenther and R. T. Hudspeth
˙ p = 4εa
3
P
1
−p(p −P
2
) sin(2q), (3.46b)
˙
P
1
=
˙
P
2
= 0, (3.46c,d)
˙
Q
1
= a
1
+εa
3
(p −P
2
)
√
P
1
−p
cos(2q), (3.46e)
˙
Q
2
= a
2
−εΩ −2εa
3
P
1
−p cos(2q), (3.46f)
where the coeﬃcients a
1
, a
2
, and a
3
are deﬁned in Appendix B. The unperturbed
vector ﬁeld α = γ = 0 in (3.46) is a three degrees of freedom Hamiltonian system
with (q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
. An important consequence of
the Hamilton–Jacobian canonical transformation is that the unperturbed Floquet
Hamiltonian 'H`(γ = 0) = H
0
(p, P
1
, P
2
) + H
ε
(q, p, P
1
, P
2
) is independent of Q
1
and Q
2
.
Hyperbolic saddle points. For every (P
1
, P
2
) ∈ R
2
, the (q − p) components of
the unperturbed vector ﬁeld α = γ = 0 in (3.46) possess a hyperbolic saddle point
that varies smoothly with P
1
and P
2
and is given by
p
0
(P
2
) = P
2
; 2q
0
= nπ; n = 0, 1, . . . , (3.47a,b)
q
0
(P
1
, P
2
) =
1
2
cos
−1
a
1
−εΩ
2εa
3
√
P
1
−P
2
, (3.47c)
P
1
≥ P
2
+
a
1
− εΩ
2εa
3
2
. (3.47d)
The ﬁxed point (q
0
, p
0
) is a hyperbolic saddle point if the determinant satisﬁes
∂ ˙ q
∂q
∂ ˙ q
∂p
∂ ˙ p
∂q
∂ ˙ p
∂p
(q=q
0
,p=p
0
)
= (a
1
−εΩ)
2
−4(P
1
−P
2
)ε
2
a
2
3
< 0. (3.48)
The symmetry properties of Hamiltonian systems imply that the stable and the
unstable manifolds of the hyperbolic saddle point (q
0
, p
0
) have equal dimensions in
the full 6D phase space (q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
. The unperturbed
system α = γ = 0 has a 4D (R
2
T
2
) normally hyperbolic invariant manifold given
by the union of the hyperbolic saddle points (q
0
, p
0
) according to
M = ¦(q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
: q = q
0
(P
1
, P
2
); p = p
0
(P
2
)¦.
(3.49)
The normally hyperbolic invariant manifold Mhas 5D (R
1
R
2
T
2
) stable manifold
W
s
(M) and unstable manifold W
u
(M) that coincide along the 5D heteroclinic
manifold H = W
s
(M) ∩ W
u
(M) − M, where W
s
(M) and W
u
(M) are the set of
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 77
initial conditions that approach the hyperbolic saddle points as t →±∞ under the
action of the unperturbed ﬂow.
Dynamics on M. The unperturbed vector ﬁeld α = γ = 0 restricted to the
normally hyperbolic invariant manifold M may be determined from
1,3
˙
P
1
= 0;
˙
P
2
= 0;
˙
Q
1
= a
1
;
˙
Q
2
= a
2
−a
1
. (3.50a–d)
The unperturbed vector ﬁeld α = γ = 0 restricted to M has the form of two degrees
of freedom, 4D phase space (T
1
R
1
T
2
), completely integrable Hamiltonian
system with the Hamiltonian given by the level energy surfaces
H(P
1
, P
2
) = 'H`(q = q
0
, p = p
0
; γ = 0) = a
1
P
1
+ (a
2
−a
1
)P
2
= E, (3.51)
where E is a constant energy that allows the phase space motion to be reduced from
four dimensions (R
2
T
2
) to three dimensions. Because P
1
= P
2
= 0, the constant
energy surfaces may be reduced to a 1D surface in the 3D constant energy space.
On this 1D surface, the angular motion is parameterized by the two frequencies
σ
1
= a
1
,
σ
2
= a
2
−a
1
.
(3.52a,b)
The angular components of the motion on the normally hyperbolic invariant
manifold M are given by
Q
1
(t) = a
1
t +Q
1
(0); Q
2
(t) = (a
2
−a
1
)t +Q
2
(0). (3.53a,b)
The normally hyperbolic invariant manifold M is a twoparameter (P
1
, P
2
) family
of 2D tori shown in Fig. 3.4. A 2D torus on M may be deﬁned by
Υ(
¯
P
1
,
¯
P
2
) =
(q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
: q = q
0
(
¯
P
1
,
¯
P
2
);
p = p
0
(
¯
P
2
); P
1
=
¯
P
1
; P
2
=
¯
P
2
¸
,
(3.54)
where each 2D torus is invariant. The level energy surface is a family of concentric
tori as shown in Fig. 3.5. Because the frequencies of motion σ
1
and σ
2
in (3.52) are
independent of P
1
or P
2
, they do not change from one concentric torus to another
concentric torus. The motion on the surface of the invariant torus Υ ⊂ M is quasi
periodic. When the frequency ratio σ
1
/σ
2
= a
1
/(a
2
− a
1
) is an irrational number,
the motion on the surface of the 2D invariant nonresonant torus may no longer be
periodic; i.e., trajectories wind densely on the surface of the torus and never close
on themselves. This 2D torus has a 3D (R
1
T
1
) stable manifold W
S
and unstable
manifold W
U
that coincide along the 3D (R
1
T
1
) heteroclinic manifold H deﬁned
for ﬁxed values of P
1
and P
2
, as shown in Fig. 3.4. On a constant level energy surface,
neither the nonresonant invariant tori nor the stable and unstable manifolds may be
isolated. In addition, the 2D nonresonant invariant torus has a 2D center manifold
corresponding to nonexponentially expanding or contracting directions tangent to
the normally hyperbolic invariant manifold M.
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
78 R. B. Guenther and R. T. Hudspeth
Fig. 3.4. 3D unperturbed heteroclinic manifold.
1
Fig. 3.5. Motion of a phase space point for an integrable Hamiltonian system with two degrees
of freedom. (a) Invariant tori in a 3D constant energy space E; and (b) the ﬂow on a 2D torus.
1,3
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 79
Heteroclinic orbits. For ﬁxed values of (P
∗
1
, P
∗
2
) ∈ R
2
, the (q, p) compo
nents of the unperturbed vector ﬁeld α = γ = 0 in (3.46) possess a 1D
heteroclinic orbit connecting the hyperbolic saddle points (q
0
, p
0
). The hetero
clinic orbit lies on each of the level energy surfaces deﬁned by (3.51) and are
solutions to
'H`(γ = 0) −(a
1
P
∗
1
+ (a
2
−a
1
)P
∗
2
) = 0. (3.55)
Values for q on the heteroclinic orbit are given by
q =
1
2
cos
−1
(a
1
− εΩ)/2εa
3
P
∗
1
−p
. (3.56)
In the full 6D phase space (q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
, the heteroclinic
manifold H may be determined by substituting
cos[2q] =
a
1
−εΩ
2εa
3
√
P
1
−p
, sin[2q] =
1 −
a
1
−εΩ
2εa
3
√
P
1
−p
2
, (3.57a,b)
into the unperturbed vector ﬁeld α = γ = 0 in (3.46), and integrating to
obtain
1,3
p
h
(t) = P
2
+
A
B
sech
2
[
√
At] +p(0), (3.58a)
q
h
(t) =
1
2
tan
−1
¸
−
√
A
a
1
−εΩ
tanh(
√
At)
¸
+q(0), (3.58b)
P
1h
(t) = P
1
(0), P
2h
(t) = P
2
(0), (3.58c,d)
Q
1h
(t) = a
1
t −q
h
(t) +Q
1
(0) +q(0)
= a
1
t −
1
2
tan
−1
¸
−
√
A
a
1
−εΩ
tanh(
√
At)
¸
+Q
1
(0), (3.58e)
Q
2h
(t) = (a
2
−a
1
)t +Q
2
(0), (3.58f)
A = [−(a
1
−εΩ)
2
+ (P
1
−P
2
)B] > 0, B = 4ε
2
a
2
3
, (3.58g,h)
p(0) = 0, q = q
n
(0) = (2n + 1)
π
2
; n = 0, 1, 2, . . . . (3.58i,j)
The trajectories of the unperturbed system α = γ = 0 along the 5D phase space
(R
1
R
2
T
2
) heteroclinic manifold H may be expressed as
Ψ
(P
1
,P
2
)
= ¦q
h
(t), p
h
(t), P
1
(0), P
2
(0), Q
1h
(t), Q
2h
(t)¦. (3.59)
The 6D phase space (q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
is a direct product
between a 4D space with coordinates (q, p, P
1
, P
2
) and a 2D torus with angular
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
80 R. B. Guenther and R. T. Hudspeth
Fig. 3.6. Unperturbed 4D phase space.
coordinates (Q
1
, Q
2
). Because P
1
and P
2
are constants, the motion in the 6D phase
space is reduced to the four dimensions (q, p, Q
1
, Q
2
) shown in Fig. 3.6. Small per
turbations are anticipated to break up the geometric structure of the unperturbed
system α = γ = 0 and to separate the manifolds. The behavior of the perturbed
systems γ > 0 and α > 0 near the unperturbed heteroclinic manifold H is required
in order to apply the GMM. The distance between the stable and unstable mani
folds of any surviving invariant set in the perturbed system must be computed at
a point { on the unperturbed heteroclinic manifold H. Two perturbed systems are
evaluated: (1) γ > 0, α = 0 and (2) γ > 0 and α > 0.
3.4.1. Geometric structure for γ > 0 and α = 0
The perturbed vector ﬁeld for nonzero wavemaker forcing γ > 0 and no dissipation
α = 0 is given by
˙ q = −a
1
+εΩ +εa
3
2P
1
+P
2
−3p
√
P
1
−p
cos[2q]
−γc
1
sin[2Q
1
]
ε
√
P
1
−p
−γc
2
sin[2(q +Q
1
)], (3.60a)
˙ p =
4εa
3
P
1
−p(p −P
2
) sin[2q] − 2γc
2
(P
2
−p) cos[2(q + Q
1
)]
+αd
4
P
2
−
α
τ
1
p + 2αεd
2
P
1
−p(p −P
2
) cos[2q]
¸
¸
, (3.60b)
˙
P
1
= −
4γ
ε
c
1
P
1
−p cos[2Q
1
] −2γc
2
(P
2
−p) cos[2(q +Q
1
)], (3.60c)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 81
˙
P
2
= 0, (3.60d)
˙
Q
1
= a
1
+εa
3
(p −P
2
)
√
P
1
−p
cos[2q] +γc
1
sin[2Q
1
]
ε
√
P
1
−p
, (3.60e)
˙
Q
2
= a
2
−εΩ −2εa
3
P
1
−p cos[2q] +γc
2
sin[2(q +Q
1
)], (3.60f)
where (q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
. The perturbed vector ﬁeld is
computed from 5D level energy surfaces deﬁned by
'H` = H
0
(p, P
1
, P
2
) + H
ε
(q, p, P
1
, P
2
) +H
γ
(q, p, P
1
, P
2
, Q
1
). (3.61)
Because the perturbation is Hamiltonian, the 3D level energy surfaces are preserved.
In the 4D normally hyperbolic invariant manifold of the unperturbed space, the
locally stable and unstable manifolds and the ﬂow describe the geometric structure
of the perturbed phase space given by the perturbed normally hyperbolic locally
invariant manifold, the locally stable and unstable manifolds, and the persistence
of the 2D nonresonant invariant tori Υ
γ
(P
1
, P
2
).
Persistence of M. The perturbed system γ > 0 and α = 0 possesses a 4D normally
hyperbolic locally invariant manifold M
γ
given by
M
γ
=
(q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
:
q = ˜ q
0
(P
1
, P
2
, Q
1
, Q
2
; γ) = q
0
(P
1
, P
2
) +O(γ);
p = ˜ p
0
(P
1
, P
2
, Q
1
, Q
2
; γ) = p
0
(P
2
) +O(γ)
. (3.62)
On M
γ
there are locally stable and unstable manifolds that are of equal dimensions
and are close to the unperturbed locally stable and unstable manifolds. The per
turbed normally hyperbolic locally invariant manifold M
γ
intersects each of the 5D
level energy surfaces in a 3D set of which most of the twoparameter family of 2D
nonresonant invariant tori persist by the KAM theorem.
1,3
The Melnikov integral
may be computed to determine if the stable and unstable manifolds of the KAM
tori intersect transversely.
KAM Theorem.
1,3
The KAM theorem determines whether the recurrent motions
occur on the perturbed normally hyperbolic locally invariant manifold M
γ
and
whether any of the two parameter families of 2D nonresonant invariant tori survive
the Hamiltonian perturbation. The unperturbed Floquet Hamiltonian 'H`(γ = 0) =
H
0
(p, P
1
, P
2
) + H
ε
(q, p, P
1
, P
2
) satisﬁes the following nondegeneracy (or nonreso
nance) condition:
∂
2
'H`
∂P
2
1
∂
2
'H`
∂P
1
∂P
2
∂
2
'H`
∂P
2
∂P
1
∂
2
'H`
∂P
2
2
(q=q
0
,p=p
0
;γ=0)
= −
(a
1
−εΩ)
2
4(P
1
−P
2
)
2
< 0. (3.63)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
82 R. B. Guenther and R. T. Hudspeth
Most of the 2D nonresonant invariant tori Υ(P
1
, P
2
)) that persist are only
slightly deformed on the perturbed normally hyperbolic locally invariant manifold
M
γ
and are KAM tori. In the phase space of the perturbed system γ > 0 and α = 0,
there are invariant tori that are densely ﬁlled with winding trajectories that are
conditionally periodic with two independent frequencies σ
1
and σ
2
. The resulting
conditionallyperiodic motions of the perturbed system are smooth functions of the
perturbation γ. A generalization of the KAM theorem states that the KAM tori
have both stable and unstable manifolds by the invariance of manifolds.
1,3
In order
to determine if chaos exists, two measurements are required in order to determine
whether or not W
s
(Υ
γ
) and W
u
(Υ
γ
) intersect transversely.
Melnikov integral. The distance between W
s
(Υ
γ
) and W
u
(Υ
γ
) at any point
{ ∈ H may be computed from
M(Q
1
(0)) = − lim
α
n
,β
n
→∞
β
n
−α
n
∂H
γ
∂Q
1
(Ψ
(P
1
,P
2
)
)dt, (3.64a)
where
H
γ
=
2c
1
ε
P
1
−p sin(2Q
1
) +c
2
(P
2
−p) sin[2(q +Q
1
)], (3.64b)
where H
γ
= γH
γ
in (3.61), and where
1,3
M(Q
1
(0)) = cos[2(Q
1
(0) + q(0))]
lim
α
n
,β
n
→∞
β
n
−α
n
−
4c
1
ε
P
1
−p
h
(cos(2q
h
−2a
1
t))
−2c
2
(P
2
−p
h
) cos(2a
1
t)
¸
¸
dt
+ sin[2(Q
1
(0) +q(0))]
lim
α
n
,β
n
→∞
β
n
−α
n
−
4c
1
ε
P
1
−p
h
(sin(2q
h
−2a
1
t))
+2c
2
(P
2
−p
h
) sin(2a
1
t)
¸
¸
dt
,
(3.64c)
that after retaining only even integrands reduces to
1,3
M(Q
1
(0)) = cos[2(Q
1
(0) +q(0))][I
1
+I
2
+I
3
], (3.65a)
I
1
= lim
n→∞
−2c
1
(a
1
−εΩ)
ε
2
a
3
β
n
−α
n
cos(2a
1
t)dt
¸
, (3.65b)
I
2
=
2c
2
A
B
∞
−∞
cos(2a
1
t)sech
2
(
√
At)dt
, (3.65c)
I
3
=
−2c
1
√
A
ε
2
a
3
∞
−∞
sin(2a
1
t)tanh(
√
At)dt
¸
. (3.65d)
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 83
The Melnikov integral M(Q
1
(0)) = 0 when
Q
1
(0) =
¯
Q
1n
(0) = (2n + 1)
π
4
−q(0); n = 0, 1, 2, . . . , (3.66a)
∂M(
¯
Q
1n
(0))
∂Q
1
(0)
= −2 sin[2(
¯
Q
1n
(0) +q(0))][I
1
+I
2
+I
3
] = 0, n = 0, 1, 2, . . . .
(3.66b)
Consequently, the stable W
s
(Υ
γ
(P
1
, P
2
)) and unstable W
u
(Υ
γ
(P
1
, P
2
)) manifolds
of the KAM tori Υ
γ
(P
1
, P
2
) intersect transversely yielding Smale horseshoes
1,3
on
the appropriate 5D level energy surfaces. This implies multiple transverse intersec
tions and the corresponding existence of chaotic dynamics in the perturbed system
γ > 0 and α = 0.
3.4.2. Geometric structure for γ > 0 and α > 0
The perturbed dissipative system α > 0 and γ > 0 possesses a 4D normally hyper
bolic locally invariant manifold M
γα
that is given by
M
αγ
=
(q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
:
q = ˜ q
0
(P
1
, P
2
, Q
1
, Q
2
; γ) = q
0
(P
1
, P
2
) +O(γ),
p = ˜ p
0
(P
1
, P
2
, Q
1
, Q
2
; γ) = p
0
(P
2
) +O(γ)
. (3.67)
The manifold M
γα
has locally stable and unstable manifolds that are close to
the unperturbed locally stable and unstable manifolds; and if these manifolds
intersect transversely, then the Smale–Birkhoﬀ theorem predicts the existence of
horseshoes and their chaotic dynamics in the perturbed dissipative system. A 2D
hyperbolic invariant torus Υ
γα
(P
1
, P
2
) may be located on M
γα
by averaging the
perturbed dissipative vector ﬁeld γ > 0 and α > 0 restricted to M
γα
over the
angular variables Q
1
and Q
2
. The averaged equations have a unique stable hyper
bolic ﬁxed point (P
1
, P
2
) = (0, 0) with two negative eigenvalues provided that the
determinant
1,3
∂'
˙
P
1
`
∂P
1
∂'
˙
P
1
`
∂P
2
∂'
˙
P
2
`
∂P
1
∂'
˙
P
2
`
∂P
2
=
ν
2
γ
2
(λ
4
r
ξτ
λ
−τ)
2λ
8
r
ξτ
2
1
τ
λ
> 0. (3.68)
The Melnikov method fails to predict chaos for this dissipative system when
α > 0.
1,3
Liapunov characteristic exponents (LCE). Dissipative systems are charac
terized by the attraction of all trajectories passing through a certain domain toward
an invariant surface or an attractor of lower dimensionality than the original space.
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
84 R. B. Guenther and R. T. Hudspeth
For the 6D phase space (q, p, P
1
, P
2
, Q
1
, Q
2
) ∈ T
1
R
1
R
2
T
2
, there are six
real exponents that may be ordered as µ
1
≥ µ
2
≥ µ
3
≥ µ
4
≥ µ
5
≥ µ
6
, where
µ
1
is the largest Liapunov characteristic exponent (LCE). The LCE is calculated
from the ﬁrst variation of the perturbed dissipative vector ﬁeld γ > 0 and α > 0
according to
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
δ ˙ q
δ ˙ p
δ
˙
P
1
δ
˙
P
2
δ
˙
Q
1
δ
˙
Q
2
¸
=
∂ ˙ q
∂q
∂ ˙ q
∂p
∂ ˙ q
∂P
1
∂ ˙ q
∂P
2
∂ ˙ q
∂Q
1
∂ ˙ q
∂Q
2
∂ ˙ p
∂q
∂ ˙ p
∂p
∂ ˙ p
∂P
1
∂ ˙ p
∂P
2
∂ ˙ p
∂Q
1
∂ ˙ p
∂Q
2
∂
˙
P
1
∂q
∂
˙
P
1
∂p
∂
˙
P
1
∂P
1
∂
˙
P
1
∂P
2
∂
˙
P
1
∂Q
1
∂
˙
P
1
∂Q
2
∂
˙
P
2
∂q
∂
˙
P
2
∂p
∂
˙
P
2
∂P
1
∂
˙
P
2
∂P
2
∂
˙
P
2
∂Q
1
∂
˙
P
2
∂Q
2
∂
˙
Q
1
∂q
∂
˙
Q
1
∂p
∂
˙
Q
1
∂P
1
∂
˙
Q
1
∂P
2
∂
˙
Q
1
∂Q
1
∂
˙
Q
1
∂Q
2
∂
˙
Q
2
∂q
∂
˙
Q
2
∂p
∂
˙
Q
2
∂P
1
∂
˙
Q
2
∂P
2
∂
˙
Q
2
∂Q
1
∂
˙
Q
2
∂Q
2
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
δq
δp
δP
1
δP
2
δQ
1
δQ
2
¸
. (3.69)
For 20 diﬀerent values of the dimensionless damping parameter and of the dimen
sionless Floquet parametric forcing parameter, the largest LCEs were calculated and
a chaos diagram for the positive values of the largest Liapunov exponents identiﬁed
the parameter space in which chaotic motions exist.
1,3
Appendix A
The extension of the Herglotz algorithm to nonautonomous dynamical systems
(GHA) signiﬁcantly reduces the eﬀort required to suspend the nonautonomous
Hamiltonian component in (3.33d).
Generalized Herglotz algorithm (GHA)
The Herglotz algorithm for autonomous dynamical systems
4
may be generalized
by: (1) including time t; and (2) deﬁning a generating function with a nonzero
determinant of second derivatives. The GHA transforms a set of 2N variables (u,
v) to a set of 2N new variables (U, V) by choosing N new variables U
i
and then
computing the remaining N new variables V
i
uniquely from the chosen U
i
so that
the transformation (u, v) →(U, V) is canonical as shown by satisfying the Poisson
bracket conditions.
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
Analyses of Damped Parametrically Excited Cross Waves 85
GHA:Type I
Equate N new variables to N old variables U
i
= U
i
(u, v, t) such that the Poisson
brackets satisfy [[U
i
, U
j
]]
uv
= 0 and the determinant
∂U
1
∂v
1
∂U
1
∂v
N
∂U
i
∂v
j
∂U
N
∂v
1
∂U
N
∂v
N
= 0. (3.A.1)
Equate N Herglotz auxiliary functions X
i
(u, v) to the absolute value of the ratios
of the old variables in either of the following forms:
X
i
= X
i
(u, v) =
[u
i
/v
i
[
[v
i
/u
i
[
; i = 1, 2, . . . , N, (3.A.2a,b)
such that the Jacobian of the new variables U and the Herglotz auxiliary func
tions X is nonzero ∂(U, X)/∂(u, v) = 0. Solve (3.A.2) for v
i
= v
i
(u, X); substitute
v
i
(u, X) into U
i
(u, v, t) and invert to obtain u
i
= u
i
(U, X, t). Compute the gener
ating function F(u, U, t) from
dF(u(U, X, t), U, t) =
N
¸
i=1
(v
i
du
i
−V
i
dU
i
), (3.A.3)
N
¸
i=1
∂F
∂X
i
dX
i
+
∂F
∂U
i
dU
i
=
N
¸
i=1
¸
N
¸
j=1
v
j
∂u
j
∂X
i
dX
i
+
N
¸
j=1
v
j
∂u
j
∂U
i
dU
i
−V
i
dU
i
¸
. (3.A.4)
Equate the coeﬃcients of like diﬀerentials and compute: (1) the generating function
F(u, U, t) =
N
¸
i=1
X
i
N
¸
j=1
v
j
∂u
j
∂X
i
dX
i
+C(U, t), (3.A.5)
and (2) the new N variables from
V
i
(u, v, t) =
N
¸
j=1
v
j
∂u
j
∂U
i
−
∂F
∂U
i
; i = 1, 2, . . . , N. (3.A.6)
In order to compute the transformed Hamiltonian in terms of the new variables (U,
V), the inverse canonical transformation (u(U, V, t), v(U, V, t)) must be computed,
July 31, 2009 8:18 9.75in x 6.5in b684ch03 FA
86 R. B. Guenther and R. T. Hudspeth
and the new Hamiltonian K(Q, P, t) is given by
K(Q, P, t) = H(q(Q, P, t), p(Q, P, t), t) +
∂F(u, U, t)
∂t
. (3.A.7)
The GHA may also be applied to equate N old variables u
i
= u
i
(U, V, t) to obtain
N new variables U
i
= U
i
(u, v, t).
1,3
Appendix B
τ
1
=
√
1 +τ κ
τ
= κ(1 +τ) τ
λ
= 1 + (τ/λ
4
r
)
f
1
=
0
−h
f(z) exp
¸
z
λ
2
r
ds =
λ
2
r
1 −exp
¸
−
h
λ
2
r
; for a full draft piston,
λ
4
r
h
h
λ
2
r
−1 + exp
¸
−
h
λ
2
r
; for a full draft hinge
a
1
= −
1
−8βξ
+
βτ
2ξλ
6
r
τ
2
1
c
1
=
√
bf
1
λ
r
τ
1
4
2ξβ
3
d
1
=
1
4
2
τ
2
1
+
τ −λ
4
r
ξτ
λ
r
βλ
6
r
ξτ
2
λ
a
2
= −
1
β
+
1
8βξ
−
βτ
2ξλ
6
r
τ
2
1
c
2
=
2β −1
4βξ
d
2
=
1
8
2bβξ
3
λ
3
r
τ
2
1
a
3
=
2βτ
2
−ξλ
4
r
τ
λ
(λ
2
r
−βτ)
8
2bβξ
5
λ
9
r
τ
1
τ
λ
d
3
=
τ(2τ +λ
4
r
ξτ
λ
r
)
8
2bβξ
5
λ
9
r
τ
2
λ
d
4
=
τ +λ
4
r
ξτ
λ
(λ
2
r
−1)
λ
6
r
ξτ
1
τ
λ
d
5
=
ττ
2
1
−λ
2
r
τ
λ
2
2bβξ
3
λ
5
r
τ
λ
τ
2
1
d
6
=
−τ +λ
4
r
ξτ
λ
4βλ
6
r
ξτ
2
λ
d
7
=
1
4
−2
τ
2
1
+
τ +λ
4
r
ξτ
λ
βλ
6
r
ξτ
2
λ
References
1. R. T. Hudspeth, Waves and Wave Forces on Coastal and Ocean Structures (World
Scientiﬁc, Singapore, 2006).
2. C. M. Bowline, R. T. Hudspeth and R. B. Guenther, Applicable Anal. 72, 287 (1999).
3. R. T. Hudspeth, R. B. Guenther and S. Fadel, Acta Mechanica 175, 139 (2005).
4. R. B. Guenther, H. Schwerdtfeger and G. Herglotz, Vorlesungen ¨ uber die Mechanik der
Kontinua (1985).
5. S. Fadel, Application of the generalized Melnikov method to weakly damped para
metrically excited cross waves with surface tension, PhD dissertation, Oregon State
University, USA (1998).
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Chapter 4
Random Wave Breaking and Nonlinearity
Evolution Across the Surf Zone
Yoshimi Goda
ECOH CORPORATION, 264 KitaUeno
Taitoku, Tokyo 1100014, Japan
goda@ecoh.co.jp
A review is made on the statistical features of breaking waves in the nearshore
waters. Inherent variability of the breaker index for regular waves is examined
with the revised Goda’s formula. The incipient breaking height of signiﬁcant wave
is about 30% lower than that of regular waves. Nonlinearity of random waves
is strongest at the outer edge of surf zone, but it is destroyed by wavebreaking
process inside the surf zone. The wave height distribution is the narrowest in the
middle of the surf zone, but it returns to the Rayleigh near the shoreline. Large
diﬀerences among various wave models are noted for prediction of wave heights
in the surf zone.
4.1. Introduction
Breaking waves exert strong actions on maritime and coastal structures, while wave
dissipation through breaking plays a major role in the generation of nearshore cur
rents. Without good understanding of wavebreaking process, we cannot pursue
any study for coastal engineering works. Nevertheless, wave breaking is an elusive
phenomenon. Not many people spend enough time to observe wonderful pictures
of wave deformation by breaking and regeneration after breaking. One needs some
kind of a pier at a beach to have a good look of waves that break and rush toward
the shore. Otherwise, one should work for some hours in a laboratory to measure
waves at various locations along a wave ﬂume.
Our knowledge on wave breaking mostly comes from various literature based on
previous research works. Quite a number of people use the formulas, diagrams, and
other information listed there without examining the credibility of the information.
For example, many people regard the breaker index, or the ratio of wave height to
water depth at breaking, as a deterministic value without paying consideration to
the fact that the breaker height exhibits large ﬂuctuations even for a given wave
87
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
88 Y. Goda
condition and that the breaker index has been obtained by drawing a mean curve
among scattered data.
Nowadays, many researchers are developing various numerical models of random
wave transformations. They have to adopt some kind of wavebreaking criteria and
energy dissipation mechanism so that the model can reproduce wave deformation
in the nearshore waters. However, modelers seem to pick up whatever is readily
available without deliberation on the physical features of the wavebreaking process
and the appropriateness of the breaking model.
The present chapter is a revised version of the author’s paper
1
presented
at the 4th International Conference on Asian and Paciﬁc Conference 2007, in
Nanjin, China. It aims at providing coastal engineers and researchers with the most
advanced knowledge on the statistics of wave breaking in the nearshore waters so
that they can make a correct approach to the problems related to wavebreaking
processes.
4.2. Physical Deﬁnitions of Wave Breaking
Waves are deﬁned as breaking when the crest starts to contain foams or when
water particles jump out from the wave crest. It is a disruption of smooth water
surface, and breaking waves are often classiﬁed into three types: spilling, plunging,
and collapsing breakers. In the sea, there is no pure spilling breaker, because it is
always accompanied with a small portion of plunging water. For example, we can
observe from a window of an airplane such a scene of wave breaking in deep water
that a crest of large wave makes plunging, leaves a patch of white foam behind, and
moves forward with blue color. The following wave grows in height and breaks. It
is a manifestation of wave energy being transported by the group velocity, which is
onehalf the phase velocity in deep water.
Theoretically, three criteria can be cited. The ﬁrst is the condition that the
horizontal velocity of the water particle at the wave crest becomes equal to or greater
than the phase speed of wave proﬁle. The second is the upward vertical acceleration
of the water particle at the wave crest to be equal to or to exceed the gravitational
acceleration. The third is the vertical gradient of the total pressure at the wave
crest to be zero or negative.
The ﬁrst criterion has been employed by mathematicians to ﬁnd out the lim
iting waves of permanent form on a horizontal bed. Conventional perturbation
techniques are ineﬀective to derive the limiting waves, and the approach speciﬁc
to the limiting form needs to be employed. These limiting waves are characterized
with the angular crest having the angle of 120
◦
. Yamada and Shiotani
2
have pro
duced the most reliable computation results so far, which are summarized by
Goda
3
as reproduced in Table 4.1. The symbol H denotes the wave height, h is
the water depth, L is the wavelength, L
0
is the deepwater wavelength, C is the
wave celerity, and η
c
is the crest elevation. The subscripts “b” and “A” refer to
the quantities at breaking and those of the small amplitude waves (Airy’s wave),
respectively.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 89
Table 4.1. Characteristics of breaking waves of permanent type (after Goda
3
based on
Yamada and Shiotani
2
).
H
b
/L
0
h
b
/L
A
a
h
b
/L
b
C
b
/C
A
a
H
b
/L
b
H
b
/h
b
η
c
/H
b
inﬁnity inﬁnity inﬁnity 1.193 0.1412 0
0.935 0.935 0.7686 1.189 0.1409 0.1791 0.6706
0.471 0.474 0.4011 1.181 0.1386 0.3456 0.6765
0.286 0.300 0.2597 1.154 0.1277 0.4919 0.6908
0.1856 0.216 0.1885 1.143 0.1115 0.5912 0.7165
0.1117 0.1510 0.1331 1.134 0.08997 0.6683 0.7619
0.0763 0.1198 0.1050 1.141 0.07410 0.7059 0.7939
0.0474 0.0915 0.07915 1.156 0.05771 0.7293 0.8392
0.0284 0.0694 0.05909 1.174 0.04430 0.7496 0.8766
0.01669 0.0525 0.04398 1.193 0.03371 0.7666 0.9061
0.01095 0.0422 0.03499 1.207 0.02720 0.7774 0.9242
0.00575 0.0306 0.2483 1.231 0.01962 0.7904 0.9453
0.00239 0.01953 0.01570 1.244 0.01260 0.8028 0.9649
0.001144 0.01351 0.01075 1.257 0.00871 0.8099 0.9757
0.000437 0.00833 0.00660 1.263 0.00538 0.8160 0.9849
0 0 0 1.285
b
0 0.8261
b
1.0000
b
Notes:
a
L
A
and C
A
denote the wavelength and celerity of small amplitude waves.
b
These values are those of solitary wave computed by Yamada et al.
4
The limiting height of solitary wave is (H/h)
b
= 0.8261 by Yamada et al.,
4
instead of the oftenquoted value of (H/h)
b
= 0.78 by McCowan.
5
The limiting
wave steepness of (H/L)
b
= 0.142 by Miche
6
is calculated for waves of nonlinear
waves, and it becomes H
b
/(L
0
)
A
= 0.1684 when the linear deepwater wavelength
(L
0
)
A
is used.
The second criterion is for the breaking of standing waves, but no theoretical
computation of limiting standing waves has been made.
The third criterion has been proposed by Nadaoka et al.,
7
for deﬁning wave
breaking in numerical timedomain computation. Zero vertical gradient of total
pressure implies no presence of water below the wave crest, i.e., wave breaking.
4.3. Parameters Governing Breaker Index
Because wave breaking attracts attention of many researchers, there have been pro
posed a number of formulas to describe the ratio of wave height to water depth.
Kaminsky and Kraus
8
called this ratio as the breaker heighttodepth index, but
the present chapter employs the term of breaker index for the ratio of wave height
to water depth at breaking for the sake of simplicity. Kamphuis
9
has compared 11
formulas for testing of goodnessofﬁt with his 225 sets of hydraulic model tests. Rat
tanapitikon and Shibayama
10
have collected 574 data points from 24 papers/reports
and calculated the rootmeansquare errors of 24 breaking index formulas. Both
authors have proposed their own formulas by modifying some of the previous ones.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
90 Y. Goda
Rattanapitikon et al.
11
have further added 121 data points in largescale ﬂume tests
and proposed revised formulas.
Similar to the approach by Rattanapitikon et al.,
11
the breaker index formulas
can be categorized into the following four functional forms:
H
b
/h
b
= f
1
(0) = constant , (4.1)
H
b
/h
b
= f
2
(h
b
/L
0
or h
b
/L
b
) , (4.2)
H
b
/h
b
= f
3
(s) , (4.3)
H
b
/h
b
= f
4
(s, h
b
/L
0
or h
b
/L
b
) , (4.4)
where s denotes the bed slope. The wavelength L
0
is the deepwater wavelength given
by L
0
= g/(2π)T
2
, where T is the wave period. Because the relative water depth
h
b
/L
0
is easily converted to h
b
/L
b
through the dispersion relationship, the two
relative depths h
b
/L
0
and h
b
/L
b
are interchangeable. The formula of H
b
/h
b
= 0.78
is a typical example of Eq. (4.1).
There are some other formulas using the parameter of deepwater wave steepness
H
0
/L
0
. They can be useful for predicting the breaker height of regular waves. Nev
ertheless, they cannot be applied for breaking of random waves, because individual
zerocrossing waves in a train of random waves are unrelated to individual waves in
deepwaters. Thus, there is little room for the parameter H
0
/L
0
to function in the
breaker index for random waves.
Performance of a breaker index formula can be judged by the magnitude of the
bias of the predicted breaker height from the observed heights. It should also be
examined with either the rootmeansquare error of predicted breaker heights or the
correlation coeﬃcient between prediction and observation. The rootmeansquare
error analysis by Rattanapitikon and Shibayama
10
is not conclusive in diﬀerentiating
the merits of four functional forms, but they recommend a certain modiﬁcation of
the slope eﬀect in the function f
4
(s, h
b
/L
0
), apparently indicating their preference
of this functional form.
Kamphuis
9
calculated the correlation coeﬃcients between 11 formulas and his
laboratory data. By assigning the bestﬁtting value to the proportionality coeﬃcient
of each formula, he obtained the determination coeﬃcient R
2
= 0.69 to f
1
(0),
R
2
= 0.67 to f
2
(h
b
/L
0
), R
2
= 0.84 to f
3
(s), and R
2
= 0.88 to f
4
(s, h
b
/L
0
). His
result clearly suggests the necessity to include both the parameters of bed slope and
relative water depth in the breaker index formula.
4.4. Breaker Index for Regular Waves and Its Scatter
4.4.1. Scatter of regular breaking waves
In 1970, Goda
3
presented a diagram of breaker index curves of regular waves for
four bed slopes, based on the laboratory data from eight sources, which included
his own largescale tests with H
b
= 0.43 to 0.93 m; Rattanapitikon et al.
11
did not
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 91
analyze this data set. Later, Goda
12,13
approximated the breaker index curves with
the following empirical formula:
H
b
h
b
=
A
h
b
/L
0
1 −exp
−1.5π
h
b
L
0
1 + 15s
4/3
: A = 0.17 . (4.5)
Equation (4.5) was derived by a graphical curveﬁtting technique without direct
comparison with the original laboratory data. Rattanapitikon and Shibayama
10
have recommended a modiﬁcation of the slope eﬀect term of (1 + 15s
4/3
)
into (1.033 + 4.71s − 10.46s
2
). Upon reexamination of the original laboratory
data, the author has also recognized a necessity of modifying the slope eﬀect term.
The revised formula is as follows:
H
b
h
b
=
A
h
b
/L
0
1 −exp
−1.5π
h
b
L
0
1 + 11s
4/3
: A = 0.17 . (4.6)
Muttray and Oumeraci
14
found the bestﬁtting coeﬃcient of Eq. (4.5) being
0.167 instead of 0.170 for the slope of 1/30. When Eq. (4.6) is applied to their data,
the coeﬃcient would have the value of 0.173.
Comparison of the laboratory data of breaker index with Eq. (4.6) is shown in
Fig. 4.1 for ﬁve groups of bed slopes, i.e., 1/7 to 1/12, 1/20, 1/30, 1/50, and 1/200
to 1/200. (See Goda
3
for description of the laboratory data employed here.) Recent
data sets by Li et al.
15
for s = 1/30 and s = 1/50, Li et al.
16
for s = 1/200, and
Lara et al.
17
for s = 1/20 are also added.
It is clear in Fig. 4.1 that the value of the breaker index increases as the bed
slope becomes steep. Thus, it is absolutely necessary to incorporate the slope eﬀect
into the breaker index formula. Because the experimental data are scattered around
the index curves, the upper and lower bound curves with the range of 87%–115%
of the value by Eq. (4.6) are drawn in Fig. 4.1.
A quantitative evaluation of the degree of the scatter is made by means of
the relative error of the breaker index, i.e., E = (1 − γ
meas
/γ
est
), where γ
meas
is
the measured value of H
b
/h
b
and γ
est
is the predicted value by the breaker index
formula of Eq. (4.6). The mean and the standard deviation of the relative error are
calculated for each group of the bed slope. The mean E
mean
indicates a bias of the
breaker index and the standard deviation of E represents the degree of scatter of
the breaker index. Because E is deﬁned as the relative error, the standard deviation
σ(E) is equivalent to the coeﬃcient of variation (CoV). A positive bias indicates a
tendency of overestimate, while a negative bias shows an underestimate. Table 4.2
lists the bias and CoV of the breaker index of Eq. (4.6) for the data of various bed
slopes. The slope data of 1/9 and 1/12 are excluded from the analysis because of
their small sample sizes.
The bias varies from −2.9% to +6.2% depending on the bed slope, but the
formula of Eq. (4.6) can be regarded as yielding reasonable estimates of the breaker
heights. The scatter of the data as expressed by CoV is about 5% to 7% for the
bed slope of 1/200 to 1/50; it increases as the slope becomes steep, and it takes the
value of 14% for the bed slope of 1/10. Such scatter of data represents an inherent
stochastic nature of wavebreaking phenomenon. It resides in the data set itself,
being independent of the breaker index formula being applied.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
92 Y. Goda
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.001 0.01 0.1
KishiIohara 1/9
Iversen 1/10
Goda et al. 1/10
Galvin 1/10
Bowen et al. 1/12
Index curve 1/10
Upper 115%
Lower 87%
B
r
e
a
k
e
r
I
n
d
e
x
,
=
γ
γ
γ
H
b
/
h
b
Relative water depth, hb/L0
Slope = 1/9 to 1/12
2.0
(a) Breaker index data for the slope of 1/9−1/12.
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.001 0.01 0.1
Iversen 1/20
Galvin 1/20
Toyoshima 1/20
Lara 1/20
Index curve 1/20
Upper 115%
Lower 87%
B
r
e
a
k
e
r
I
n
d
e
x
,
=
H
b
/
h
b
Relative water depth, hb/L0
Slope = 1/20
1.5
(b) Breaker index data for the slope of 1/20.
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.001 0.01 0.1
Iversen 1/30
Mitsuyasu 1/30
Toyoshima 1/30
Li Y.C. et al. 1/30
Index curve 1/30
Upper 115%
Lower 87%
B
r
e
a
k
e
r
I
n
d
e
x
,
=
H
b
/
h
b
Relative water depth, hb/L0
Slope = 1/30
1.2
(c) Breaker index data for the slope of 1/30.
Fig. 4.1. Comparison of breaker index formula Eq. (4.6) with laboratory data of regular waves.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 93
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.001 0.01 0.1
Iversen 1/50
Mitsuyasu 1/50
Li Y.C. et al. 1/50
Index curve 1/50
Upper 115%
Lower 87% B
r
e
a
k
e
r
I
n
d
e
x
,
=
γ
γ
H
b
/
h
b
Relative water depth, hb/L0
Slope = 1/50
(d) Breaker index data for the slope of 1/50.
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.001 0.01 0.1
Goda 1/100
Li Y.C. et al. 1/200
Index curve 1/100
Index curve 1/200
Upper 115%
Lower 87% B
r
e
a
k
e
r
I
n
d
e
x
,
=
H
b
/
h
b
Relative water depth, hb/L0
Slope = 1/100 to 1/200
(e) Breaker index data for the slope of 1/100–1/200.
Fig. 4.1. (Continued )
Table 4.2. Bias and CoV of the breaker index formula of Eq. (4.6).
Bed No. of Bias = E
mean
CoV = σ(E)
slope data (%) (%) Sources
1/10 29 −0.5 14.0 Iversen, Goda et al.
1/20 47 +3.9 11.3 Iversen, Galvin, Toyoshima et al., Lara et al.
1/30 73 +1.4 8.6 Iversen, Mitsuyasu, Toyoshima et al., Li et al.
1/50 28 −2.9 4.8 Iversen, Mitsuyasu, Li et al.
1/100 32 +6.2 5.5 Goda
1/200 19 +3.5 7.4 Li et al.
Even under a wellcontrolled laboratory test, the breaking point ﬂuctuates over
some distance and the breaker height varies from wave to wave. Smith and Kraus
18
reported on their regular wave tests that “despite care in conducting the tests and
use of the average value of the given quantity (i.e., over 10 waves), wide scatter
appeared in some quantities and must be considered inherent to the breaking process
of realistic waves.” One cause of the data scatter is the presence of smallamplitude,
longperiod oscillations of water level in a laboratory ﬂume, but the breaking
process itself is triggered by many small factors beyond the control of experi
menters. We should regard the wavebreaking phenomenon as stochastic one and
accept a certain range of natural ﬂuctuation. As listed in Table 4.2, the coeﬃcient of
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
94 Y. Goda
variation is large for steep slope and it becomes small for gentle slope. In any research
work involving wavebreaking phenomena, due consideration should be given to
such stochastic nature of breaker heights.
4.4.2. Scatter of random breaking wave heights
The ﬁrst eﬀort to identify the breaker heights of individual waves among trains
of irregular waves was undertaken by Kimura and Seyama
19
and Seyama and
Kimura,
20
who analyzed about 1000 individual breaking waves for each slope of
1/10, 1/20, 1/30, and 1/50 with the aid of a video camera and wave gauges. The
breaker index of individual waves varied over a wide range, which was equivalent
to CoV of 18% to 23% (the present author’s visual inspection of scatter diagrams).
In order to reduce the scatter of data, they proposed to employ an artiﬁcial water
depth below the midlevel between the wave crest and trough of individual breaking
waves and succeeded in reducing CoV to the values between 8% and 11%. Because
Eq. (4.5) yielded the breaker index larger than most of the observed value, they
proposed its modiﬁed version. When the revised breaker index formula of Eq. (4.6)
is employed, however, the center line of the scatted data appears to be at the level
of 85% (s = 1/10) to 95% (for other slopes) of the predicted value.
Black and Rosenberg
21
made observation of individual breaking waves on a
natural beach with a depth of 1.0–1.5 m at Apollo Bay in southern Australia. The
median value of the breaker index was about 84% and 87% of those by Eqs. (4.5)
and (4.6), respectively.
Another observation was made by Kriebel
22
in a large wave ﬂume with a bed
slope of 1/50 for waves with a signiﬁcant height of 0.46 m and peak period of 2.9 s.
When he applied Eq. (4.5) to individual breaker heights, he found that the value of
the proportionality coeﬃcient A ﬁtted to the data varied from about 0.09 to 0.18
(against the original value of A = 0.17) with the mean of 0.142 (84% of Eq. (4.5))
and the standard deviation of 0.017 (equivalent to CoV of 12%). Li et al.
23
also
reported the result of their measurements of random breaking waves on the slope of
1/50 and 1/200, recommending the coeﬃcient value of A = 0.150 with the standard
deviation of 0.031.
It should be recalled that the random wave breaking model by Goda
24
in 1975
had already incorporated the variability of breaker heights by assigning a variable
probability of individual wave breaking. The probability was assumed to increase
linearly from 0 to 1 in the range of the wave height from 71% to 106% of the height
predicted by Eq. (4.5), corresponding to A = 0.12 to 0.18. Therefore, it is expected
that the median value of individual breaker heights would be smaller than those
predicted by Eq. (4.5) or (4.6).
4.5. Breaker Index for Random Waves
4.5.1. Incipient breaking index of signiﬁcant wave
Equations (4.5) and (4.6) are examples of the breaker index for regular waves.
There are some people who try to apply such breaker index formulas to coastal
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 95
waves or random waves, but such application does not yield correct answers. The
breaker index for regular waves may be utilized for the highest wave in an irregular
wave train, but it cannot be applied for the signiﬁcant wave, the rootmeansquare
wave, or any other characteristic wave. When random waves approach the shore,
breaking of individual waves occurs gradually with large waves ﬁrst at the far dis
tance, medium waves next at some distance from the shore, and small waves near
the shoreline. The variation of signiﬁcant wave height from the oﬀshore toward the
shore is so gradual that we cannot employ the concept of wavebreaking line, which
is so obvious in the case of regular waves.
Against such diﬃculty of deﬁning the breaking point of signiﬁcant wave,
Kamphuis
9
introduced the deﬁnition of “incipient wave breaking.” He measured
crossshore variations of signiﬁcant wave height beyond and across the surf zone,
drew a curve of wave shoaling trend in the outside of the surf zone and a curve of
wave height decay within the surf zone, and called the condition at the crosspoint
of the two curves as the incipient wave breaking. By using the data of the signif
icant wave height at incipient breaking, he calibrated 11 breaker index formulas and
determined the bestﬁtting proportionality coeﬃcient. For the formula of Eq. (4.5),
he obtained the proportionality coeﬃcient of A = 0.12 for signiﬁcant wave height.
Li et al.
23
have also presented a data of the breaker index of (H
1/3
/h)
b
on the
slope of 1/200, which is ﬁtted to Eq. (4.5) with a modiﬁed constant value of 0.12
for the initial stage of breaking. Their breaking condition was some observation of
large individual breaking waves in an irregular wave train. Goda
24
has prepared a
set of diagrams depicting variations of signiﬁcant wave height across the surf zone
(reproduced as Figs. 3.29 to 3.32 in Ref. 25). The boundary lines of 2% decay in
these diagrams approximately correspond to the breaker index with A = 0.11, and
the water depth (h
1/3
)
peak
at which the signiﬁcant wave takes a peak value within
the surf zone (Fig. 3.34 in Ref. 25) corresponds to A = 0.11–0.13. Therefore, the
incipient breaker index of the signiﬁcant wave can be expressed with the following
formula:
H
1/3, b
h
b
incipient
=
0.12
h
b
/L
0
1 −exp
−1.5π
(h
b
)
incipient
L
0
1 + 11s
4/3
. (4.7)
Thus, the incipient breaker index of signiﬁcant wave is about 30% lower than
that of regular waves. The incipient breaking of signiﬁcant wave corresponds to the
condition that the high waves of upper several percent among individual waves have
begun to break.
4.5.2. Laboratory data of breaker index of random waves
After incipient breaking, the percentage of wave breaking increases as waves proceed
across the surf zone. The ratio of the signiﬁcant wave height to the water depth
gradually increases toward the shoreline. Ting
26,27
made detailed laboratory inves
tigations of random wave deformations on a uniform slope of 1/35, using frequency
spectra of broad and narrowband with the peak enhancement factor of 3.3 and
100, respectively. Waves had the signiﬁcant height of H
s
= 0.15 m and the spectral
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
96 Y. Goda
0.10
1.0
0.01 0.1
Broad spec. H1/3/h
Narrow spec. H1/3/h
Breaker envlp. H1/3/h
Broad spec. Hrms /h
Narrow spec. Hrms /h
Breaker envlp. Hrms /h
B
r
e
a
k
e
r
i
n
d
i
c
e
s
,
H
1
/
3
/
h
&
H
r
m
s
/
h
Relative depth, h/L0
0.8
0.6
0.4
0.3
0.2
0.03 0.3
Slope = 1/35
Fig. 4.2. Breaker indices for H
1/3
and H
rms
on s = 1/35 with the data by Ting.
26,27
peak period of T
p
= 2.0 s. He recorded wave proﬁles at an oﬀshore station with the
depth of 0.457 m and at six stations on the slope with the depth of 0.27–0.0625m.
Waves at the six stations on the slope had the percentage of breaking ranging from
5% to 94% (the case of broadband spectrum).
Wave records were analyzed by the zerodowncrossing method, and calculated
results of characteristic wave heights and periods are presented in tabular forms.
From these results, the ratios of H
1/3
and H
rms
to the local depth (inclusive of
mean water level change) are calculated and plotted against the relative water depth
h/L
0
, as shown in Fig. 4.2.
The curves denoted as breaker envelopes are calculated by Eq. (4.6) for s = 1/35
with the proportionality coeﬃcient of A = 0.145 (85% of regular waves) for the
signiﬁcant height H
1/3
and to A = 0.111 (65%) for the rootmeansquare height
H
rms
. Because the percentage of breaking waves is high in these data, an A value
higher than that for incipient breaking ﬁts to the data.
The breaker index data for H
rms
by Tick is higher than the value proposed by
Sallenger and Holman,
28
who gave an expression of H
rms
/h = 3.2s + 0.32 without
inclusion of the relative depth (h/L
0
) term. They converted the orbital velocity
spectra to the surface wave spectra with the transfer function based on the linear
theory, and estimated the energybased H
rms
, which must have been smaller than
the statistical H
rms
value based on direct measurement data of surface proﬁles.
4.5.3. Description of ﬁeld data employed for analysis
In the ﬁeld wave observation at a ﬁxed station, it is not feasible to judge whether
individual waves are at the stage of breaking or not, unless simultaneous measure
ments with video cameras are taken. However, we may ﬁnd out an upper limit
of signiﬁcant wave height for a given water depth by taking an envelope of many
data at diﬀerent relative water depths. For this type of analysis, stationary wave
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 97
Table 4.3. Summary of stationary coastal wave data employed in the present analysis.
Type of Sampling Signiﬁcant Signiﬁcant
Observation wave Water interval height period No.
station gauge depth (m) ∆t (s) H
1/3
(m) T
1/3
(s) of data
Rumoi Port Stepresistance −11.5 0.5 2.2–7.1 5.9–11.7 44
YamaseDomari
Port
” −12.7 0.5 1.9–6.2 7.7–15.6 9
Tomakomai Port ” −10.8 1.0 2.9–5.8 7.7–10.9 9
” −13.8 0.5 2.6–2.8 6.7 – 7.5 2
Ultrasonic −20 0.5 2.4–2.5 6.9–7.4 2
Kanazawa Port ” −20 1.0 1.0–6.8 1.0–6.8 13
Caldera Port,
Costa Rica
” −18 0.5 1.5–3.6 14.2–18.4 50
Sakata Port Pressure −14.5 0.5 1.7–9.7 6.3–13.4 123
−10.5 0.5 1.7–6.1 6.5–15.0 123
(Source: Goda and Nagai,
29
Goda
24,30
).
observation data analyzed by Goda and Nagai
29
and the data of longtraveled swell
recorded with an ultrasonic wave sensor reported by Goda
30
were utilized. Table 4.3
lists the characteristics of these ﬁeld data. Waves were recorded by means of either
stepresistance gauges or ultrasonic wave sensors so that they were reliable regis
tration of surface wave proﬁles. The data at Tomakomai and Kanazawa as well as
Caldera were measured with ultrasonic wave sensors. They are not analyzed for
breaker limits but for wave nonlinearity eﬀects to be described in Sec. 4.6.
Table 4.3 also lists the wave records at Sakata Port measured by means of
pressure gauges, which were utilized by Goda
24
for calibration of his random wave
breaking model. Although there remains a problem of pressure conversion to surface
proﬁles, the conversion error would have been small because of the relatively shallow
water depth at Sakata stations (10.5 and 14.5 m). They were included in the present
analysis to increase the size of database.
Other sources of nearshore waves are the photogrametric measurement data by
Hotta and Mizuguchi
31,32
as well as by Ebersole and Hughes.
33
Hotta and Mizuguchi
mobilized 11 motionpicture cameras set on top of a coastal observation pier at
Ajigaura Beach, Ibaragi, Japan. They took ﬁlm pictures of instantaneous water
surfaces simultaneously at some 60 surveyor’s poles erected in the nearshore waters
on a line perpendicular to the shoreline stretched over a distance of about 120 m.
Films of surface wave records were taken every 0.2 s for an eﬀective duration of 760s.
The beach proﬁle in September 1978 had a trough at the distance of 25 m from the
shoreline and the slope of about 1/60 oﬀshore of the trough. The beach proﬁle in
December 1978 was somewhat uniform without any bar or trough, and the slope was
about 1/70. The water depth inclusive of tides at the poles varied from 0.1 to 2.7 m.
Photogrametric measurements of nearshore waves were also executed by
Ebersole and Hughes
33
during the DUCK85 campaign in Duck, North Carolina,
USA with the cooperation of Dr. Hotta who brought twelve cameras with him and
took charge of ﬁlming. They referred to this technique as “the photopole method”;
this terminology is employed in the present chapter. Over the distance of 64 m,
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
98 Y. Goda
Table 4.4. Summary of photopole measurements.
Name Date No. of data h (m) H
0
(m) T
1/3
(s) H
0
/L
0
Ajigaura 1978/09/05 54 0.6–2.7 0.7 8.4 0.0064
1978/12/13–14 175 0.1–1.8 0.5–0.7 7.2–7.9 0.0059–0.0081
DUCK85 1985/09/04–05 99 0.5–2.4 0.3–0.5 10.3–11.1 0.0017–0.0028
SUPERDUCK 1986/09/11–19 140 0.4–3.7 0.5–1.1 5.4–11.5 0.0027–0.0254
12 poles were erected in the initial depth ranging from 0.4 to 1.9 m. Measurements
were taken for nine runs during the fourth and ﬁfth days of September 1985. With
the variation of the tide level, the actual water depth varied from 0.5 to 2.4 m. The
eﬀective duration of wave recording was about 650 s, judging from the number of
waves and average periods listed by Ebersole and Hughes.
33
The beach proﬁle was nearly ﬂat for about 25 m from the shoreline with the
depth of about 0.5 m below the mean sea level, and it had the slope of about 1/30
beyond that. Another series of photopole measurements were carried out during the
SUPERDUCK campaign in 1986. Dr Hughes kindly supplied the author with the
data ﬁles of measured wave statistics. The number of poles was increased to 20 and
the water depth inclusive of tides varied from 0.4 to 3.7 m. The beach proﬁle during
SUPERDUCK is not known, but it would have been nearly the same as DUCK85
because of the same season. All the photopole measurement data were analyzed
by the zerodowncrossing method, and various statistical wave characteristics were
calculated.
Table 4.4 lists the summary of the photopole wave measurement conditions.
The signiﬁcant wave period T
1/3
has been converted from the poleaveraged values
of either the mean period T
mean
or the spectral period T
p
by assumption of
T
1/3
= 1.05T
mean
or T
1/3
= 0.95T
p
, which would be appropriate for swell of very low
steepness. The oﬀshore wave height H
0
was converted from the signiﬁcant height
H
1/3
measured at the most oﬀshore pole using the shoaling coeﬃcient; no refraction
eﬀect was taken into consideration as no information of wave direction was available.
All the waves were swell of very low steepness ranging from 0.0017 to 0.0081,
except one case of SUPERDUCK with H
0
/L
0
= 0.0254.
4.5.4. Field data of breaker index for energybased signiﬁcant
waves
Coastal surface wave data, pressureconverted wave data, and three sets of photopole
data are plotted together in Fig. 4.3 in the form of H
m0
/h versus h/L
0
. The energy
based signiﬁcant wave height H
m0
deﬁned by Eq. (4.8) is employed here instead of
the zerocrossing height H
1/3
, because the latter is greatly enhanced over H
m0
by
strong eﬀects of wave nonlinearity and it is not representative of breakingdissipated
wave energy level; this aspect is discussed in Sec. 4.6.
H
m0
= 4.0η
rms
= 4.0
√
m
0
, (4.8)
where m
0
denotes the zeroth moment of frequency spectrum.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 99
0.1
1.0
0.001 0.01 0.1 1
Stepresistance ( h=1114m)
Presssure gage (h=1014m)
Photopole Ajigaura (h=0.12.7m)
Photopole DUCK85 (h=0.52.4m)
Photopole SUPERDUCK (h=0.53.2m)
Breaker envelope (80% limit)
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
,
H
m
0
/
h
Relative depth, h/L0
0.8
0.6
0.4
0.2
Fig. 4.3. Breaker index for H
m0
based on the ﬁeld wave data.
Coastal wave data, on the other hand, were not much aﬀected by wave non
linearity eﬀects, and the zerocrossing signiﬁcant height H
1/3
was almost the same
as H
m0
.
Many data points in Fig. 4.3 belong to nonbreaking condition, but what interests
us is the upper envelope which provides an estimate of the upper limit of breaking
wave height. The curve of dashdot line in Fig. 4.3 has been calculated by Eq. (4.6)
for the slope of s = 0.0143 (1/70) with the coeﬃcient being reduced to A = 0.136
(80%). Similar with the laboratory data shown in Fig. 4.2, the wave height ratio
H
m0
/h is higher than the incipient breaker index of signiﬁcant wave expressed by
Eq. (4.7). It is because the breaker index increases inside the surf zone as the
percentage of breaking waves increase.
It is seen that the energybased signiﬁcant wave height H
m0
on gentle slopes
does not exceed 0.7 times the local water depth except for the lowsteepness swell
in very shallow water. For the range of h/L
0
> 0.03, the upper limit of signiﬁcant
wave height is about 0.6 times the local water depth. Some data points above the
dashdot curve are those of DUCK85 and SUPERDUCK, which were conducted on
the beach steeper than the beach in Ajigaura.
4.6. Evolution of Wave Nonlinearity Across Surf Zone
4.6.1. Variations of skewness and kurtosis across the surf zone
Ocean waves are characterized with almost linear property, as evidenced by the
Gaussian distribution of surface elevation. Wave linearity is the basis of spectral
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
100 Y. Goda
representation and analysis. Deviation from the linearity is measured with the values
of the skewness and kurtosis of the surface elevation with reference to the mean
water level during a wave record.
The skewness is zero when a distribution is symmetric with respect to the mean,
and takes a positive value when a distribution is asymmetric with a long tail toward
the right side (large value). The kurtosis takes a value of 3.0 for the Gaussian
distribution. When the mode of distribution has a sharp peak and the distribution
has long tails in both the left and right sides, the value of kurtosis becomes much
larger than 3.0. The degree of positive skewness and the deviation of kurtosis from
3.0 are the measure of the strength of wave nonlinearity. The skewness of ocean
waves is less than 0.5, and the kurtosis is below 4.0 for most cases, and thus the
nonlinearity of waves in deepwater is weak.
The variations of the skewness and kurtosis of ﬁeld waves are examined with
coastal surface waves listed in Table 4.3 (excluding the pressure sensor data of
Sakata Port), and the photopole data in Ajigaura, DUCK85, and SUPERDUCK
listed in Table 4.4. The data of skewness and kurtosis of September 5th in Ajigaura
were kindly provided by Dr Hotta. The data of the photopole measurements of
SUPERDUCK provided by Dr Hughes had the data of skewness only.
As waves approach the shore, wave nonlinearity is enhanced and both the
skewness and kurtosis increase signiﬁcantly. Figure 4.4(a) exhibits the increase of
the skewness with the nonlinearity parameter Π
1/3
, which was introduced by Goda
34
with the following deﬁnition:
Π
1/3
=
H
1/3
L
A
coth
3
2πh
L
A
, (4.9)
where L
A
denotes the wavelength calculated by small amplitude wave theory or
Airy’s theory.
The data are grouped by the range of the oﬀshore wave steepness H
0
/L
0
: the ﬁrst
group for 0.001 < H
0
/L
0
< 0.0029, the second group for 0.003 < H
0
/L
0
< 0.0049,
the third group for 0.0050 < H
0
/L
0
< 0.0099, the fourth group for 0.010 < H
0
/L
0
<
0.029, and the ﬁfth group for 0.030 < H
0
/L
0
< 0.049 (Legends are shown with
abbreviated ﬁgures).
The data shown in the left diagram of Fig. 4.4 are those outside the surf zone.
Because the boundary of surf zone is diﬃcult to be set for random waves, an arbi
trary boundary of h/H
0
= 2.5 is employed here to separate the wave data outside
and inside the surf zone. As shown in the left diagram, the skewness outside the surf
zone shows a clear correlation with the wave nonlinearity parameter. The skewness
begins from the value of zero at Π
1/3
= 0, increases almost linearly with Π
1/3
,
and attains the value of 2.0 around Π
1/3
= 4. The dashed line represents a semi
theoretical relationship, which is based on the analysis of ﬁnite amplitude regular
wave proﬁles by Goda
34
with the consideration of the probability of individual wave
heights according to the Rayleigh distribution.
Variation of the skewness inside the surf zone (h/H
0
< 2.5) is shown in
Fig. 4.4(b). The ordinate is the ratio of oﬀshore wave height to water depth, H
0
/h,
which increases rapidly as waves approach the shore. There is a clear trend of
skewness decreasing toward β
1
= 0 with the increase of the heighttodepth ratio
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 101
0.5
0.0
0.5
1.0
1.5
2.0
2.5
0.01 0.1 1 10
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
0.030<H0/L0<0.049
Semitheoretical
S
k
e
w
n
e
s
s
,
β β
1
Nonlinearity parameter, Π1/3
h/H0 > 2.5
(a) O I utside the surf zone (h/H
0
> 2.5) (b) nside the surf zone (h/H
0
< 2.5)
0.5
0.0
0.5
1.0
1.5
2.0
2.5
0.4 0.6 0.8 1 1.2 1.4
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
0.030<H0/L0<0.049
S
k
e
w
n
e
s
s
,
1
Heighttodepht ratio, H0/h
h/H0 < 2.5
Fig. 4.4. Variation of the skewness of surface elevation outside and inside the surf zone.
1.5
3.0
4.5
6.0
7.5
9.0
10.5
0.01 0.1 1 10
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
0.030<H0/L0< 0.049
Empirical
K
u
r
t
o
s
i
s
,
β
β
2
Nonlinearity parameter, Π1/3
h/H0 > 2.5
(a) Outside the surf zone (h/H
0
> 2.5) (b) nside the surf zone (h/H
0
< 2.5)
1.5
3.0
4.5
6.0
7.5
9.0
10.5
0.4 0.6 0.8 1.0 1.2 1.4
0.001<H0/L0<0.002
0.005<H0/L0<0.009
0.010<H0/L0<0.029
0.030<H0/L0<0.049
K
u
r
t
o
s
i
s
,
2
Heighttodepht ratio, H0/h
h/H0 < 2.5
I
Fig. 4.5. Variation of kurtosis of surface elevation outside and inside the surf zone.
H
0
/h. Use of the heighttodepth ratio in Fig. 4.4(b) is to provide a kind of contrast
of the increase and decrease of skewness in the outside and inside of the surf zone,
respectively. While the increase of skewness outside the surf zone seems indiﬀerent
to the wave steepness, the value of skewness inside the surf zone is much aﬀected by
the wave steepness; waves of low steepness maintain large values of skewness, while
waves of high steepness have small values only.
Variation of the kurtosis of surface elevation is shown in Fig. 4.5; Fig. 4.5(a)
shows kurtosis outside the surf zone and Fig. 4.5(b) shows kurtosis inside the surf
zone. The pattern of variation is the same as that of skewness, though the available
number of kurtosis data is smaller than the skewness data. The kurtosis starts from
the value of 3.0 at Π
1/3
= 0, increases as the nonlinearity parameter increases, and
attains the value of 9 around Π
1/3
= 4. In Fig. 4.5(b), the kurtosis inside the surf
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
102 Y. Goda
zone decreases toward β
2
= 3.0 as waves approach the shoreline (i.e., increase of
H
0
/h). Waves with the steepness larger than 0.01 exhibit the kurtosis value much
smaller than swell with very low steepness. It is because waves with large steepness
cannot experience the full process of nonlinear shoaling owing to the early start of
wave breaking.
4.6.2. Variations of zerocrossing wave heights across the surf zone
Wave nonlinearity is also reﬂected in individual wave heights deﬁned by the zero
crossing method. In the oﬀshore, individual wave heights almost follow the Rayleigh
distribution. Close examination of wave height distribution has revealed a distri
bution slightly narrower than the Rayleigh with the result of the mean ratio of
H
1/3
/H
m0
being 0.95. The diﬀerence is due to the energy spread over frequency
as represented by wave spectrum, the degree of which can be measured with the
spectral shape parameter as discussed by Goda and Kudaka.
35
As waves propagate toward the shore, however, waves undergo nonlinear
shoaling. In this process, wave proﬁles become skewed with high and sharp crests
and low and ﬂat troughs. Because of skewed wave proﬁle, the potential energy con
tained in such a proﬁle is smaller than the energy of sinusoidal wave with the same
height. In other words, nonlinear waves can have the height much larger than the
linear (sinusoidal) waves for the same potential energy.
Owing to the nonlinear shoaling, individual wave heights increase rapidly with
the rate greater than that of linear shoaling. Signiﬁcant wave heights and other
characteristic wave heights also grow rapidly. The nonlinearity eﬀect becomes most
conspicuous around the outer edge of the surf zone. After waves enter the surf
zone and begin to be attenuated through breaking process, the wave nonlinearity
is gradually lessened. The degree of nonlinear shoaling eﬀect may be judged by
a departure of the statistical wave height from the theoretical prediction by the
Rayleigh distribution, which is expressed as below:
H
1/10
= 1.27H
m0
, H
1/3
= H
m0
, H
rms
= 0.707 H
m0
. (4.10)
Figures 4.6–4.8 show the evolutions of the wave height ratios H
1/10
/H
m0
,
H
1/3
/H
m0
, and H
rms
/H
m0
, respectively. The diﬀerence between H
1/3
and H
m0
in
the nearshore waters has been pointed by several researchers such as Thompson and
Vincent
36
and Ebersole and Hughes.
33
In Figs. 4.6–4.8, the (a) parts are for those
outside the surf zone and the (b) parts are for those inside the surf zone, similar to
Figs. 4.4 and 4.5. The boundary of the surf zone is subjectively set at h/H
0
= 2.5
in Figs. 4.6 and 4.7 and h/H
0
= 1.5 in Fig. 4.8. In case of the rootmeansquare
wave height H
rms
, the surf zone should be deﬁned in a shallower area than for the
signiﬁcant wave, because H
rms
is calculated with all individual waves. Thus, the
boundary of H
rms
was set at h/H
0
= 1.5.
The dashed lines in the left diagrams are semitheoretical predictions based
on the potential energy calculation of ﬁnite amplitude waves and the Rayleigh
distribution of wave heights, by referring to the methodology employed by
LonguetHiggins.
37
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 103
(a) Outside the surf zone (h/H
0
> 2.5)
1.0
1.2
1.4
1.6
1.8
2.0
2.2
0.01 0.1 1 10
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
0.030<H0/L0<0.049
Rayleigh
Semitheoretical
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
,
H
1
/
1
0
/
H
m
0
Nonlinearity parameter, Π1/3
h/H0 > 2.5
(b) Inside the surf zone (h/H
0
< 2.5)
1.0
1.2
1.4
1.6
1.8
2.0
2.2
10
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0< 0.009
0.010<H0/L0<0.029
0.030<H0/L0<0.049
Rayleigh
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
,
H
1
/
1
0
/
H
m
0
Heighttodepht ratio, H0/h
h/H0 < 2.5
0.4 1 4
Fig. 4.6. Variation of wave height ratio H
1/10
/H
m0
outside and inside the surf zone.
(a) utside the surf zone ( O h/H
0
> 2.5)
0.75
1.00
1.25
1.50
1.75
0.01 0.1 1 10
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
0.030<H0/L0<0.049
Rayleigh
Semitheoretical
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
,
H
1
/
3
/
H
m
0
Nonlinearity parameter, Π1/3
h/H0 > 2.5
(b) Inside the surf zone (h/H
0
< 2.5)
10 1 4
0.75
1.00
1.25
1.50
1.75
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
0.030<H0/L0<0.049
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
,
H
1
/
3
/
H
m
0
Heighttodepht ratio, H0/h
0.4
h/H0 < 2.5
Fig. 4.7. Variation of wave height ratio H
1/3
/H
m0
outside and inside the surf zone.
Outside the surf zone shown in the left diagrams of Figs. 4.6–4.8, the wave height
ratios H
1/10
/H
m0
, H
1/3
/H
m0
, and H
rms
/H
m0
increase with the increase of the wave
nonlinearity parameter. Although there is much scatter of data, they follow the
trend of semitheoretical curves. The onetenth highest wave height H
1/10
in Fig. 4.6
is given the initial value of 1.20H
m0
based on the trend of the data in consideration
of the spectral width eﬀect.
The maximum values of the ratios of the statistical wave heights to the energy
based signiﬁcant wave height for swell of very low steepness are around 2.1 for
H
1/10
/H
m0
, 1.6 for H
1/3
/H
m0
, and 1.15 for H
rms
/H
m0
. With reference to the
wave height ratios at the weak nonlinearity state, the statistical wave heights are
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
104 Y. Goda
(a) O utside the surf zone (h/H
0
> 1.5)
0.6
0.7
0.8
0.9
1.0
1.1
1.2
0.01 0.1 1 10
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
Rayleigh
Semitheoretical
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
,
H
r
m
s
/
H
m
0
Nonlinearity parameter, Π1/3
h/H0 > 1.5
I (b) nside the surf zone (h/H
0
< 1.5)
0.6
0.7
0.8
0.9
1.0
1.1
1.2
0.001<H0/L0<0.002
0.003<H0/L0<0.004
0.005<H0/L0<0.009
0.010<H0/L0<0.029
Rayleigh
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
,
H
r
m
s
/
H
m
0
Heighttodepht ratio, H0/h
0.4
h/H0 < 1.5
10 1 4
Fig. 4.8. Variation of wave height ratio H
rms
/H
m0
outside and inside the surf zone.
enhanced by 1.75 times for H
1/10
, 1.68 times for H
1/3
, and 1.64 times for H
rms
. Such
enhancement of statistical wave heights are apparent ones without real increase of
wave energy as discussed earlier.
Inside the surf zone shown in the right diagrams of Figs. 4.6–4.8, the data exhibit
large scatter but they indicate a clear tendency of decrease with the heighttodepth
ratio H
0
/h; the wave height ratios H
1/10
/H
m0
, H
1/3
/H
m0
, and H
rms
/H
m0
converge
to 1.27, 1.0, and 0.70, respectively, as H
0
/h becomes larger than 2. As for the eﬀect
of wave steepness, waves with low steepness in the range of 0.001 to 0.0049 have the
mean value of H
1/3
/H
m0
= 1.4 around H
0
/h = 0.4, while waves with the medium
steepness of 0.005 to 0.009 starts from the mean value of H
1/3
/H
m0
= 1.2. Waves
with a high steepness of 0.010 to 0.049 have the H
1/3
/H
m0
value less than 1.1.
Decrease of the wave height ratio with decreasing water depth has been pointed out
by Thompson and Vincent,
36
who suggested earlier start of decrease for waves of
high steepness compared with waves of low steepness.
Such eﬀects of wave steepness on the nonlinear features of waves inside the surf
zone are originated from the fact that waves of low steepness experience a high
degree of nonlinear wave shoaling before they are attenuated by breaking, while
waves of high steepness are attenuated much earlier before they experience strong
nonlinear shoaling. It is seen in the left diagrams of Figs. 4.6–4.8 that waves with the
steepness of 0.010–0.049 have the nonlinearity parameter up to Π
1/3
= 0.3 only and
waves are transferred into the right diagrams which represent waves inside the surf
zone. Waves with the steepness of 0.005 to 0.0099 have the nonlinearity parameter
up to Π
1/3
= 1.2 and moves into the group of those inside the surf zone. Waves with
the steepness below 0.0029 can have the nonlinearity parameter up to Π
1/3
= 4
before they enter into the surf zone.
Thus, the evolution of wave nonlinearity expressed in terms of the skewness
β
1
, kurtosis β
2
, and wave height ratios H
1/10
/H
m0
, H
1/3
/H
m0
, and H
rms
/H
m0
is summarized as follows. It is weak in the oﬀshore, increases gradually as waves
propagate shoreward, becomes strongest in the outer half of the surf zone, begins
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 105
to decrease as breaking process of random waves become intensive, and is nearly
lost at the shoreline.
4.7. Changes of Wave Height Distribution and Probability
Density Function Across Surf Zone
4.7.1. Histogram of wave height distribution
The ﬁrst formulation of random wavebreaking process in a numerical model was
made by Collins
38
by modifying the probability density function (pdf ) of the
Rayleigh distribution. He assumed that all waves beyond the breaking limit will
reduce their heights to the breaker height H
b
, i.e., the pdf is truncated at H = H
b
,
and it has an inﬁnite intensity of delta function there.
An early work of Battjes
39
adopted the same approach, while Kuo and Kuo
40
redistributed the portion of broken waves over the range of 0 < H ≤ H
b
in pro
portion to the remaining probability density. On the other hand, Goda
24
proposed
to make a gradual cut over an upper onethird range of wave height below the
breaker height as explained in Sec. 4.4.2.
Goda’s proposal was based on the ﬁeld data of wave height histograms, which
was analyzed by himself. Figure 4.9 is a replotting of the data of surface wave records
of Rumoi, YamaseDomari, Tomakomai, and Kanazawa Ports listed in Table 4.3.
The data is presented in four groups according to the value of relative water height
H
1/3
/h. The group with H
1/3
/h < 0.4 demonstrates the wave heights being in good
agreement with the Rayleigh distribution. Though not discernible in the ﬁgure,
there is a small number of waves in the class of H/H
mean
= 3.25–3.50, corresponding
to the probability density of the Rayleigh. The next group with 0.4 ≤ H
1/3
/h <
0.5 does not have any wave in the class of H/H
mean
> 3.25, and the histogram
indicates a slight leftward shift. As the relative wave height increases, disappearance
of large individual waves becomes clearer and so is the deviation from the Rayleigh
distribution. However, there is no concentration of wave heights at the upper end
as assumed by Collins
38
and Battjes.
39
Changes of the pdf of wave heights within the surf zone have been observed in
laboratory data such as Goda,
24
Baldock et al.,
41
and Ting.
42,43
Figure 4.10 is an
example of the cumulative distribution of wave heights in the surf zone, which has
been prepared for the data presented in Fig. 4.5 of Ting.
42
The abscissa of Fig. 4.10 is
the relative wave height H/H
1/3
and the ordinate is the nonexceedance probability.
When wave heights follow the Rayleigh distribution, their cumulative distribution
is aligned on the heavy straight line of Fig. 4.10. The data at the relative depth
of h/H
0
= 1.90 deviates only slightly from the straight line of the Rayleigh. As
the relative depth decreases, the data of cumulative distribution shift leftward with
the largest relative height being H
max
/H
1/3
= 1.3 or less. At the relative depth of
h/H
0
= 0.69 and 0.48, however, the cumulative distribution exhibits a tendency of
returning to the Rayleigh with H
max
/H
1/3
= 1.5.
By referring to the original diagram of Fig. 4.5 of Tick,
42
it is observed that
deviation from the Rayleigh distribution is largest around h/H
0
= 1.0. Closer to the
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
106 Y. Goda
0
0.2
0.4
0.6
0.8
1.0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Observation
Rayleigh
P
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y
,
p
(
x
)
Normalized wave height, x = H/Hmean
H1/3/h = 0.400  0.499
17 records with 1398 waves
0
0.2
0.4
0.6
0.8
1.0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Observation
Rayleigh
P
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y
,
p
(
x
)
Normalized wave height, x = H/Hmean
H1/3/h = 0  0.399
52 records with 5111 waves
0
0.2
0.4
0.6
0.8
1.0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Observation
Rayleigh
P
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y
,
p
(
x
)
Normalized wave height, x = H/Hmean
H1/3/h = 0.500  0.599
7 records with 534 waves
0
0.2
0.4
0.6
0.8
1.0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Observation
Rayleigh
P
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y
,
p
(
x
)
Normalized wave height, x = H/Hmean
H1/3/h = 0.600  0.630
3 records with 188 waves
Fig. 4.9. Histograms of ﬁeld wave data in four groups of relative water depth (after Goda
24
).
shore, the range of wave height begins to spread owing to regeneration of waves after
breaking, and the wave heights show a tendency to approach the Rayleigh again.
Some researchers such as Battjes and Groenendijk
44
have tried to apply certain
statistical distributions to the wave heights in the surf zone and to empirically
formulate the parameter values. As demonstrated in Figs. 4.9 and 4.10, however, the
wave height distribution continues to vary throughout the surf zone, and any eﬀort
to describe it with a combination of multiple distribution functions will be futile.
4.7.2. Variation of wave height ratios across the surf zone
Change in the pdf of wave heights also yields variation of the ratios among char
acteristic wave heights. General tendency in the surf zone is that the values of
H
max
/H
1/3
, H
1/20
/H
1/3
, and H
1/10
/H
1/3
decrease as waves enter the surf zone and
take the minimum values around at its middle, while the values of H
rms
/H
1/3
and
H
mean
/H
1/3
increase. However, the wave height ratios come back to the values pre
dicted by the Rayleigh distribution as waves approach the shoreline.
Goda
45
has made a prediction of such variation of wave height ratios by means
of a numerical model based on the Parabolic Equation with a Gradational Breaker
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 107
0
0.5
1.0
1.5
2.0
2.5
0 0.5 1.0 1.5 2.0
N
o
n
e
x
c
e
e
d
a
n
c
e
P
r
o
b
a
b
i
l
i
t
y
,
P
Rayleigh
h/H0' =1.90
h/H0' =1.60
h/H0' =1.28
h/H0' =0.97
h/H0' =0.69
h/H0' =0.48
R
e
d
u
c
e
d
v
a
r
i
a
t
e
,
S
Q
R
T
[

l
n
(
1

P
)
]
Relative wave height, H/H1/3
0.99
0.95
0.90
0.75
0.50
0.25
0.10
0.05
0.01
Ting's Lab. Data
Fig. 4.10. Evolution of cumulative distribution of wave heights in the surf zone based on the data
by Ting.
42
Index for Spectral waves (PEGBIS). It has been demonstrated that the diﬀerence
among characteristic wave heights gradually decreases as the relative water depth
becomes shallow down to around h/H
0
= 1, but the range of wave heights expands
again as the water becomes much shallower toward the shoreline.
Laboratory tests by Ting
42,43
on the wave breaking on a uniform slope of
1/35 show a variation of the wave height ratio H
rms
/H
1/3
, following the general
trend described above. Similar variations of the wave height ratios H
1/10
/H
1/3
and H
rms
/H
1/3
are found in the photopole data at Ajigaura, DUCK85, and
SUPERDUCK. Figure 4.11 presents a compilation of the laboratory and ﬁeld data
concerning the wave height ratio changes. Also included in Fig. 4.11 is the pre
diction of the wave height ratios across the surf zone by means of the PEGBIS model
applied for the Ajigaura beach condition. Although there are large scatters of data,
the ratio H
1/10
/H
1/3
takes the minimum value of about 1.15 around h/H
0
= 1.4
for the ﬁeld data, which indicate slightly larger values than the PEGBIS prediction,
probably owing to the wave nonlinearity eﬀect. The ratio H
rms
/H
1/3
takes the
maximum value of about 0.81 around h/H
0
= 1.0 for both the laboratory and ﬁeld
data. The DUCK85 data exhibit the H
rms
/H
1/3
being closer to 1 than the Ajigaura
data. It might be due to lower wave steepness of the swell at DUCK85.
Such changes of the wave height ratios across the surf zone are the result of the
evolution of the pdf of wave heights discussed in Sec. 4.7.1. Narrowing of the wave
height distribution at the middle of the surf zone provides a larger safety margin
in the design of breakwaters and other structures than the case of applying the
Rayleigh distribution, because the design wave height is speciﬁed with either H
max
(composite breakwaters) or H
1/20
(mound breakwaters) for a prescribed oﬀshore
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
108 Y. Goda
0.50
0.75
1.00
1.25
1.50
0.0 1.0 2.0 3.0 4.0 5.0 6.0
H1/10/H1/3: Ajigaura
H1/10/H1/3: DUCK85
H1/10/H1/3: SUPERDUCK
H1/10/H1/3: PEGBIS
Hrms /H1/3: Ajigaura
Hrms /H1/3: DUCK85
Hrms /H1/3: SUPERDUCK
Hrms /H1/3: Ting's Lab.
Hrms /H1/3: PEGBIS
W
a
v
e
h
e
i
g
h
t
r
a
t
i
o
s
,
H
1
/
1
0
/
H
1
/
3
&
H
r
m
s
/
H
1
/
3
Relative depth, h/H0
Fig. 4.11. Variations of wave height ratios H
1/10
/H
1/3
and H
rms
/H
1/3
by photopole ﬁeld data
and H
rms
/H
1/3
in laboratory tests by Ting together with prediction by PEGBIS model.
signiﬁcant wave height. Therefore, reliable information on the evolution of char
acteristic wave heights, not just H
m0
or H
rms
, needs to be provided by numerical
models of wave transformation with random breaking.
4.8. Incorporation of Wave Breaking Process in Numerical Wave
Transformation Models
For engineering applications, numerical wave transformation models are indis
pensable tools, and there have been many models, proposed by various researchers
and institutions. The models must be capable of dealing with random waves and the
process of individual wave breaking, because the majority of engineering problems
are concerned with the situations within the surf zone.
For the simplest case of random waves on a uniform slope, Goda
46
compared per
formance of seven models listed in Table 4.5, which include those by Battjes,
39
Kuo
and Kuo,
40
Goda,
24,45
Battjes and Janssen,
47
Thornton and Guza,
48
and Larson
and Kraus,
49
though there are many more models for random wave breaking. All
the models introduced the breaker index H
b
/h
b
, but the degree of inclusion of the
parameters of the relative depth h
b
/L
0
and the slope s diﬀers among them. The
inherent variability of breaker heights is taken into account by the models by
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 109
Table 4.5. Main characteristics of random wavebreaking models for comparative study.
Factors aﬀecting breaking wave height
Models
Energy
dissipation
mechanism
PDF of broken
wave H
b
/h h/L
Bed
slope
Breaker
vari
ability
Wave
setup
Waves
spec
trum
Battjes
39
pdf
deformation
Delta function
at breaker
height
Yes Yes No No Yes No
Kuo and
Kuo
40
” Remove and
adjust the
remainder
Yes No No No No No
Goda
24
” ” Yes Yes Yes Yes Yes No
Battjes
and
Janssen
47
Bore model Delta function
at breaker
height
Yes Yes No No Yes No
Thornton
and
Guza
48
” Adjust with
weight
function
Yes No No No No No
Larson and
Kraus
49
Dally
model
Delta function
at breaker
height
Yes No No No Yes No
Goda
45
(PEGBIS)
” Remove and
adjust the
remainder
Yes Yes Yes Yes Partially Yes
Goda
24,45
only. The energy dissipation is dealt directly with either the bore analogy
or the model by Dally et al.
50
or indirectly through modiﬁcation of the pdf of wave
heights.
Figure 4.12 is an example of diﬀerences in the variations of the rootmeansquare
wave height in the nearshore waters predicted by the seven models on the slope of
s = 0.02 (Goda
46
). Diﬀerences are very large beyond expectation and Goda
46
shows
other examples on the slope of s = 1/10, 1/30, and 1/100. Because the generation of
longshore currents is governed by the spatial gradient of the radiation stresses, these
models produce much diﬀerent crossshore proﬁles of longshore current velocities as
demonstrated by Goda.
46
Smith
51
has also demonstrated diﬀerent performances of ﬁve wavebreaking
models with diﬀerent breaker indices against ﬁeld measurement of spectral waves.
Because of the large diﬀerence in the prediction of nearshore wave heights, one
should be careful in selecting a numerical model for engineering applications.
Development of numerical models for timedomain wave transformation has been
tried by many researchers, but random wavebreaking process has not been well
reproduced in these models. One of the exceptions is the Boussinesqtype model
developed by Hirayama et al.
52
and Hirayama and Hiraishi,
53
who employed the
breaking criterion of the vertical pressure gradient by Nadaoka et al.
7
They raised
the threshold gradient from 0 to 0.5 to compensate the insuﬃciency in numerical
accuracy due to the features of weak nonlinearity inherent to the Boussinesq
equation. They have succeeded in reproducing the pdf variation across the surf zone.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
110 Y. Goda
0
0.5
1.0
1.5
2.0
0 1.0 2.0 3.0 4.0 5.0 6.0
Battjes '72
Kuo & Kuo '74
Goda '75
Battjes & Janssen '78
Thornton & Guza: M1
Larson & Kraus '91
Goda '04 spectr
R
M
S
W
a
v
e
H
e
i
g
h
t
,
H
r
m
s
(
m
)
Depth, h (m)
Bottom Slope: s = 0.02
Fig. 4.12. Variations of wave heights H
rms
in the surf zone on planar beaches computed by seven
random wavebreaking models for (H
1/3
)
0
= 2.0 m, T = 8.004 s, and θ
0
= 30
◦
(Goda
46
).
Good reproduction of random wavebreaking process having the features dis
cussed in the present chapter is a key factor for future development of reliable and
practical numerical models for engineering applications.
4.9. Summary
Wavebreaking phenomenon in the nearshore waters is characterized by the fol
lowing features:
(1) The breaker index formula by Goda
13
for regular waves is revised by reducing
the bed slope eﬀect for better agreement with laboratory data.
(2) Depthlimited breaker heights have inherent variability with the coeﬃcient of
variation of about 6% for the slope of 1/100 and 14% for the slope of 1/10,
which increases as the bed slope becomes steep.
(3) Breaker index or the ratio of breaker height to water depth is governed by
both the relative water depth h
b
/L
0
and the bed slope.
(4) Incipient breaker index for the signiﬁcant height of random waves is smaller
than that for regular waves by about 30%, and the signiﬁcant wave height has
an upper limit of about 0.6 times the water depth on gentle slopes.
(5) Wave nonlinearity expressed in terms of skewness, kurtosis, and the wave
ratios H
1/10
/H
m0
, H
1/3
/H
m0
, and H
rms
/H
m0
increases with wave propagation
toward the shore and is most enhanced just outside the surf zone. For swell of
very low steepness, the skewness, kurtosis, and wave height ratio H
1/3
/H
m0
may go up to 2.0, 9.0, and 1.6, respectively, at the wave nonlinearity parameter
around 4.0.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 111
(6) For wind waves and swell of relatively large steepness, random wave breaking
begins at large depths compared with swell of small steepness. Thus, the former
waves do not exhibit conspicuous wave nonlinearity.
(7) Wave nonlinearity is weakened by wave breaking inside the surf zone and is
eventually lost near the shoreline with return to linear wave features.
(8) Probability density function (pdf ) of wave heights gradually deviates from
the Rayleigh as waves propagate toward the shore, with the largest deviation
taking place at the middle of the surf zone. Near the shoreline, the wave height
distribution approaches the Rayleigh again.
(9) Diﬀerences between characteristic wave heights such as H
1/10
, H
1/3
, and H
rms
decrease in the middle of the surf zone, but they return to those by the Rayleigh
near the shoreline. The PEGBIS model can simulate such changes of charac
teristic wave heights.
(10) Various existing models of random wave breaking produce quite diﬀerent
prediction of wave heights in the surf zone. Careful selection of a random
wavebreaking model should be made so that the model will yield reliable
information for engineering applications.
Acknowledgment
The author much appreciates kind cooperation by Dr Steven Hughes and
Dr Shintaro Hotta for providing him with their valuable wave data.
References
1. Y. Goda, How much do we know about wave breaking in the nearshore waters, Proc.
4th Int. Conf. Asian and Paciﬁc Coast (APAC 2007), Nanjin, China, CDRM (2007).
2. H. Yamada and T. Shiotani, On the highest water waves of permanent type, Bull.
Disaster Prevention Res. Inst., Kyoto Univ. 18–2(135), 1–22 (1968).
3. Y. Goda, A synthesis of breaker indices, Trans. Japan Soc. Civil Eng. 2(2), 39–49
(1970).
4. H. Yamada, G. Kimura and J. Okabe, Precise determination of the solitary waves of
extreme height on water of a uniform depth, Rep. Res. Inst. Appl. Mech., Kyushu
Univ. XVI(52), 15–32 (1968).
5. J. McCowan, On the highest waves in water, Phil. Mag. Ser. 5, 36, 351–358 (1894).
6. R. Miche, Mouvements ondulatoires de lamer en profondeur ou d´ecroissante, Annales
de Ponts et Chauss´ees 19, 370–406 (1944).
7. K. Nadaoka, O. Ono and H. Kurihara, Nearcrest pressure gradient of irregular water
waves approaching to break, Proc. Coastal Dynamics ’97 (1997), pp. 255–264.
8. G. M. Kaminsky and N. C. Kraus, Evaluation of depthlimited wave breaking criteria,
Second Int. Symp. Ocean Waves Measurement and Analysis, WAVES 1995 (1995),
pp. 180–193.
9. J. W. Kamphuis, Incipient wave breaking, Coastal Eng. 15, 185–203 (1991).
10. W. Rattanapitikon and T. Shibayama, Veriﬁcation and modiﬁcation of breaker height
formulas, Coastal Eng. J. 42(4), 389–406 (2000).
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
112 Y. Goda
11. W. Rattanapitikon, T. Vivattanasirisak and T. Shibayama, A proposal of new breaker
height formula, Coastal Eng. J. 45(1), 29–48 (2003).
12. Y. Goda, A new method of wave pressure calculation for the design of composite
breakwaters, Rept. Port Harbour Res. Inst. 14(3), 59–106 (1973) (in Japanese).
13. Y. Goda, New wave pressure formulae for composite breakwaters, Proc. 14th Int.
Conf. Coastal Eng., Copenhagen, ASCE (1974) pp. 1702–1720.
14. M. Muttray and H. Oumeraci, Wave transformation on the foreshore of coastal
structure, Coastal Eng. 2000 (Proc. ICCE) (2000) pp. 2178–1191.
15. Y. C. Li, G. H. Dong and B. Teng, Wave breaker indices in ﬁnite water depth, China
Ocean Eng. 5(1), 51–64 (1991).
16. Y. C. Li, Y. Yu, L. F. Cui and G. H. Dong, Experimental study of wave breaking on
gentle slope, China Ocean Eng. 14(1), 59–67 (2000).
17. J. L. Lara, I. J. Losada and P. L.F. Liu, Breaking waves over a mild gravel slope:
Experimental and numerical analysis, J. Geophys. Res. 111(C11019), 1–26 (2006).
18. J. W. Smith and N. C. Kraus, Laboratory study of wavebreaking over bars and arti
ﬁcial reefs, J. Waterway, Port, Coastal, Ocean Eng. ASCE, 117(4), 307–325 (1991).
19. A. Kimura and A. Seyama, Breaking limit of irregular waves on slopes, Proc. 33rd
Japanese Coastal Eng. Conf. (1986), pp. 174–178 (in Japanese).
20. A. Seyama and A. Kimura, The measured properties of irregular wave breaking and
wave height change after breaking on the slope, Proc. 21st Int. Conf. Coastal Eng.
(1988), pp. 419–497.
21. K. P. Black and M. A. Rosenberg, Semiempirical treatment of wave transformation
outside and inside the breaker line, Coastal Eng. 16, 313–345 (1992).
22. D. Kriebel, Breaking waves in intermediatedepths with and without current, Coastal
Eng. 2000 (Proc. ICCE), (2000) pp. 203–215.
23. Y. C. Li, Y. Yu, L. F. Cui and G. H. Dong, Transformation and breaking of irregular
waves on very gentle slope, China Ocean Eng. 14(3), 261–278 (2000).
24. Y. Goda, Irregular wave deformation in the surf zone, Coastal Eng. Jpn, JSCE 18,
13–26 (1975).
25. Y. Goda, Random Seas and Design of Maritime Structures, 2nd Edn. (World
Scientiﬁc, Singapore, 2000).
26. F. C. K. Ting, Laboratory study of wave and turbulence velocities in a broadbanded
irregular wave surf zone, Coastal Eng. 43, 183–208 (2001).
27. F. C. K. Ting, Laboratory study of wave and turbulence characteristics in a narrow
banded irregular breaking waves, Coastal Eng. 46, 291–313 (2002).
28. A. H. Sallenger and R. A. Holman, Wave energy saturation on a natural beach of
variable slope, J. Geophys. Res. 90(C6), 11939–11944 (1985).
29. Y. Goda and K. Nagai, Investigation of the statistical properties of sea waves with ﬁeld
and simulation data, Rept. Port Harbour Res. Inst. 13(1), 3–37 (1974) (in Japanese).
30. Y. Goda, Analysis of wave grouping and spectra of longtravelled swell, Rept. Port
Harbour Res. Inst. 22(1), 3–41 (1983).
31. S. Hotta and M. Mizuguchi, A ﬁeld study of waves in the surf zone, Coastal Eng. Jpn,
JSCE 23, 59–79 (1980).
32. S. Hotta and M. Mizuguchi, Statistical properties of ﬁeld waves in the surf zone, Proc.
33rd Jpn Coastal Eng. Conf. (1986) pp. 154–157 (in Japanese).
33. B. A. Ebersole and S. A. Hughes, DUCK85 photopole experiment, US Army Corps
of Engrs., WES, Misc. Paper, CERC8718, (1987) pp. 1–165.
34. Y. Goda, A uniﬁed nonlinearity parameter of water waves, Rept. Port Harbour Res.
Inst. 22(3), 3–30 (1983).
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
Random Wave Breaking and Nonlinearity Evolution Across the Surf Zone 113
35. Y. Goda and M. Kudaka, On the role of spectral width and shape parameters in
control of individual wave height distribution, Coastal Eng. J. 49(3), 311–335 (2007).
36. E. F. Thompson and C. L. Vincent, Signiﬁcant wave height for shallow water design,
J. Waterway, Port, Coastal, Ocean Eng. ASCE 111(5), 828–841 (1985).
37. M. S. LonguetHiggins, On the distribution of the heights of sea waves: Some eﬀects
of nonlinearity and ﬁnite band eﬀects, J. Geophys. Res. 85(C3), 1519–1523 (1980).
38. J. I. Collins, Probabilities of breaking wave characteristics, Proc. 12th Int. Coastal
Eng. Conf., Washington, DC, ASCE (1970) pp. 399–414.
39. J. A. Battjes, Setup due to irregular wave, Proc. 13th Int. Conf. Coastal Eng., Van
couver, ASCE (1972) pp. 1993–2004.
40. C. T. Kuo and S. T. Kuo, Eﬀect of wave breaking on statistical distribution of wave
heights, Proc. Civil Eng. Ocean, ASCE (1974) pp. 1211–1231.
41. T. E. Baldock, P. Holmes, S. Bunker and P. van Weert, Crossshore hydrodynamics
with an unsaturated surf zone, Coastal Eng. 34(3–4), 173–196 (1998).
42. F. C. K. Ting, Laboratory study of wave and turbulence velocities in a broadbanded
irregular wave surf zone, Coastal Eng. 43, 183–208 (2001).
43. F. C. K. Ting, Laboratory study of wave and turbulence characteristics in a narrow
banded irregular breaking waves, Coastal Eng. 46, 291–313 (2002).
44. J. A. Battjes and H. W. Groenendijk, Wave height distributions on shallow foreshores,
Coastal Eng. 40, 161–182 (2000).
45. Y. Goda, A 2D random wave transformation model with gradational breaker index,
Coastal Eng. J. 46(1), 1–38 (2004).
46. Y. Goda, Examination of the inﬂuence of several factors on longshore current compu
tation with random waves, Coastal Eng. 53, 156–170 (2006).
47. J. A. Battjes and J. P. F. M. Janssen, Energy loss and setup due to breaking of
random waves, Proc. 16th Int. Conf. Coastal Eng., Hamburg, ASCE (1978), pp. 1–19.
48. E. B. Thornton and R. T. Guza, Transformation of wave height distribution,
J. Geophys. Res. 88(C10), 5925–5938 (1983).
49. M. Larson and N. C. Kraus, Numerical model of longshore current for bar and trough
beaches, J. Waterway, Port, Coastal, Ocean Eng., ASCE 117(4), 326–347 (1991).
50. W. R. Dally, R. G. Dean and R. A. Darlymple, Wave height variation across beaches
of arbitrary proﬁle, J. Geophys. Res. 90(C6), 11917–11927 (1985).
51. J. M. Smith, Breaking in a spectral wave model, Ocean Wave Measurement and
Analysis (Proc. WAVES 2001), ASCE (2001) pp. 1022–1031.
52. K. Hirayama, N. Hara and T. Hiraishi, Application of bore model to nonlinear wave
transformation, Proc. 13th Int. Oﬀshore Polar Eng. Conf. (2003), pp. 796–801.
53. K. Hirayama and T. Hiraishi, A Boussinesq model for wave breaking and runup in
a coastal zone, 1D, Fifth Int. Symp. Ocean Waves Measurement Analysis, WAVES
2005, CDROM, No. 151 (2005), pp. 1–10.
July 31, 2009 8:18 9.75in x 6.5in b684ch04 FA
This page intentionally left blank This page intentionally left blank
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Chapter 5
Aeration and Bubbles in the Surf Zone
Nobuhito Mori
Disaster Prevention Research Institute
Kyoto University, Gokasho, Uji, Kyoto 6110011, Japan
mori@oceanwave.jp
Shohachi Kakuno
Osaka City University
33138 Sugimoto, Sumiyoshiku, Osaka 5588585, Japan
hachi@civil.eng.osakacu.ac.jp
Daniel T. Cox
Oregon State University, Corvallis, OR 973314501, USA
dan.cox@oregonstate.edu
This chapter presents a brief summary of aeration in the surf zone, beginning
with a review of air–water characteristics in surf zone waves. Second, measure
ments techniques of the bulk of air and bubbles induced by breaking waves in
the surf zone are described, and third, the bulk of air and bubble characteristics
are summarized based on the in situ and visualization laboratory measurements.
Finally, the gas transfer in the surf zone is described and related to the wave
characteristics.
5.1. Introduction
Ocean surface gravity wave propagation from oﬀshore to shoreline is often regarded
as a single phase ﬂow using potential ﬂow theories or solving the Navier–Stokes
equation. Generally, the single phase ﬂow approach to ocean waves is successful for
simulating wave transformation in the coastal area. However, this assumption breaks
down because waves become steep and start to break owing to the bottom bathy
metric eﬀects in the near shore. The wavebreaking process creates dense bubble
plumes, dissipates energy and generates turbulence in the surf zone. Figure 5.1
shows an example of the crest of wave breaking on a gentle slope. Most of the waves
115
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
116 N. Mori, S. Kakuno and D. T. Cox
Fig. 5.1. Image of breaking wave crest at an initiation of breaking in the surf zone.
can be regarded as a single (liquid) phase ﬂow, and only the nearcrest region is
aerated by the jet attachment to the water surface. An accurate estimation of the
bulk of the void fraction, bubble size, and population distributions in the surf zone
are important for understanding the twophase ﬂow characteristics, solving engi
neering problems, and elucidating environmental mechanisms of the coastal area.
1
For example, Peregrine
2
summarized that the entrapped air of breaking wave
reduces the wave impact pressure owing to its greater compressibility compared
with the singlephase approach. There is also a signiﬁcant diﬀerence between salt
water and fresh water experiments on the wave impact pressure (bubbles generated
by turbulence in saltwater are smaller than fresh water), and the entrapped air
reduced the pressure by approximately 10%.
3
The compressibility owing to air–
water mixture decreases the velocity of sound and is being used to estimate large
scale prototype impacts, since the usual Froude scaling is unlikely to be correct
for engineering problems. Therefore, the connection between airmixture, bubble
distribution, and wavebreakinginduced turbulence is essential to understand gas–
liquid interactions in the surf zone.
The air bubble plume is always present in the surf zone breaking waves. Exper
imental data show that the entrained air is proportional to the energy dissipation.
4
For these breaking waves, entrained air enhances the gasexchange at the water
surface. For example, the dissolved oxygen is associated with the breaking waves in
the coastal waters,
5,6
although the carbon dioxide gas transfer can be regarded as
a function of wind speed.
Understanding air entrainment and bubble distribution for the depthlimited
breaking of surf zone waves is limited in comparison with the windwave breaking.
The air bubble distributions under the windwave breaking in the upper layer of the
ocean surface has been summarized in detail by Thorpe.
7
The windwave bubble
formation and its dependence on the turbulent dissipation rate of ﬂuids have been
discussed,
8
and it is proposed a −2 powerlaw scaling with bubble diameter. The
−10/3 powerlaw scaling is alternatively proposed based on the discussion of bubble
fragmentation due to strong turbulent shear ﬂow. These bubble size scaling laws are
still under development.
9
However, the knowledge of windwave breaking induced
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Aeration and Bubbles in the Surf Zone 117
bubbles and associated gas transfer has become wellknown issue over the last
decade.
10
Despite the fruitful knowledge of air entrainment of windwave breaking, few
studies of air entrainment for surf zone wave breaking exist. It has been found that
there is little diﬀerence in the bubble populations beneath the mechanically gen
erated surface waves in saltwater and fresh water.
11
On the other hand, it has been
reported that the void fraction depends on the turbulent intensity in the bore region
of surf zone waves.
12
The largescale bubbles entrained by breaking waves in the
surf zone are split by strong local turbulent shear at the bubble scale.
13
However,
neither qualitative nor quantitative bubble characteristics in the surf zone and con
nections between bubble characteristics and wave breaking are well known owing
to the lack of detailed observations. This detailed information of twophase ﬂow
characteristics is required for mathematical modeling to solve both basic scientiﬁc
interests and applied engineering problems.
5.2. Measurements Techniques
There are several measurement methods for bubble population and size distribu
tions, i.e., in situ measurements,
14,15
video or photographic measurements,
16
laser
measurements,
17
and acoustic measurements.
18
The bubble size measurements using
lasers probe such as a phase Doppler anemometer show high accuracy in smallscale
bubble measurements smaller than 500 µm but is of limited use in the presence
of high void fractions and largescale bubble measurements due to fundamental
instrument limitations. Acoustic measurements of bubbles are useful in deepwater
but have limitation for very shallow water region owing to the multireﬂection of
sound beams. Therefore, conventional optical or resistivity type void probes are
useful for the surf zone breaking waves. The electrical conductivity probe (socalled
void probe) has been generally used for the high void fractions.
19
This type of void
probe is robust and easy to handle for both wave ﬂume experiments and ﬁeld obser
vations. Therefore, the void probe measurements have been widely used for the two
phase ﬂow measurement in the surf zone.
12,14,20
However, the in situ measurement
has clear limitation for inhomogeneous surf zone breaking phenomena.
The photographic or imaging technique has been advanced to in situ measure
ments which can cover the spatial and temporal variations of the twophase ﬂow
characteristics in the surf zone. Recent photographic studies have illustrated the dis
integration of entrapped air cavities divided into bubbles.
13,21
Several imaging tech
niques have been proposed to measure the bubble characteristics in the surf zone.
One is an application of Particle Image Velocimetry (PIV) method with forward and
backward lighting.
22,23
This method basically correlates the scattered light intensity
and bubble plume intensity, and it can be used for high void fraction region over 20%,
although it cannot measure the bubble shape itself. The other imaging technique is a
bubble shape tracking method with backlighting.
24
This method is a socalled shadow
graph method because the bubble shape is enhanced by a backlight source toward a
camera. The shadowgraphmethodcanmeasure bothbubble shape andvelocity simul
taneously. However, the density of the bubbles has to be within a certain limit to sep
arate individual bubbles fromthe image by the shadowgraph method.
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
118 N. Mori, S. Kakuno and D. T. Cox
In summary, there is no ideal method that can be used for universal purpose of
bubble measurements and one can choose appropriate method(s) for a particular
purpose.
5.3. Bulk Aeration Characteristics in the Surf Zone
Figure 5.2 shows successive images of breaking wave taken by highspeed video
camera in the surf zone. Figure 5.2(a) is just after attachment of jet on the water
surface which corresponds to the initial stage of bubble generation. Figures 5.2(b)
and 5.2(c) are 42 ms and 108 ms after image (a), respectively. The lifetime of the
large volume of entrained air can be categorized into two phases. The ﬁrst phase
is the initial stage of breaking waves [Figs. 5.2(a) and 5.2(b)]. The newly injected
air column produces pluses of sound and, therefore, this phase is regarded as the
acoustically active phase.
21
In the active phase, the trapped volume of air between
the overturning jet and water surface has fragmented and injected into the deeper
region, and the air pocket is fragmented into small size bubbles. This phase occurs
within 1 s of the initial stage of wave breaking. The second phase is the acoustically
silent phase. Most of the bubbles are advected by the surface roller of the carrier
waves with buoyancy, and some of them are trapped by oblique descending eddy
in this phase [Figs. 5.2(c)]. In the active phase, the void fraction ratio is relatively
high and there is no clear formation of bubbles in that region. Therefore, the only
possible approach to this phase is to account for the total amount of air, which is
the summation of the air pocket and bubbles. The silent phase has a lower void
fraction and is more dispersive, and it is possible to discriminate individual bubbles.
These injected air penetrate into a deeper region during the active phase. After
that, the volume of air and/or bubbles are advected by the wave motion due to their
buoyancy. Therefore, the gas phase characteristics in the surf zone, such as the void
fraction, are time–space variables and are inhomogeneous in both time and space.
The next part of this chapter discusses the total amount of air, bubble size
distributions, and gastransfer in the surf zone, respectively.
5.3.1. Temporal characteristics
The aeration process by breaking wave in the surf zone is obviously timedependent.
Therefore, the temporal transition is important in understanding the phenomena in
the beginning. Figure 5.3 shows an example of phaseaveraged time series of surface
elevation and turbulence components of the mechanically generated periodic wave
breaking on a slope.
20
The notations z and H
b
denote the vertical axis positive
upward from the mean water level and wave height at the breaking point, respec
tively. The initial wave height and period are 16.3 cm and 1.6 s, respectively. Using
the phaseaveraged method, the velocity data were divided into the mean, periodic,
and turbulence components. The notation of σ
2
u
, σ
2
v
, and σ
2
w
are the turbulent
stresses for the velocity components (u, v, w). The peak of the turbulent stress is
located at the front of the crest and decreases linearly as time passes. This is a
typical time series of turbulence characteristics of depthlimited breaking waves.
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Aeration and Bubbles in the Surf Zone 119
(a)
(b)
(c)
Fig. 5.2. Formation of bubble plume under the breaking wave in the surf zone. (a) Attachment
of jet on the surface (initiation of bubble plume generation); (b) Active phase of bubble plume
generation [42 ms after (a)]; (c) Bubble plume in silent phase [108 ms after (a)].
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
120 N. Mori, S. Kakuno and D. T. Cox
5
0
5
10
0
500
1000
1500
0
500
1000
1500
0
500
1000
1500
z/H =0.033
z/H =0.013
z/H =0.0
<
>
[
c
m
]
u 2
,
v 2
,
w 2
[
c
m
σ
α
η
σ
σ
2
/
s
2
]
(a)
(b)
(c)
(d)
t/T
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.1
0.2
<
>
z/H =0.033
z/H =0.013
z/H =0.0
(e)
b
b
b
b
b
b
Fig. 5.3. Example of phaseaveraged surface elevation η and turbulence components 1.0 m from
B.P. in Case 1. (a) Free surface η (solid) and standard deviation η±σ
η
; (b)–(d) temporal variation
of σ
u
(thick solid), σ
v
(thin solid), and σ
w
(dotted) and (e) the phaseaveraged void fraction α.
20
The highest void fraction occurs at the wave front, and the void fraction is decreased
rapidly in time. The temporal variation of the void fraction depends on the air
supply from the surface, advection, and buoyancy, and these eﬀects are diﬃcult
to discriminate from each other. Although the process is complex, the normalized
temporal characteristics show a universal behavior. Similar laboratory experiments
were observed by several researchers.
12,25,26
Furthermore, Cox and Shin
12
found the
selfsimilar exponential decay of the temporal series of void fraction normalized by
the averaged value in diﬀerent types of breaking waves. In addition, they reported
that the maximum void fraction in the time series was three to four times higher
than its average value.
5.3.2. Vertical distribution of void fraction
The vertical distribution of void fraction of the surf zone breaking waves was ﬁrst
systematically measured by Hoque and Aoki.
27
They showed a simple description
of vertical void fraction distribution as follows. If a void fraction α simply follows a
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Aeration and Bubbles in the Surf Zone 121
8
6
4
2
0
0 0.1 0.2
C0
z (cm)
xxb = 0.4 m
12
10
8
6
4
2
0
0 0.1 0.2
C0
xxb = 0.6 m
z (cm)
12
10
8
6
4
2
0
0 0.1 0.2
C0
z (cm)
xxb = 0.8 m
12
10
8
6
4
2
0
0 0.1 0.2
C0
z (cm)
xxb = 1.1 m
12
10
8
6
4
2
0
0 0.1 0.2
C0
z (cm)
xxb 1.35 m
2
Fig. 5.4. Vertical distribution of void fraction for plunging breaker.
27
diﬀusion equation in steady state, we have
∂(αu)
∂z
=
∂
∂z
D
∂α
∂z
, (5.1)
where z is the vertical axis, u is the water velocity, and D is the turbulent diﬀusivity
of air bubble plume. Assuming a uniform velocity distribution and D independent
in the vertical direction, we have
∂
2
α
∂z
2
=
u
0
D
∂α
∂z
, (5.2)
where u
0
denotes the uniform velocity. The solution of the above equation is
α(z) = α
0
exp(kz), (5.3)
where k = u
0
/D is the newly introduced decay parameter of the vertical void
fraction distribution and α
0
denotes the void fraction at the mean water level
(z = 0). Figure 5.4 shows the comparison between experimental data and Eq. (5.3)
for plunging breakers on a gentle slope. The experimental data shows an expo
nential decay of the void fraction in the vertical direction. Furthermore, introducing
k
0
= kh, where h is the local depth, k
0
can be regarded as a constant value ranging
approximately from 3.75 to 4.0, independent of breaker type. Similar results were
measured by a visualization technique
23
after that. In addition, the linear decay
of the maximum water depth of penetrated bubble shoreward is also observed as
shown in Fig. 5.4.
27
5.3.3. Horizontal distribution of void fraction
The characteristics of the volume of air changes both horizontally, vertically and
temporally. The void fraction and total kinetic energy show similar distributions,
and Cox and Shin
12
also reported the dependence of void fraction on turbulent
intensity in the bore region of surf zone waves. The entrapped air induced by wave
breaking generates strong turbulence near the free surface. Therefore, the rela
tionship between void fraction and turbulence should be connected to each other.
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
122 N. Mori, S. Kakuno and D. T. Cox
0
5
10
15
0
5
10
15
0
5
10
15
0 2 4 6 8 10 12 14
0
5
10
15
×10
2
x
b
/H
b
α
×
1
0
4
,
k
[
c
m
2
/
s
2
]
z/H
b
=0.033
z/H
b
=0.013
z/H
b
=0.0
z/H
b
=0.013
⋅ 10
2
⋅ 10
2
⋅ 10
2
Fig. 5.5. Horizontal relationship between timeaveraged void fraction α (solid line) and time
averaged TKE k (dotted line).
20
Figure 5.5 shows experimental results of void fraction and total kinetic energy
(TKE) variations from the breaking point x
b
to the shore line.
20
The spatial varia
tions of the timeaveraged void fraction α and timeaveraged TKE, k show similar
tendency, although the peak of α is located approximately 0.2 × H
b
∼ 2 × H
b
seaward of the peak of k, as shown in Fig. 5.5. The short spatial lag between α and
k might be associated with the spatial diﬀerence of entrapped air and associated
turbulence. The spatial variations of the turbulence and void fraction are highly cor
related near the mean water level. The entrained bulk of air and bubbles generate
turbulence, and turbulence splits the coarse bubbles into ﬁne bubbles in this area.
Moreover, the TKE and void fraction increased by an order of magnitude from the
trough level (z/H
b
= −0.013) to above the mean water level (z/H
b
= 0.033); thus,
the twophase ﬂow characteristics are remarkably strong near the wave crest.
Neglecting the short spatial lag between the void fraction and TKE, the
direct comparison between the timeaveraged void fraction α and normalized time
averaged TKE k of all the measuring locations is shown in Fig. 5.6. For small
values of α or k, the relationship between the two shows a high correlation and
becomes scattered at high void fraction or TKE region. Overall, the relationship
between timeaveraged void fraction and timeaveraged TKE shows high corre
lation
12
as does the incident wavedependence on the relationship between α and k.
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Aeration and Bubbles in the Surf Zone 123
0 0.1 0.2 0.3 0.4
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
( k/gh )
1/2
α
Case1
Case3
Case2
Fig. 5.6. Relationship between timeaveraged TKE k and timeaveraged void fraction α. (•: Case
1, +: Case 2, : Case 3).
The relation between the void fraction and TKE, the scaledependence of void
fraction, and further theoretical considerations will be required to ﬁnd the universal
relationship between the turbulence characteristics of depthlimited breaking waves
and air entrainment.
5.4. Bubble Characteristics in the Surf Zone
Bubble generation is a smallscale process of aeration in the surf zone.
28
There is
a large body of literature dealing with the distribution of bubbles in windwave
breaking.
7,8,29,30
From the knowledge of windwave breaking, there are two modes
of bubble generation.
31
One is entrapment of air under the falling water jet by the
wave overturning and another is air entrainment at the point where the falling water
jet meets the vortex formed at the rear of the splashup.
Figure 5.7 shows the time series of bubble diameter distribution just beneath
the mean water level near the breaking point under the same wave condition as
in Fig. 5.5.
24
The bubble diameter here means the averaged value of the major
and minor diameter. The horizontal axis is the characteristic bubble diameter, the
vertical axis is the temporal change, and the contour indicates number of bubbles
per unit area. The plunging wave breaking not only entrains the bulk of air but also
generates a large number of bubbles in comparison with the spilling breaker. As
discussed earlier, the void fraction decreases exponentially in time. The time series
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
124 N. Mori, S. Kakuno and D. T. Cox
Fig. 5.7. Time series of bubble size distribution.
24
(a) Spilling breaker; (b) Plunging breaker.
of the bubble number density distribution (Fig. 5.7) shows a similar tendency, but
the smallscale bubbles survive longer than largescale bubbles.
Generally, entrapped air by breaking waves is split into small bubbles owing
to turbulence shear of surrounding ﬂows. The early work of this problem for a
general condition is discussed by Kolmogorov
32
and Hinze
33
in the middle of the
last century, and a large bibliography has been generated.
34
LonguetHiggins
35
pro
posed a crushing air cavity model for the bubble distributions after the formation
of a bubble plume and shows good agreement of bubble aeration in general con
dition, qualitatively. Figure 5.8 shows the instantaneous and timeaveraged bubble
size spectra with the estimated power law measured by an imaging technique.
24
The instantaneous bubble size spectra corresponds to the time of the maximum
void fraction observed. There is no signiﬁcant diﬀerence of bubble size distribution
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Aeration and Bubbles in the Surf Zone 125
10
−1
10
0
10
1
10
−4
10
−2
10
0
10
2
diameter [mm]
N
u
m
b
e
r
o
f
b
u
b
b
l
e
s
/
c
m
2
timeaveraged
instantaneous
d
−1.0
d
−3.4
d
−0.9
Fig. 5.8. Power law of bubble size distribution.
between instantaneous and timeaveraged spectra at small scale (diameter < 1 mm).
On the other hand, two scaling laws can be seen in the bubble size distribution. The
small and large size bubble power scaling laws are d
−1
and d
−3–−4
, respectively.
Garrett et al.
9
proposed semiempirical −10/3 powerlaw scaling for the bubble
which is larger than the Hinze scale, based on the discussion of bubble fragmentation
in the strong turbulence ﬂow below the trough level. On the other hand, Deane and
Stokes
21
empirically proposed a −3/2 powerlaw scaling, smaller than the Hinze
scale in the acoustically active phase near the crest. The node point between the
two power scaling corresponds to the Hinze scale of the bubble splitting theory.
33
The measured powerlaw scalings of the small size bubble shown in Fig. 5.8 are close
to −3/2 and −10/3 power laws, respectively. These powerlaw scalings can be seen
in diﬀerent conditions in mechanically generated waves in the ﬂume.
13,20,21
The Hinze scale d
H
21
of bubble splitting can be given by
d
H
= 2
−8/5
c
ρ
w
γ
ε
−2/5
, (5.4)
where ρ
w
is the water density, γ is the surface tension coeﬃcient, ε is turbulence
energy dissipation at the bubble diameter, and c = (We
c
)
3/5
, where We
c
is the
critical Weber number. The Weber number We is deﬁned as
We =
ρ
w
γ
δu
2
d, (5.5)
where δu is turbulence velocity at the bubble scale d. The turbulence velocity δu at
the bubble scale in the inertial range can be given by δu = 2ε
2/3
d
2/3
. If the energy
dissipation rate ε can be estimated by wave energy dissipation, the Hinze scale can
be obtained. For example, if ε = 12 W/kg, the Hinze scale is 2 mm in diameter.
These theoretical predictions agree fairly well with the experimental data,
20,21
although the Hinze scale is formulated under the assumption of isotropic and
homogeneous turbulence ﬂow.
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
126 N. Mori, S. Kakuno and D. T. Cox
5.5. Gas Exchange in the Surf Zone
For moderately solvable gases such as carbon dioxide and oxygen, the gas transfer
across the air–sea interface is controlled by the aqueous condition, and is parame
terized as a function of a gas transfer velocity k
L
k
L
= αSc
−1/2
u
∗
, (5.6)
where Sc = νκ is the Schmidt number, α is the empirical coeﬃcient, and u
∗
is
the friction velocity, respectively.
36
The role of bubbles in the exchange of gas in
the surf zone is poorly understood in comparison to the windwave breaking. An
analogy of windwave breaking to surf zone breaking waves is similar, but surface
renewal process and the bubblemediated gas transfer are diﬀerent from each other.
For example, the surface renewal by the wind stress is important for the wind
wave scenario, but the turbulence generated by the wave overturning is dominant
in the surf zone. Moreover, the bubblemediated gas transfer process depends on
the bubble population and size distribution in the surf zone. For the windwave
case, the largescale bubbles with a diameter greater than approximately 1 mm play
an important role in air–sea gas exchange.
37,38
However, there are a few conﬁdent
prediction models that exist for macro or microscopic bubblemediated gas transfer
for surf zone breaking waves.
The oxygen gas transfer is important for ﬁsh and marine mammals near the
coastal region. Therefore, oxygen gas transfer, rather than carbon dioxide, plays a
vital role in aquatic environment in the surf zone.
39
There are signiﬁcant diﬀerence
between the windwaves and surf zone waves generally. For example, the spatial
homogeneity can be assumed in windwaves, but there is a strong spatial dependence
for surf zone waves. Therefore, a diﬀerent approach to predict the gas transfer is
required. One may assume that the gas transfer is related to the total water mass
volume of surf zone (Fig. 5.9). The general expression of the gas transfer in an
aeration tank of constant volume can be formulated:
∂C
∂t
=
A
V
k
L
∆C = k
2
∆C, (5.7)
where C is the gas concentration, ∆C is the concentration diﬀerence between liquid
and gas phase, A is the aeration (surface) area, k
L
is the gas transfer velocity,
Fig. 5.9. Illustration of gas transfer in the surf zone.
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Aeration and Bubbles in the Surf Zone 127
and k
2
is the reaeration coeﬃcient. Eckenfelder
40
proposed the bubblemediated gas
transfer velocity k
L
for aeration tank as
k
L
=
αh
2/3
Q
a
Ad
m
Sc
1/2
, (5.8)
where α is the empirical coeﬃcient, h is the water depth of aeration tank, Q
a
is
the total amount of air ﬂow rate, d
m
is mean bubble diameter. Multiplying the surf
zone area A by Eq. (5.8), it can be formulated for aeration by surf zone breaking
waves as
41
k
L
A =
αH
2/3
b
V
g
d
m
S
1/2
c
T
, (5.9)
where H
b
is the wave height at breaking point, V
g
is the total entrained air volume
per wave period, and T is the wave period (Fig. 5.10). In this formulation, the gas
transfer in the surf zone linearly depends on the total entrained air volume. Now,
we introduce the gas transfer ﬂow rate k
L
A instead of the gas transfer velocity k
L
in Eq. (5.9). This is because the aeration area A is diﬃcult to be estimated in the
surf zone, but the total water volume V is easily obtained. Assuming a uniform
bottom slope, Kakuno
41
empirically obtained a relationship between V
g
and surf
zone parameter as V
g
∝ H
b
h
b
/ tanθ, where h
b
is the water depth at the breaking
point, and θ is the bottom slope. Substituting this expression into Eq. (5.9), we
have the gas transfer ﬂow rate coeﬃcient k
L
A for slope beach
41
:
k
L
A =
α
2
H
5/3
b
h
b
Td
m
S
1/2
c
tan θ
. (5.10)
Kakuno compared Eq. (5.10) with the fresh water experimental data and found
that the empirical coeﬃcient α
2
is approximately 2.3 ×10
−3
∼ 2.8 ×10
−3
as shown
in Fig. 5.11.
41
Equation (5.10) expresses the gas transfer ﬂow rate as a function
of breaking wave height H
b
, water depth at breaking point h
b
, wave period T,
mean bubble diameter d
m
, bottom slope θ, and Schmidt number, although salinity
and scaling eﬀects are neglected. Furthermore, Kakuno
41
obtained an empirical
relation between the gas transfer velocity and wave energy dissipation as k
L
∝ ε
1/3
.
wave3 wave2 wave1
X
Ti
X
b
V
g
Fig. 5.10. Total bubble volume distribution in the surf zone (X
b
: length of surf zone, V
g
: bubble
volume per wave period).
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
128 N. Mori, S. Kakuno and D. T. Cox
Fig. 5.11. Gas transfer ﬂow rate and bottom bathymetry in the surf zone (•: 1/10 slope, ◦: 1/20
slope, : 1/30 slope).
Research to estimate the gas transfer velocity k
L
from the wave characteristics
through the wave energy dissipation is ongoing.
5.6. Conclusions and Future Study
In this chapter, we summarize our general review of gas–water twophase ﬂow char
acteristics in the surf zone. There are several measurement methods for the aeration
in the surf zone, such as in situ measurements, video or photographic measure
ments, laser measurements, and acoustic measurements. There is no one appropriate
method to measure the volume of air and bubbles in the surf zone. The spatial and
temporal characteristics of void fraction are summarized, and the bubble size dis
tribution is described with simple theories. Furthermore, the gas transfer prediction
method is presented as a function of the bottom slope and wave characteristics for
the surf zone waves.
There are many other problems such as salinity eﬀect, scale eﬀects, and more
mathematical modeling should be investigated in the future.
References
1. M. Donelan, W. Drennan, E. Saltzman and R. Wanninkhof, Gas Transfer at Water
Surfaces (AGU, Washington, DC, 2002).
2. D. Peregrine, Water wave impact on walls, Ann. Rev. Fluid Mechanics 35, 23–43
(2003).
3. B. G. N. A. Crawford, P. Hewson and P. Bird, Experiments on fullscale wave impact
pressures, Coastal Eng. 8, 331–346 (1999).
4. S. Longo, M. Petti and I. J. Losada, Turbulence in the swash and surf zones: A review,
Coastal Eng. 45, 129–147 (2002).
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
Aeration and Bubbles in the Surf Zone 129
5. D. Farmer, C. McNeil and B. Johnson, Evidence for the importance of bubbles in
increasing airsea gas ﬂux, Nature 361, 620–623 (1993).
6. D. Wallace and C. Wirick, Large airsea ﬂuxes associated with breaking waves, Nature
356, 145–147 (1992).
7. S. Thorpe, On the clouds of bubbles formed by breaking wind waves in deep water,
and their role in airsea gas transfer, Phil. Trans. R. Soc. London A 304, 155–210
(1982).
8. S. Baldy, A generationdispersion model of ambient and transient bubbles in the close
vicinity of breaking waves, J. Geophys. Res. C10, 18277–18293 (1993).
9. C. Garrett, M. Li and D. Farmer, The connection between bubble size spectra and
energy dissipation rates in the upper ocean, J. Phys. Oceanogr. 30, 2163–2171 (2000).
10. B. J¨ahne and H. Haußecker, Airwater gas exchange, Ann. Rev. Fluid Mechanics 30,
443–468 (1998).
11. M. Loewen, M. O’Dor and M. Skafel, Bubbles entrained by mechanically generated
breaking waves, J. Geophys. Res. 101, 20759–20769 (1996).
12. D. Cox and S. Shin, Laboratory measurements of void fraction and turbulence in the
bore region of surf zone waves, J. Eng. Mechanics, ASCE 129(10), 1197–1205 (2003).
13. G. B. Deane and M. D. Stokes, Air entrainment processes and bubble size distributions
in the surf zone, J. Phys. Oceanogr. 29, 1393–1403 (1999).
14. H. Chanson, S. Aoki and M. Maruyama, Unsteady air bubble entrainment and
detrainment at plunging breaker: Dominant time scales and similarity of water level
variation, Coastal Eng. 46, 139–157 (2002).
15. H. Chanson, Comments on ﬁber optic refectometer for velocity and fraction ratio
measurements in multiphase ﬂows, Rev. Sci. Instrum. 75(1), 284–285 (2003).
16. I. Leifer and G. de Leeuw, Bubble measurements in breakingwave generated bubble
plumes during the LUMINY windwave experiment, in Gas Transfer at Water Sur
faces, eds. M. Donelan, W. Drennan, E. Saltzman and R. Wanninkhof (AGU,
Washington, DC, 2002), pp. 303–310.
17. N. Mori, Experimental study of air bubble distributions induced by wind wave
breaking, in Proc. Asia Paciﬁc Coast 2003, JSCE, Makuhari (2003), CD–ROM.
18. G. B. Deane, Sound generation and air entrainment by breaking wave in the surf zone,
J. Acoustic Soc. Am. 102, 2671–2689 (1997).
19. S. Vagle and D. Farmer, A comparison of four methods of bubble measurement, IEEE
Oceanic Eng. 23(3), 211–222 (1998).
20. N. Mori, T. Suzuki and S. Kakuno, Experimental study of air bubbles and tur
bulence characteristics in the surf zone, J. Geophys. Res. 112(C05014) (2007),
doi: 10.1029/2006JC003647.
21. G. B. Deane and M. D. Stokes, Scale dependence of bubble creation mechanisms in
breaking waves, Nature 418, 839–844 (2002).
22. Y. Ryu, K.A. Chang and H. J. Lim, User of bubble image velocimetry for mea
surement of plunging wave impinging on structure and associated greenwater, Meas.
Sci. Technol. 16, 1945–1953 (2002).
23. O. Kimmoun and H. Branger, A particle image velocimetry investigation on laboratory
surfzone breaking waves over a sloping beach, J. Fluid Mech. 588, 353–397 (2007),
doi: 10.1017/S0022112007007641.
24. N. Mori and S. Kakuno, Aeration and bubble measurements of the coastal breaking
waves, Fluid Dynamics Res. 40(7–8), 616–626 (2008).
25. F. Ting and J. Kirby, Observation of undertow and turbulence in a laboratory surf
zone, Coastal Eng. 24, 51–80 (1994).
July 31, 2009 8:18 9.75in x 6.5in b684ch05 FA
130 N. Mori, S. Kakuno and D. T. Cox
26. D. Cox, N. Kobayashi and A. Okayasu, Vertical variations of ﬂuid velocities and shear
stress in surf zones, Proc. 24th Int. Conf. on Coastal Eng., ASCE (1994), pp. 98–112.
27. A. Hoque and S. Aoki, Distributions of void fraction under breaking waves in the surf
zone, Ocean Eng. 32, 1829–1840 (2005).
28. M. Loewen, Inside whitecaps, Nature 418, 830 (2002).
29. R. Cipriano and D. Blanchard, Bubble and aerosol spectra produced by a laboratory
breaking waves, J. Geophys. Res. C86, 8085–8092 (1981).
30. C. Garrett, M. Li and D. Farmer, The connection between bubble size and energy
dissipation rates in the upper ocean, J. Phys. Oceanogr. 30, 2163–2171 (2000).
31. P. Bonmarin, Geometric properties of deepwater breaking waves, J. Fluid Mech. 209,
405–433 (1989).
32. A. Kolmogorov, On the breakage of drops in a turbulent ﬂow, Dokl. Akad. Navk. SSSR
66, 825–828 (1949).
33. J. Hinze, Fundamentals of the hydrodynamic mechanism of splitting in dispersion
processes, Am. Inst. Chem. Eng. J. 1, 289–295 (1955).
34. C. Mart´ınezBaz´an, J. Montan´es and J. Lasheras, On the breakup of air bubble
injected into a fully developed turbulent ﬂow. Part I: Breakup frequency, J. Fluid
Mech. 401, 157–182 (1999).
35. M. LonguetHiggins, The crushing of air cavities in a liquid, Proc. Roy. Soc. London A
439, 611–626 (1992).
36. W. Melville, The role of surfacewave breaking in airsea interaction, Ann. Rev. Fluid
Mechanics 28, 279–321 (1996).
37. L. Memery and L. Marlivat, Modeling of gas ﬂux through bubbles at the airwater
interface, Tellus 37B, 272–285 (1985).
38. R. Keeling, On the role of large bubbles in airsea gas exchange and supersaturation
in the ocean, J. Marine Res. 51, 237–271 (1993).
39. C. Moutzouris and E. Daniil, Water oxygenation in the vicinity of coastal struc
tures due to wave breaking, Proc. 24th Int. Conf. Coastal Eng., ASCE (1994),
pp. 3167–3177.
40. W. Eckenfelder Jr., Absorption of oxygen from air bubbles in water, J. Sanitary Eng.
Div. Proc. ASCE, 89–99 (1959).
41. S. Kakuno, D. B. Moog, T. Tatekawa, K. Takemura and T. Yamagishi, The eﬀect of
bubble on airwater oxygen transfer in the breaker zone, in Gas Transfer at Water
Surfaces, eds. M. Donelan, W. Drennan, E. Saltzman and R. Wanninkhof (AGU,
Washington, DC, 2002), pp. 265–277.
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Chapter 6
Freak Wave
Nobuhito Mori
Disaster Prevention Research Institute
Kyoto University, Gokasho, Uji, Kyoto 6110011, Japan
mori@oceanwave.jp
In this chapter, we describe mechanisms that lead to the occurrence of freak waves
in the ocean. It is now generally recognized that freak waves can be generated
by any one of four possible mechanisms. The traditional mechanism is a simple
linear superposition of waves, the theory of which is reviewed here. The newest
mechanism attempts to include the thirdorder nonlinearities that depart from
linear wave theory. Therefore, we present here a stateoftheart review that is
based on nonlinear wave dynamics.
6.1. Introduction
The history contains many reports of encounters with monstrous waves.
1
For
example, the captain of the cargo vessel Junior reported a wave estimated to be
100 ft high and the famous reliable report was that of the wave encountered by the
U.S.S. Ramapo in the North Paciﬁc in 1993; that wave was estimated to be 112 ft
high. There are many more reports of encounter of similar waves in the history of
the seas. Over the last decade, more evidence has accumulated for such extreme
waves. This may be because of an increase in the number of measurement locations
together with the development of more accurate ﬁeld measurements, for example,
several extreme wave events, the socalled freak waves, recorded in the North Sea.
2
A sample wave proﬁle is shown in Fig. 6.1, which was extracted from the data pro
vided by wave radar shooting downward from the platform to the sea surface in the
North Sea. The data were collected over 12–15 hours at a depth of 40 m. Since the
early 1990s such an event has come to be called a freak wave or a rogue wave. More
evidence has been obtained by analyzing ﬁeld data from the North Sea,
3,4
the Sea
of Japan,
5,6
and the Gulf of Mexico.
7
The ﬁrst clear deﬁnition of a freak wave was proposed by Klinting and Sand
2
;
in its original form, the deﬁnition required a wave to satisfy three conditions to be
131
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
132 N. Mori
Fig. 6.1. Freak wave recorded on the 24th November, 1981 at the Gorm Field in the Danish
Sector of the North Sea.
8
considered “freak wave”:
(1) It has a wave height higher than twice the signiﬁcant wave height.
(2) Its wave height is larger than two times of the foregoing and the following wave
heights.
(3) Its wave crest height is larger than 65% of of its wave height.
Subsequently, these conditions were found to be overly stringent, and today the ﬁrst
condition is generally considered to be suﬃcient to identify a freak wave.
8
Freak wave research has two basic objectives. The ﬁrst is an understanding of
the mechanisms that create freak waves; several mechanisms have been proposed
to explain why extremewave events occur in the ocean. The second is to establish
a reliable statistical model for the occurrence of freak waves. However, the problem
is diﬃcult because a freak wave is only one wave, or perhaps just a few waves, in
a wave train. Therefore, its occurrence exhibits statistical sensitivity. Moreover, an
extreme wave event is transient: it forms and disappears quickly in both time and
space. To address these problems, we should study wave populations that contain
freak waves, rather than concentrate on the features of an individual wave proﬁle.
(This is not to say, however, that observations of single events are unimportant.)
Once such wave populations can be characterized, then it should be possible to
estimate how often waves of any given size will occur.
Over the last decades, researchers in coastal and ocean engineering, and the
oceanography have dedicated much eﬀorts to the study of extreme wave events.
Detail on the state of the art are available in the conference proceedings of Rogue
Waves 2004.
9
In this chapter, we provide an overview of freak waves. First, we
review the four mechanisms that are now thought to generate freak waves in Sec. 6.2.
Next, we describe the probability distribution for the populations of linear waves in
Sec. 6.3. Third, we show how that probability distribution changes when nonlinear
eﬀects are taken into account in Sec. 6.4. Fourth, we discuss the accuracy with
which extreme wave events can be predicted in Sec. 6.5. Finally, we make some brief
suggestions about future work.
6.2. Mechanisms for Generating Freak Waves
It is now accepted that at least four mechanisms can explain the formation of freak
waves. A problem common to these four is whether a freak wave obeys an ordinary
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 133
statistical law or whether other physical mechanisms come into play. That is, if we
are to be able to predict freak waves, then any underlying mechanism must lead to
reliable statistical model for extreme wave events.
First Mechanism. The ﬁrst mechanism is simply that a freak wave can be caused
by a linear superposition of waves. In this case, the probability distribution for
wave height in the limit of the narrowband approximation obeys a Rayleigh dis
tribution
10
; corrections due to ﬁnite spectral band width have been obtained.
11–13
Wave crest statistics can be obtained by using a secondorder theory.
14
The proba
bility distribution for wave crests has been found in a narrowband approximation
15
and for ﬁnite band width.
16
This mechanism is theoretically well established and
relatively easy to follow; the basic theory will be reviewed in Sec. 6.3.
Second mechanism. The second mechanism is that freak wave can result when
waves interact with currents. When this situation occurs, ray theory can be used
to explain the formation of extreme waves. However, it is not yet known how the
statistical properties of surface are related to the properties of currents. From the
studies of freak waves generated over a random current, it has been found that wave
focusing with 1 m/s velocity ﬂuctuations occurs in the uniform current.
17
Third Mechanism. A third mechanism has its basis in crossingsea states (i.e.,
twosea systems). An example is a swell and a wind sea in which the two have
diﬀerent directions coexisting over some region of the ocean, or diﬀraction of waves
behind an island. Usually, both the wavecurrent mechanism and the twosea state
mechanism occur in speciﬁc locations, such as along the Angus current; therefore,
freak waves caused by either of these mechanisms are relatively predictable.
Fourth Mechanism. In the fourth mechanism, extreme wave events result from
the highorder nonlinear interactions; speciﬁcally, the underlying mechanism is four
wave interactions.
18,19
Numerical and experimental studies have demonstrated that
freaklike waves can be generated frequently in a twodimensional wave ﬂume,
even in the absence of a current, or wave refraction, or wave diﬀraction
18,20–22
;
see Fig. 6.2. Further, numerical studies clearly indicate that a freak wave having
a single, steep crest can be generated in deepwater by the thirdorder nonlinear
interactions.
18
Freak wave generation is sometimes discussed in the context of the Benjamin–
Feir instability in deepwater waves because of the similarity in the steep wave
Fig. 6.2. Freak wave simulated by the highorder spectrum method.
18
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
134 N. Mori
proﬁle.
18,22
Over the last two decades, the Benjamin–Feir instability of deep
water gravity waves has been studied using the nonlinear Schr¨ odingertype
equations,
23–25
modecoupling equations,
26
pseudospectral methods,
27
and experi
ments.
28
However, the Benjamin–Feir instability vanishes when the wave spectrum
is suﬃciently broad. This means, there is a discrepancy between the nonlinear
behaviors of periodic waves and random waves. Janssen
19
investigated freak waves
that occur as a consequence of fourwave interactions, including the eﬀects of non
resonant fourwave interactions. Using Monte Carlo simulations of the Zakharov
equation, he found that homogeneous nonlinear interactions give rise to deviations
from the Gaussian distribution for the surface elevation.
The fourth mechanism for freak wave generation requires additional knowledge
before predictions can be made. Relationship between kurtosis and freak wave events
has been studied using laboratory data,
29
numerical results,
30
and ﬁeld data.
7
North
Sea data shows that the maximum crest, normalized by its wave heights, correlates
with skewness, while the wave heights, normalized by signiﬁcant wave height, cor
relate with the kurtosis.
7
This is because the kurtosis is the fourth moment of the
probability density function; hence, it measures the relevance of tails in a distri
bution.
These results explain the general characteristics of a freak wave. (a) A second
order nonlinearity causes wave asymmetry over the shorttime evolution, O(1/ε
2
),
of a wave train (Here ε is a small parameter that corresponds to the steepness of
deepwater waves). (b) But the kurtosis is associated with the thirdorder nonlinear
interactions.
14
Therefore, it may be that thirdorder nonlinear interactions are the
fundamental factors contributing to the occurrence of freak waves. However, the
timescales for such interactions are longer than those for secondorder interactions,
O(1/ε
4
), and they are two orders of magnitude longer than those for secondorder
nonlinearities.
The formal relation between the kurtosis and the exceedance probability for wave
height have been discussed (Mori and Janssen
31
; see also Tayfun and Lo
32
; Mori
and Yasuda
33
). The kurtosis enters into the distribution function as a correction to
the Rayleigh distribution: when the kurtosis tends to three, the expected Gaussian
values, and then the distribution function tends to the Rayleigh distribution. In this
context, changes in the kurtosis can be evaluated once the evolution of the wave
spectrum is known.
19
The wave height exceedance probability has been derived as a
function of the number of waves.
31
This relation will be explained in a later part in
this chapter. In all these studies, an ultimate goal is the determination of the shape
of the exceedance probability function P(H). Once an analytical form for P(H) is
found for all sea conditions (for example, as a function of the wave spectrum), then
the probability of a wave exceeding a certain height can be estimated directly.
One diﬃculty in studying extreme wave events is the veriﬁcation of theory.
Statistical theories are evaluated under the assumptions of stationarity; however,
the real seas change states both spatially and temporally. Therefore, statistical
theories are diﬃcult to verify.
34
In addition, freak wave prediction requires us to
predict the maximum wave height distribution. However, although study of the
maximum wave height distribution is necessary to verify a theory for freak wave
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 135
45000 45100 45200 45300 45400 45500 45600
−4
−2
0
2
4
6
8
Time (s)
S
u
r
f
a
c
e
E
l
e
v
a
t
i
o
n
(
m
)
Yura Y88121401 Surface Waves
10 20 30 40 50 60 70 80 90 100
0.1
0.2
0.3
0.4
0.5
0.6
F
r
e
q
u
e
n
c
y
(
H
z
)
Wavelet Spectrum of Yura Y88121401 Surface Waves
79400 79500 79600 79700 79800 79900 80000
−5
0
5
Time (s)
S
u
r
f
a
c
e
E
l
e
v
a
t
i
o
n
(
m
)
Yura Y88010901 Surface Waves
10 20 30 40 50 60 70 80
0.1
0.2
0.3
0.4
0.5
0.6
F
r
e
q
u
e
n
c
y
(
H
z
)
Wavelet Spectrum of Yura Y88010901 Surface Waves
Fig. 6.3. Freak wave recorded in 1988 in the Sea of Japan.
34
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
136 N. Mori
prediction, veriﬁcation of wave height distribution is not suﬃcient to completely
verify a theory.
3,31
Samples of freak wave proﬁles and their wavelet spectrum, recorded in the Sea
of Japan, are shown in Fig. 6.3.
34
The wave proﬁle shown in Fig. 6.3(a) is a 10 min
timeseries segment that contains a freak wave along with the corresponding contour
plot for the wavelet spectrum. It appears that for the welldeﬁned freak wave shown
in Fig. 6.3(a), the occurrence of the freak wave can be readily identiﬁed from the
wavelet spectrum: the strong energy density in the spectrum instantly increases
at the onset of the freak wave, at the same time, that increase in energy appears
to carry over into high frequency components. However, for another freak wave
time series shown in Fig. 6.3(b), no corresponding focused surge in energy appears
in the wavelet spectrum. So in this case, it is uncertain whether a freak wave is
begun, generated by linear or nonlinear wave focusing. Nevertheless, at the onset
of the event in Fig. 6.3(a), the freak wave proﬁle appears rather asymmetric with
respect to the mean level, whereas the freak wave proﬁle in Fig. 6.3(b) is generally
symmetric. So, these diﬀerences in wavelet spectra might result from diﬀerences in
the processes that generated these freak waves. In any case, time series alone may
or may not be reliable guides for identifying freak wave events. These observations
suggest that linear and nonlinear focusing eﬀects are coupled (that is, linear and
nonlinear processes do not belong to totally diﬀerent populations), and this coupling
must be understood if we are to be able to predict freak waves.
6.3. Linear Wave
The surface gravity waves’ celerity depends on the wave length owing to dispersive
eﬀects in deepwater. The nonlinear eﬀects slightly modify the dispersion relation
in the order of wave steepness.
35
The order of nonlinear eﬀects, both dispersion
and wave proﬁle, are square or cubic of the wave steepness if there is no resonant
interactions in the system. Therefore, in general, the ocean waves can be regarded
as a linear, ergodic, homogeneous, stationary, Gaussian processes. Because of these
assumptions, the probability density function for water surface elevation of unidi
rectional waves can be expressed as the Gaussian distribution
p(η) =
1
√
2πm
0
exp
−
η
2
m
0
, (6.1)
σ
2
= η
2
rms
= m
0
=
∞
0
S(ω)dω , (6.2)
and η is the surface elevation while σ =
√
m
0
is its variance. Let η(t) be the sea
surface elevation at time t and let ζ(t) be an auxiliary variable such that η(t) and
ζ(t) are not correlated. Assuming both η(t) and ζ(t) are real zeromean functions
with variance σ, we have
Z(t) = η(t) +iζ(t) = A(t)e
iφ(t)
, (6.3)
A(t) =
η
2
(t) +ζ
2
(t) , (6.4)
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 137
φ(t) = tan
−1
ζ(t)
η(t)
, (6.5)
where A is the envelope of the wave train and φ is the phase. As there is no corre
lation between η and ζ, the joint probability density function of η and ζ becomes
p(η, ζ) =
1
2π
exp
¸
−
1
2
(η
2
+ζ
2
)
. (6.6)
All variables will be normalized by the variance of the surface elevation σ =
√
m
0
.
The probability density function (PDF) for the envelope A can now be obtained
by integrating the joint probability distribution
p(A, φ) = Ap(η, ζ) , (6.7)
over φ; hence,
p(A) =
2π
0
dφ p(A, φ). (6.8)
The results are the usual Rayleigh distributions for wave amplitude
p(A) = Ae
−(1/2)A
2
(6.9)
and wave height
p(H) =
1
4
He
−(1/8)H
2
. (6.10)
In the narrowband approximation, the wave height H = 2A. The exceedance prob
ability P
H
(H) for wave height now follows by integrating (6.10) from H to ∞:
P
H
(H) = e
−(1/8)H
2
. (6.11)
In an exact, linear narrowband random wave theory, H
1/3
= 4.004
√
m
0
, but for
simplicity we will assume H
1/3
= 4
√
m
0
. In this notation, the freak wave condition
H/H
1/3
> 2 corresponds to H/
√
m
0
> 8. According to Eq. (6.11), the probability
of occurrence for a freak wave in a wave train is P(8) = e
−8
= 3.35 ×10
−4
. This is
equivalent to a freak wave occurring once in every 2981 waves. Assuming a charac
teristic wave period of 10 s, this corresponds to a storm having a duration of 8.3 h.
When this linear wave theory is extended to 3D space, the appearance probability
for fully directional conditions (S
max
= 10) increases by 45% over that for the uni
directional condition.
36
The maximum wave height distribution becomes important when we examine
statistical properties of freak waves over a series of recorded data. This is because
a freak wave prediction is equivalent to estimating the maximum wave height and
the exceedance probability (that is, Eq. (6.11) is insuﬃcient for prediction). The
PDF for the maximum wave height p
m
in wave trains can be obtained once the
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
138 N. Mori
PDF of wave height p(H) and exceedance probability of the wave height P(H) are
known.
37
Thus,
p
m
(H
max
)dH
max
= N[1 −P(H
max
)]
N−1
p(H
max
)dH
max
, (6.12)
where N is the number of waves in the wave train. For suﬃciently large N, we may
use the approximation
N
lim
N→∞
[1 −P(H
max
)]
N
N
lim
N→∞
exp [−NP(H
max
)] . (6.13)
Substituting Eqs. (6.11) and (6.13) into Eq. (6.12), gives the PDF for the maximum
wave height p
m
,
p
m
(H
max
)dH
max
=
N
4
H
max
e
−H
2
max
/8
exp
−Ne
−H
2
max
/8
¸
dH
max
, (6.14)
and the exceedance probability of maximum wave height P
m
,
P
m
(H
max
) = 1 −exp
−Ne
−H
2
max
/8
¸
. (6.15)
Note that Eqs. (6.14) and (6.15) are evaluated as functions of N number of waves
in the train. The freak wave condition in this study becomes H
max
/
√
m
0
> 8, and
we obtain from Eq. (6.15) the following simple formula for probability of a freak
wave occurring as a function of N,
P
freak
= 1 −exp
−e
−8
N
. (6.16)
This is a general expression for the expected probability of freak waves in unidi
rectional linear random waves. Therefore, in the linear wave case, the occurrence
probability for a freak wave depends only on the number of waves N. For example,
if the number of waves equals 2981, then Eq. (6.16) gives P
freak
= 0.632, that is, the
expected probability of a freak wave is 63.2% when we randomly sample from a pop
ulation of 2981 waves. For N = 120 waves, the expected probability of a freak wave
is 3.9% and for N = 10,000 waves, it is 96.5%. These results show that the number
of waves in a wave train (or, equivalently, the duration of a storm) is important for
predicting freak waves.
6.4. Nonlinear Interactions
6.4.1. Role of nonlinear interactions
Since the beginning of freak wave research, highorder nonlinear eﬀects have been
considered.
18,30
As a result, the nonlinear occurrence of freak waves, including
their mechanisms and detailed dynamic properties, are becoming better under
stood.
21,22,34,38–40
The state of the art has been summarized at the Rogue Wave
Conference, held in 2000 and 2004.
9,41
The consensus from those conferences is that
thirdorder nonlinear interactions not only enhance the appearance of freak wave,
but they are also the primary cause of freak waves. The only exceptions occur during
strong wave–current interactions or wave diﬀraction behind an island.
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 139
6.4.2. Quasiresonant fourwave interactions
and highorder moments
Investigations have been carried out on the creation of freak waves in response
to fourwave interactions, including the eﬀects of nonresonant fourwave interac
tions.
19,31
Those studies culminated in the formulation of a probability for the
occurrence of freak waves in nonlinear wave ﬁelds.
However, the modulational instability is a quasiresonant interaction process,
i.e., wave numbers and frequencies satisfy the following conditions:
k
1
+k
2
−k
3
−k
4
= 0 , (6.17)
ω(k
1
) +ω(k
2
) −ω(k
3
) −ω(k
4
) ≤
2
, (6.18)
where is a small parameter that corresponds to the steepness of deepwater waves.
The modulational instability takes place when two wave numbers are the same,
k
1
= k
2
, while k
3
and k
4
are side bands separated from k
1
by ∆k, which should be
small in order to satisfy the conditions Eqs. (6.17) and (6.18). The standard kinetic
equation that describes the evolution of the wave spectrum in time
42
is formally
valid only for exact resonances and large times, O(
−4
ω
−1
0
). It has been extended
to include quasiresonant interactions,
19
where a kinetic equation, which should
also be valid on the timescale of the modulational instability, has been derived. If
we then consider the evolution of higher order moments, such as the kurtosis µ
4
,
we ﬁnd that the quasiresonant interactions are responsible for deviations from the
Gaussian distribution. The explicit relation between nonlinear interactions of free
waves and the fourthorder moment of the surface elevation η(x, t) in deepwater
has been investigated.
19
The result can be expressed in terms of κ
40
, which is the
fourthorder cumulant of the surface elevation (κ
40
= µ
4
−3),
κ
40
=
η
4
m
2
0
−3
=
12
g
2
m
2
0
dk
1,2,3,4
T
1,2,3,4
√
ω
1
ω
2
ω
3
ω
4
×δ
1+2−3−4
R
r
(∆ω, t)N
1
N
2
N
3
. (6.19)
Here,
√
m
0
is the variance of the surface elevation η, g is the acceleration of gravity, k
is the wave number, ω is the angular frequency, N is the wave action spectral density,
and T
1,2,3,4
is the coupling coeﬃcient in the Zakharov equation (see Krasitskii
43
for
its analytical form), δ
1+2−3−4
= δ(k
1
+ k
2
− k
3
− k
4
), dk
1,2,3,4
= dk
1
dk
2
dk
3
dk
4
.
The resonant function is
R
r
=
1 − cos(∆ωt)
∆ω
, (6.20)
with
∆ω = ω
1
+ω
2
−ω
3
−ω
4
. (6.21)
In the limit of large times, R
r
→P/∆ω, where P denotes the principle value of the
integral to avoid singularity in the integral.
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
140 N. Mori
In the narrowband approximation, we assume that the spectrum E(ω) has a
Gaussian shape,
E(ω) =
m
0
σ
ω
√
2π
e
−(1/2)ν
2
, (6.22)
where ν = (ω −ω
0
)/σ
ω
and σ
ω
is the spectral band width; the integral in Eq. (6.19)
for large times becomes
κ
40
=
24
2
∆
2
P
dν
1,2,3
(2π)
3/2
e
−(1/2)[ν
2
1
+ν
2
2
+ν
2
3
]
(ν
1
+ν
2
−ν
3
)
2
−ν
2
1
−ν
2
2
+ν
2
3
. (6.23)
Here, = k
0
√
m
0
is the steepness parameter and ∆ = σ
ω
/ω
0
is the relative spectral
band width. The integral can be evaluated analytically to obtain
κ
40
=
π
√
3
BFI
2
, (6.24)
where BFI is deﬁned as by Janssen
19
:
BFI =
∆
√
2 . (6.25)
Equation (6.24) is the simpliﬁed predictive equation for the kurtosis of the surface
elevation assuming a narrowband, unidirectional wave train, but a full description
requires evaluation of sixdimensional integral in wave number space Eq. (6.19).
6.4.3. Wave height distributions in a nonlinear wave ﬁeld
To include nonlinear eﬀects in the wave height distribution function, thereby pos
sibly obtaining deviations from Rayleigh statistics, the standard approach is to use
the Edgeworth series.
44
The theory for wave height has been described,
31–33
and
the resulting distribution has been named the Modiﬁed Edgeworth Rayleigh (MER)
distribution. The MER wave height and exceedance wave height distributions are
given by Mori and Janssen
31
:
p(H)dH =
1
4
He
−(1/8)H
2
[1 +κ
40
A
H
(H)] dH , (6.26)
P
H
(H) = e
−(1/8)H
2
[1 +κ
40
B
H
(H)] . (6.27)
Here, H is the wave height normalized by η
rms
=
√
m
0
, κ
40
is deﬁned in Eq. (6.23),
while A
H
(H) and B
H
(H) are polynomials deﬁned as
A
H
(H) =
1
384
(H
4
−32H
2
+ 128) , (6.28)
B
H
(H) =
1
384
H
2
(H
2
−16) . (6.29)
Note that the distributions describe the deviations from linear statistics under the
hypothesis of a narrowband, weakly nonlinear wave train. Further, in the case of
linear waves (κ
40
= 0), these distributions are equivalent to Eqs. (6.10) and (6.11).
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 141
Secondorder contributions, which are important for the distribution of wave crests,
can be included by using an approach like that used by Tayfun.
15,45
The probability distribution function, p
m
, and the exceedance probability, P
m
,
of the maximum wave height can also be expressed as a function of the fourth
cumulant of the surface elevation κ
40
and the number of waves N in the wave train,
p
m
(H
max
) =
N
4
H
max
e
−H
2
max
/8
[1 +κ
40
A
H
(H
max
)]
×exp
−Ne
−H
2
max
/8
[1 +κ
40
B
H
(H
max
)]
¸
, (6.30)
P
m
(H
max
) = 1 −exp
−Ne
−H
2
max
/8
[1 +κ
40
B
H
(H
max
)]
¸
, (6.31)
where H
max
is the maximum wave height normalized by η
rms
. Qualitative agreement
has been found between the theoretical wave height distribution, given by Eq. (6.30),
and ﬁeld data.
31
Equations (6.30) and (6.31) are evaluated as functions of N and
κ
40
(or µ
4
). When κ
40
= 0, Eqs. (6.30) and (6.31) reduce to Eqs. (6.14) and (6.15),
which are provided by the Rayleigh distribution.
The occurrence probability of a freak wave as a function of N and κ
40
, taking
into account nonlinear eﬀects, now becomes
P
freak
= 1 − exp
−e
−8
N(1 + 8κ
40
)
. (6.32)
Using Eq. (6.31) we see that the eﬀect of kurtosis becomes of the same order as the
linear theory for κ
40
= 1/8. This corresponds to µ
4
= 3.125, which is not a strong
nonlinear condition. Hence, both the eﬀects of ﬁnite kurtosis and the number of
waves N are important for determining the probability of maximum wave height in
the nonlinear wave train. It should be emphasized here that the above discussion
does not require any empirical, ad hoc parameters. A diﬀerent approach, also based
on the dynamics of a cubic nonlinearity, has been used for ﬁnding the exceedance
probability for wave height.
46
That approach involves an optimization problem,
requiring that at a speciﬁc space and time, the Fourier phases are all equal, under
the constraints that the conservation laws of the Zakharov equation are satisﬁed.
To summarize, quasiresonant fourwave interactions introduce deviations from
linear statistics for surface elevation; in particular, for weakly nonlinear, narrow
banded and longcrested wave trains, the kurtosis evolves according to Eq. (6.19).
In the narrowband approximation, the kurtosis is related to the BFI [deﬁned in
Eq. (6.25)]; the tail of the wave height distribution depends on the kurtosis/BFI and
it increases as the kurtosis increases. Finally, the maximum wave height distribution
depends on the number of waves in the wave train (record length) and on the
kurtosis of the nonlinear wave ﬁeld, as shown in Eq. (6.30).
6.5. Accuracy of Freak Wave Prediction
Figure 6.4 compares the linear (Rayleigh) theory with those from the present theory
for the occurrence probability of a freak wave, P
freak
; the plot shows how P
freak
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
142 N. Mori
10
2
10
3
10
4
10
−2
10
−1
10
0
N
P
f
r
e
a
k
Rayleigh
µ
4
= 3.1
µ
4
= 3.2
µ
4
= 3.3
µ
4
= 3.4
µ
4
= 3.5
Fig. 6.4. Occurrence probability for freak waves as a function of number of waves N and
kurtosis µ
4
.
31
responds to changes in the number of waves N and in the strength of nonlinearities,
as measured by the kurtosis µ
4
. For N = 100 waves, the occurrence probability of a
freak wave is predicted to be 3.3% by the linear theory, while it is 15.4% according
to Eq. (6.30) with µ
4
= 3.5. For N = 1000 waves, the occurrence probability of
a freak wave is 28.5% according to linear theory, while it is 81.3% according to
Eq. (6.31) with µ
4
= 3.5. The number of waves N = 1000 corresponds to a duration
of about 3 hours, assuming T
1/3
= 10 s, which is not unusual in stormy conditions.
Alternatively, deﬁning the threshold value for the occurrence probability of a freak
wave as 50%, the expected number of waves that would include at least one freak
wave is predicted to be 2000 waves from linear theory, but only 500 waves when
Eq. (6.31) is used with µ
4
= 3.5. Thus, in a strong nonlinear wave ﬁeld, freak waves
can occur several times more frequently than in a linear ﬁeld.
Figure 6.5 shows freak wave occurrence probability as a ratio R
freak
of the
value predicted by the present approach to that predicted by the Rayleigh theory
(κ
40
= 0); the ratio depends on the kurtosis µ
4
,
R
nonlinear
=
P
freak
P
freak

κ
40
=0
−1 . (6.33)
For a small number of waves N ≤ 250, the ratio R
freak
depends linearly on µ
4
. If
µ
4
= 3.1 and N ≤ 500, the occurrence probability of freak waves is 50% more from
the nonlinear theory than from the linear theory. However, the rate of increase in
R
freak
decreases as the number of waves increases. This is because, for very large
number of waves, the maximum wave height almost always exceeds 2H
1/3
, even in
linear wave trains.
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 143
3 3.1 3.2 3.3 3.4 3.5
0
50
100
150
200
250
300
350
µ
4
R
f
r
e
a
k
[
%
]
N=100
N=250
N=500
N=1000
N=5000
Fig. 6.5. Ratio of freak wave occurrence predicted by Eq. (6.31) to the Rayleigh theory.
31
The consequences of the maximum wave height distribution [Eq. (6.30)] are
generally hard to verify, not only because Eq. (6.30) depends on the number of
waves but also because large amount of data are required to characterize nonlinear
eﬀects. In the following discussion, we compare the above theories with a set of long
timeseries experimental data
47
; these data appear to be suﬃciently long to make
useful comparisons.
Figure 6.6 shows the probability of occurrence of freak waves H
max
/η
rms
≥ 8
for short time records, as a function of kurtosis. We assumed 100 waves per record
and counted the number of freak waves. In the experimental data, the occurrence
probability for freak waves shows a clear dependence on the kurtosis (this eﬀect of
course is not described by the Rayleigh distribution). But the experimental data
appear to depend quadratically on the kurtosis, while the nonlinear theory Eq. (6.31)
predicts a linear dependence. This is probably because the nonlinear theory includes
only the lowestorder correction for nonlinearity: it excludes highorder cumulants.
Figure 6.7 compares the maximum wave height distribution from the experi
mental data with linear, Eq. (6.14), and nonlinear theories, Eq. (6.30), for N = 150.
In Fig. 6.7(a), the comparison between theory and experiment is shown for a nearby
linear wave condition (µ
4
= 3.06). The peak in the observed maximum wave height
distribution is larger than the Rayleigh H
max
distribution and the MER H
max
dis
tribution with µ
4
= 3.06, but the observed distribution is not as wide (we ascribe
this diﬀerence to the eﬀects of the ﬁnite width of the spectrum). The maximum
wave height distribution from the laboratory data departs from the Rayleigh dis
tribution [Figs. 6.7(a)–6.7(c)] at some distance from the wavemaker, and the peak
of the observed distribution shifts to larger wave heights. While the Rayleigh H
max
distribution, Eq. (6.14), is independent of the kurtosis, the MER H
max
distribution,
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
144 N. Mori
3 3.5 4 4.5
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
µ
4
P
r
o
b
a
b
i
l
i
t
y
o
f
F
r
e
a
k
W
a
v
e
[
%
]
N=11900
Exp.
MER
Rayleigh
Fig. 6.6. Occurrence probability for freak waves in long recorded wave data: N = 11, 900
(•: Experimental data, solid line: MER distribution from Eq. (6.31), dashed line: Rayleigh
distribution).
Eq. (6.30), follows the behavior of the experimental data. The MER H
max
distri
bution seems to reproduce the experimental data reasonably well; for very small
kurtosis, it overestimates the experimental data, and for large kurtosis it slightly
underestimates the data. This is consistent with what we have seen in Fig. 6.6 where
the occurrence probability of freak waves is plotted as a function of the kurtosis.
Similar results have been obtained in other studies.
31
We also consider the general behavior of the probability density function of
maximum wave height in a nonlinear wave ﬁeld. To do so, we plot in Fig. 6.8
the expected value of the maximum wave height, indicated by angle brackets,
as a function of µ
4
and N. In the ﬁgure, we show the numerically integrated
value of Eq. (6.30) together with the ensemble average of the experimental data.
The quantity H
max
/H
1/3
in Fig. 6.8 from the Rayleigh theory corresponds to
H/
√
m
0
> 8. The dependence of H
max
/H
1/3
on µ
4
and N is clear in both the
experimental data and Eq. (6.30). Overall, H
max
/H
1/3
from the experimental data
is smaller than that from the MER H
max
distribution; nevertheless, the theoretical
and experimental data show similar trends.
The above evidence shows that freak waves can be predicted for given strengths
of nonlinearities, as measured by the kurtosis. The remaining problems, therefore,
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 145
4 5 6 7 8 9 10 11 12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
H
max
/η
rms
P
(
H
m
a
x
/
η
r
m
s
)
Exp.
MER
Rayleigh
(a) µ
4
=3.06
4 5 6 7 8 9 10 11 12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
H
max
/η
rms
P
(
H
m
a
x
/
η
r
m
s
)
Exp.
MER
Rayleigh
(b) µ
4
=3.43
Exp.
MER
Rayleigh
4 5 6 7 8 9 10 11 12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
H
max
/η
rms
P
(
H
m
a
x
/
η
r
m
s
)
(c) µ
4
=4.10
Fig. 6.7. Comparison of maximum wave height distribution H
max
/η
rms
with N = 150 (•: exper
imental data, solid line: MER H
max
distribution, dashed line: Rayleigh H
max
distribution).
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
146 N. Mori
3
3.2
3.4
3.6
3.8
4
200
400
600
800
1000
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
µ
4
N
<
H
m
a
x
/
H
1
/
3
>
3
3.2
3.4
3.6
3.8
4
200
400
600
800
1000
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
µ
4
N
<
H
m
a
x
/
H
1
/
3
>
Fig. 6.8. Comparison of expected value of H
max
/H
1/3
.
48
are developing accurate methods for estimating the kurtosis and learning how to
take directional eﬀects into account for the application in the ocean.
6.6. Conclusion and Future Study
This chapter summarized generally accepted mechanisms that cause freak waves
and we have described linear and nonlinear mechanisms in more detail. Fourwave
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 147
quasiresonant interactions have been shown to play an important role in deter
mining the statistical properties of surface elevation. This discovery, which is fairly
new, is of some relevance to extreme wave forecasting. Previously, deviations from
Gaussian behavior have been attributed only to bound modes. However, even in
the most severe sea states, the kurtosis obtained from secondorder theory rarely
reaches values above 3.15. Moreover, contributions to wave height distribution from
secondorder theory are practically negligible (in the narrowband approximation
that are exactly zero). If waves are longcrested and suﬃciently steep, the dynamics
of the free waves can cause very strong departures from Gaussian behavior. Large
values of the kurtosis can substantially change the tails of the probability density
function for wave height and can increase the occurrence of freak waves.
Although nonlinear eﬀects have been found to be important, some issues remain
unresolved. First, we need an evaluation of the accuracy with which the kurtosis is
estimated from spectra. BFI originally describes the asymptotic behavior, given an
initial value, but in reality, there is no initial condition in the ocean. This should
be veriﬁed in the near future. Second, we need to account for wave directionality.
In real sea states, directional eﬀects are important. Directional contributions to the
evolution of the kurtosis, using a modiﬁed nonlinear Schr¨ odinger equation, have
been described,
49
and the eﬀects of fourwave interactions only appear close to long
crested waves conditions.
50
A quantitative discussion of directional eﬀects should
be presented in the near future.
References
1. L. Draper, “Freak” ocean waves, Marine Observer 35, 193–195 (1965).
2. P. Klinting and S. Sand, Analysis of prototype freak waves, in Coastal Hydrodynamics,
ed. R. Darlymple (ASCE, 1987), pp. 618–632.
3. P. Stansell, J. Wolfram and S. Zachary, Horizontal asymmetry and steepness distri
butions for winddriven ocean waves, Appl. Ocean Res. 25, 137–155 (2003).
4. G. Guedes Soares, Z. Cherneva and E. Ant¨ao, Characteristics of abnormal waves in
North Sea storm sea states, Appl. Ocean Res. 25, 337–344 (2003).
5. T. Yasuda and N. Mori, Occurrence properties of giant freak waves the sea area around
Japan, J. Waterway, Port, Coastal Ocean Eng. 123(4), 209–213 (1997).
6. T. Yasuda, N. Mori and S. Nakayama, Characteristics of giant freak waves observed
in the Sea of Japan, in Waves97 (ASCE, 1997), pp. 482–495.
7. G. Guedes Soares, Z. Cherneva and E. Antao, Abnormal waves during hurricane
Camille, J. Geophys. Res. Ocean 109, C08008 (2004).
8. S. Sand, N. O. Hansen, P. Klinting, O. Gudmestad and M. Sterndorﬀ, Freak Wave
Kinematics (Kluwer Academic Pub., 1990), pp. 535–549.
9. M. Olagnon (ed.), Rogue Waves 2004 (IFREMER, 2004).
10. M. LonguetHiggins, On the statistical distribution of the heights of sea waves,
J. Marine Res. 11, 245–266 (1952).
11. A. Næss, On the distribution of cresttotrough wave heights, Ocean Eng. 12(3),
221–234 (1985).
12. P. Boccotti, On Mechanics of Irregular Gravity Waves, Chap. VIII (Atti Accademia
Nazionale dei Lincei, Memorie, 1989), pp. 111–170.
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
148 N. Mori
13. M. A. Tayfun, Distribution of cresttotrough wave heights, J. Waterway, Port,
Coastal Ocean Eng. 107(3), 149–158 (1981).
14. M. LonguetHiggins, The eﬀect of nonlinearities on statistical distirbutions in the
theory of sea waves, J. Fluid Mechanics 17, 459–480 (1963).
15. M. A. Tayfun, Narrow band nonlinear sea waves, J. Geophys. Res. Ocean 85(C3),
1548–1552 (1980).
16. F. Fedele and F. Arena, Weakly nonlinear statistics of high nonlinear random waves,
Phys. Fluids 17, 026601 (2005).
17. B. White and B. Fornberg, On the chance of freak wave at sea, J. Fluid Mechanics
355, 113–138 (1998).
18. T. Yasuda, N. Mori and K. Ito, Freak waves in a unidirectional wave train and
their kinematics, Proc. 23rd Int. Conf. Coastal Eng., Vo. 1, ASCE, Venice (1992),
pp. 751–764.
19. P. Janssen, Nonlinear fourwave interactions and freak waves, J. Phys. Oceanogr.
33(4), 863–884 (2003).
20. C. Stansberg, Extreme waves in laboratory generated irregular wave trains, in
Water Wave Kinematics, eds. A. Tørum and O. Gudmestad (Kluwer Academic Pub.,
1990), pp. 573–590.
21. K. Trulsen and K. Dysthe, Freak waves — A threedimensional wave simulation, Proc.
21st Symp. Naval Hydrodynamics (National Academy Press, 1997), pp. 550–558.
22. M. Onorato, A. Osborne, M. Serio and S. Bertone, Freak waves in random oceanic
sea states, Phys. Rev. Lett. 86(25), 5831–5834 (2001).
23. H. C. Yuen and B. M. Lake, Nonlinear dynamics of deepwater gravity waves, Adv.
Appl. Mech. 22, 67–229 (1982).
24. E. Caponi, P. Saﬀman and H. Yuen, Instability and conﬁned chaos in a nonlinear
dispersive wave system, Phys. Fluids 25(12), 2159–2166 (1982).
25. K. Dysthe, Note on modiﬁcation to the nonlinear Schr¨odinger equation for application
to deep water waves, Proc. Roy. Soc. Lond. A 369, 105–114 (1979).
26. M. Stiassnie and L. Shemer, Energy computations for evolution of class I and II
instabilities of Stokes waves, J. Fluid Mechanics 174, 299–312 (1987).
27. T. Yasuda and N. Mori, Roles of sideband instability and mode coupling in forming
a water wave chaos, Wave Motion 26(2), 163–185 (1997).
28. M. Su, Threedimensional deepwater waves. Part 1. Experimental measurement of
skew and symmetric patterns, J. Fluid Mechanics 124, 73–108 (1982).
29. N. Mori, T. Yasuda and K. Kawaguchi, Breaking eﬀects on waves statistics for deep
water random wave train, in Waves97 (ASCE, 1997), pp. 316–328.
30. T. Yasuda and N. Mori, High order nonlinear eﬀects on deepwater random wave
trains, Int. Symp. WavesPhysical and Numerical Modelling, Vol. 2, Vancouver (1993),
pp. 823–832.
31. N. Mori and P. Janssen, On kurtosis and occurrence probability of freak waves,
J. Phys. Oceanogr. 36(7), 1471–1483 (2006).
32. M. A. Tayfun and J.M. Lo, Nonlinear eﬀects on wave envelope and phase,
J. Waterway, Port, Coastal Ocean Eng. 116(1), 79–100 (1990).
33. N. Mori and T. Yasuda, A weakly nonGaussian model of wave height distribution for
random wave train, Ocean Eng. 29(10), 1219–1231 (2002).
34. N. Mori, P. Liu and T. Yasuda, Analysis of freak wave measurements in the Sea of
Japan, Ocean Eng. 29(11), 1399–1414 (2002).
35. M. LonguetHiggins and O. Phillips, Phase velocity eﬀects internary wave interactions,
J. Fluid Mechanics 12, 133–135 (1962).
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
Freak Wave 149
36. A. Kimura, Probability of the freak wave appearance in a 3dimensional sea condition,
Proc. 24th Int. Conf. Coastal Eng., Vol. 1, ASCE (1994), pp. 356–369.
37. Y. Goda, Random Seas and Design of Maritime Structures, 2nd Edn. (World Scientiﬁc,
2000).
38. I. Lavrenov, The wave energy concentration at the Agulhas current oﬀ South Africa,
Natural Hazard 17, 117–127 (1998).
39. A. Osborne, M. Onorato, M. Serio and S. Bertone, The nonlinear dynamics of rogue
waves and holes in deep water gravity wave trains, Phys. Lett. A 275(5–6), 386–393
(2000).
40. S. Haver, Evidences of the existence of freak waves, in Rogue Waves, eds. M. Olagnon
and G. Atanassoulis (IFREMER, 2001), pp. 129–140.
41. M. Olagnon and G. Athanassoulis (eds.), Rogue Waves 2000 (IFREMER, 2000).
42. K. Hasselmann, On the nonlinear energy transfer in gravitywave spectrum. I. General
theory, J. Fluid Mechanics 12, 481–500 (1962).
43. V. Krasitskii, On reduced equations in the Hamiltonian theory of weakly nonlinear
surface waves, J. Fluid Mechanics 272, 1–20 (1994).
44. F. Edgeworth, On the representation of statistical ferquency by a series, J. Roy. Statist.
Soc. 70, 102–106 (1907).
45. M. A. Tayfun, Statistics of nonlinear wave crests and groups, Ocean Eng. 33(11–12),
1589–1622 (2006).
46. F. Fedele and M. A. Tayfun, Explaining freak waves by a stochastic theory of wave
groups, 25th Int. Conf. Oﬀshore Mechanics and Arctic Engineering 1 (2006).
47. M. Onorato, A. R. Osborne, M. Serio, L. Cavaleri, C. Brandini and C. T. Stansberg,
Extreme waves, modulational instability and second order theory: Wave ﬂume exper
iments on irregular waves, Eur. J. Mech. B/Fluids 25(5), 586–601 (2006).
48. N. Mori, M. Onorato, P. A. E. M. Janssen, A. R. Osborne and M. Serio, On the extreme
statistics of long crested deep water waves: Theory and experiments, J. Geophys. Res.,
Ocean 112, C09011 (2007), doi: 10.1029/2006JC004024.
49. H. SocquetJuglard, K. Dysthe, K. Trulsen, H. Krogstad and J. Liu, Probability dis
tribution of surface gravity waves during spectral changes, J. Fluid Mechanics 542,
195–216 (2005).
50. O. Gramstad and K. Krulsen, Inﬂuence of crest and group length on the occurrence
of freak waves, J. Fluid Mechanics 582, 463–472 (2007).
July 31, 2009 8:18 9.75in x 6.5in b684ch06 FA
This page intentionally left blank This page intentionally left blank
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
Chapter 7
ShortTerm Wave Statistics
Akira Kimura
Department of Social Systems Engineering
Faculty of Engineering, Tottori University, 6808552
Koyama Minami 4101, Tottori, Japan
kimura@sse.tottoriu.ac.jp
This chapter deals with the shortterm statistical properties of irregular sea waves.
It begins with a description of an early study by Rice [Bell Syst. Tech. J. 23,
282–332; 24, 46–156 (1944, 1945)] about the relation between wave spectra and
statistics of wave amplitudes. Next, an explanation is given of studies which deal
with the eﬀect of spectrum width and wave nonlinearity on wave statistics as well
as with joint statistical properties between wave heights and periods. Finally, the
chapter closes with a look at the statistics of wave direction, length of wave crest,
and spatial maximum amplitude (which are the 3D wave’s properties) as well as
the time series of wave height.
7.1. Introduction
Waves in sea conditions have certain irregular properties. These comprise properties
such as wave heights and wave periods. For many years, engineers have known
that it is not suﬃcient to know only statistically representative values such as the
mean value. Larger wave heights can induce damage on coastal structures while
longer wave periods may induce resonance phenomena. To understand their overall
statistical properties, engineers and researchers have tried to clarify the statistical
characteristics of populations for these properties with the intention of ﬁnding out
their probability density function (pdf). In particular, their pdf have been studied
in relation to the wave spectrum.
The study of this development was initiated primarily by Rice.
20
The appli
cation of this idea in coastal engineering was started notably by LonguetHiggins.
15
He developed pdf for surface elevation and gradient, maxima and minima of water
surface, number of zerocrossings along a line, etc. Of course, sailors have carried out
visual observations of waves since ancient times. From the theoretical and empirical
knowledge of wave statistics, two types of statistical study have appeared: short
term and longterm wave statistics. Since waves must be stationary stochastic to
151
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
152 A. Kimura
study shortterm statistics, the duration of concern is 1 or 2 days at most. On the
other hand, longterm statistics cover waves for time periods of at least dozens of
years. In this type, the time series of water surface elevation is not used. Instead,
representative wave properties such as signiﬁcant wave height and period are uti
lized. The wave height/period is estimated for coastal structures within their lifetime
from longterm statistics, for example. The eﬀectiveness of results from longterm
statistics is heavily dependent on shortterm statistics. Both are inseparable for the
eﬃcient use of wave statistics in coastal engineering.
7.2. Rice Model
Rice
20
proposed a model of the Fourier sum for irregular signals. The irregular signal
I(t) is expressed by the summation of trigonometric functions as
I(t) =
N
¸
n=1
(a
n
cos σ
n
t +b
n
sin σ
n
t) =
N
¸
n=1
c
n
cos(σ
n
t +ε
n
), (7.1)
in which σ
n
is the angular frequency and ε
n
(= tan
−1
(−b
n
/a
n
)) is the phase of the
nth component wave, respectively. a
n
, b
n
, and c
n
are amplitudes of nth component
waves which are distributed normally with zero mean. Their variances are
'a
2
n
` = 'b
2
n
` =
c
2
n
2
= S(σ)dσ , (7.2)
where ' ` means the ensemble mean, S(σ) is the energy spectrum of I(t). Rice
deﬁned the envelope function for I(t) as
I(t) =
N
¸
n=1
c
n
cos(σ
n
t +ε
n
) = I
c
(t) cos σt −I
s
(t) sin σt , (7.3)
where σ is the mean frequency of S(σ), and I
c
(t) and I
s
(t) are
I
c
(t) =
N
¸
n=1
c
n
cos(σ
n
t −σt +ε
n
), I
s
(t) =
N
¸
n=1
c
n
sin(σ
n
t −σt +ε
n
) . (7.4)
The envelope function R(t) is
R(t) = ¦I
2
c
(t) +I
2
s
(t)¦
1/2
. (7.5)
The nth component wave in Eq. (7.4), c
n
cos(σ
n
t −σt +ε
n
) or c
n
sin
(σ
n
t −σt +ε
n
), has a Ushape pdf of ﬁnite width.
14
From the central limit theorem,
the inﬁnite sum (N →∞) of the trigonometric functions, I
c
(t) and I
s
(t), are dis
tributed normally. Their joint pdf is
p(I
c
, I
s
)dI
c
dI
s
=
1
2πϕ
0
exp
−
I
2
c
+I
2
s
2ϕ
0
dI
c
dI
s
, (7.6)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 153
since I
c
(t) and I
s
(t) are independent ('I
c
(t) I
s
(t)` = 0). ϕ
0
is the variance of I
c
(t),
I
s
(t), and I(t), respectively:
'I
2
c
` = 'I
2
s
` = 'I
2
` = ϕ
0
. (7.7)
The variables R(t), I(t), I
c
(t), and I
s
(t) are represented as R, I, I
c
, and I
s
, respec
tively. It is convenient to introduce the variable θ, where
I
c
= Rcos θ, I
s
= Rsin θ . (7.8)
By substituting Eq. (7.8) into Eq. (7.6), the joint pdf of R and θ becomes
p(R, θ)dRdθ =
R
2πϕ
0
exp
−
R
2
2ϕ
0
dRdθ . (7.9)
Since R and θ are independent and θ is distributed uniformly over the range 0 to
2π, the pdf of R is obtained as
p(R)dR =
R
ϕ
0
exp
−
R
2
2ϕ
0
dR. (7.10)
Figure 7.1(a) shows the relation between I and R. The spectrum S(σ) for I is the
following normalized Wallops type (r = 10)
8
:
S(σ) =
σ
σ
p
−r
exp
¸
r
4
1 −
σ
σ
p
−4
¸¸
. (7.11)
The maxima or minima of I almost coincide with R within a “zeroup/down
crossing” interval of I. If I takes extremes (∂I/∂t = 0) at t = t
∗1
, t
∗2
, . . . , wave
amplitude can be approximated by R[
t
∗i
, (i = 1, 2, . . .) [Fig. 7.1(b)]. The intervals
between t
∗1
, t
∗2
, . . . are almost constant when the spectrum is very narrow. The pdf
for R[
t
∗i
is also the Rayleigh distribution approximately, since the local maxima/
minima is equal to the wave amplitude. Equation (7.10) becomes the pdf for wave
amplitude
14
approximately.
7.3. Pdf of Wave Heights
7.3.1. Nonlinear eﬀect on the pdf
Since the study by LonguetHiggins,
14
the pdf for wave amplitudes has been used
extensively as pdf for wave heights. Many measurements and numerical simulations
have supported the Rayleigh distribution for the wave height distribution. The
standard form of the widely used Rayleigh distribution is
p(H)dH =
π
2
H
H
2
exp
−
π
4
H
2
H
2
dH , (7.12)
in which H is the mean wave height. When the wave spectrum is very narrow,
wave height distribution can be the Rayleigh distribution, since wave height can be
approximated as
H ≈ 2R. (7.13)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
154 A. Kimura
Fig. 7.1. (a) Wave proﬁle I(t) (solid line) and its envelope R(t) (dotted line). (b) Wave proﬁle
I(t) (solid line) and envelope R(t) (dotted line) (small discrepancies between I(t) and R(t) at
t = t
∗1
, t
∗2
, · · · ).
However, the zerocrossing wave height is deﬁned by
H = I
l max
−I
l min
,
where I
l max
and I
l min
are the consecutive local maximum and minimum of the
wave proﬁle, respectively. The pdf for H may not be the Rayleigh distribution if
the pdf for I
l max
and I
l min
are the Rayleigh distribution. Field measurements have
shown that there are small discrepancies between the Rayleigh distribution p(H)
and frequency distributions of measured data F(H): F(H) < p(H) for H < H
p
and F(H) > p(H) for H > H
p
, where p(H
p
) is a peak of p(H). In particular, as
nonlinearity on the wave proﬁle becomes prominent, this characteristic may bring
a considerable eﬀect on the pdf. Forristall
3
used the following Weibull distribution
to improve the goodness of ﬁt:
P(ς) = exp
−ς
α
β
, (7.14)
where P(ς) is the Weibull distribution function and α = 2.126, β = 8.42, and
ς =
H
∗
m
1/2
0
, (7.15)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 155
in which H
∗
is the diﬀerence between the heights of consecutive wave crest and
trough within a zerocrossing interval of the wave proﬁle. LonguetHiggins
17
pointed
out that the discrepancy should be improved using rms (root mean square) wave
amplitude.
The ith moment of a spectrum is determined as
m
i
=
∞
0
σ
i
S(σ)dσ (i = 1, 2, . . .) . (7.16)
From the Khintchine theorem, the 0th moment is also deﬁned in another way as
m
0
= η
2
rms
, (7.17)
where η is the wave proﬁle and η
rms
is its rms. When the spectrum is narrow and
the nonlinearity on the wave proﬁle is negligible,
m
0
=
(a
rms
)
2
2
, (7.18)
in which a
rms
is the rms of wave amplitude a. The Rayleigh distribution function
given by LonguetHiggins is
P(a) = exp
−
a
2
a
2
rms
. (7.19)
a
2
rms
can be replaced by 2m
0
if the nonlinearity is negligible. He discussed the non
linearity eﬀect on the wave proﬁle as follows. Applying the thirdorder Stokes wave
theory, the 2nd moment of the periodic wave proﬁle is derived fromEq. (7.17) as
m
0
=
1
2
a
2
1 −
1
2
(ak)
2
−
19
12
(ak)
4
−
3077
720
(ak)
6
−
, (7.20)
in which k is the wave number. m
0
for irregular waves is calculated by
m
0
=
∞
0
2V (a)dp(a) , (7.21)
in which V (a) denotes the potential energy (2V = η
2
= m
0
) of the zerocrossing
wave with an amplitude a given as Eq. (7.20), p(a) is the pdf of a. Applying
Eq. (7.19), m
0
becomes
m
0
=
a
max
0
2V (a)d
exp
−
a
2
a
2
rms
+δ , (7.22)
in which a
max
is the maximum wave amplitude of nonbreaking waves. He used the
same V for the waves (ak)
2
≥ 0.196. δ is the remainder and is exponentially small
for the large value of a
2
max
/a
2
rms
. The new m
0
is
m
0
=
1
2
a
2
rms
1 −(a
rms
k)
2
−
19
2
(a
rms
k)
4
−
3077
30
(a
rms
k)
6
−
. (7.23)
Since a
rms
k for irregular waves distributes roughly 0.158 < a
rms
/k < 0.203,
13
0.935 <
2m
0
a
2
rms
< 0.968 . (7.24)
If wave proﬁles have a nonlinearity, rms wave amplitude between 2.066m
0
< a
2
rms
<
2.140m
0
may give a correction for the Rayleigh distribution of wave amplitudes.
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
156 A. Kimura
7.3.2. Eﬀect of spectrum width
Since the zerocrossing method for irregular waves does not have a physical basis,
a mathematical treatment is diﬃcult except for the very narrow band spectrum.
Cartwright and LonguetHiggins
2
derived a spectrum width parameter as
ε =
1 −
m
2
2
m
0
m
4
: 0 < ε < 1 , (7.25)
in which m
0
, m
2
, and m
4
are determined by Eq. (7.16). Pdf of amplitudes (local
maxima) is the Rayleigh distribution when ε →0. If ε = 0, however, waves must be
periodic. LonguetHiggins
17
investigated the eﬀect of the spectrum width on pdf of
wave amplitudes. He used a periodic wave proﬁle with small disturbance given as
y = b cos(σt +ε) +
¸
n
a
n
cos(σ
n
t +ε
n
) , (7.26)
in which a
n
, σ
n
, and ε
n
are the amplitude, angular frequency, and phase of the
disturbance, respectively. Their amplitudes are b >> a
n
. The spectrum of the dis
turbances S
(σ) is deﬁned as
σ
n
+dσ
¸
σ
n
1
2
a
2
n
= S
(σ
n
)dσ . (7.27)
The rms of y is
'y
2
rms
` = b
2
rms
+
∞
0
sin
2
πσ
2σ
+
σ
σ
2
S
(σ)dσ , (7.28)
and the 0th moment of the spectrum of y is
m
0
=
∞
0
¦S(σ) +S
(σ)¦dσ =
b
2
rms
2
+
∞
0
S
(σ)dσ , (7.29)
where S(σ) is the spectrum of the periodic wave and b
rms
is the rms of the periodic
wave proﬁle. From Eqs. (7.28) and (7.29), y
rms
is
'y
2
rms
` = 2m
0
+
∞
0
σ
2
σ
2
−1
−cos
2
πσ
2σ
S
(σ)dσ. (7.30)
Using the relations
σ
2
σ
2
−1
= 2
σ
σ
−1
+
σ
σ
−1
2
cos
2
πσ
2σ
=
1 −cos π(σ/σ −1)
2
=
π(σ/σ −1)
2
4
+
µ
i
=
∞
0
(σ −σ)
i
S
(σ)dσ ,
(7.31)
the next relation is obtained.
'y
2
rms
`
2m
0
= 1 +
µ
1
m
0
σ
−
(π
2
/8 −1/2)µ
2
m
0
σ
2
. (7.32)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 157
If the spectrum is symmetry for ¯ σ,
'y
2
rms
`
2m
0
= 1 −
π
2
8
−
1
2
ν
2
, (7.33)
where
ν
2
=
µ
2
m
0
σ
2
=
m
0
m
2
−m
2
1
m
2
1
, (7.34)
in which ν is also the spectrum width parameter.
18
Since b a
n
, y
rms
≈ b
rms
,
Eq. (7.33) may give a correction for the Rayleigh distribution of wave amplitudes
when the spectrum is not a narrow band.
Tayfun
25
also investigated the eﬀect of spectrum width. A better deﬁnition for
the zerocrossing wave height than Eq. (7.13) may be given as
H
i
= R
m
i
+R
m
i+1
, (7.35)
where R
m
i
(i = 1, 2, . . .) are the amplitudes of the wave envelope when the wave
proﬁle I takes the local extreme values (t = t
∗1
, t
∗2
, . . .) [Fig. 1(b)]. The joint pdf
for the consecutive amplitudes R
m
i
, R
m
i+1
is
p(ξ
1
, ξ
2
) =
π
2
ξ
1
ξ
2
4(1 −κ
2
)
exp
¸
−
π(ξ
2
1
+ξ
2
2
)
4(1 −κ
2
)
I
0
¸
πκξ
1
ξ
2
2(1 −κ
2
)
, (7.36)
in which I
0
[ ] is the 0th order modiﬁed Bessel function, ξ
i
= R
m
i
/(πm
0
/2)
1/2
,
ξ
i+1
= R
m
i
+1
/(πm
0
/2)
1/2
. They are replaced as ξ
i
= ξ
1
and ξ
i+1
= ξ
2
, respectively
for simplicity. κ is the correlation parameter given by
k =
ρ
2
+λ
2
1/2
, (7.37)
in which
ρ =
1
m
0
∞
0
S(σ) cos(σ −σ)
τ
2
dσ ,
λ =
−1
m
0
∞
0
S(σ) sin(σ −σ)
τ
2
dσ ,
τ =
m
1
m
0
,
where τ is the mean wave period. Normalized wave height is given as
ς =
H
H
m
=
ξ
1
+ξ
2
2
. (7.38)
Substituting the relation ξ
2
= 2ς − ξ
1
into Eq. (7.36) and the integration in terms
of ξ
1
from 0 to 2ς gives the pdf for ς:
p(ς) =
2ς
0
p(ξ
1
, 2ς −ξ
1
)dξ
1
. (7.39)
Figure 7.2 shows p(ς) for two diﬀerent wave spectra [Wallopstype, Eq. (7.11):
r = 5 (dotted line), r = 10 (broken line)] and the Rayleigh distribution (solid line).
Kimura et al.
10
showed that the discrepancy in Fig. 7.2 comes from the diﬀerence
between I and R at (t = t
∗1
, t
∗2
, . . .) [Fig. 7.1(b)].
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
158 A. Kimura
1.2
1.0
0.8
0.6
0.4
0.2
0.0
p
(
ζ
)
3.0 2.5 2.0 1.5 1.0 0.5 0.0
ζ
Rayleigh
r = 5
r = 10
Fig. 7.2. Pdf of wave height.
7.4. Maximum Wave Height
Zerocrossing wave amplitude in Eq. (7.19) extends the range from 0 to inﬁnity if
the number of waves is inﬁnity. When the population of waves is ﬁnite, such as data
from 20 min measurements every 2 h, maximum wave amplitude may be ﬁnite. If a
time series of wave amplitudes
a
1
, a
2
, . . . , a
N
. .. .
Group No. 1
, a
N+1
, . . . , a
2N
. .. .
2
, a
2N+1
, . . . , a
3N
. .. .
3
, a
3N+1
. . . , a
4N
. .. .
4
,
is divided into groups with N waves, the maximum values (a
max
) from individual
groups show irregular properties. The pdf of the maximum wave amplitude is derived
by LonguetHiggins.
14
The probability that one wave amplitude a is greater than a
0
is P(a
0
)
[Eq. (7.19)]. When the size of the group is N, the probability that all the wave
amplitudes are a < a
0
is ¦1 − P(a
0
)¦
N
. The probability that one amplitude, at
least, exceeds a
0
is 1 − ¦1 −P(a
0
)¦
N
. Therefore, the probability that maximum
wave amplitude lies in the range a
max
∼ a
max
+da is
1 −¦1 −P(a
max
)¦
N
−1 +¦1 −P(a
max
+da)¦
N
= −d[1 −¦1 −P(a
max
)¦
N
]
= −N¦1 −P(a
max
)¦
N−1
dP(a
max
)
da
da . (7.40)
The pdf for the maximum wave amplitude is
p(a
max
) = 2N
¸
1 −exp
−
a
max
a
rms
2
¸¸
N−1
a
max
a
rms
exp
−a
2
max
a
2
rms
. (7.41)
Its mean value is derived as
E(a
max
)
a
rms
= (ln N)
1/2
+γ(ln N)
−1/2
+ (7.42)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 159
with an approximation
1 −e
−(a
max
/a
rms
)
2
N
≈ e
−e
−θ
, (7.43)
in which
θ
=
a
max
a
rms
2
−ln N , (7.44)
and γ is the Euler constant (γ = 0.5772 . . .). Maximum wave height E(H
max
)/H
rms
is usually compared also with Eq. (7.42).
7.5. ZeroCrossing Wave Period
The problem of deriving the pdf of zerocrossing wave periods is very diﬃcult.
Rice derived a pdf for the zerocrossing interval. He treated the interval where I(t)
[Eq. (7.1)] crosses the 0 line upward at t = 0 and crosses the 0 line downward
between τ and τ + dτ.
p(τ) = p
0
(τ) −
1
1!
τ
0
p
1
(r, τ)dr +
1
2!
τ
0
τ
0
p
2
(r, s, τ)drds
−
1
3!
τ
0
τ
0
τ
0
p
3
(r, s, u, τ)drdsdu + , (7.45)
where p
0
(τ) is the probability that I(t) = 0 at t = 0 and I(t) = 0 lies between
τ < t < τ + dτ, p
1
(r, τ) is the probability that I(t) = 0 at t = 0 and I(t) = 0 lies
between r < t < r + dr and τ < t < τ + dτ (r < τ), p
2
(r, s, τ) is the probability
that I(t) = 0 at t = 0, and I(t) = 0 lies between s < t < s +ds, r < t < r +dr, and
τ < t < τ +dτ(s < r < τ), and so on. Since this series of integrations is inexcusable,
Rice
21
showed p(τ) only applying the ﬁrst term of Eq. (7.45) as
p(τ) =
dτ
2π
−
ψ
0
ψ
0
1/2
M
23
H
ψ
2
0
−ψ
2
τ
−3/2
¦1 +H cot
−1
(−H)¦ , (7.46)
with
H = M
23
¸
M
2
22
−M
2
23
¸
−1/2
,
where M
22
and M
23
are the cofactors of elements µ
22
and µ
23
of the matrix,
M =
¸
¸
¸
µ
11
µ
12
µ
13
µ
14
µ
21
µ
22
µ
23
µ
24
µ
31
µ
32
µ
33
µ
34
µ
41
µ
42
µ
43
µ
44
¸
=
¸
¸
¸
ψ
0
0 ψ
τ
ψ
τ
0 −ψ
0
−ψ
τ
−ψ
τ
ψ
τ
−ψ
τ
−ψ
0
0
ψ
τ
−ψ
τ
0 ψ
0
¸
(7.47)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
160 A. Kimura
and
ψ
τ
=
1
2π
∞
0
S(σ) cos στ dσ ,
ψ
τ
= −
1
2π
∞
0
σS(σ) sin στ dσ ,
(7.48)
ψ
τ
= −
1
2π
∞
0
σ
2
S (σ) cos στ dσ ,
ψ
0
= ψ
τ
[
τ=0
, ψ
0
= ψ
τ
[
τ=0
.
Equation (7.46) has a peak around the mean value of τ. The second and third peaks
appear for larger τ. This is because the higher terms in Eq. (7.45) are neglected.
No successful analytical solution for the zerocrossing wave period has been derived
at the present time. Instead of wave periods, LonguetHiggins
16
derived the pdf for
wave lengths. The wave proﬁle of unidirectional irregular waves is expressed as
ς = Rcos ¦χ¦ , (7.49)
where
χ = ux +φ, χ
x
= u +φ
x
, (7.50)
in which R is the amplitude of the wave envelope, φ is the phase, u is the mean
wave number in xdirection, and subscript x means a diﬀerentiation in terms of x.
Joint pdf for χ and χ
x
was derived by LonguetHiggins
15
as
p (χ, χ
x
) =
(m
∗
0
/µ
∗
2
)
1/2
4π
1 + (χ
x
−u)
2
m
∗
0
/µ
∗
2
¸
3/2
, (7.51)
with
m
∗
n
=
∞
0
E (u) u
n
du , (n = 0, 1, . . .) (7.52)
µ
∗
2
=
∞
0
E (u) (u −u)
2
du , (7.53)
u = m
∗
1
/m
∗
0
, (7.54)
in which u is the wave number in the xdirection and E (u) is its spectrum. The
event ς = 0 is equivalent to χ = (2r −1/2) π or χ = (2r + 1/2) π (r: integer). The
probability that χ = (2r −1/2) π between x and x +dx is
H(χ)dx =
∞
−∞
p(χ, χ
x
)
χ=(2r−1/2)π
[χ
x
[ dxdχ
x
=
1
2π
m
∗
2
m
∗
0
1/2
dx. (7.55)
Since the probability that χ = (2r + 1/2)π between x and x + dx is given by the
same equation, Eq. (7.55), the probability that ς = 0 between x and x + dx is
2H(χ)dx. The average zerocrossing interval is
= π
m
∗
0
m
∗
2
1/2
=
π
u
1 +ν
∗2
1/2
, (7.56)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 161
with
ν
∗2
=
m
∗
0
m
∗
2
−m
∗2
1
m
∗2
1
=
µ
∗
2
u
2
µ
∗
0
. (7.57)
If ν
∗
is small,
≈
π
u
. (7.58)
The zeroupcrossing event of ς takes place when χ = (2r −1/2)π with ∂χ/∂x > 0
or χ = (2r + 1/2)π with ∂χ/∂x < 0. The probability of the latter condition is
∞
−∞
p (χ, χ
x
)
χ=(2r+1/2)π
[χ
x
[ dxdχ
x
=
1
4
m
∗
2
m
∗
0
1/2
1 −
1 +ν
∗2
−1/2
dx. (7.59)
This probability is negligible for small values of ν
∗
. Between any two successive
zerocrossing points x
1
and x
2
, χ must increase by π. Hence,
π = χ(x
2
) −χ(x
1
) =
¸
χ
x
+
1
2
2
χ
xx
+
.
The terms on the left of this equation, Taylor series expansion of χ(x
2
) around
x = x
1
is used. Since χ
xx
= O(ν
2
), χ
xxx
= O(ν
3
), . . . and they are very small,
neglecting these terms,
≈
π
χ
x
. (7.60)
Applying the same method by Rice,
20
a mean zerocrossing number of ς per unit
length, the conditional pdf of χ
x
is
p(χ
x
)
χ
=
p(χ, χ
x
)[χ
x
[
H(χ)
=
1
2
m
∗
0
m
∗
2
1/2
[χ
x
[(m
∗
0
/µ
∗
2
)
1/2
¦1 + (χ
x
− u)
2
(m
∗
0
/µ
∗
2
)¦
3/2
, (7.61)
in which [χ
x
[/H(χ) is the probability that χ falls between x and x + dx. Since
Eq. (7.61) does not involve χ, p(χ
x
)
χ
directly becomes the pdf for χ
x
. Substituting
Eqs. (7.57), (7.58), and (7.60) into Eq. (7.61),
p(ξ) =
1
2ν¦1 + (ξ/ν
∗
)
2
¦
3/2
, (7.62)
in which ξ = ( −)/. If the frequency spectrum S(σ) is used instead of the wave
number spectrum E(u), the pdf for the zerocrossing interval (halfwave period) is
derived following the same method.
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
162 A. Kimura
7.6. Joint Pdf for Wave Height and Period
LonguetHiggins
18
derived the joint pdf for wave amplitude and period. He used
the following expression for the wave proﬁle:
ς = Re ¦Aexp (iσt)¦ , (7.63)
where Re means a real part and
A = a exp¦iφ¦ , (7.64)
in which a is the wave amplitude, φ is the phase of A, and σ is the mean angular
frequency given by
σ =
m
1
m
0
, (7.65)
where m
0
and m
1
are calculated by Eq. (7.16). The phase of ς is χ = σt + φ.
Diﬀerentiation of the phase with t becomes
χ
t
= φ
t
+σ . (7.66)
When the spectrum is very narrow, the envelope of ς changes very slowly and
φ
t
σ. The wave period can be approximated as
τ =
2π
χ
t
=
2π
σt +φ
t
. (7.67)
LonguetHiggins derived a joint pdf for a and φ
t
:
p(a, φ
t
) =
a
2
(2πµ
2
0
µ
2
)
1/2
exp
−
a
2
(1/µ
0
+φ
2
t
/µ
2
)
2
, (7.68)
where µ
0
and µ
2
are calculated with Eq. (7.31), and ν is calculated by Eq. (7.34).
With new variables
R =
a
(2m
0
)
1/2
, T =
τ
τ
,
Eq. (7.68) is expressed as
p(R, T) = p(a, φ
t
)
∂(a, φ
t
)
∂(R, T)
=
2R
2
L
π
1/2
νT
2
exp
−
R
2
[1 + (1 −1/T)
2
]
ν
2
, (7.69)
in which L is the correction factor to neglect the p(R, T) in the region T < 0,
L =
2
1 + (1 +ν
2
)
−1/2
. (7.70)
Figure 7.3 shows the contour map of p(R, T). Solid curves are contours for
p(R, T)/p
max
= 0.99, 0.90, 0.70, 0.50, 0.30, 0.10 from inside to outside, respectively.
The marginal distribution for normalized wave amplitudes and periods are
p(R) =
2L
π
1/2
Rexp(−R
2
)
R/ν
−∞
exp(−β
2
)dβ, (7.71)
p(T) =
L
2νT
2
1 +
(1 −1/T)
2
ν
2
−3/2
. (7.72)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 163
Fig. 7.3. Joint pdf for wave amplitude and period Eq. (7.69) (after LonguetHiggins
18
).
Fig. 7.4. Pdf for wave amplitude p(R) Eq. (7.71) (after LonguetHiggins
18
).
Figures 7.4 and 7.5 show p(R) and p(T), respectively. If ν = 0, p(R) is not the
Rayleigh distribution. Peak of the p(R) slightly shifts to the larger side of R with
increasing spectrum width ν. When ν = 0, p(T) becomes a straight line (broken
line, T: constant).
Figure 7.6 shows the comparison of data and Eq. (7.69) by Goda.
8
The number
of waves within a grid of ∆H = 0.2H and ∆T = 0.2T are shown. H and T are
the mean values of wave height and period, respectively. Many data sets, their
correlation coeﬃcients between H and T distributed from 0.19 to 0.25, are used.
The spectrum width parameter in Eq. (7.69) is v = 0.26.
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
164 A. Kimura
Fig. 7.5. Pdf for wave period p(T) Eq. (7.72) (after LonguetHiggins
18
).
Fig. 7.6. Joint distribution of wave height and period (after Goda
5
).
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 165
7.7. Statistical Properties of 3D Irregular Waves
Statistical properties of waves, such as length and direction of wave crest, are
important in the design of coastal structures. Wave height and period are deter
mined from a record of point measurement in the ordinary method. In contrast, a
3D wave proﬁle is necessary for the deﬁnition of wave crest length, direction, etc.
As a result, very few studies have examined their properties.
7.7.1. Pdf for wave directions
Kimura
12
introduced a theoretical pdf for the wave direction. Figure 7.7 shows
axes and wave crest lines (dotted line). Two wave gauges are installed along the
yaxis. One is on the origin y
1
and the other is on y
2
(0, y
0
). Figure 7.8 shows the
wave proﬁles measured at these points. ∆t
1
, ∆t
2
, ∆t
3
, . . . are the time lags between
corresponding wave crests. The phase lag of individual waves is determined as
ε
i
= 2π
∆t
i
T
i
, (i = 1, 2, . . .) , (7.73)
in which T
i
is the zerocrossing wave period. The angle between the crest line and
yaxis (Fig. 7.7) is
θ
i
= sin
−1
ε
i
L
i
/2π
y
0
, (i = 1, 2, . . .) , (7.74)
where L
i
is the wave length corresponding to the zerocrossing wave period T
i
and
θ
i
is the direction of wave propagation measured from the xaxis.
Fig. 7.7. Wave crest lines (dotted line) and axes.
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
166 A. Kimura
Fig. 7.8. Wave proﬁles (solid line) at y
1
and y
2
.
Assuming y
0
u < 1 in which u is the mean wave number in the x direction, wave
proﬁles at these points are
η
1
(t) =
∞
¸
i=1
c
i
cos(σ
i
t +δ
i
) = η
c1
cos σt −η
s1
sinσt , (7.75)
η
2
(t) =
∞
¸
i=1
c
i
cos(σ
i
t +δ
i
+v
i
y
0
) =η
c2
cos σt −η
s2
sin σt , (7.76)
in which c
i
, σ
i
, δ
i
, and v
i
are the amplitude, the angular frequency, the initial phase,
and the wave number in the ydirection of the ith component wave, respectively. σ
is the mean angular frequency. η
c1
, η
s1
, η
c2
, and η
s2
are derived in the same way as
Eq. (7.4). Their pdfs are the normal distribution with 0 mean, respectively. Their
joint pdf is
p(η
c1
, η
s1
, η
c2
, η
s2
) =
1
(2π)
2
(m
2
0
−µ
2
3
−µ
2
4
)
exp
−
1
2(m
2
0
−µ
2
3
−µ
2
4
)
(m
0
η
2
c1
−2µ
3
η
c1
η
c2
−2µ
4
η
c1
η
s2
+ m
0
η
2
s1
+ 2µ
4
η
s1
η
c2
−2µ
3
η
s1
η
s2
+m
0
η
2
c2
+m
0
η
2
s2
)
,
(7.77)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 167
with
m
0
= 'η
2
c1
` = 'η
2
s1
` = 'η
2
c2
` = 'η
2
s2
` ,
µ
3
= 'η
c1
η
c2
` = 'η
s1
η
s2
` ,
(7.78)
µ
4
= 'η
c1
η
s2
` = −'η
s1
η
c2
` ,
'η
c1
η
s1
` = 'η
s2
η
c2
` = 0 .
From the translations
η
c1
= R
1
cos(σt +φ
1
) = R
1
cos(χ
1
) ,
η
s1
= R
1
sin(σt +φ
1
) = R
1
sin(χ
1
) ,
(7.79)
η
c2
= R
2
cos(σt +φ
2
) = R
2
cos(χ
2
) ,
η
s2
= R
2
sin(σt +φ
2
) = R
2
sin(χ
2
) ,
the joint pdf for R
1
, R
2
, χ
1
, and χ
2
is
p(R
1
, R
2
, χ
1
, χ
2
) =
1
(2π)
2
(m
2
0
−µ
2
3
−µ
2
4
)
exp
−
1
2(m
2
0
−µ
2
3
−µ
2
4
)
(m
0
(R
2
1
+R
2
2
)
− 2R
1
R
2
[µ
3
cos(χ
1
−χ
2
) −µ
4
sin(χ
1
−χ
2
)])
, (7.80)
in which R
i
and χ
i
are the amplitude and phase of η
i
(i = 1, 2), respectively.
15
The crest of the wave proﬁle passes the origin of axes when χ
1
= 2nπ
(n = 0, 1, 2, . . .). The probability of the event that χ
1
= 2nπ between t and t + dt
is H(χ
1
)dt = m
1
/m
0
dt. The conditional pdf for R
1
, R
2
, and χ
2
when χ
1
= 2nπ is
p(R
1
, R
2
, χ
2
[ χ
1
= 2nπ) =
σp(R
1
, R
2
, χ
1
, χ
2
)
H(χ
1
)
χ
1
=2nπ
=
R
1
R
2
2πm
2
0
(1 −κ
2
)
exp
−
1
2 m
2
0
(1 −κ
2
)
[m
0
(R
2
1
+R
2
2
)
−2R
1
R
2
(µ
3
cos(2nπ −χ
2
) −µ
4
sin(2nπ −χ
2
))]
,
(7.81)
in which
κ
2
=
µ
2
3
+ µ
2
4
m
2
0
≈ 1 −y
2
0
u
2
γ
2
∗
(7.82)
and γ
∗
is the longcrestedness parameter.
15
Putting χ
2
= 2nπ +ε, pdf for ε is
p(ε[R
1
, R
2
) =
p(R
1
, R
2
, ε[χ
1
= 2nπ)
p(R
1
, R
2
)
= exp
R
1
R
2
κ
m
0
(1 −κ
2
)
cos(ε −β)
2πI
0
¸
R
1
R
2
κ
m
0
(1 −κ
2
)
, (7.83)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
168 A. Kimura
where I
0
[ ] is the 0th order modiﬁed Bessel function and β = tan
−1
(µ
4
/µ
3
). β is an
angle of the dominant wave direction from xaxis. From Eq. (7.74), the pdf for the
wave direction from xaxis becomes
p(θ[R
1
, R
2
) =
2π y
0
L
cos θ
exp
R
1
R
2
m
0
y
2
∗
1 −y
2
∗
cos
2πy
0
L
sin θ −β
2πI
0
¸
R
1
R
2
m
0
y
2
∗
1 −y
2
∗
,
(7.84)
in which L is the wave length corresponding to σ, and
y
∗
≈
1 −κ
2
=
1 −
µ
2
3
+µ
2
4
m
2
0
. (7.85)
If the interval between two wave gauges is inﬁnitesimally small, i.e., y
∗
→ 0, R
1
=
R
2
→R. Putting β = 0 and applying the relation
z →∞⇒I
0
(z) →
exp(z)
2πz
, (7.86)
Eq. (7.84) reduces to
p(θ[R)dθ =
Rcos θ
γ
∗
√
2πm
0
exp
−
1
2γ
2
∗
R
2
m
0
sin
2
θ
dθ . (7.87)
Figure 7.9 shows p(θ[R) for several R. R is the mean of R. The 3D spectrum is
the Bretschneider–Mitsuyasu type with Mitsuyasudirectional function modiﬁed by
Goda.
8
The peakedness parameter for the directional function is S
max
= 10. The
pdf for larger R shows a narrower width.
7.7.2. Mean length of 3D wave crest
The length of 3D irregular wave crests may aﬀect the total wave force on coastal
structures with ﬁnite width such as caissontype breakwaters, for example. Since
a deﬁnition of the wave crest length is diﬃcult, there are very few studies about
its statistical properties. Goda
6
studied its statistics numerically. He calculated a
spatial wave proﬁle η(x, y) of 3D irregular waves. The black and white pattern in
Fig. 7.10 shows
η(x, y) > η
th
, Black area,
η(x, y) < η
th
, Other ,
(7.88)
in which η
th
is the threshold value (η
th
= 0.1H
1/3
in Fig. 7.10). The 3D wave
spectrum is the Bretschneider–Mitsuyasu type (H
1/3
= 2.0 m, T
1/3
= 8.01 s) with
Mitsuyasudirectional function modiﬁed by Goda, and S
max
= 25.
The pattern recognition method is applied to determine the wave crest lengths.
Four threshold values η
th
= 0.2 m, 0.5 m, 1.0 m, and 2.0 m were used. The calculated
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 169
Fig. 7.9. Pdf for wave directions: Eq. (7.87).
crest lengths were arranged in descending order. The top 0.4%, 1%, 5%, and 10%
of total data for each η
th
were used in the analysis, respectively. Figure 7.11 shows
the distribution of the mean length of wave crests. The four lines in the ﬁgure show
the results from the above four diﬀerent data.
7.7.3. Pdf of the local maximum of 3D wave amplitudes
Figure 7.12 shows a short crested spatial wave proﬁle and a wave gauge. The ﬁxed
type wave gauges usually fail to record the peak wave proﬁle in the neighborhood.
Wave height changes along a large object such as a ship, caissontype breakwater,
etc. If a large wave hits any part of these objects, it is recognized that the large
wave attacked it regardless of its position. Since local damage on the caissontype
breakwater, for example, may lead to the whole failure, the size eﬀect on the pdf
becomes important. Kimura et al.
11
showed the pdf of the maximum wave ampli
tudes within a ﬁnite interval.
x
, y
, and vertical z
axes are taken in the 3D wave ﬁeld. The vertical plane H
is installed parallel to x
axis. Figure 7.13 schematically shows the envelope R(x
)
(solid line) for the cross section of the 3D wave proﬁle on H. A wave gauge is
installed at x
= 0. The width of the object is 2∆L (−∆L < x
< ∆L). R
∗
is the local
maximum amplitude of R(x
) (−∆L < x
< ∆L), and R(0) is the amplitude of the
envelope at x
= 0. ∆R is the diﬀerence between R
∗
and R(0). ∆R is approximately
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
170 A. Kimura
Fig. 7.10. Wave crests (after Goda
6,7
).
Fig. 7.11. Mean length of wave crest (after Goda
6,7
).
given as
R
(0)
R
(0)
≤ ∆L : ∆R = −
¦R
(0)¦
2
2R
(0)
, (7.89)
R
(0)
R
(0)
> ∆L
R
(0) ≥ 0 : ∆R = R
(0)∆L +
R
(0)(∆L)
2
2
, (7.90)
R
(0) < 0 : ∆R = −R
(0)∆L +
R
(0)(∆L)
2
2
, (7.91)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 171
Fig. 7.12. 3D wave proﬁle and the vertical plane H.
Fig. 7.13. Wave envelope (solid line) for 3D wave proﬁle.
in which R
(0) and R
(0) are the ﬁrst and second derivatives of R(x
) at x
= 0,
respectively. The probability that R
∗
falls between c < R
∗
< c + dR on condition
R(0) is
Prob(c < R
∗
< c +dR)[
R(0)
=
S
p[R
(0), R
(0) : R(0)]dS , (7.92)
in which p[R
(0), R
(0) : R(0)] is a conditional joint pdf of R
(0) and R
(0) on
condition R(0), and S is the region in which c < R
∗
< c +dR in the R
∼ R
plane.
The probability that R
∗
falls between c < R
∗
< c +dR regardless of R(0) is
Prob(c < R
∗
< c +dR) =
c
0
¦Prob(c < R
∗
< c +dR)[
R(0)
¦p[R(0)]dR(0) , (7.93)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
172 A. Kimura
Fig. 7.14. Pdf of local maximum wave amplitudes (dotted lines) within three ﬁnite width ∆L = 0
(Rayleigh dist.), 0.05L
1/3
, and 0.10L
1/3
.
where p[R(0)] is the pdf of R at x
= 0 (Rayleigh distribution). Connecting
Eqs. (7.92) and (7.93), the pdf of the local maximum amplitudes within the ﬁnite
width is given as
p(R
∗
)dR
∗
=
R
∗
0
G
p[R(0), R
(0), R
(0)]dGdR(0) , (7.94)
in which p[R(0), R
(0), R
(0)] is the joint pdf of R(0), R
(0), and R
(0), and G is
the region in which the local maximum amplitude falls within R
∗
and R
∗
+dR
∗
on
condition R(0). Rice
20
formulated p[R(0), R
(0), R
(0)].
Figure 7.14 shows p(R
∗
) (dotted lines). The spectrum is the Bretschneider–
Mitsuyasutype with H
1/3
= 5.5 m and T
1/3
= 10 s and Goda’s directional function
with S
max
= 10, the water depth is h/L
1/3
= 1.0, and the halfobject width is ∆L =
0.050L
1/3
and ∆L = 0.10L
1/3
, respectively. The Rayleigh distribution (solid line
∆L = 0) is also shown in the ﬁgure. Broken lines are the Weibull distribution with
m ≈ 2.5 (shape factor) shows good agreement when ∆L/L
1/3
> 0.05 (h/L
1/3
= 1.0,
S
max
= 10).
7.8. Statistics of the Time Series of Wave Heights
Sawhney
23
and Goda
4
started to investigate the statistical properties of the time
series of irregular wave heights,
. . . , H
i−2
, H
i−1
, H
i
, H
i+1
, H
i+2
, . . . , (7.95)
since this property may aﬀect wave overtopping, slow drift oscillation of ﬂoating
bodies, etc. This property is analyzed by applying the concept of “run.”
Figure 7.15 shows a time series of wave heights. The run of high waves is
deﬁned by waves, which are H
i
> H
∗
consecutively, in which H
∗
is the threshold
wave height. In the case shown in Fig. 7.15, the length of the high wave run is 3.
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 173
Fig. 7.15. Run of wave heights (high waves, low waves, and total run).
Goda
4
gave the pdf for the run of high waves as
p(j
1
) = p
j
1
−1
q , (7.96)
in which j
1
is the length of the high wave run, p is the probability that the event
H > H
∗
takes place and q = 1 −p.
Since the pdf for wave heights is the Rayleigh distribution, p and q are given as
p = exp
−
H
2
∗
8m
0
,
(7.97)
q = 1 −p .
The mean and standard deviation of j
1
are
j
1
=
1
q
, σ(j
1
) =
√
p
q
, (7.98)
respectively. The recurrence interval (total run) of high waves is deﬁned as the sum
of the high wave run and the consecutive low wave run. The run of low waves is
deﬁned by waves, which are H
i
< H
∗
, consecutively. The total run in Fig. 7.15 is
j
t
= 3 (j
1
: high wave run) + 4 (j
2
: low wave run) = 7.
The pdf for the total run is
p(j
t
) =
pq
p −q
(p
j
t
−1
−q
j
t
−1
) . (7.99)
The mean and standard deviation of j
t
are
j
t
=
1
p
+
1
q
, σ(j
t
) =
p
q
2
+
q
p
2
, (7.100)
respectively. Equations from (7.96) to (7.100) are derived assuming that the wave
height is independent:
γ(H
m
H
m+n
) = 0, m, n = 1, 2, 3, . . . , (7.101)
where γ( ) is the correlation coeﬃcient of wave heights. From obser
vation, Rye
22
showed that there is a correlation between consecutive wave
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
174 A. Kimura
heights γ(H
m
H
m+1
) = 0. From numerical calculations, Kimura
9
showed that
γ(H
m
H
m+1
) = 0 but γ(H
m
H
m+n
) ≈ 0 (n ≥ 2). He assumed the time series of wave
heights from a Markov chain
9
and introduced the pdf for the high wave run:
p(j
1
) = p
j
1
−1
22
(1 −p
22
) . (7.102)
p
22
is the transition probability from the event H
i
> H
∗
to H
i+1
> H
∗
. Its mean
and standard deviation are
j
1
=
1
1 −p
22
, σ(j
1
) =
√
p
22
1 −p
22
. (7.103)
The pdf for the total run is
p(j
t
) =
(1 −p
22
)(1 −p
11
)
p
22
−p
11
(p
j
t
−1
22
−p
j
t
−1
11
) . (7.104)
p
11
is the transition probability from the event H
i
< H
∗
to H
i+1
< H
∗
. The mean
and standard deviation of j
t
are
j
t
=
1
1 −p
22
+
1
1 −p
11
,
(7.105)
σ(j
t
) =
p
11
(1 − p
11
)
2
+
p
22
(1 −p
22
)
2
1/2
,
where p
11
and p
22
are given by
p
11
=
H
∗
0
H
∗
0
p(H
1
, H
2
)dH
1
dH
2
/
H
∗
0
p(H
1
)dH
1
,
p
22
=
∞
H
∗
∞
H
∗
p(H
1
, H
2
)dH
1
dH
2
/
∞
H
∗
p(H
1
)dH
1
,
(7.106)
with
p(H
1
, H
2
) =
4H
1
H
2
(1 −κ
2
)H
4
rms
exp
H
2
1
+H
2
2
(κ
2
− 1)H
2
rms
I
0
¸
2κH
1
H
2
(1 −κ
2
)H
2
rms
, (7.107)
p(H
1
) =
2H
1
H
2
rms
exp
−
H
2
1
H
2
rms
. (7.108)
Equations (7.107) and (7.108) are the 2D Rayleigh and Rayleigh distributions,
respectively. k is the correlation parameter. The relation between κ and the corre
lation coeﬃcient γ(H
m
H
m+1
) is
γ(H
m
H
m+1
) =
1
1 −π/4
E(κ) −
1
2
(1 −κ
2
)K(κ) −
π
4
, (7.109)
in which E( ) and K( ) are complete elliptic integrals of the ﬁrst and second kind,
respectively.
Battjes et al.
1
and LonguetHiggins
19
showed κ and the relation between κ and
γ(H
m
H
m+1
) as
κ
2
=
1
m
0
¸
∞
0
S (σ) cos στ
0
dσ
2
+
¸
∞
0
S (σ) sinστ
0
dσ
2
¸
(7.110)
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 175
Fig. 7.16. (a) Mean length of high wave runs (solid line: Eq. (7.103); circles: after Goda
8
).
(b) Mean length of total runs (solid line: Eq. (7.105); circles: Goda used j
2
for the total run, after
Goda
8
).
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
176 A. Kimura
and
κ
2
≈ γ(H
m
H
m+1
) κ
2
< 0.6 ,
γ(H
m
H
m+1
) ≈ 1 −
1 −κ
2
4 −π
κ: very close to 1 .
Figures 7.16(a) and 7.16(b) show the comparison between j
1
and Eq. (7.103),
and j
t
and Eq. (7.105), respectively. The threshold wave height is H
1/3
.
7.9. Remarks
Statistical studies have shown us the entire structures of wave irregularities in sea
conditions. A knowledge about their pdf has brought us an advance not only for
designing coastal structures but also for understanding nearshore hydraulics of the
wave and current system. However, we have to keep several things in mind when
we use these wave statistics.
The Rayleigh distribution for wave height is derived on the condition that the
wave spectrum is narrow. The problem still remains to clarify a good correspon
dence between measured distributions and the Rayleigh distribution regardless of
the spectrum.
The assumption of the narrow band spectrum gives a considerable eﬀect on the
pdf of wave periods. Although the theoretical models have the spectrum parameter
in the equation, we have to be careful about the stretch use for the spectrum with
ﬁnite band width.
The range of the pdf may be changed by wave dynamics such as wave breaking.
Extensive application of the theoretical pdf is sometimes risky.
The applications of the periodic wave theory for zerocrossing wave properties
raise problems. There have been very few studies about the physical properties of
zerocrossing waves, of wave period–wavelength/celerity, of wave height/period —
water particle velocities, etc.
Finally, we have to take into account the statistical properties for 3D irregular
waves. Common deﬁnitions for wave properties, ﬁrst of all, become necessary: for
wave height, wave period, wavelength, celerity, wave direction, wave crest length, etc.
References
1. J. A. Battjes and G. Ph. van Vledder, Veriﬁcation of Kimura’s theory for wave group
statistics, Proc. 19th ICCE (1984), pp. 642–648.
2. D. E. Cartwright and M. S. LonguetHiggins, The statistical distribution of the
maxima of a random function, Proc. R. Soc. Lond. A 237, 212–232 (1956).
3. G. Z. Forristall, On the statistical distribution of wave heights in a storm, J. Geophys.
Res. 83(C5), 2353–2358 (1978).
4. Y. Goda, Numerical experiments on wave statistics with spectral simulation, Report
PHRI 9(3), 3–57 (1970).
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
ShortTerm Wave Statistics 177
5. Y. Goda, On the joint distribution of measured wave heights and periods, Tec. Note
of PHRI, No. 272, (1977) 19pp. (in Japanese).
6. Y. Goda, Statistics of wave crest lengths based on directional wave simulations, J. Oﬀ
shore Mech. and Arctic Eng., Trans. ASME 116, 239–245 (1994).
7. Y. Goda, Numerical simulation of ocean waves for statistical analysis, J. Marine Res.
33, 5–14 (1999).
8. Y. Goda, Random Sea and Design of Maritime Structures, 2nd Edn. (World Scientiﬁc,
2000), p. 443.
9. A. Kimura, Statistical properties of random wave groups, Proc. 17th ICCE (1980),
pp. 2955–2973.
10. A. Kimura and T. Ohta, Statistical property of the irregular wave heights deﬁned by
the zerocrossing method, Proc. Coastal Eng. 40, 146–150 (1993) (in Japanese).
11. A. Kimura and T. Ohta, Probability distribution of the maximum wave height along
a sea wall with ﬁnite width, Proc. 25th ICCE (1996), pp. 2272–2283.
12. A. Kimura, Probability distribution of 3D irregular wave directions, Proc. Waves 97,
295–306 (1997).
13. B. M. Lake and H. C. Yuen, A new model for nonlinear wind waves, 1, Physical model
and experimental evidence, J. Fluid Mech. 88, 33–62 (1978).
14. M. S. LonguetHiggins, On the statistical distribution of the heights of sea waves,
J. Marine Res. IX(3), 245–266 (1952).
15. M. S. LonguetHiggins, The statistical analysis of a random, moving surface, Phil.
Trans. Roy. Soc. Ser. A 249(966), 321–387 (1957).
16. M. S. LonguetHiggins, On the intervals between successive zeros of a random function,
Proc. Roy. Soc. Ser. A 246, 99–118 (1958).
17. M. S. LonguetHiggins, On the distribution of the heights of sea waves: Some eﬀects
of nonlinearity and ﬁnite band width, J. Geophys. Res. 85(C3), 1519–1523 (1980).
18. M. S. LonguetHiggins, On the joint distribution of wave periods and amplitudes in a
random wave ﬁeld, Proc. Roy. Soc. Ser. A 389, 241–258 (1983).
19. M. S. LonguetHiggins, Statistical properties of wave groups in a random sea state,
Phil. Trans. Roy. Soc. Ser. A 312, 219–250 (1984).
20. S. O. Rice, The mathematical analysis of random noise, Bell System Tech. J. 23,
282–332, 24, 46–156 (1944).
21. S. O. Rice, Selected papers on Noise and Stochastic Processes, ed. N. Wax (Dover,
1945), pp. 133–294.
22. H. Rye, Wave group formation among storm waves, Proc. 14th ICCE (1974),
pp. 164–183.
23. M. D. Sawhney, A study of ocean wave amplitudes in terms of the theory of runs and
a Markov chain process, Tech. Rep. of New York University (1962), p. 29.
24. R. J. Sobey, Correlation between individual waves in a real sea states, Coastal Eng.
27, 223–242 (1996).
25. M. A. Tayfun, Eﬀects of spectrum band width on the distribution of wave heights and
periods, Ocean Eng. 10(2), 107–118 (1983).
July 31, 2009 8:18 9.75in x 6.5in b684ch07 FA
This page intentionally left blank This page intentionally left blank
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
Chapter 8
Generation and Prediction of Seiches
in Rotterdam Harbor Basins
M. P. C. de Jong
Formerly at Environmental Fluid Mechanics Section
Delft University of Technology; Presently at Deltares Delft Hydraulics
P. O. Box 177, 2600 MH, Delft, The Netherlands
J. A. Battjes
Environmental Fluid Mechanics Section
Delft University of Technology, P. O. Box 5048, 2600 GA
Delft, The Netherlands
A resume is presented of recent work that has identiﬁed the mechanism through
which the seiches are generated that occasionally occur in Rotterdam Harbor
basins, and that has led to the development and implementation of an operational
warning system for the occurrence of signiﬁcant seiche events. The generation
is due to moving systems of atmospheric convection cells that can arise over
the relatively warm water of the North Sea, behind a cold front. These cause
ﬂuctuations in wind speed and atmospheric pressure that can generate long waves
at sea, as these cells move toward the coast. The average variance spectrum of
these longwave surface elevations is found to have an approximately f
−1.5
tail.
As these long waves at sea approach the harbor mouth, they can be resonantly
ampliﬁed inside certain semiclosed basins. The ratio of seiche amplitudes in two
diﬀerent basins, derived through numerical simulations based on a 2DH mild
slope model and an incident f
−1.5
longwave elevation spectrum, is in excellent
agreement with the mean ratio derived from observations for the same locations.
Finally, the principle of a possible system of prediction of the occurrence of seiche
events is described. Its predictive potential is proven through a series of hindcasts.
Based on these promising results, an operational forecasting system has been
developed and implemented by the authorities involved.
8.1. Introduction
Resonant harbor oscillations (seiches) occur in a number of basins of the Port of
Rotterdam, The Netherlands, during rough weather episodes. Within the Rotterdam
179
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
180 M. P. C. de Jong and J. A. Battjes
Fig. 8.1. The Port of Rotterdam.
Harbor system, the highest seiches occur in the Caland Canal, with measured crest
heights over 1 m at the closed end near Rozenburgse Sluis, ROZ (see Fig. 8.1).
This semiclosed basin has a length of approximately 20 km and a depth of
approximately 20 m. For almost all of the observed seiche events at ROZ, the
dominant frequency (determined through wavelet analysis) equals the lowest eigen
frequency (quarterwavelength mode), equivalent to a period of approximately
90 min. Smalleramplitude seiches, though still quite signiﬁcant, also occur in other
harbor basins.
The seiche oscillations in Rotterdam Harbor give rise to some practical concerns.
First, water levels used for the design of dikes (in particular their crest height) are
required by law to contain an allowance for seiches, superposed on the astronomical
high water level (spring tide) and wind setup. The appropriate measure of this
allowance is not well known.
Second, seiches aﬀect the design and management of a movable storm surge
barrier (MSSB) in Rotterdam Waterway (Fig. 8.1), which protects the low areas
behind it against ﬂooding. The MSSB must be closed during extreme storm condi
tions. In that case a semiclosed basin is created on its seaside. Seiches are expected
to occur in this temporary basin, with a dominant eigenperiod of approximately
30 min. These contribute to the failure probability of the barrier, but their expected
heights were not known.
Third, the Rotterdam Port Management needs to guarantee a minimum depth
(“nautical depth”) in the approach channel and throughout the harbor for deep
draught vessels. To this end, water levels are predicted operationally including tides
and windinduced setup. Until recently, no method was available for inclusion of
seiche events in this shortrange prediction, despite their practical importance: the
trough of a seiche could cause ships to run aground in case of a narrow predicted
tidal window.
The aforementioned practical applications in which seiches play a role have led to
a recent study on the origin of the seiche events in Rotterdam. The present chapter
summarizes the ﬁndings. Results are also presented of an analysis of the response
of Rotterdam Harbor basins to incoming wave energy in the seiche frequency band
using measured and theoretical ratios of seiche amplitudes at the closed ends of
two basins. Finally, a new system of routine prediction of seiche events is described,
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
Generation and Prediction of Seiches 181
based on the newly identiﬁed generation mechanism, as well as an indication of its
forecasting skill. The abovementioned research and results have been more fully
described by de Jong and Battjes.
1,2
8.2. Measurements
8.2.1. Data acquisition
The sea surface elevation measurements used in this study were obtained at three
oﬀshore platforms located approximately 20–40km from the harbor mouth: Euro
platform (EUR), Lichteiland Goeree (GOE), and Meetpost Noordwijk (MPN).
Inside the harbor, surface elevation measurements are available from multiple
locations (more than 10 keylocations are monitored). These have been taken con
tinuously since 1995 with ﬂoaters, sampled at a 60 s interval.
In addition, sea surface temperature measurements have been obtained from
1995 to 2001 at platforms located several hundreds of kilometers to the north and
northwest of the harbor mouth in the North Sea (along the path of the relevant
storm systems approaching the Port of Rotterdam). These were available as 24h
averaged daily values. Vertical proﬁles of air temperature and other meteorological
parameters, obtained from balloon measurements, were available at the North Sea
at Ekoﬁsk platform every 12 or 24 h.
8.2.2. Data analysis
To identify seiche events, records of the surface elevations at ROZ were Fourier
bandpassed ﬁltered, passing the seiche frequency band (0.1–2.0 mHz). A wavelet
analysis based on the Morlet wavelet
3−6
has been applied to the ﬁltered surface
elevation data and meteorological data. This technique is suited for the detection of
ﬂuctuations that come in bursts. It has been used to identify the seiche events inside
the harbor and the corresponding periods of increased levels of lowfrequency energy
at sea. Following ﬁrst identiﬁcation from the wavelet spectra, the time intervals
with seiche events were reviewed in more detail.
8.3. Origin of the Seiches in Rotterdam Harbor
8.3.1. Generation of seiche events by atmospheric
convection cells
In this section, a short description is given of the mechanism generating the long
waves at sea that cause seiche episodes inside the harbor (as mentioned above, a
more detailed description can be found in the work of de Jong and Battjes
1,2
).
Meteorological measurements and visual inspection of weather charts show that
all seiche episodes in Rotterdam Harbor (arbitrarily deﬁned as a seicheevent in
which the crest height exceeded 0.25 m at ROZ, a criterion that was applied in
several previous studies of seiches at this location) coincided with lowpressure
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
182 M. P. C. de Jong and J. A. Battjes
systems with cold fronts crossing the southern North Sea toward the Dutch coast.
However, the converse does not hold: not all such cold front passages give rise to
seiche events.
Analysis of measurements and numerical simulations showed that the seiches in
Rotterdam Harbor result from resonant ampliﬁcation of oscillations at sea, which in
turn are generated by ﬂuctuations in atmospheric pressure and wind speed following
a moving cold front. For the large majority of the seiche events, these ﬂuctuations
are caused by atmospheric convection cells that can arise in the area behind a cold
front. These cells are a mesoscale atmospheric phenomenon, with typical horizontal
spatial scales of 30–100km.
7
Conditions conducive to the formation of convection cells occur when cold
(polar) air moves over relatively warm sea water, which is likely to occur in fall
and winter (also the seasons of signiﬁcant seiche episodes). Especially in the area
behind the cold front, the relatively warm sea water causes heating of the air in the
lower layers, which results in an unstable lower atmosphere that causes air to rise
to higher altitudes in narrow plumes, (self)organized into a more or less regular
pattern of convection cells.
If suﬃcient moisture is present in the air, condensation of water vapor can
locally cause a release of latent heat, further driving the circulation of air. Narrow
bands of clouds can then be formed surrounding clear skies, which can be seen
on satellite images (an example is shown in Fig. 8.2) and, in case of precipitation,
Fig. 8.2. Satellite image of infrared frequency band (10:12 GMT, 11 January 1995), taken during
a seiche event in the Port of Rotterdam. The image shows a pattern of convection cells that passes
the southern North Sea from the northwest. The location of the Port of Rotterdam is indicated
by the black and white dot. Source image: Dundee Satellite Receiving Station, Dundee University,
Scotland.
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
Generation and Prediction of Seiches 183
cold air
cloud formation
2
−
4
k
m
30−100 km
warm air
cold front
advection speed
Fig. 8.3. Sketch of convection cells in the area behind a cold front together with a theoretical
surface elevation response.
also on radar precipitation images. These satellite and precipitation images showed
that the episodes of convection cells passing the southern North Sea more or less
coincided with the seiche episodes. The convection cells cause more or less spatially
periodic ﬂuctuations in atmospheric pressure and wind speed, which in a ﬁxed point
will manifest themselves as temporal ﬂuctuations as the cells are advected with the
weather system.
Because the system of convection cells moves, these ﬂuctuations can generate
lowfrequency waves at sea, which in turn can cause seiches in the harbor basins.
A sketch of the sea surface elevation response to the wind speed changes caused by
the convection cells is depicted in Fig. 8.3.
A 1D model based on the linearized shallow water equations was applied for a
transect across the North Sea along the main path of the convection cells, forced by
sinusoidal variations in atmospheric pressure and wind speed translating at a con
stant speed, with values that are typical for those observed. Negligibly low responses
were found in the more remote, deeper parts of the North Sea. However, the model
showed that the ﬂuctuations in wind speed and atmospheric pressure caused by
the passing convection cells are particularly eﬃcient at generating long waves with
signiﬁcant amplitudes in the relatively shallow part of the North Sea, the south
ernmost 400 km near the Dutch coast. This is because in this region, with typical
depths of approximately 30 m, the phase speed of shallowwater waves is approx
imately 17 m/s, i.e., near the advection velocity of the system of cold front and
convection cells (typically 15–20 m/s). The result is a nearresonant response of the
long waves at sea to the moving pattern of atmospheric convection cells, with ampli
tudes up to 1–2 dm near the Dutch coast. These waves can in turn be ampliﬁed in
semiclosed harbor basins through (a quarterwave) resonance, occasionally forming
seiches with observed amplitudes (crest heights) up to a meter.
8.3.2. Energy density spectrum of surface elevation at sea
Most previous studies of the seiches in the Port of Rotterdam focused on the rel
ative responses in diﬀerent basins. To this end, and for lack of better information,
those studies assumed a uniform incident energy density spectrum (a “white noise”
spectrum). Veraart
8
ﬁrst studied the spectra of measured surface elevations at the
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
184 M. P. C. de Jong and J. A. Battjes
10
−1
10
0
10
1
10
2
10
3
10
4
Frequency (mHz)
v
a
r
i
a
n
c
e
d
e
n
s
i
t
y
(
c
m
2
/
H
z
)
n
o
r
m
a
l
i
z
e
d
b
y
m
0
(
c
m
2
)
23/24 February 2001 (m0 = 1.5 cm
2
)
30 October 2000 (m0 = 3.4 cm
2
)
21 March 2001 (m0 = 0.40 cm
2
)
f
−1.5
trend
Fig. 8.4. Three normalized variance density spectra of surface elevation measurements at the
EUR platform at sea. Nominal cutoﬀ frequencies are indicated by the thick dashed vertical lines
at 0.1 mHz and 2.0 mHz. The thick line represents the bestﬁt power law of f
−1.5
.
southern North Sea in the seiche frequency band, obtained during seiche events
at ROZ. Based on a relatively small number of seiche episodes he found that, on
average, the variance density spectrum in the seiche frequency band approximately
followed a f
−1.2
trend. From June 2000, numerous processed and validated surface
elevation data have become available for a number of storm events. These have
now been used to determine the spectrum in the seiche frequency band at sea in
more detail (results described here are based on measurements from the platform
EUR). The variance density spectra have been estimated from ﬁltered records (0.1–
2.0 mHz) of 4096min (2
12
values) approximately centered around the seicheevent,
covering the complete storm duration of each event. Spectra of a number of events
normalized by the energy content of the seiche frequency band are shown in Fig. 8.4.
A leastsquares ﬁt of a power law to the averaged normalized spectral densities
results in a f
−1.5
trend.
8.4. Seiche Amplitudes at Diﬀerent Locations
8.4.1. Seiche events in diﬀerent basins of the Port of Rotterdam
A total of 49 seiche events were registered at ROZ in the studied time interval
(1995–2001). For this location, each such event is deﬁned as a period of enhanced
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
Generation and Prediction of Seiches 185
seiche activity with amplitudes (crest heights) exceeding 0.25 m. During most of
these events, the response at the lowest eigenperiod of the Caland Canal (90 min)
was dominant.
Surface elevation measurements at Europe Harbor (EH, see Fig. 8.1) have also
been studied for the same time interval as the observations at ROZ (1995–2001).
These time series were ﬁltered for the frequency band 0.3–2.0 mHz in order to isolate
the main eigenfrequencies of this (shorter) harbor basin. Based on a wavelet analysis,
and using 0.1 m as an amplitude threshold, 46 seiche events at EH were identiﬁed,
all at the dominant eigenperiod of 50 min.
Measurement data at ROZ were available during 43 of the 46 events that were
identiﬁed at EH. Of these 43 events at EH, 86% coincided with a seicheevent at
ROZ. This percentage depends to some extent on the thresholds for the identiﬁ
cation of a seicheevent. These have been chosen subjectively: in case a threshold
of 0.15 m is chosen for the identiﬁcation of seiche events at EH, 100% of the events
at EH coincide with a seicheevent at ROZ (with amplitude exceeding 0.25 m).
Conversely, of the 49 events at ROZ, 82% coincided with a seicheevent at EH.
In this case too, this result depends on the threshold of the deﬁnition of a seiche
event. In case a threshold of 0.40 m is applied at ROZ, 100% of the events at ROZ
coincided with an event at EH (with a threshold of 0.10 m).
The fact that largeamplitude seiche events simultaneously occur in basins of
diﬀerent dimensions indicates that the same mechanism generates seiches in basins
of diﬀerent natural frequencies.
8.4.2. Empirical amplitude ratio
Surface elevation signals from ROZ and EH have been ﬁltered for the complete
seiche frequency band (0.1–2.0mHz) in order to compare the crest heights during
coinciding seiche events in the Caland Canal and the Europe Harbor. For this
purpose, a typical local seiche amplitude could be based on, for example, the
standard deviation of the ﬁltered surface elevation signals. However, the time
span for which this value needs to be determined (for each event separately) is
not obvious. This is problematic in view of the transient nature of the seiche
events. Therefore, the maximum crest height of an event at each location has been
chosen. Moreover, in order to get a more robust estimate, the average of the ﬁve
largest crest heights and the ﬁve largest trough depths of an event has also been
used. The latter parameter is taken as a typical average seiche amplitude of an
event.
Figure 8.5 shows these average seiche amplitudes for the 30 coinciding events in
1995–2001 for ROZ and EH for which the average amplitude could be determined
from at least ﬁve crests and ﬁve troughs. These data have a correlation coeﬃcient
of 0.92. A linear trend line was leastsquares ﬁtted to the data, indicating that on
average the seiche amplitudes at EH are a factor of approximately 0.56 smaller
than those at ROZ. A scatter plot based on the single maximum crest height for all
coinciding events (not shown) gave a similar result, though with some more scatter
(correlation coeﬃcient 0.7).
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
186 M. P. C. de Jong and J. A. Battjes
0 10 20 30 40 50 60 70
0
10
20
30
40
average seiche amplitude ROZ (cm)
a
v
e
r
a
g
e
s
e
i
c
h
e
a
m
p
l
i
t
u
d
e
E
H
(
c
m
)
y=0.5596x
Fig. 8.5. Scatter plot of average seiche amplitudes (the average of the ﬁve highest crests and ﬁve
deepest troughs of an event) of coinciding seiche events at ROZ and Europe Harbor. Trend line
based on leastsquares estimate.
8.4.3. Theoretical amplitude ratio
The calculation of theoretical seiche amplitude ratios (r) at two locations a and
b is based on a combination of the energy supply at the harbor mouth and its
ampliﬁcation according to
r
ab
=
S(f)R
2
a
(f)df
S(f)R
2
b
(f)df
, (8.1)
where S(f) is the (normalized) energy spectrum incident from the sea at the harbor
mouth and R
a
(f) and R
b
(f) are the amplitude ampliﬁcation spectra for the con
sidered locations a and b. These had been determined previously using a 2DH mild
slope numerical wave model (PHAROS).
9
Eﬀects of bottom stress, inﬂow losses,
and seaward radiation were included. Applying these response functions and the
empirical seiche energy spectrum with a f
−1.5
tail, the theoretical amplitude ratio
between EH and ROZ is found to be 0.57, nearly equal to the value of 0.56 that was
estimated from the measurements. This result, based on 30 seiche events, supports
the validity of the theoretical approach, and it conﬁrms that the energy density at
sea in the seiche frequency band typically follows a f
−1.5
trend, not just in the three
events from which it was estimated.
8.5. Prediction of Occurrence of Seiche Events
8.5.1. Principle of prediction and ﬁrst veriﬁcation
As noted in the Introduction, practical needs related to shipping or to the movable
storm surge barrier require realtime knowledge of expected occurrences of seiches
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
Generation and Prediction of Seiches 187
in Rotterdam Harbor. However, operational prediction of these cannot be based on
measurements at available platforms at the North Sea, because the response at the
remote stations is too low, whereas the nearby platforms are too close to allow a
useful prediction window.
Since moving convection cells are at the basis of the generating mechanism of
seiche events in Rotterdam Harbor, circumstances needed for these cells to arise
can, in principle, be used to predict the occurrence, and also the absence, of seiche
events. The feasibility of this approach was shown by de Jong and Battjes
2
using
historical data. Based on this promising academic result, a pilot study has been
set up using realtime data and weather predictions, as a precursor of a fully oper
ational system. This conﬁrmed that the method was indeed feasible and that it
had suﬃcient forecasting skill to warrant implementation of an operational version.
Such a system, in which predictions with windows of 12 h and 24 h are provided to
relevant parties throughout the harbor, has now been in operation since 2005 and
so far has functioned to satisfaction.
The basis for the prediction system is as follows. Convection cells of suﬃcient
strength arise only if the lower atmosphere is (suﬃciently) unstable. The criterion
for this is that the adiabatic lapse rate (rate of decrease in temperature from the sea
surface upward) exceeds some threshold, the value of which depends on the moisture
content because of its consequence for possible cloud formation. In case the lapse
rate is smaller than 5.5 to 7.5
◦
C/km, the lower atmosphere is stable; here 6.5 C/km
is used as a representative critical value. Between approximately 10
◦
C/km and (say)
6.5
◦
C/km, it is conditionally unstable (depending on the moisture content), and a
temperature lapse rate larger than 10
◦
C/km results in an unstable atmosphere.
10
Prediction of the lapse rate and moisture content using multilayered atmospheric
models is part of routine weather forecast systems. This, therefore, also allows oper
ational prediction of the occurrence of signiﬁcant seiche events on a routine basis,
just like common weather prediction.
To test the feasibility of this approach, several months of atmospheric temper
ature and pressure data as well as sea surface water temperature data have been
used for “predictions” in a hindcast mode, comparing the outcomes to observations
of occurrences (or not) of seiche events. Results of a threemonth period are shown
here as a typical example of the outcome.
The lapse rate can in practice be estimated using the air temperature at sea
level (provisionally taken equal to the sea surface water temperature) and at some
higher elevation. In meteorology, standard reference levels in the boundary layer
are used corresponding to an atmospheric pressure of 925hPa (average altitude of
about 700 m) and 850 hPa (average altitude of about 1500 m). At the latter altitude,
cloud formation can occur, which would result in the release of latent heat, implying
inﬂuence of moisture content on the possible generation of convection cells. Indeed,
preliminary tests showed that correct predictions as well as false predictions are
found if the average altitude of the reference pressure of 850 hPa is used and moisture
is not taken into account.
However, generally speaking, the air will not become saturated below the level
of the 925 hPa pressure. Therefore, cloud formation will not occur below this level.
In this case, no information about moisture content of the air is required because
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
188 M. P. C. de Jong and J. A. Battjes
the uncertain range of conditional stability is excluded. A lower limit of the lapse
rate for unconditional instability of 10
◦
C/km then applies. The following “predic
tions” use this criterion and the estimated temperature gradient between sea level
and the observed level of 925 hPa pressure. For comparison, results based on the
average elevation of this pressure level (approximately 700 m) are shown as well.
The latter would be simpler operationally, albeit at the cost of enhanced errors in
the veriﬁcation of the temperature diﬀerence criterion.
The results of the aforementioned approach are illustrated by presenting the
data of January till March 1999. This time interval was selected because the quality
of the data was relatively high during these months (only a small percentage of
missing data due to, e.g., atmospheric soundings that were not available on certain
days). Weather maps have been used to identify the days during which cold fronts
have passed the North Sea that resulted in a ﬂow toward the Dutch coast (including
possible successive days during which this ﬂow direction continued). This is obvi
ously needed for seiche events to be generated in Rotterdam Harbor.
Figure 8.6 shows the results. The bars in the top panel show the observed dif
ference between the air temperature at the (observed) altitude of the reference
pressure 925 hPa (at platform EKO) and the sea surface water temperature (at
platform AUK) for each day (not restricted to days of a front passage). Gaps are
present for those days for which no data were available. The lower panel shows the
5 10 15 20 25 30 5 10 15 20 25 5 10 15 20 25 30
0
5
10
15
t
e
m
p
.
d
i
f
f
e
r
e
n
c
e
(
°
C
)
unstable
stable
5 10 15 20 25 30 5 10 15 20 25 5 10 15 20 25 30
0
5
10
15
t
e
m
p
.
d
i
f
f
e
r
e
n
c
e
(
°
C
)
stable
unstable
Jan Feb Mar
Jan Feb Mar
Fig. 8.6. The bars indicate the observed temperature diﬀerence between the surface sea water
at AUK and the air at 925 hPa level at Ekoﬁsk for January till March 1999. Top panel: all days;
bottom panel: days of occurrence of a cold front that moved from the North Sea toward the Dutch
coast. Straight dashed lines: critical temperature diﬀerence for an unstable atmosphere based on
average elevation of 925 hPa. Wiggly curves: values of the critical temperature diﬀerence for an
unstable atmosphere from day to day, based on an observed elevation of 925 hPa.
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
Generation and Prediction of Seiches 189
same data, now restricted to days during which cold fronts have passed the North
Sea toward the Dutch coast.
The straight dashed lines in both panels indicate the critical temperature dif
ference for an unstable atmosphere based on the average altitude of the pressure
level of 925hPa (∆T = 7
◦
C). The wiggly curves in both panels indicate the value
of the critical temperature diﬀerence for an unstable atmosphere from day to day,
based on the temperature at the measured altitude of the reference pressure of
925 hPa for each day.
A seicheevent is predicted for those days in the lower panel for which the
observed temperature diﬀerence (the bar) exceeds the critical one (indicated by the
wiggly curve). The asterisks indicate the days for which seiche events have been
identiﬁed at ROZ. (Consecutive days with seiche events indicated represent a single
event that started during the ﬁrst day and ended on the last day indicated.)
The bottom panel shows that an unstable atmosphere is found during a number
of potentially seicheprone coldfront situations, correctly predicting a seiche
episode. All the seiche events that occurred in this time interval are correctly pre
dicted (no misses). A stable atmosphere is found during most other days indicated in
the bottom panel and no convection cells are expected during these days (regardless
of the availability of moisture in the air), which agrees with the fact that no seiche
events occurred.
Nevertheless, a limited number of false predictions is also found (17 January and
28 February). On these days, the temperature diﬀerence just slightly exceeded the
critical value. The resulting false predictions could therefore easily be due to the
criterion not being strict enough, or to the spatial variability of the temperature
of the sea surface water and the air at the reference levels: although the nominal
criterion is met locally, this does not necessarily have to be representative for the
complete southern North Sea region since only observations from sparse locations
are used. This will not be the case during the correctly predicted seiche events
in view of the fact that the prediction on these days is based on relatively large
temperature diﬀerences, which far exceed the critical values.
8.5.2. Operational seiche prediction
The evidence presented above indicates that conditions for the existence of con
vection cells, together with the information regarding the passage and direction
of cold fronts, can be used to identify and predict seicheprone situations. In an
operational setting, the predicted weather chart for the next 24 h is available with
relatively high reliability, and the predicted air temperature can be obtained from
numerical weather prediction models. The moisture content of the air could also
be taken into account. Moreover, a more thorough analysis of the stability of the
lower atmosphere is possible, since information of the complete vertical and for
the complete southern North Sea area is available (possibly avoiding false alarms
due to criteria that are only met locally). This makes the use of speciﬁc reference
levels unnecessary. Therefore, all relevant parameters are available for the prediction
of seicheprone situations with a 24h prediction window, suﬃcient for the port
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
190 M. P. C. de Jong and J. A. Battjes
authorities and for the closure management of the movable storm surge barrier of
Rotterdam (which typically require prediction windows of, say, 6 h).
Based on the aforementioned ﬁndings, a pilot project was implemented, set up
by the Ministry of Public Works and Waterways in cooperation with the Royal
Dutch Meteorological Institute (KNMI) for operational prediction of the occurrence
of seiche events in the Port of Rotterdam, using parameter values derived from oper
ational numerical weather prediction models. This pilot project has shown that the
system has suﬃcient predictive skill for operational use. Such a system has since
been implemented on an operational basis. Experience shows that all major seiche
events (above the seicheamplitude threshold) have been successfully predicted.
However, in some cases a false prediction was made in the sense that a warning was
issued while no signiﬁcant seiches occurred. This is a drawback that needs repair,
but even as it is, the system has merit because it does not miss signiﬁcant events.
The false alarms mentioned above are related to an important limitation of the
abovedescribed method, viz, the fact that so far it only predicts seicheprone sit
uations but not the expected energy level of the seicheevent. In case additional
parameters are taken into account, such as the (range in) sizes of the atmospheric
convection cells, the advection velocity, the excess of the temperature diﬀerence
above the critical value (not just the question whether it exceeds a critical value
or not), and the direction of movement of the weather systems, prediction of the
seiche energy level might become possible, for example, through the use of a neural
network that can be trained with a set of observed values of these quantiﬁable
parameters. This may open the possibility of operational forecasting of seicheprone
conditions together with expected seiche crest heights. In that case, a seiche pre
diction (warning) would only be issued when seicheconducive atmospheric condi
tions are expected that can lead to a seicheevent with amplitudes which justify a
warning.
8.6. Conclusions
Fluctuations in atmospheric pressure and wind speed, induced by mesoscale atmo
spheric convection cells that can arise following a cold front passing southward
over the North Sea, generate lowfrequency gravity waves in the North Sea. These
lowfrequency waves can subsequently induce a resonant response in coastal harbor
basins, thus causing the socalled seiche events.
During episodes of enhanced seiche activity in Rotterdam Harbor, the surface
elevation energy spectrum at sea in the frequency band corresponding to the typical
eigenfrequencies in Rotterdam Harbor basins, say, 0.1–2.0mHz, varies with fre
quency (f) approximately as f
−1.5
. An episode of enhanced wave energy at sea in
the seiche frequency band causes simultaneous resonant seiche responses in basins
with diﬀerent natural frequencies. The ratio between seiche amplitudes in two dif
ferent basins can be accurately predicted using an f
−1.5
excitation spectrum and
numerically calculated spectral transfer functions.
Operational prediction of the occurrence of a seicheepisode cannot be based on
measurements at available platforms at sea, because at the more remote stations
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
Generation and Prediction of Seiches 191
the signal is too weak, and for the nearby stations the available prediction window
is too short to be of practical use.
A hindcast study has shown that the occurrence of signiﬁcant seiche episodes
in Rotterdam Harbor can be accurately predicted on the basis of the prediction of
the occurrence of meteorological conditions conducive to the formation of mesoscale
atmospheric convection cells behind a cold front. This is presently realized using
output of air temperatures and moisture content from numerical atmospheric
models that are used in operational weather forecasting. The outcomes of this pre
diction method are used by the port authorities for ship traﬃc control and, after
further development and testing, can eventually be incorporated into the closure
management system of the storm surge barrier in Rotterdam Waterway.
Acknowledgments
The ﬁrst author acknowledges the ﬁnancial support from the Dr Ir Cornelis Lely
Foundation. A part of this study was ﬁnanced by the DirectorateGeneral for Public
Works and Water Management, SouthHolland Directorate (Contract number: ZHA
23735). Furthermore, the authors thank the DirectorateGeneral for Public Works
and Water Management and the Municipal Harbor Organization, Rotterdam for
providing the surface elevation measurements at sea and in the harbor, respectively.
The authors thank Dr A. J. van Delden of Utrecht University, The Netherlands,
for his extensive information on convection cells. Atmospheric sounding data
have been obtained from the Department of Atmospheric Science, University of
Wyoming, USA. Wavelet software was provided by Torrence and Compo,
6
through
URL: http://paos.colorado.edu/research/wavelets/. The ﬁgures in this chapter are
reprinted from de Jong and Battjes,
2
with permission from Elsevier.
References
1. M. P. C. de Jong and J. A. Battjes, Lowfrequency sea waves generated by
atmospheric convection cells, J. Geophys. Res. 109(C1), C01011 (2004), doi:
10.1029/2003JC001931.
2. M. P. C. de Jong and J. A. Battjes, Seiche characteristics of Rotterdam Harbour,
Coastal Eng. 51(5–6), 373–386 (2004).
3. J. Morlet, G. Arens, E. Furgeau and D. Giard, Wave propagation and sampling
theory — Part 2: Sampling theory and complex waves, Geophysics 47(2), 222–236
(1982).
4. M. Farge, Wavelet transforms and their applications to turbulence, Ann. Rev. Fluid
Mech. 24, 395–457 (1992).
5. S. Mallat, A Wavelet Tour of Signal Processing (Academic Press, San Diego, 1998).
6. C. Torrence and G. P. Compo, A practical guide to wavelet analysis, Bull. Am.
Meteorol. Soc. 79, 61–78 (1998).
7. B. W. Atkinson and J. W. Zhang, Mesoscale shallow convection in the atmosphere,
Rev. Geophys. 34(4), 403–431 (1996).
July 31, 2009 8:18 9.75in x 6.5in b684ch08 FA
192 M. P. C. de Jong and J. A. Battjes
8. C. Veraart, Events of Long Waves in the North Sea, Technical Report RIKZ94.034,
Rijkswaterstaat/RIKZ and Delft University of Technology, Delft, The Netherlands
(1994).
9. P. Van den Bosch and J. J. Veldman, EuropoortkeringBeerdam, Technical Report
H2438, WL Delft Hydraulics, Delft, The Netherlands (1995).
10. J. R. Holton, An Introduction to Dynamic Meteorology, 3rd edn. (Academic Press,
London, 1992).
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Chapter 9
Seiches and Harbor Oscillations
Alexander B. Rabinovich
P.P. Shirshov Institute of Oceanology, Russian Academy of Sciences
36 Nakhimovsky Prosp., Moscow, 117997 Russia
Department of Fisheries and Oceans, Institute of Ocean Sciences
9860 West Saanich Road, Sidney, BC, Canada V8L 4B2
a.b.rabinovich@gmail.com
alexander.rabinovich@dfompo.gc.ca
This chapter presents an overview of seiches and harbor oscillations. Seiches are
longperiod standing oscillations in an enclosed basin or in a locally isolated part
of a basin. They have physical characteristics similar to the vibrations of a guitar
string or an elastic membrane. The resonant (eigen) periods of seiches are deter
mined by basin geometry and depth and in natural basins may range from tens of
seconds to several hours. The set of seiche eigen frequencies (periods) and asso
ciated modal structures are a fundamental property of a particular basin and
are independent of the external forcing mechanism. Harbor oscillations (coastal
seiches) are a speciﬁc type of seiche motion that occur in partially enclosed
basins (bays, fjords, inlets, and harbors) that are connected through one or more
openings to the sea. In contrast to seiches, which are generated by direct external
forcing (e.g., atmospheric pressure, wind, and seismic activity), harbor oscillations
are mainly generated by long waves entering through the open boundary (harbor
entrance) from the open sea. Energy losses of seiches in enclosed basins are mostly
due to dissipative processes, while the decay of harbor oscillations is primarily due
to radiation through the mouth of the harbor. An important property of harbor
oscillations is the Helmholtz mode (pumping mode), similar to the fundamental
tone of an acoustic resonator. This mode is absent in a closed basin.
Harbor oscillations can produce damaging surging (or range action) in some
ports and harbors yawing and swaying of ships at berth in a harbor. A property
of oscillations in harbors is that even relatively small vertical motions (sea level
oscillations) can be accompanied by large horizontal motions (harbor currents),
resulting in increased risk of damage of moored ships, breaking mooring lines
as well as aﬀecting various harbor procedures. Tsunamis constitute another
important problem: catastrophic destruction may occur when the frequencies of
arriving tsunami waves match the resonant frequencies of the harbor or bay.
Seiches, as natural resonant oscillations, are generated by a wide variety of
mechanisms, including tsunamis, seismic ground waves, internal ocean waves, and
193
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
194 A. B. Rabinovich
jetlike currents. However, the two most common factors initiating seiches are
atmospheric processes and nonlinear interaction of wind waves or swell. At certain
places in the World Ocean, waves due to atmospheric forcing (atmospheric gravity
waves, pressure jumps, frontal passages, squalls) can be responsible for signiﬁcant,
even devastating harbor oscillations, known as meteorological tsunamis. They have
the same temporal and spatial scales as typical tsunami waves and can aﬀect
coasts in a similar damaging way.
9.1. Introduction
Seiches are longperiod standing oscillations in an enclosed basin or in a locally
isolated part of a basin (in the Japanese literature they are commonly known as
“secondary oscillations (undulations) of tides”).
30,56,59
The term “seiches” appar
ently originated from the Latin word siccus which means dry or exposed (from
the exposure of the littoral zone at the downswing).
31,95
Freesurface oscillations,
known as seiches or seiching in lakes and harbors or as sloshing in coﬀee cups,
bathtubs, and storage tanks, have been observed since very early times; a vivid
description of seiching in Lake Constance, Switzerland, was given in 1549, and the
ﬁrst instrumental record of seiches obtained in 1730 in Lake Geneva.
46,95
Korgen
34
describes seiches as “the rhythmic, rocking motions that water bodies undergo after
they have been disturbed and then sway backandforth as gravity and friction grad
ually restore them to their original, undisturbed conditions.” These oscillations occur
at the natural resonant periods of the basin (socalled “eigen periods”) and phys
ically are similar to vibrations of a guitar string and an elastic membrane. The
resonant (eigen) periods of seiches are determined by the basin geometry and
depth
94,95
and in natural basins may be from a few tens of seconds to several
hours. The oscillations are known as natural (or eigen) modes. The mode with the
lowest frequency (and thus, the longest period) is referred to as the fundamental
mode.
41
The set of seiche eigen frequencies (periods) and associated modal structures are
a fundamental property of a particular basin and are independent of the external
mechanism forcing the oscillations. In contrast, the amplitudes of the generated
seiches strongly depend on the energy source that generates them, and can therefore
have pronounced variability.
31
Resonance occurs when the dominant frequencies of
the external forcing match the eigen frequencies of the basin.
Harbor oscillations (coastal seiches according to Ref. 19 are a speciﬁc type of
seiche motion that occur in partially enclosed basins (gulfs, bays, fjords, inlets,
ports, and harbors) that are connected through one or more openings to the sea.
41,94
Harbor oscillations diﬀer from seiches in closed water bodies (for example, in lakes)
in three principal ways
68
:
(1) In contrast to seiches generated by direct external forcing (e.g., atmospheric
pressure, wind, and seismic activity), harbor oscillations are mainly generated
by long waves entering through the open boundary (harbor entrance) from the
open sea.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 195
(2) Energy losses of seiches in closed basins are mostly associated with dissipation,
while the decay of harbor oscillations is mainly due to radiation through the
mouth of the harbor.
(3) Harbor oscillations have a speciﬁc fundamental mode, the Helmholtz mode,
similar to the fundamental tone of an acoustic resonator.
54
This modes is absent
in closed basins.
Because harbor oscillations can produce damaging surging (or range action) —
yaw and swaying of ships at berth in a harbor — this problem has been extensively
examined in the scientiﬁc and engineering literature.
4,11,12,36,41,46,47,62,67,68,70,78,79,
94,95
One of the essential properties of oscillations in harbors is that even rela
tively small vertical motions (sea level oscillations) can be accompanied by large
horizontal water motions (harbor currents); when the period of these motions coin
cides with the natural period of sway, or yaw of a moored ship, further resonance
occurs, which can result in considerable motion and possible damage of a moored
ship.
82,94
Harbor oscillations can also break mooring lines, cause costly delays in
loading and unloading operations at port facilities, and seriously aﬀect various
harbor procedures.
79,80
Tsunamis constitute another important problem that have greatly stimulated
investigations of harbor oscillations. Professor Omori (Japan) was likely the ﬁrst
to notice in 1902 that the dominant periods of observed tsunami waves are nor
mally identical to those caused by ordinary long waves in the same coastal basin
(see Ref. 30). His explanation was that the bay or portion of the sea oscil
lates like a ﬂuid pendulum with its own period, i.e., the arriving tsunami waves
generate similar seiches as those generated by atmospheric processes and other
types of external forcing (see also Ref. 30). Numerous papers on the spectral
analysis of tsunami records for various regions of the world ocean have conﬁrmed
this conclusion.
13,48,69,74,75,84,90
Catastrophic destruction may occur when the fre
quencies of arriving tsunami waves match the resonant frequencies of the harbor or
bay. One of the best examples of strong tsunami ampliﬁcation due is the resonant
response of Port Alberni (located at the head of long Alberni Inlet on the Paciﬁc
coast of Vancouver Island, Canada) to the 1964 Alaska tsunami.
28,54
9.2. Hydrodynamic Theory
The basic theory of seiche oscillations is similar to the theory of free and forced
oscillations of mechanical, electrical, and acoustical systems. The systems respond
to an external forcing by developing a restoring force that reestablishes equilibrium
in the system. A pendulum is a typical example of such a system. Free oscillations
occur at the natural frequency of the system if the system disturbed beyond its
equilibrium. Without additional forcing, these free oscillations retain the same fre
quencies but their amplitudes decay exponentially due to friction, until the system
eventually comes to rest. In the case of a periodic continuous forcing, forced oscilla
tions are produced with amplitudes depending on friction and the proximity of the
forcing frequency to the natural frequency of the system.
84
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
196 A. B. Rabinovich
9.2.1. Long and narrow channel
Standing wave heights in a closed, long and narrow nonrotating rectangular basin
of length, L, and uniform depth, H, have a simple trigonometric form
35,95
:
ζ(x; t) = Acos kxcos ωt, (9.1)
where ζ is the sea level elevation, A is the wave amplitude, x is the alongbasin
coordinate, t is time, k = 2π/λ is the wave number, λ is the wavelength, ω = 2π/T
is the angular wave frequency, and T is the wave period. The angular frequency
and wave number (or the period and wavelength) are linked through the following
wellknown relationships:
ω = kc, (9.2a)
T =
λ
c
, (9.2b)
where c =
√
gH is the longwave phase speed and g is the gravitational acceleration.
The condition of noﬂow through the basin boundaries (x = 0; x = L) yields
the wave numbers:
k =
π
L
,
2π
L
,
3π
L
, . . . ,
nπ
L
, (9.3)
which are related to the speciﬁc oscillation modes [Fig. 9.1(a)], i.e., to the various
eigen modes of the water basin. The fundamental (n = 1) mode has a wavelength
equal to twice the length of the basin; a basin oscillating in this manner is known
as a halfwave oscillator.
34
Other modes (overtones of the main or fundamental
n = 1
n = 2
n = 3
n = 4
n = 0
n = 1
n = 2
n = 3
Closed basin
Openended basin
0.00 0.00 1.00 1.00
L L
Fig. 9.1. Surface proﬁles for the ﬁrst four seiche modes in closed and openended rectangular
basins of uniform depth.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 197
Table 9.1. Normalized periods, T
∗
n
= T
n
√
gH/(2L), for a
closed and openmouth rectangular basin of uniform depth.
Mode
Basin n = 0 n = 1 n = 2 n = 3 n = 4
Closed — 1 1/2 1/3 1/4
Openmouthed 2 2/3 2/5 2/7 2/9
“tone”) have wavelengths equal to onehalf, onethird, onefourth, and so on, of the
wavelength of the fundamental mode (Fig. 9.1(a), Table 9.1).
The fundamental mode is antisymmetric: when one side of the water surface
is going up, the opposite side is going down. Maximum sea level oscillations are
observed near the basin borders (x = 0; x = L), while maximum currents occur at
the nodal lines, i.e., the lines where ζ = 0 for all time. Positions of the nodal lines
are determined by
x
m
n
=
(2m−1)L
2n
, m, n = 1, 2, . . . ; m ≤ n. (9.4)
Thus, for n = 1, there is one nodal line: x
1
1
= L/2 located in the middle of the
basin; for n = 2, there are two lines: x
1
2
= L/4 and x
2
2
= 3L/4; for n = 3: x
1
3
= L/6,
x
2
3
= 3L/6 = L/2 and x
3
3
= 5L/6 and so on. The number of nodal lines equals
the mode number n [Fig. 9.1(a)], which is why the ﬁrst mode is called the uninodal
mode, the second mode is called binodal mode, the third mode the trinodal mode,
etc.
31,95
The antinode positions are those for which ζ attains maximum values, and
are speciﬁed as
x
j
n
=
jL
n
, j = 0, 1, 2, . . . , n. (9.5)
For example, for n = 2 there are three antinodal lines: x
0
2
= 0, x
1
2
= L/2, and
x
2
2
= L. Maximum currents occur at the nodal lines, while minimum currents occur
at the antinodes. Water motions at the seiche nodes are entirely horizontal, while
at the antinodes they are entirely vertical.
The relationships (9.2) and (9.3) yield the wellknown Merian’s formula for the
periods of (natural) in a rectangular basin of uniform depth
68,78
:
T
n
=
2L
n
√
gH
, (9.6)
where n = 1, 2, 3, . . .. Merian’s formula (9.6) shows that the longer the basin length
(L) or the shallower the basin depth (H), the longer the seiche period. The fun
damental (n = 1) mode has the maximum period; other modes — the overtones
of the main fundamental — “tone” — have periods equal to onehalf, onethird,
onefourth, and so on, of the fundamental period (Fig. 9.1(a), Table 9.1). The fun
damental mode and all other odd modes are antisymmetric, while even modes are
symmetric; an antinode line is located in the middle of the basin.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
198 A. B. Rabinovich
The structures and parameters of openmouth basins are quite diﬀerent from
those of closed basins. Standing oscillations in a rectangular bay (harbor) with
uniform depth and open entrance also have the form (9.1) but with a nodal line
located near the entrance (bay mouth). In general, the approximate positions of
nodal lines are determined by the following expressions (Fig. 9.1(b), Table 9.1):
x
m
n
=
(2m+ 1)L
2n + 1
, m, n = 0, 1, 2, . . . ; m ≤ n, (9.7)
while antinodes are located at
x
j
n
=
2jL
2n + 1
, j, n = 0, 1, 2, . . . ; j ≤ n. (9.8)
In particular, for n = 1 there are two nodal lines: x
0
1
= L/3 and x
1
1
= L and two
antinodal lines: x
0
1
= 0 and x
1
1
= 2L/3; for n = 2 there are three nodal lines:
x
0
2
= L/5, x
1
2
= 3L/5 and x
2
2
= L, and three antinodal: x
0
2
= 0, x
1
2
= 2L/5, and
x
2
2
= 4L/5.
The most interesting and important mode is the lowest mode, for which n = 0.
This mode, known as the Helmholtz mode, has a single nodal line at the mouth of the
bay (x = L) and a single antinode on the opposite shore (x = L). The wavelength
of this mode is equal to four times the length of the bay; a basin oscillating in
this manner is known as a quarterwave oscillator.
34
The Helmholtz mode, which is
also called the zeroth mode,
a
the gravest mode and the pumping mode (because it is
related to periodic mass transport — pumping — through the open mouth,
36,41
is of
particular importance for any given harbor. For narrowmouthed bays and harbors,
as well as for narrow elongated inlets and fjords, this mode normally dominates.
The periods of the Helmholtz and other harbor modes can be approximately
estimated as
84,95
T
n
=
4L
(2n + 1)
√
gH
, for mode n = 0, 1, 2, 3, . . . . (9.9)
Using (9.9) and (9.6), the fundamental (Helmholtz) mode in a rectangular open
mouth basin of uniform depth H is found to have a period, T
0
= 4L/
√
gH, which is
double the period of the gravest mode in a similar but closed basin, T
1
= 2L/
√
gH.
Normalized periods of various modes (for n ≤ 4) are shown in Table 9.1.
Expressions (9.4)–(9.9), Table 9.1, and Fig. 9.1 are all related to the idealized
case of a simple rectangular basin of uniform depth. This model is useful for
some preliminary estimates of seiche parameters in closed and semiclosed natural
and artiﬁcial basins. Analytical solutions can be found for several other basins of
simple geometric form and nonuniform depth. Wilson
95
summarizes results that
involve common basin shapes (Tables 9.2 and 9.3), which in many cases are quite
a
In many papers and text books,
8,94,95
this mode is considered the “ﬁrst mode”. However, it is
more common to count nodal lines only inside the basin (not at the entrance) and to consider the
fundamental harbor mode as the “zeroth mode”.
41,68,78,80,84
This approach is physically more
sound because this mode is quite speciﬁc and markedly diﬀerent from the ﬁrst mode in a closed
basin.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 199
good approximations to rather irregular shapes of natural lakes, bays, inlets, and
harbors.
The main concern for port operations and ships and boats in harbors is not
from the sea level seiche variations but from the strong currents associated with the
seiche. As noted above, maximum horizontal current velocities occur at the nodal
lines. Therefore, its locations in the vicinity of the nodes that are potentially most
risky and unsafe. Maximum velocities, V
max
, can be roughly estimated as
84
:
V
max
= A
n
g
H
, (9.10)
where A
n
is the amplitude of the sea level oscillation for the mode. For example, if
A
n
= 0.5 m and H = 6 m, V
max
≈ 0.64 m/s.
9.2.2. Rectangular and circular basins
If a basin is not long and narrow, the 1D approach used above is not appropriate. For
such basins, 2D eﬀects may begin to play an important role, producing compound
or coupled seiches.
95
Two elementary examples, which can be used to illustrate the
2D structure of seiche motions, are provided by rectangular and circular basins of
uniform depth (H). Consider a rectangular basin with length L (x = 0, L) and
width l (y = 0, l). Standing oscillations in the basin have the form
35,41
ζ(x, y; t) = A
mn
cos
mπx
L
cos
nπy
l
cos ωt, (9.11)
where m, n = 0, 1, 2, 3, . . .. The eigen wave numbers (k
mn
) are
k
mn
=
¸
mπ
L
2
+
nπ
l
2
1/2
, (9.12)
and the corresponding eigen periods are
78
T
mn
=
2
√
gh
¸
m
L
2
+
n
l
2
−1/2
. (9.13)
For n = 0 expression (9.13) becomes equivalent to the Merian’s formula (9.6); the
longest period corresponds to the fundamental mode (m = 1, n = 0) which has one
nodal line in the middle of the basin. In general, the numbers m and n denote the
number of nodal lines across and along the basin, respectively. The normalized eigen
periods T
∗
mn
= T
mn
/T
10
and spatial structure for the diﬀerent modes are shown in
Table 9.4.
For oscillations in a circular basin of radius r = a, it is convenient to use a polar
coordinate system (r, θ) with the origin in the center:
x = r cos θ, y = r sin θ,
where θ is the polar angle. Standing oscillations in such basins have the form
ζ(x, y; t) = J
s
(kr)(A
s
cos sθ +B
s
sinsθ) cos ωt, (9.14)
J
u
l
y
3
1
,
2
0
0
9
8
:
1
8
9
.
7
5
i
n
x
6
.
5
i
n
b
6
8
4

c
h
0
9
F
A
2
0
0
A
.
B
.
R
a
b
i
n
o
v
i
c
h
Table 9.2. Modes of free oscillations in closed basins of simple geometric shape and constant width (after Ref. 95).
Periods of free oscillation
Basin type Mode ratios T
n
/T
1
values for n
Description Dimensions Proﬁle equation Fundamental T
1
1 2 3 4
Rectangular
h h x ( )
0
0
x
L
h(x) = h
0
2L/(gh
0
)
1/2
1.000 0.500 0.333 0.250
Triangular
(isosceles)
h
h x ( )
0
0
2
∠
x
L
h(x) = h
0
(1 −2x/L) 1.305[2L/(gh
0
)
1/2
] 1.000 0.628 0.436 0.343
Parabolic
h
h x ( )
0
0
2
∠
x
L
h(x) = h
0
(1 −4x
2
/L
2
) 1.110[2L/(gh
0
)
1/2
] 1.000 0.577 0.408 0.316
Quartic
h
h x ( )
0
0
2
∠
x
L
h(x) = h
0
(1 −4x
2
/L
2
)
2
1.242[2L/(gh
0
)
1/2
] 1.000 0.686 0.500 0.388
(Continued)
J
u
l
y
3
1
,
2
0
0
9
8
:
1
8
9
.
7
5
i
n
x
6
.
5
i
n
b
6
8
4

c
h
0
9
F
A
S
e
i
c
h
e
s
a
n
d
H
a
r
b
o
r
O
s
c
i
l
l
a
t
i
o
n
s
2
0
1
Table 9.2. (Continued)
Periods of free oscillation
Basin type Mode ratios T
n
/T
1
values for n
Description Dimensions Proﬁle equation Fundamental T
1
1 2 3 4
Triangular
(rightangled)
h
h x ( )
1
0
x
L
h(x) = h
1
x/L 1.640[2L/(gh
1
)
1/2
] 1.000 0.546 0.377 0.288
Trapezoidal
h
h
h x ( ) 0
1
0
x
L
h(x) = h
0
+ mx
m = (h
1
−h
0
)/L
1.000 0.546 0.377 0.288
Coupled,
rectangular
h
h
h x ( )
1
2
2 1
0
x
L L
h(x) = h
1
(x < 0)
h(x) = h
2
(x < 0)
L
1
/L
2
= 1/2 4L
2
/(gh
2
)
1/2
1.000 0.500 0.250 0.125
h
1
/h
2
= 1/4 L
1
/L
2
= 1/3 3.13L
2
/(gh
2
)
1/2
1.000 0.559 0.344 0.217
L
1
/L
2
= 1/4 2.73L
2
/(gh
2
)
1/2
1.000 0.579 0.367 0.252
L
1
/L
2
= 1/8 2.31L
2
/(gh
2
)
1/2
1.000 0.525 0.371 0.279
J
u
l
y
3
1
,
2
0
0
9
8
:
1
8
9
.
7
5
i
n
x
6
.
5
i
n
b
6
8
4

c
h
0
9
F
A
2
0
2
A
.
B
.
R
a
b
i
n
o
v
i
c
h
Table 9.3. Modes of free oscillations in semiclosed basins of simple geometric shape (modiﬁed after Ref. 95).
Periods of free oscillation
Basin type Mode ratios T
s
/T
1
[n = (s + 1)/2]
Description Dimensions Proﬁle equation Fundamental T
0
n = 0 1 2 3
Rectangular
b x ( ) b
1
0 x
L
Rectangular
h x ( ) h
1
0
x
L
h(x) = h
1
2.000[2L/(gh
1
)
1/2
] 1.000 0.333 0.200 0.143
Rectangular
b x ( ) b
1
0 x
L
Triangular
h x ( )
h
1
0
x
L
h(x) = h
1
x/L 2.618[2L/(gh
1
)
1/2
] 1.000 0.435 0.278 0.203
Rectangular
b x ( ) b
1
0 x
L
Semiparabolic
h x ( )
h
1
0
x
L
h(x) = h
1
(1 −x
2
/L
2
) 2.220[2L/(gh
1
)
1/2
] 1.000 0.409 0.259 0.189
Triangular
b x ( )
b
1
0 x
L
Rectangular
h x ( ) h
1
0
x
L
b(x) = b
1
x/L 1.308[2L/(gh
1
)
1/2
] 1.000 0.435 0.278 0.230
h(x) = h
1
(Continued)
J
u
l
y
3
1
,
2
0
0
9
8
:
1
8
9
.
7
5
i
n
x
6
.
5
i
n
b
6
8
4

c
h
0
9
F
A
S
e
i
c
h
e
s
a
n
d
H
a
r
b
o
r
O
s
c
i
l
l
a
t
i
o
n
s
2
0
3
Table 9.3. (Continued)
Periods of free oscillation
Basin type Mode ratios T
s
/T
1
[n = (s + 1)/2]
Description Dimensions Proﬁle equation Fundamental T
0
n = 0 1 2 3
Triangular
b x ( )
b
1
0 x
L
Triangular
h x ( )
h
1
0
x
L
b(x) = b
1
x/L 1.653[2L/(gh
1
)
1/2
] 1.000 0.541 0.374 0.283
h(x) = h
1
x/L
Semielliptic
b
1
0 x
L
Semiparaboloidal
h
1
0
x
L
b
1
/L = 2 2.220[2L/(gh
1
)
1/2
] 1.000 0.707 0.578 0.378
= 4/3 0.554 0.493 0.323
= 1 0.447 0.468 0.264
= 2/3 0.317 0.455 0.185
Semicircular
b
1
0
L
h x ( )
Semiparaboloidal
h
1
0
r
r
L
h(x) = h
1
(1 −r
2
/L
2
) 2.220[2L/(gh
1
)
1/2
] 1.000 0.707 0.578 0.500
J
u
l
y
3
1
,
2
0
0
9
8
:
1
8
9
.
7
5
i
n
x
6
.
5
i
n
b
6
8
4

c
h
0
9
F
A
2
0
4
A
.
B
.
R
a
b
i
n
o
v
i
c
h
Table 9.4. Mode parameters for free oscillations in uniform depth basins of rectangular and circular geometric shape.
Rectangular basin (l = 0.5L) Circular basin
Mode Mode Circular Normalized Relative
numbers Mode forms Relative period numbers Mode forms nodal lines frequency period
m n L T
mn
/T
10
s m 2a r
1
r
2
ωa/c T
sm
/T
10
1 0
+
l
1.000 1 0
+
— — 1.841 1.000
2 0
+
0.500 2 0
+
+ — — 3.054 0.603
0 1
+
0.500 0 1
+
0.628 — 3.832 0.480
1 1
+
+
0.447 1 1
+ +
0.719 — 5.331 0.345
2 1
+
+ +
0.354 2 1
+
+
+
+
0.766 — 6.706 0.275
0 2
+
0.250 0 2
+ +
0.343 0.787 7.016 0.262
1 2
+
+
+
0.243 1 2
+ +
+
0.449 0.820 8.536 0.216
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 205
where J
s
is the Bessel function of an order s, A
s
and B
s
are arbitrary constants,
and s = 0, 1, 2, 3, . . .
35,41
These oscillations satisfy the boundary condition:
J
s
(kr)
r=a
= J
s
(ka) = 0. (9.15)
The roots of this equation determine the eigen values k
sm
(m, s = 0, 1, 2, 3, . . .), with
corresponding eigen modes described by Eq. (9.14) for various k = k
sm
. Table 9.4
presents the modal parameters and the free surface displacements of particular
modes.
As illustrated by Table 9.4, there are two classes of nodal lines, “rings” and
“spokes” (diameters). The corresponding mode numbers m and s give the respective
exact number of these lines. Due to mass conservation, the mode (0, 0) does not
exist in a completely closed basin.
41
For the case s = 0, the modes are symmetrical
with respect to the origin and have annular crest and troughs.
35
In particular, the
ﬁrst symmetrical mode (s = 0, m = 1) has one nodal ring r = 0.628a (Table 9.4).
When the central part of the circular basin (located inside of this ring) is going up,
the marginal part (located between this ring and the basin border) is going down,
and vice versa. The second symmetrical mode (s = 0, m = 1) has two nodal rings:
r = 0.343a and r = 0.787a.
For s > 0, there are s equidistant nodal diameters located at an angle ∆θ = π/s
from each other; i.e., 180
◦
for s = 1, 90
◦
for s = 2, 60
◦
for s = 3, etc. Posi
tions of these diameters are indeterminate, since the origin of θ is arbitrary. The
indetermatability disappears if the boundary deviates even slightly from a circle.
Speciﬁcally, the ﬁrst nonsymmetrical mode (s = 1, m = 0) has one nodal diameter
(θ = π/2), whose position is undeﬁned; but if the basin is not circular but elliptical,
the nodal line would coincide with either the major or minor axis, and the corre
sponding eigen periods would be unequal.
35
The ﬁrst unsymmetrical mode has the
lowest frequency and the largest eigen period (Table 9.4); in this case, the water
sways from one side to another relative to the nodal diameter. This mode is often
referred to as the “sloshing” mode.
78
Most natural lakes or water reservoirs can support rather complex 2D seiches.
However, the two elementary examples of rectangular and circular basins help to
understand some general properties of the corresponding standing oscillations and
to provide rough estimates of the fundamental periods of the basins.
9.2.3. Harbor resonance
Let us return to harbor oscillations and consider some important resonant properties
of semiclosed basins. First, it is worthy to note that expressions (9.7)–(9.9) and
Table 9.3 for openmouth basins give only approximate values of the eigen periods
and other parameters of harbor modes. Solutions of the wave equation for basins of
simple geometric forms are based on the boundary condition that a nodal line (zero
sea level) is always exactly at the entrance of a semiclosed basin that opens onto a
much larger water body. In this case, the free harbor modes are equivalent to odd
(antisymmetric) modes in a closed basin, formed by the openmouth basin and its
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
206 A. B. Rabinovich
mirror image relative to the mouth.
b
However, this condition is not strictly correct
because it does not take into account wave energy radiation through the mouth into
the open sea. The exact solutions may be obtained based on the Sommerfeld radi
ation condition of free wave radiation through the open boundary.
35,41
Following
application of the appropriate mouth correction (α), the nodal line is located close
to but outside the entrance. In other words, the eﬀect of this correction is to increase
the eﬀective length of the basin.
95
The mouth correction depends on two param
eters: the basin aspect ratio q = l/L, which relates the width of the basin (l) to
its length (L); and the aperture ratio ϑ = b/l, in which b is the actual width of
the mouth.
Mathematical determination of α is rather complicated but, as a rule, it increases
with increases of q and ϑ. For example, the fractional correction to Eq. (9.9) for
the fundamental mode in a rectangular basin of uniform depth and open mouth
(ϑ = 1.0) is determined as
30,95
α =
q
π
¸
3
2
−γ −ln
πq
4
, (9.16)
where γ = 0.5772 is Euler’s constant. Roughly speaking, radiation into the external
basin and the mouth correction are important when the semiclosed basin is broad
and has a large open entrance, and negligible when the basin is long and narrow
(i.e., when q is small); in the latter case, expressions (9.7)–(9.9), as well as those
presented in Table 9.3, are quite accurate.
The character of natural oscillations in a bay or harbor is strongly controlled
by the aperture ratio ϑ = b/l, which can vary from ϑ = 1.0 to ϑ = 0.0. These two
asymptotic cases represent a fully open harbor and a closed basin, respectively. It is
evident that the smaller is ϑ (i.e., the smaller the width of the entrance), the slower
water from the external basin (open sea) penetrates into the harbor. Thus, as ϑ
decreases, the periods of all harbor modes for n ≥ 1 in Table 9.1 increase, tending
to the periods of the corresponding eigen modes for a closed basin, while the period
of the fundamental (Helmholtz) harbor mode tends to inﬁnity.
c
This is one of the
important properties of harbor oscillations.
Another important property is harbor resonance. The ampliﬁcation factor for
long waves impinging on a harbor from the open sea is
H
2
(f) =
1
(1 −f/f
0
)
2
+Q
−2
(f/f
0
)
2
, (9.17)
where f is the frequency of the long incoming waves, f
0
is the resonant frequency of
the harbor, and Q is the quality factor (“Qfactor”), which is a measure of energy
damping in the system.
47,95
Speciﬁcally,
Q
−1
=
dE/dt
ω E
= 2β, (9.18)
b
This approach is used for numerical computation of eigen modes in natural 2D basins.
70
c
This is the reason for calling this the “zeroth mode.”
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 207
where E = E
0
e
−2βωt
is the energy of the system as it decays from an initial value
E
0
, β is a dimensionless damping coeﬃcient, and ω = 2πf is the angular frequency.
The power ampliﬁcation factor attains the value Q
2
at resonance (f = f
0
), decreases
to unity at f = 0 and goes to zero as f goes to inﬁnity. Therefore, Q for harbor
oscillations plays a double role: as a measure of the resonant increase of wave heights
for waves arriving from the open ocean and as an index of the time decay rate of wave
heights inside the harbor. The higher the Q, the stronger will be the ampliﬁcation
of the incoming waves and the slower the energy decay, i.e., the longer the “ringing”
of seiche oscillations inside the harbor.
In closed basins, like lakes, bottom friction is the main factor controlling energy
decay. Normally, it is quite small, so in lakes with fairly regular topographic fea
tures (low damping), a high Qfactor may be expected. Consequently, even a small
amount of forcing energy at the resonant frequency can produce signiﬁcant seiche
oscillations that persist for several days.
31,95
In contrast, the main factor of energy
decay in semiclosed water basins, such as gulfs, bays, fjords, inlets and harbors, is
wave radiation through the entrance. In their pioneering work, Miles and Munk
47
concluded that narrowing the harbor entrance would increase the quality factor Q
and, consequently, the ampliﬁcation of the arriving wave. This means that the con
struction of dams, dikes, and walls to protect the harbor from wind waves and swell
could so constrict the entrance width that it leads to strong ampliﬁcation of the
resonant seiche oscillations inside the harbor. Miles and Munk
47
named this harbor
paradox.
As pointed out by Miles and Munk,
47
there are two limitations to the previous
conclusions:
(1) A time of order Q/π cycles is necessary for the harbor oscillations to adjust
to the external forcing. This means that harbors with high Q would not respond to a
strong but shortlived incoming disturbance. In most cases, this limitation is not of
major concern because atmospheric disturbances (the major source of opensea long
waves inducing harbor oscillations) are likely to last at least for several hours. Even
tsunami waves from distant locations “ring” for many hours, resonantly “feeding”
harbor seiches and producing maximum oscillations that have long (12–30 h) dura
tions that persist well after the arrival of the ﬁrst waves.
74,75
This contrasts with the
case for nearﬁeld sites, where tsunamis normally arrive as shortduration impulsive
waves. Such tsunamis are much more dangerous at open coastal regions than in
bays or harbors, as was observed for the coast of Thailand after the 2004 Sumatra
tsunami.
86
(2) As the harbor mouth becomes increasingly narrower, the internal harbor
dissipation eventually exceeds energy radiation through the mouth. At this stage,
further narrowing does not lead to a further increase in the Qfactor. However,
normally internal dissipation is small compared to the typical radiative energy losses
through the entrance.
Originally, Miles and Munk
47
believed that their “harbor paradox” concept was
valid for every harbor mode provided the corresponding spectral peak was sharp
and well deﬁned. Further thorough examination of this eﬀect
37,46,78
indicated that
the harbor paradox is only of major importance for the Helmholtz mode, while
for higher modes frictional and nonlinear factors, not accounted for in the theory,
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
208 A. B. Rabinovich
dampen this eﬀect.
95
However, the Helmholtz mode is the most important mode in
natural basins and is normally observed in bays, inlets, and harbors with narrow
entrance, i.e., in semiclosed basins with high Qfactor. Signiﬁcant problems with
the mooring and docking of ships (and the loading and unloading of their cargo)
in ports and harbors are often associated with this fundamental mode and most
typically occur in ports with high Q.
4,6,41,62,66–68,78–80
Rabinovich
67
suggested reducing these negative eﬀects in ports by artiﬁcially
increasing the internal dissipation. The idea is the same as that widely used in
rocket technology to damp eigen oscillations in fuel tanks.
44,45
Radial piers in ports
and harbors play the same role as internal rings and ribs in rocket tanks, eﬃciently
transforming wave energy into vortical motions which reduce the wave energy and
therefore the intensity of the seiches and their associated horizontal currents. As
shown by Rabinovich,
67
the logarithmic attenuation factor, δ
0
= π/Q, for the
Helmholtz mode associated with the jth pier, is given by
δ
j
0
=
∆E
j
0
2E
0
=
1
6
C
j
x
b
j
r
0
ˆ
ζ
0
h
0
ω
0
σ
0
, (9.19)
where E
0
is the energy of the mode inside the harbor, ∆E
j
0
is the energy dissipated
at the pier over the mode period (T
0
= 2π/ω
0
), b
j
is the length of the pier, r
0
and h
0
are the mean radius and depth of the harbor,
ˆ
ζ
0
is the mean amplitude
of the Helmholtz mode in the harbor, C
j
x
is a dimensionless resistance coeﬃcient,
and σ
0
= (gh
0
)
1/2
/(πr
0
). Thus, the rate of damping of oscillations in a harbor
depends on the number of piers (N) and a number of dimensionless parameters:
speciﬁcally, the relative amplitudes of the oscillations, ξ
0
=
ˆ
ζ
0
/h
0
; the normalized
harbor frequency, Ω
0
= ω
0
/σ
0
; the relative lengths of the piers, B
j
= b
j
/r
0
; and the
coeﬃcient C
j
x
. The parameter ξ
0
depends on the intensity of the external forcing
while the two other parameters Ω
0
and B
j
do not depend on forcing but only on the
characteristics of the harbor. The coeﬃcient C
j
x
strongly depends on the Keulegan–
Carpenter (KC) number which relates hydraulic resistance in oscillating ﬂows to
those for stationary currents.
32
For typical values B
j
= 0.3, ξ
0
= 0.1, Ω
0
= 1.0,
N = 8, and C
j
x
= 10, we ﬁnd δ
0
≈ 0.4 and Q ≈ 8.
Another important aspect of the harbor oscillation problem is that changes in
port geometry, and the construction of additional piers and dams can signiﬁcantly
change the natural (eigen) periods of the port, thereby modifying considerably the
resonant characteristics of the basin.
6
Helmholtz resonators in acoustics are used to
attenuate sound disturbances of long wavelengths, which are diﬃcult to reduce using
ordinary methods of acoustical energy dissipation. Similarly, side channel resonators
are suggested as a method for attenuating incident wave energy in harbors.
6,66,78
In general, estimation of the Qfactor is a crucial consideration for ports, harbors,
bay, and inlets. For a rectangular basin of uniform depth and entirely open mouth
(ϑ = b/l = 1.0), this factor is easily estimated as:
Q =
L
l
, (9.20)
which is inversely proportional to the aspect ratio q = l/L. This means that
high Qfactors can be expected for long and narrow inlets, fjords, and waterways.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 209
Honda et al.
30
and Nakano and Unoki
59
examined coastal seiches at more than
110 sites on the coast of Japan and found that strong and highly regular seiche
oscillations are most often observed in such elongated basins and that the periods
of these oscillations are in good agreement with the approximate period (9.9) for
the Helmholtz mode (n = 0):
T
0
=
4L
√
gH
. (9.21)
If the aperture ratio ϑ < 1.0, corresponding to a partly closed entrance, it is
more diﬃcult to estimate the Q value and the resonant mode periods analytically. In
practice, special diagrams for a rectangular basin with various q and ϑ are used for
these purposes.
80,84
For natural basins, these parameters can be estimated numer
ically or from direct observations. If the respective spectral peak in observational
data is isolated, sharp and pronounced enough, then we can assume that Q 1.
In this case, it follows from (9.17) that the halfpower frequency points (f
1/2
) are
given by the following expression
47
:
f
±
1/2
= f
0
1 ±
1
2Q
, (9.22)
and the relative frequency bandwidth is simply
∆f
f
0
= Q
−1
, (9.23)
where ∆f = f
+
1/2
− f
−
1/2
and f
0
= 1/T
0
is the resonant frequency. This is a
useful practical method for estimating the Qfactor and ampliﬁcation for coastal
basins based on results of spectral analysis of observational data. However, the
spatial structure of diﬀerent modes, the distribution of currents, and sea levels
inside a natural basin, inﬂuence harbor reconstruction based on changes in these
characteristics, and many other aspects of harbor hydrodynamics, are diﬃcult
to estimate without numerical computations. Numerical modeling has become a
common approach that is now widely used to examine harbor oscillations.
4,14,70,92
9.2.4. Harbor oscillations in a natural basin
Some typical features of harbor oscillations are made more understandable using a
concrete example. Figures 9.2 and 9.3 illustrate properties of typical harbor oscil
lations and results of their analysis and numerical modeling. Several temporary
cable bottom pressure stations (BPS) were deployed in bays on the northern coast
of Shikotan Island, Kuril Islands in 1986–1992.
13,14,68,70
All BPSs were digital
instruments that recorded long waves with 1min sampling. One of these stations
(BPS1) was situated at the entrance of False Bay, a small bay with a broad
open mouth [Fig. 9.2(a)]. The oscillations recorded at this site were weak and
irregular; the respective spectrum [Fig. 9.2(b)] was “smooth” and did have any
noticeable peaks, probably because of the closeness of the instrument position to
the position of the entrance nodal line. Two more gauges (BPS2 and BPS3) were
located inside Malokurilsk Bay, a “bottlelike” bay with a maximum width of about
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
210 A. B. Rabinovich
140° 145° 150°
40°
Pacific Ocean
Hokkaido I.
Iturup I.
Kunashir I.
Shikotan I.
45°
Sea of
Okhotsk
Shikotan I.
Malokurilsk Bay
False Bay
1
2
3
5
4
K
u
r
i
l
I
s
l
a
n
d
s
(d) (c) (b)
(a)
10 10 10
10
10
10
10
10
10
10
10
10 10
10
10 10 10 10 10 10 10 10
4 4 4
5
3
3
2
2
2
1
1
2 0
0
2 2 3 2 3 1 1 1
Frequency (cpm)
95% 95%
3
1
2
4
18.6
18.6
95%
S
p
e
c
t
r
a
(
c
m
c
p
m
)
2

1
0.5 km
Fig. 9.2. (a) Location of cable bottom pressure stations near the northern coast of Shikotan
Island (Kuril Islands) and sea level spectra at (b) BPS1, (c) BPS2 (both in autumn 1986) and
(d) BPS3 and BPS4 (October–November 1990).
1300m and a narrow neck of 350 m [Fig. 9.2(a)]. The oscillations recorded by these
instruments were signiﬁcant, highly regular and almost monochromatic; the corre
sponding spectra [Figs. 9.2(c) and 9.2(d)] have a prominent peak at a period of
18.6 min. An analogue tide gauge [#5 in Fig. 2(a)] situated on the coast of this
bay permanently measure oscillations with exactly the same period.
70
It is clear
that this period is related to the fundamental mode of the bay. The Qfactor of the
bay, as estimated by expression (9.23) based on spectral analysis of the tide gauge
data for sites BPS2 and BPS3, was 12–14 and 9–10, respectively. The high Q
factors are likely the main reason for the resonant ampliﬁcation of tsunami waves
that arrive from the open ocean. Such tsunami oscillations are regularly observed in
this bay.
13,69
In particular, the two recent Kuril Islands tsunamis of 15 November
2006 and 13 January 2007 generated signiﬁcant resonant oscillations in Malokurilsk
Bay of 155 cm and 72 cm, respectively, at the same strongly dominant period
of 18.6 min.
76
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 211
n = 0
18.9 min
n = 3
3.4 min
n = 1
6.5 min
n = 4
2.6 min
n = 2
3.8 min
n = 5
2.3 min
Fig. 9.3. Computed eigen modes and periods of the ﬁrst six modes in Malokurilsk Bay (Shikotan
Island). Black triangles indicate positions of the BPS2 and BPS3 gauges (from Ref. 70).
Figure 9.3 shows the ﬁrst six eigen modes for Malokurilsk Bay.
70
The compu
tations were based on numerical conformal mapping of the initial mirror reﬂected
domain on a circular annulus (for details, see Ref. 77) and the following application
of Ritz’s variational method to solve the eigenvalue problem. The computed period
of the fundamental (Helmholtz) mode (18.9 min) was close to the observed period of
18.6 min. The spectra at BPS2 and BPS2 indicate weak spectral peaks (three
orders of magnitude less than the main peak) with periods 4.1, 3.3, and 2.9 min
(the latter only at BPS3), thought to be related to modes n = 2, 3, and 4. The
ﬁrst mode (n = 1), with period of 6.5 min, was not observed at these sites appar
ently because the nodal line for this mode passes through the positions of BPS2
and BPS3.
Thus, the computed periods of the bay eigen modes are in good agreement with
observation; plots in Fig. 9.3 give the spatial structure of the corresponding modes.
However, this approach does not permit direct estimation of the bay response to
the external forcing and the corresponding ampliﬁcation of waves arriving from
the open ocean. In actuality, the main purpose of the simultaneous deployments
at sites BPS3 and BPS4 [Fig. 9.2(a)] in the fall of 1990 was to obtain observed
response parameters that could be compared with numerically evaluated values.
14
The spectrum at BPS4, the station located on the outer shelf of Shikotan Island
near the entrance to Malokurilsk Bay [Fig. 9.2(d)], contains a noticeable peak with
period of 18.6 min associated with energy radiation from the bay. This peak is about
1.5 orders of magnitude lower than a similar peak at BPS3 inside the bay. The
ampliﬁcation factor for the 18.6 min period oscillation at BPS4 relative to that at
BPS3 was found to be about 4.0. Numerical computations of the response charac
teristics for Malokurilsk Bay using the HNmethod
14
gave resonant periods which
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
212 A. B. Rabinovich
were in close agreement with the empirical results of Rabinovich and Levyant
70
(indicated in Fig. 9.3). Resonant ampliﬁcation of tsunami waves impinging on the
bay was found to be 8–10.
9.2.5. Seiches in coupled bays
A wellknown physical phenomenon are the oscillations of two simple coupled pen
dulums connected by a spring with a small spring constant (weak coupling). For
such systems, the oscillation energy of the combined system systematically moves
from one part of the system to the other. Every time the ﬁrst pendulum swings,
it pulls on the connecting string and gives the second pendulum a small tug, so
the second pendulum begins to swing. As soon as the second pendulum starts to
swing, it begins pulling back on the ﬁrst pendulum. Eventually, the ﬁrst pendulum is
brought to rest after it has transferred all of its energy to the second pendulum. But
now the original situation is exactly reversed, and the ﬁrst pendulum is in a position
to begin “stealing” energy back from the second. Over time, the energy repeatedly
switches back and forth until friction and air resistance eventually remove all of the
energy out of the pendulum system.
A similar eﬀect is observed in two adjacent bays that constitute a coupled system.
Nakano
56
was probably the ﬁrst to investigate this phenomenon based on observa
tions for Koaziro and Moroiso bays located in the Miura Peninsula in the vicinity
of Tokyo. The two bays have similar shapes and nearly equal eigen periods. As was
pointed out by Nakano, seiches in both bays are very regular, but the variations
of their amplitudes are such that, while the oscillations in one bay become high,
the oscillations in the other become low, and vice versa. Nakano
56
explained the
eﬀect theoretically as a coupling between the two bays through water ﬂowing across
the mouths of each bay. More than half a century later Nakano returned to this
problem
60
and, based on additional theoretical studies and hydraulic model exper
iments, demonstrated that two possible regimes can exist in the bays: (1) cophase
oscillations when seiches in the two bays have the same initial phase; and (2) contra
phase when they have the opposite phase. The superposition of these two types of
oscillations create beat phenomenon of timemodulated seiches, with the opposite
phase modulation, such that “while one bay oscillates vigorously, the other rests”.
Nakano and Fujimoto suggested the term “liquid pendulums” for the coupled inter
action of two adjacent bays.
A more complicated situation occurs when the two adjacent bays have sig
niﬁcantly diﬀerent eigen periods. For example, Ciutadella and Platja Gran are
two elongated inlets located on the west coast of Menorca Island, one of the
Balearic Islands in the Western Mediterranean [the inlets are shown in the inset
of Fig. 9.5(a)]. Their fundamental periods (n = 0) are 10.5 min and 5.5 min,
respectively.
50,71,73
As a result of the interaction between these two inlets, their
spectra and admittance functions have, in addition to their “own” strong resonant
peaks, secondary “alien” peaks originating from the other inlet.
38
This means that
the mode from Ciutadella “spills over” into Platja Gran and vice versa. The two
inlets are regularly observed to experience destructive seiches, locally known as
“rissaga”.
18,22,49–51,85
Speciﬁc aspects of rissaga waves will be discussed later (in
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 213
Sec. 9.4); however, it is worth noting here that the coupling between the two inlets
can apparently amplify the destructive eﬀects associated with each of the inlets
individually.
38
9.3. Generation
Because they are natural resonant oscillations, seiches are generated by a wide
variety of mechanisms (Fig. 9.4), including tsunamis,
14,28,54,69
seismic ground
waves,
2,15,34,40
internal ocean waves,
8,19,21
and jetlike currents.
30,54,57
However, the
two most common factors initiating these oscillations in bays and harbors are atmo
spheric processes and nonlinear interaction of wind waves or swell (Fig. 9.4).
62,68,95
Seiches in lakes and other enclosed water bodies are normally generated by direct
external forcing on the sea surface, primarily by atmospheric pressure variations
and wind.
31,95
In contrast, the generation of harbor oscillations is a twostep process
involving the generation of long waves in the open ocean followed by forcing of the
sec
Wavelength (m)
Meteorological waves
Wind waves
Storm
surges
Scattering
nonstationarity Nonlinearity
Infragravity waves
Tsunami waves
Seismicity
O C E A N
B O T T O M
10 10 10
10
10
10
10
10
10
10
10
10
6 0 1
1
2
2
3
3
4
4
5
5
Period (sec)
Swell
A T M O S P H E R E
Long waves
Fig. 9.4. Sketch of the main forcing mechanisms generating long ocean waves.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
214 A. B. Rabinovich
harbor oscillations as the long waves arrive at the harbor entrance where they lead
to resonant ampliﬁcation in the basin.
Seiche oscillations produced by external periodic forcing can be both free and
forced. The free oscillations are true seiches (i.e., eigen oscillations of the cor
responding basin). However, if the external frequency (σ) diﬀers from the eigen
frequencies of the basin (σ = ω), the oscillations can be considered forced seiches.
95
Openocean waves arriving at the entrance of a speciﬁc openmouth water body
(such as a bay, gulf, inlet, fjord, or harbor) normally consist of a broad frequency
spectrum that spans the response characteristics of the water body from resonantly
generated eigenfree modes to nonresonantly forced oscillations at other frequencies.
Following cessation of the external forcing, forced seiches normally decay rapidly,
while free modes can persist for a considerable time.
Munk
53
jokingly remarked that “the most conspicuous thing about long waves in
the open ocean is their absence.” This is partly true: the longwave frequency band,
which is situated between the highly energetic tidal frequencies and swell/wind wave
frequencies, is relatively empty (Fig. 9.5). For both swell/wind waves and tides,
the energy is of order 10
4
cm
2
, while the energy contained throughout the entire
intermediary range of frequencies is of order 1–10cm
2
. However, this particular fre
quency range is of primary scientiﬁc interest and applied importance (Walter Munk
himself spent approximately 30 years of his life working on these “absent” waves!).
Long waves are responsible for formation and modiﬁcation of the coastal zone and
shore morphology;
5,68
they also can strongly aﬀect docking and loading/unloading
of ships and construction in harbors, causing considerable damage.
41,78,79,96
Finally,
10
10
10
10
10
10 10 10 10 10 10 10
10
10
3
1
2
2
5
2
1
2 1 0 1 2 3 4
0
3
hr min sec
Period
100 2412 5 2 1 30 10 5 2 1 30 10 5 2 1
Tides
Swell, wind waves
Storm surges
Meteorological
waves
IG waves
Frequency (cpm)
V
a
r
i
a
n
c
e
(
c
m
)
2
Tsunami
Fig. 9.5. Spectrum of surface gravity waves in the ocean (modiﬁed from Ref. 68). Periods (upper
scale) are in hours (h), minutes (min), and seconds (s).
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 215
and probably the most important, are tsunamis and other marine hazardous long
waves, which are related to this speciﬁc frequency band. The recent 2004 Sumatra
tsunami in the Indian Ocean killed more than 226,000 people, triggering the largest
international relief eﬀort in history and inducing unprecedented scientiﬁc and public
interest in this phenomenon and in long waves in general.
86
Because of their resonant properties, signiﬁcant harbor seiches can be produced
by even relatively weak open ocean waves. In harbors and bays with high Q
factors, seiches are observed almost continuously. However, the most destructive
events occur when the incoming waves have considerable energy at the resonant
frequencies, especially at the frequency of the fundamental mode. Such a situation
took place in Port Alberni located in the head of long Alberni Inlet on Vancouver
Island (Canada) during the 1964 Alaska tsunami, when resonantly generated seiche
oscillations in the inlet had troughtocrest wave heights of up to 8 m, creating total
economic losses of about $10 million (1964 dollars).
28,54
9.3.1. Meteorological waves
Long waves in the ocean are the primary factor determining the intensity of harbor
oscillations. If we ignore tsunamis and internal waves, the main source of background
long waves in the ocean are atmospheric processes (Fig. 9.4).
10,53
There are three
major mechanisms to transfer the energy of atmospheric processes into long waves
in the ocean
68
:
(1) Direct generation of long waves by atmospheric forcing (pressure and wind) on
the sea surface.
(2) Generation of lowfrequency motions (for example, storm surges) and subse
quent transfer of energy into higher frequencies due to nonlinearity, topographic
scattering and nonstationarity of the resulting motions.
(3) Generation of highfrequency gravity waves (wind waves and swell) and sub
sequent transfer of energy into larger scale, lower frequency motions due to
nonlinearity.
Long waves generated by the ﬁrst two mechanisms are known as atmospher
ically induced or meteorological waves.
d
Typical periods of these waves are from
a few minutes to several hours, typical scales are from one to a few hundreds of
kilometers. The ﬁrst mechanism is the most important because it is this mech
anism that is responsible for the generation of destructive seiche oscillations (mete
orological tsunamis) in particular bays and inlets of the World Ocean (Sec. 9.4).
“Meteorological waves” can be produced by the passages of typhoons, hurricanes
or strong cyclones. They also have been linked to frontal zones, atmospheric
pressure jumps, squalls, gales, wind gusts and trains of atmospheric buoyancy
waves.
10,59,68,71,87,95
The most frequent sources of seiches in lakes are baro
metric ﬂuctuations. However they can also be produced by heavy rain, snow,
d
The Russian name for these waves is “anemobaric”
68
because they are induced by atmospheric
pressure (“baric”) and wind (“anemos”) stress forcing on the ocean surface.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
216 A. B. Rabinovich
or hail over a portion of the lake, or ﬂood discharge from rivers at one end of
the lake.
27,31,95
9.3.2. Infragravity waves
Long waves generated through the nonlinear interaction of wind waves or swell are
called infragravity waves.
5,63
These waves have typical periods of 30 s to 300–600s
and length scales from 100 m to 10 km. The occurrence of relatively highfrequency
long waves, highly correlated with the modulation of groups of wind or swell waves,
was originally reported by Munk
52
and Tucker.
89
Because the waves were observed
as sea level changes in the nearshore surf zone, they became known as surf beats.
Later, it was found that these waves occur anywhere there are strong nonlinear
interacting wind waves. As a result, the more general term infragravity waves (pro
posed by Kinsman
33
) became accepted for these waves. Recent ﬁeld measurements
have established that infragravity waves (IG waves) dominate the velocity ﬁeld
close to the shore and consist of superposition of free edge waves propagating along
the shore, free leaky waves propagating in the oﬀshore direction, and forced bound
waves locked to the groups of wind waves or swell propagating mainly onshore.
3,5,68
Bound IG waves form the setdown that accompanies groups of incident waves,
having troughs that are beneath the high short waves of the group and crests in
between the wave groups.
39
They have the same periodicity and the same lengths
as the wave groups and travel with the group velocity of wind waves, which is signif
icantly smaller than the phase speed of free long waves with the same frequencies.
Free edge IG waves arise from the trapping of swell/wind wave generated oscilla
tions over sloping coastal topography, while free leaky waves are mainly caused by
the reﬂection of bound waves into deeper water.
5,63
The general mechanisms of the
formation of IG waves are shown in Fig. 9.6.
e
IG waves are found to be responsible for many phenomena in the coastal zone,
including formation of rip currents, wave setup, crescentic bars, beach cusps and
other regular forms of coastal topographies, as well as transport of sediment mate
rials. Being of highfrequency relative to meteorological waves, IG waves can induce
seiches in comparatively smallscale semiclosed basins, such as ports and harbors,
which have natural periods of a few minutes and which may pose a serious threat
for large amplitude wave responses.
Certain harbors and ports are known to have frequent strong periodic hori
zontal water motions. These include Cape Town (South Africa), Los Angeles (USA),
Dakar (Senegal), Toulon and Marseilles (France), Alger (Algeria), Tuapse and Sochi
(Russia), Batumi (Georgia) and Esperance (Australia). Seiche motions in these
basins create unacceptable vessel movement which can, in turn, lead to the breaking
of mooring lines, fenders and piles, and to the onset of large amplitude ship oscilla
tions and damage.
62,67,68,82,94–96
Known as surging or range action,
78,79
this phe
nomenon has wellestablished correlations with (a) harbor oscillations, (b) natural
oscillations of the ship itself, and (c) intensive swell or wind waves outside the
e
Figure 9.6 does not include all possible types of IG waves and mechanisms of their generation; a
more detailed description is presented by Bowen and Huntley
5
and Battjes.
3
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 217
Fig. 9.6. Generation mechanisms for infragravity waves in the coastal zone.
harbor. Typical eigen periods of a harbor or a moored ship are the order of minutes.
Therefore, they cannot be excited directly by wind waves or swell, having typical
periods on the order of seconds.
96
However, these periods exactly coincide with the
periods of wave groups and IG waves. So, it is conventional wisdom that surging
in harbors is the result of a triple resonance of external oscillations outside the
harbor, natural oscillations within the harbor, and natural oscillations of a ship.
The probability of such triple resonance is not very high, thus surging occurs only
in a limited number of ports. Ports and harbors having large dimensions and long
eigen periods (>10 min) are not aﬀected by surging because these periods are much
higher than the predominant periods of the IG waves and the surging periods of the
vessels. On the other hand, relatively small vessels are not aﬀected because their
natural (eigen) periods are too short.
82
The reconstruction of harbors and the cre
ation of new harbor elements, can signiﬁcantly change the harbor resonant periods,
either enhancing or, conversely, reducing the surging.
f
Another important aspect of
the problem is that ship and mooring lines create an entirely separate oscillation
system.
79
Changing the material and the length of the lines and their position,
f
A famous example of this kind is the French port Le Havre. Before World War II it was known
for very common and strong surging motions that created severe problems for ships. During the
war a German submarine torpedoed by mistake a riprap breakwater, creating a second harbor
opening of 20–25 m width. After this, the surging in the port disappeared.
68
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
218 A. B. Rabinovich
changes the resonant properties of the system (analogous to changing the material
and the length of a pendulum).
It is important to keep in mind that each oscillation mode has a speciﬁc
spatial distribution of sea level variability and associated current (as emphasized in
Sec. 9.2.1, maximum currents are observed near the nodal lines). The intensity of
the currents varies signiﬁcantly from place to place. Moreover, topographic irregu
larities within the harbor and the presence of structure elements (dams, dykes, piers
and breakwaters) can create intense local vortexes that may signiﬁcantly aﬀect the
ships.
67
So, the eﬀect of surging on a ship strongly depends on the exact location
of the ship, and even on its orientation, in the harbor.
In summary, harbor oscillations arise through cooscillation of sea surface eleva
tions and currents in the harbor with those at the entrance to the harbor. Seiche
generating motions outside the harbor typically have periods of several minutes
and most commonly arise from bound and free long waves that are incident on the
harbor entrance.
9.3.3. Tsunami
Tsunami waves are the main factor creating destructive seiche oscillations in bays,
inlets and harbors.
30,41,53,54,95
Tsunamis can produce “energies” of 10
3
–10
5
cm
2
,
although such events are relatively rare (depending on the region, from once every
1–2 years to once every 100–200 years). The main generation mechanisms for
tsunamis are major underwater earthquakes, submarine landslides and volcanic
explosions. Great catastrophic transoceanic tsunamis were generated by the 1946
Aleutian (magnitude M
w
= 7.8), 1952 Kamchatka (M
w
= 9.0), 1960 Chile
(M
w
= 9.5), and 1964 Alaska (M
w
= 9.2) earthquakes. The events induced strong
seiche oscillations in bays, inlets, and harbors throughout the Paciﬁc Ocean.
90
The magnitude M
w
= 9.3 earthquake that occurred oﬀshore of Sumatra in
the Indian Ocean on 26 December 2004 generated the most destructive tsunami
in recorded history. Waves from this event were recorded by tide gauges around
the world, including nearsource areas of the Indian Ocean (Fig. 9.7), and remote
regions of the North Paciﬁc and North Atlantic, revealing the unmatched global
reach of the 2004 tsunami.
42,74,86,88
In general, the duration of tsunami “ringing”
increased with increasing oﬀsource distance and lasted from 1.5 to 4 days.
74,75
The
recorded oscillations were clearly polychromatic, with diﬀerent periods for diﬀerent
sites, but with clear dominance of 40–50min waves at most sites. The analysis of
various geophysical data from this event indicates that the initial tsunami source
had a broad frequency spectrum, but with most of the energy within the 40–50 min
band. Therefore, although tsunami waves at diﬀerent sites induced local eigen modes
with a variety of periods, the most intense oscillations were observed at sites having
fundamental periods close to 40–50 min.
Diﬀerences in spectral peaks among the various tide gauge records are indicative
of the inﬂuence of local topography. For example, for the Paciﬁc coast of Van
couver Island (British Columbia), the most prominent peaks in the tsunami spectra
were observed for Winter Harbor (period ∼30–46 min) and Toﬁno (∼50 min). In
fact, the frequencies of most peaks in the tsunami spectra invariably coincide with
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 219
December 2004 (UTC)
26 27 28
2
0
0
c
m
West and Central Indian Ocean
S
e
a
l
e
v
e
l
Hanimaadhoo (Md)
Male (Md)
Colombo (SL)
Salalah (Om)
Pointe La Rue (Sl)
Port Louis (Mt)
E
Fig. 9.7. Tsunami records in the Indian Ocean for the 2004 Sumatra tsunami for six selected sites:
Colombo (Sri Lanka); Male and Gan (both Maldives); Salalah (Oman); Pointe La Rue (Seychelles);
and Port Louis (Mauritius). Solid vertical line labeled “E” denotes the time of the main earthquake
shock (from Ref. 74).
corresponding peak frequencies in the background spectra. This result is in good
agreement with the wellknown fact that periods of observed tsunami waves are
mainly related to the resonant properties of the local/regional topography rather
than to the characteristics of the source, and are almost the same as those of ordinary
(background) long waves for the same sites. For this reason, the spectra of tsunamis
from diﬀerent earthquakes are usually similar at the same location.
30,48,69,g
It is
therefore diﬃcult to reconstruct the source region spectral characteristics based on
data from coastal stations.
Rabinovich
69
suggested a method for separating the eﬀects of the local topog
raphy and the source on the resulting tsunami wave spectrum. This method can be
used to reconstruct the openocean spectral characteristics of tsunami waves. The
approach is based on the assumption that the spectrum S(ω) of both the tsunami
and background sea level oscillations near the coast can be represented as
S(ω) = W(ω)E(ω), (9.24)
g
The resonant characteristics of each location are always the same; however, diﬀerent sources
induce diﬀerent resonant mode, speciﬁcally, large seismic sources generate lowfrequency modes
and small seismic sources generate highfrequency modes.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
220 A. B. Rabinovich
where W(ω) = H
2
(ω), H(ω) is the frequency admittance function describing the
linear topographic transformation of long waves approaching the coast, and E(ω) is
the source spectrum. It is assumed that the sitespeciﬁc properties of the observed
spectrum S
j
(ω) at the jth site are related to the topographic function H
j
(ω) for that
site, while all mutual properties of the spectra at all sites are associated with the
source (assuming that the source is the same for all stations). For typical background
oscillations the source spectrum has the form, E(ω) = S
0
(ω), where S
0
(ω) = Aω
−2
,
and A = 10
−3
−10
−4
cm
2
.
68,69
During tsunami events, sea level oscillations observed
near the coast can be represented as
ζ
obs
(t) = ζ
t
(t) +ζ
b
(t), (9.25)
where ζ
t
are the tsunami waves generated by an underwater seismic source and ζ
b
are the background surface oscillations. If the spectra of both tsunami, S
t
(ω), and
background oscillations,
ˆ
S
b
(ω) and S
b
(ω) (during and before the tsunami event,
respectively) have the form (9.24), and the admittance function, W(ω), is the same
for the observed tsunami and the background long waves, then the spectral ratio
R(ω), is estimated as
R(ω) =
S
t
(ω) +
ˆ
S
b
(ω)
S
b
(ω)
=
E(ω) +
ˆ
S
0
(ω)
S
0
(ω)
= A
−1
ω
2
E(ω) + 1.0. (9.26)
The function R(ω), which is independent of local topographic inﬂuence, is deter
mined solely by the external forcing (i.e., by tsunami waves in the open ocean near
the source area) and gives the ampliﬁcation of the longwave spectrum during the
tsunami event relative to the background conditions. The close similarity of R
j
(ω)
for various sites conﬁrms the validity of this approach.
69
The topographic admittance function H
j
(ω), which is characteristic of the res
onant properties of individual sites, can be estimated as
H
j
(ω) =
¸
S
j
(ω)
E(ω)
1/2
=
¸
S
j
b
(ω)
S
0
(ω)
¸
1/2
= ω
¸
S
j
b
(ω)
A
¸
1/2
. (9.27)
The same characteristic can be also estimated numerically.
9.3.4. Seismic waves
There is evidence that seismic surface ground waves can generate seiches in both
closed and semiclosed basins. In particular, the Great 1755 Lisbon earthquake trig
gered remarkable seiches in a number of Scottish lochs, and in rivers and ponds
throughout England, western Europe and Scandinavia.
95
Similarly, the Alaska
earthquake of 27 March 1964 (M
w
= 9.2) induced seismic surface waves that took
only 14 min to travel from Prince Williams Sound, Alaska, to the Gulf Coast region
of Louisiana and Texas where they triggered innumerable seiches in lakes, rivers,
bays, harbors, and bayous.
15,34
Recently, the 3 November 2002 Denali earthquake
(M
w
= 7.9) in Alaska generated pronounced seiches in British Columbia and Wash
ington State.
2
Sloshing oscillations were also observed in swimming pools during
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 221
these events.
2,15,40
The mechanism for seiche generation by seismic waves from
distant earthquakes is not clear, especially considering that seismic waves normally
have much higher frequencies than seiches in natural basins. McGarr
40
concludes
that there are two major factors promoting eﬃcient conversion of the energy from
distant largemagnitude earthquakes into seiches:
(1) A very thick layer of soft sediments that amplify the horizontal seismic ground
motions.
(2) Deeper depths of natural basins, increasing the frequencies of eigen periods for
the respective water oscillations.
It should be noted, however, that seismic origins for seiches must be considered
as very rare in comparison, for example, with seiches generated by meteorological
disturbances.
95
9.3.5. Internal ocean waves
In some regions of the World Ocean, deﬁnitive correlation has been found between
tidal periodicity and the strong seiches observed in these regions. For example, at
Palawan Island in the Philippines, periods of maximum seiche activity are asso
ciated with periods of high tides.
21
Bursts of 75min seiches in the harbor of
Puerto Princesa (Palawan Island) are assumed to be excited by the arrival at the
harbor entrance of internal wave trains produced by strong tidal current ﬂow across
a shallow sill located about 450 km from the harbor.
8
Internal waves can have
quite large amplitudes; furthermore, they can travel over long distances without
noticeable loss of energy. Internal waves require 2.5 days to travel from their source
area in the Sulu Sea to the harbor of Puerto Princessa, resulting in a modulation
of the seiche oscillations that are similar to those of the original tidal oscillations.
Similarly, large amplitude seiches on the Caribbean coast of Puerto Rico are also
related to tidal activity and are usually observed approximately seven days after
a new or full moon (syzygy). Highest seiches in this region occur in late summer
and early fall, when thermal stratiﬁcation of the water column is at its annual
maximum. The sevenday interval between syzygy and maximum seiche activity
could be accounted for in terms of internal tidal soliton formation near the south
western margin of the Caribbean Sea.
34
A theoretical model of seiche generation
by internal waves, devised by David Chapman (Woods Hole Oceanographic Insti
tution), demonstrated that both periodic and solitary internal waves can generate
coastal seiches.
8,20
Thus, this mechanism can be responsible for the formation of
seiches in highly stratiﬁed regions.
9.3.6. Jetlike currents
Harbor oscillations (coastal seiches) can also be produced by strong barotropic tidal
and other currents. Such oscillations are observed in Naruto Strait, a narrow channel
between the Shikoku and Awaji islands (Japan), connecting the Paciﬁc Ocean and
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
222 A. B. Rabinovich
the Inland Sea. Here, the semidiurnal tidal currents move large volumes of water
back and forth between the Paciﬁc and the Inland Sea twice per day with typical
speed of 13–15 km/h. This region is one of the greatest attractions in Japan because
of the famous “Naruto whirlpool”, occurring twice a month during spring tides,
when the speed of tidal ﬂow reaches 20 km/h. Honda et al.
30
noticed that ﬂood tidal
currents generate near both coasts signiﬁcant seiche oscillations with a period of
2.5 min, which begin soon after low tide and cease near high tide; the entire picture
repeats with a new tidal cycle. No seiches are observed during ebb tidal currents (i.e.,
between high and low water) when the water is moving in the opposite direction.
Nakano
57
explained this phenomenon by assuming that a strong current passing
the mouth of a bay could be the source of bay seiches, similar to the way that a
jet of air passing the mouth piece of an organ pipe produces a standing oscillation
within the air column in the pipe. Special laboratory experiments by Nakano and
Abe
58
demonstrated that jetlike ﬂow with a speed exceeding a speciﬁc critical
number generates a chain of antisymmetric, counterrotating von Karman vortexes
on both sides of the channel. The checkerboard pattern of vortexes induce standing
oscillations in nearby bays and harbors if their fundamental periods match the
typical vortex periods,
T
vor
=
l
u
, (9.28)
where l is the distance between vortexes, and u is the speed of the vortexes (u = 0.4V
to 0.6V , where V is the speed of the tidal currents). For the parameters of the
Naruto tidal currents, the laboratory study revealed that values of T
vor
agreed with
the observed seiche period of 2.5 min. Apparently, the same mechanism of seiche
generation can also work in other regions of strong jet currents.
9.3.7. Ice cover and seiches
It seems clear that ice cannot generate seiches (except for the case of calving icebergs
or avalanches that generate tsunamilike waves). However, an ice cover can signif
icantly impact seiche motions, suppressing them and impeding their generation.
At the same time, strong seiches can eﬀectively break the ice cover and promote
polynya creation.
Little is known on the speciﬁc aspects of ice cover interaction with seiche modes.
Hamblin
26
suggested that the ice cover in Lake Winnipeg inﬂuences the character
of seiche activity. Schwab and Rao
83
assumed that absence of certain peaks in the
sea level spectra for Saginaw Bay (Lake Huron) in winter may have been due to the
presence of ice cover. Murty
55
examined the possible eﬀect of ice cover on seiche
oscillations in Kugmallit Bay and Tuktoyaktuk Harbor (Beaufort Sea) and found
that the ice cover reduces the eﬀective water depth in the bay and harbor and in
this way diminishes the frequency of the fundamental mode: in Kugmallit Bay from
0.12 cph (icefree period) to 0.087 cph (icecovered); and in Tuktoyaktuk Harbor
from 1.0 cph to 0.9 cph.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 223
9.4. Meteorological Tsunamis
As discussed in Sec. 9.3.3, tsunamis are the main source of destructive seiches
observed in various regions of the World Ocean. However, atmospheric disturbances
(atmospheric gravity waves, pressure jumps, frontal passages, squalls) can also
be responsible for signiﬁcant, even devastating, long waves, which have the same
temporal and spatial scales as typical tsunami waves. These waves are similar to
ordinary tsunami waves and can aﬀect coasts in a similar damaging way, although
the catastrophic eﬀects are normally observed only in speciﬁc bays and inlets.
Nomitsu,
61
Defant,
10
and Rabinovich and Monserrat
71,72
suggested to use the term
“meteorological tsunamis” (“meteotsunami”) for this type of waves.
At certain places in the World Ocean, these hazardous atmospherically induced
waves occur regularly and have speciﬁc local names: “rissaga” in the Balearic
Islands, “ˇs´ciga” on the Croatian coast of the Adriatic Sea, “marubbio” (“marrobio”)
in Sicily, “milghuba” in Malta, “abiki” and “yota” in Japan, “Seeb¨ar” in the Baltic
Sea, “death waves” in Western Ireland, “inchas” and “lavadiads” in the Azores and
Madeira islands. These waves are also documented in the Yellow and Aegean seas;
the Great Lakes; northwestern Atlantic; Argentina and New Zealand coastal areas;
and Port Rotterdam.
7,9–12,16,27,29,30,34,43,51,64,68,71,72,91,92
Table 9.5 gives a list of
destructive harbor oscillations, which apparently have the same atmospheric origin
and similar resonances due to similarities in the characteristics of the atmospheric
disturbances and local geometry and topography of the corresponding basins.
Because of the strong likeness between “meteotsunamis” and seismically generated
tsunamis,
51,88
it is quite diﬃcult sometimes to recognize one from another. Cata
logues of tsunamis normally contain references to numerous “tsunamilike” events
of “unknown origin” that are, in fact, atmospherically generated ocean waves.
“Rissaga” (a local Catalan word that means “drying”, similar to a Spanish
word “resaca”) is probably the best known example of meteorological tsunamis.
h
These signiﬁcant shortperiod sea level oscillations regularly occur in many bays
and harbors of the Catalan and Valencian coasts of the Iberian Peninsula, and on
the coast of the Balearic Islands. The waves in Ciutadella Harbor, Menorca Island
[Fig. 9.8(a)] are particularly high and occur more frequently than in any other
location.
18,22,49–51,71–73,81,85
Ciutadella Inlet is about 1 km long, 100 m wide, and 5 m deep; the harbor is
located at the head of the inlet [Fig. 9.8(a)]. The fundamental period of the inlet
(Helmholtz mode) is approximately 10.5 min [Figs. 9.8(b,c)]. Due to the particular
geometry of Ciutadella Inlet, it has a large Qfactor, which results in signiﬁcant
resonant ampliﬁcation of longwave oscillations arriving from the open sea. Seiche
h
For this reason Derek Goring, a wave specialist from New Zealand, suggested to apply the term
“rissaga” to all rissagalike meteorological seiches in other areas of the World Ocean.
23
However,
if we were to adopt this term, then we would loose information on the cause of the oscillations
and the fact that they are part of a family of events that include seismically generated tsunamis,
landslide tsunamis, volcanic tsunamis, and meteotsunamis.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
224 A. B. Rabinovich
Table 9.5. Extreme coastal seiches in various regions of the World Ocean.
Maximum
Typical observed
Region Local name peroid height References
Nagasaki Bay,
Japan
Abiki 35 min 4.78 m Honda et al.
30
Amano [1957],
Akamatsu
1
, Hibiya and Kajiura
29
Pohang Harbor,
Korea
— 25 min >0.8 m Chu [1976], Park et al. [1986]
Longkou
Harbor, China
— 2 h 2.93 m Wang et al.
93
Ciutadella
Harbor,
Menorca I.,
Spain
Rissaga 10.5 min >4.0 m Fontser´e,
17
Tintor´e et al.,
88
Monserrat et al.,
49−51
Gomis
et al.,
22
Garcies et al.,
18
Rabinovich and Monserrat,
71,72
Rabinovich et al.
73
Gulf of Trieste,
Italy
— 3.2 h 1.6 m Caloi [1938], Greco et al. [1957],
Defant,
10
Wilson
95
West Sicily,
Italy
Marrubio
(Marrobbio)
∼15 min >1.5 m Plattania [1907], Oddone [1908],
Defant,
10
Colucci and
Michelato,
9
Candela et al.
7
Malta,
Mediterranean
Milghuba ∼20 min ∼1.0 m Airy [1878], Drago
16
West Baltic,
Finland coast
Seeb¨ar — ∼2.0 m Doss [1907], Meissner [1924],
Defant,
10
Credner [1988],
Croatian coast
East Adriatic
ˇ
S´ciga 10–30
min
∼6.0 m Hodˇzi´c [1979/1980]; Orli´c
60
;
Vilibi´c et al.
92
; Monserrat et al.
51
Newfoundland,
Canada
— 10–40
min
2.0–3.0 m Mercer et al. [2002]
Western Ireland Death Waves ? ? Berninghausen [1964], Korgen
34
Azores Is and
Madeira Is,
East Atlantic
Inchas,
Lavadiads
? ? Berninghausen [1964], Korgen
34
Rotterdam
Harbor,
The Netherlands
85–100
min
>1.5 m de Looﬀ and Veldman [1994], de
Jong et al.,
11
de Jong and
Battjes
12
[2005]
Comment: Exact references can be found in: Wiegel (1964), Korgen (1995), Rabinovich and
Monserrat (1996), de Jong et al. (2003) and Monserrat et al. (2006).
oscillations of duration ranging from a few hours to several days and wave heights
exceeding 0.5 m recur in Ciutadella every summer. However, rissaga events (large
amplitude seiches) having wave heights more than 3–4 m, with dramatic conse
quences for the harbor, usually take place once in 5–6 years. During the rissaga of
21 June 1984 (Fig. 9.9), about 300 boats were destroyed or strongly damaged.
71
More recently, on 15 June 2006, Ciutadella Harbor was aﬀected by the most dra
matic rissaga event of the last 20 years, when almost 6m waves were observed in
the harbor and the total economic loss was of several tens millions of euros.
51
Fontser´e,
17
in the ﬁrst scientiﬁc paper on extreme seiches for the Catalan
coast, showed that these seiches always occur from June to September and ﬁrst
suggested their atmospheric origin. This origin of rissaga waves was supported by
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 225
Fig. 9.8. A map of the Balearic Islands and positions of four tide gauges (M0, M1, M2 and MW3)
deployed in Ciutadella and Platja Gran inlets and on the shelf of Menorca Island during the LAST
97 experiment.
50
The arrow shows the predominant direction of propagation of atmospheric waves
during “rissaga” events. (b) The strong “rissaga” event recorded in Ciutadella Inlet on 31 July
1998 by a tide gauge located at position M0. (c) Spectra for “rissaga” of 24 July 1997 (solid line)
and background oscillations (dashed line) for four tide gauges indicated in (a). The actual fourday
records during this event are shown in the insets.
Ramis and Jans`a
81
based on observed oscillations on the Balearic Islands. These
authors also deﬁned a number of typical synoptic atmospheric conditions nor
mally associated with rissaga events. The atmospheric source of rissaga is now well
established.
18,22,49,50,85
During late spring and summer, meteorological conditions in
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
226 A. B. Rabinovich
Fig. 9.9. Ciutadella Harbor during the rissaga of 21 June 1984. Photo by Josep Gornes (from
Ref. 71).
the western Mediterranean are favorable for the formation of highfrequency atmo
spheric pressure disturbances with parameters promoting the generation of rissaga
waves. These conditions include the entrance of warm air from the Sahara at near
surface levels, and relatively strong middle level winds from the southwest. When
this synoptic meteorological situation exists, trains of atmospheric pressure gravity
waves (with periods of minutes) are reported traveling from SW to NE.
49
If these
atmospheric pressure disturbances propagate from SW to NE with a phase speed
of about 22–30m/s, resonant conditions are set up for the southeastern shelf of
Mallorca Island (“Proudman resonance”) and dynamic energy associated with the
atmospheric waves is eﬃciently transferred into the ocean waves. When these waves
reach the coast of Menorca Island, they can generate signiﬁcant (and sometimes
even hazardous) seiche oscillations inside Ciutadella and other inlets due to harbor
resonance.
The Qfactor for the fundamental Helmholtz mode in Ciutadella Inlet (10.5
min), roughly estimated by Eq. (9.20), is about 9. Spectral estimates based on
Eq. (9.23) give a similar value, Q ≈ 10.
73
As shown in Fig. 9.8(b), rissaga oscillations
in Ciutadella Inlet have a very regular monochromatic character. Maximum wave
heights occur during the fourth to the sixth oscillations, in good agreement with the
criterion by Miles and Munk
47
that time of the order of Q/π cycles is necessary for
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 227
the harbor oscillations to adjust themselves to external forcing. The peak period of
10.5 min for the Helmholtz mode strongly dominates the spectra for the M0 and
M2 gauges located in Ciutadella Inlet [Fig. 9.8(c)] both for rissaga and background
spectra, while in the adjacent inlet Platja Gran (M1), where rissaga waves are also
observed but weaker than in Ciutadella, the dominant peak associated with the
Helmholtz mode is 5.5 min. In contrast, on the shelf (MW3) both peaks are absent
and oscillations are signiﬁcantly weaker.
Spectral analysis results [Fig. 9.8(c)] reveal that harbor resonance is a crucial
factor in the formation of rissaga waves, as well as “meteorological tsunamis” in
other bays, inlets, and harbors of the World Ocean. Barometric data from the
Balearic Islands
49–51,81
as well as from Japan,
29
and Eastern Adriatic Sea
64,91,92
demonstrate that generation of these destructive waves is associated with strong
atmospheric disturbances, e.g., trains of atmospheric gravity waves, or isolated
pressure jumps. These atmospheric disturbances may have diﬀerent origin: dynamic
instability, orographic inﬂuence, frontal passages, gales, squalls, storms, tornados,
etc.
24
However, even during the strongest events, the atmospheric pressure oscilla
tions at the meteotsunami scales (from a few minutes to a few hours) reach only
2–6 hPa, corresponding to only a 2–6 cm change in sea level. Consequently, these
atmospheric ﬂuctuations may produce signiﬁcant sea level response only when reso
nance occurs between the ocean and the atmosphere. During the resonance process,
the atmospheric disturbance propagating above the ocean surface generates signif
icant long ocean waves by continuously pumping additional energy into these waves.
Possible resonances that are responsible for the formation of meteorological
tsunamis are
68
:
• Proudman resonance,
65
when U = c, i.e., the atmospheric disturbance speed (U)
equals to the longwave speed of ocean waves c =
√
gh;
• “Greenspan resonance”,
25
when U
l
= c
j
, the alongshore component (U
l
) of the
atmospheric disturbance velocity equals the phase speed of the jth mode of edge
waves (c
j
);
• “shelf resonance,” when the atmospheric disturbance and associated atmospher
ically generated ocean wave have a period/wavelength equal to the resonant
period/length of the shelf.
These resonant eﬀects may signiﬁcantly amplify ocean waves approaching the
coast. Nevertheless, even strong resonant ampliﬁcation of atmospherically generated
ocean waves normally cannot produce waves with suﬃcient energy to extensively
aﬀect the open coast (for example, a 3–4 hPa pressure jump and a factor of 10
resonant ampliﬁcation, will only produce ocean wave heights of 30–40 cm). It is
when energetic ocean waves arrive at the entrance of a semiclosed coastal basin
(bay, inlet, fjord or harbor) that they can induce hazardous oscillations in the basin
due to harbor resonance.
On the other hand, intense oscillations inside a harbor (bay or inlet) can only
be formed if the external forcing (i.e., the waves arriving from the open sea)
are energetic enough. Seismically generated tsunami waves in the open ocean can
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
228 A. B. Rabinovich
be suﬃciently energetic even in the absence of additional resonant eﬀects (thus,
according to satellite altimetry measurements, tsunami waves generated by the 2004
Sumatra earthquake in the open Indian Ocean had troughtocrest wave heights of
approximately 1.0–1.2m,
86
while atmospherically generated tsunamilike can reach
such potentially dangerous levels only in the case of some external resonance. This
is an important diﬀerence between tsunami waves and meteotsunamis.
It follows from expression (9.17) that a large Qfactor is critical but that anoma
lously pronounced harbor oscillations can only be produced when there is resonant
matching between the dominant frequency (f) of the arriving (external) waves and
an eigen frequency f
0
of the harbor (normally, the eigen frequency of the funda
mental — Helmholtz — harbor mode). This means that catastrophic harbor oscilla
tions are the result of a double resonance eﬀect
51,68
: (a) external resonance between
the moving atmospheric disturbances and openocean waves; and (b) internal res
onance between the arriving openocean waves and the fundamental eigen mode of
the harbor (bay, inlet). An additional favorable factor is the speciﬁc direction of the
propagating atmospheric waves (and corresponding openocean waves) toward the
entrance of the harbor (bay).
Summarizing what has been presented above, we can formulate the particular
conditions promoting creation of extreme atmospherically induced oscillations near
the coast (meteotsunamis) as follows:
• A harbor (bay, inlet or fjord) with deﬁnite resonant properties and high Qfactor.
• The occurrence of strong smallscale atmospheric disturbance (a pressure jump or
a train of internal atmospheric waves).
• A propagation direction that is headon toward the entrance of the harbor.
• The occurrence of an external resonance (Proudman, Greenspan or shelf) between
the atmospheric disturbance and ocean waves.
• The occurrence of internal resonance between the dominant frequency of the
incoming openocean waves and the fundamental harbor mode frequency.
Due to these necessary levels of matching between the atmospheric disturbance,
the openocean bathymetry and the shelfharbor geometries, the direction and speed
of the atmospheric disturbance probably are even more important than the actual
energy content of the incoming waves. In any case, the necessary coincidence of
several factors signiﬁcantly diminishes the possibility of these events occurring, and
is the main reason why this phenomenon is relatively rare and restricted to speciﬁc
locations.
68
Honda et al.
30
and Nakano and Unoki
59
investigated more than 115 gulfs, bays,
inlets, and harbors of the Japanese coast and found that highly destructive seiches
(not associated with tsunami waves) occur only in a few of them. Extremely strong
seiche oscillations (socalled “abiki” waves) are periodically excited in Nagasaki Bay.
In particular, the abiki waves of 31 March 1979 with periods of about 35 min reached
wave heights of 478 cm at the northern end of the bay and killed three people.
1,29
High meteotsunami risk in certain exceptional locations mainly arises from the
combination of shelf topography and coastline geometry coming together to create a
multiple resonance eﬀect. The factors (internal and external) of critical importance
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 229
are: (1) welldeﬁned resonant characteristics of the harbor (bay, inlet, etc.); and
(2) speciﬁc properties of the shelf favorable for external resonance (between atmo
spheric and openocean waves) and internal resonance (between arriving openocean
waves and harbor oscillations). The combination of these factors for some particular
sites is like a “timebomb”: sooner or later it will explode (when the atmospheric
disturbance is strong enough and the parameters of disturbance coincide with the
resonant parameters of the corresponding topography/geometry). Locations with
known regular extreme seiches are just the places for these “timebombs”.
51
The catastrophic abiki wave event of 31 March 1979 best illustrates the physical
mechanisms responsible for the generation of meteotsunamis [Fig. 9.10(a)]. Hibiya
and Kajiura
29
(HK in the following text) examined this event in detail and con
structed an eﬀective numerical model that agrees well with observational data.
Nagasaki Bay is a narrow, elongated bay located on the western coast of Kyushu
Island, Japan [Fig. 9.10(b)]; the length of the bay is about 6 km, the width is 1 km,
and the mean depth is 20 m. The fundamental period of the bay (Helmholtz mode)
is 35 min, and this period prevails in seiche oscillations inside the bay (95% of all
observed events) and it was speciﬁcally this period that was observed on 31 March
1979.
1
HK noticed that almost all known cases of signiﬁcant abiki waves are asso
ciated with pressure jumps. In the case of the 1979 event, there was an abrupt
pressure jump (∆P
a
) of 2–6 hPa (according to the observations at several sites)
that propagated eastward (more precisely, 5.6
◦
north of east) over the East China
Sea with an approximate mean speed U = 31 m/s (Fig. 9.5). HK approximated
this jump as ∆P
a
= 3 hPa over a linear increase distance L
1
= 28 km and a linear
decrease distance L
2
= 169 km. So, the corresponding static inverted barometer
response of sea level was ∆
¯
ζ ≈ −3 cm [Fig. 9.10(a)]. Moreover, the depth of the
East China Sea between mainland China and Kyushu Island is between 50 and
150 m, and the corresponding longwave speed c ≈ 22–39 m/s. Thus, it was a clas
sical example of Proudman resonance. HK presented a simple expression describing
resonant ampliﬁcation of forced openocean long waves as:
∆ζ =
∆
¯
ζ
L
1
x
f
2
, (9.29)
where x
f
= Ut is the distance traveled by the pressure jump during time t. If
L
1
= 28 km and x
f
= 300 km [from the source area to the Goto Islands — see
Fig. 9.10(b)], then ∆ζ ≈ 16 cm. More precise numerical computation with real
istic twodimensional bathymetry gives the resonant factor ε = ∆ζ/∆
¯
ζ = 4.3 and
∆ζ ≈ 12.9 cm in good agreement with observation. Therefore, due to the reso
nance, the initial disturbance of 3 cm increased in the open sea by four to ﬁve times
[Fig. 9.10(a)]. It is interesting to note that the resonant ampliﬁcation is inversely
proportional to L
1
[see Eq. (9.29)], so the faster the change in atmospheric pressure
(the more abrupt is the pressure jump), the stronger is the ampliﬁcation of the
generated waves (HK).
According to the HK computations, the outer shelf region between the Goto
Islands and the mainland of Kyushu (“Goto Nada”) has resonant periods of 64, 36,
and 24 min. The second period (36 min) almost coincides with the fundamental
period of Nagasaki Bay (35 min). The Goto Nada shelf did not signiﬁcantly amplify
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
230 A. B. Rabinovich
1 2 3
U c
Proudman resonance
4.8 m
2.8 m
1.3 m
45 cm
Nagasaki Bay
inlet
shelf
measured at
tide gauges
harbor
resonance
shelf
amplification
16 cm
3 cm
wave amplification
through Proudman resonance
East China Sea
airpressure disturbance
3 hPa
U = 31 m/s
Korea
China
Japan
East China
Sea
1
0
0
5
0
0
P
Nagasaki
200
km
0
31 m/s
(b)
Kyushu
I.
a
1
2
3
Nagasaki
Bay
1 2
3
km
5 0
Kyushu I. Nezumi (1)
31 March 1979
Nagasaki (2)
(c)
10 12 14 16 18hr
100
50
50
0
0
50
50
100
150
S
e
a
l
e
v
e
l
(
c
m
)
(a)
Goto Is
Fig. 9.10. (a) A sketch illustrating the physical mechanism for formation of the catastrophic
meteotsunami at Nagasaki Bay (Japan) on 31 March 1979. Numbers “1”, “2”, and “3” correspond
to locations shown in (b). (b) Map of Nagasaki Bay and the initial atmospheric pressure disturbance
(shaded rectangular region). (c) Tide records of the catastrophic “abiki waves” of 31 March 1979
at Nezumi (9.1) and Nagasaki (2); positions of the tide gauges are shown in the inset in panel (b).
the incoming wave (the ﬁrst crest height was 16 cm at the shelf depth of 60 m)
but it selected and ampliﬁed waves with speciﬁc periods, in particular those with
a period of 36 min. Between the outer sea (depth 60 m) and the head of Nagasaki
Bay, the arriving waves were ampliﬁed by a factor of 2.4 due to the combined eﬀects
of topographic convergence, partial reﬂection, and shoaling inside the bay. Finally,
resonant ampliﬁcation in Nagasaki of incoming wave train with a period of about
35 min formed catastrophic oscillations within the bay with a maximum recorded
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 231
wave height of 278 cm [as measured by a tide gauge located in the middle of the
bay — see Fig. 9.10(c)] and an estimated wave height in the head of the bay of
478 cm.
1
Thus, for this extreme event, we observe the full combination of “hazardous”
conditions (factors) responsible for the formation catastrophic oscillations inside
Nagasaki Bay: (1) A pronounced atmospheric disturbance (pressure jump of
2–6hPa), (2) propagating toward the bay with (3) nearresonant phase speed of
31 m/s; this disturbance resonantly generated opensea long waves with selected
(over the shelf) 36 min period that matched (4) the fundamental 35min period of
the bay that has (5) high Qfactor and welldeﬁned resonant properties. As a result,
3 cm ocean waves in the source area resulted in 478 cm waves at the head of the bay
(Fig. 9.4).
Analysis of destructive meteotsunami events in the Mediterranean
18,22,50,51,64,
71,72,91,92
indicate that the physical mechanisms of these events were similar to
those for Nagasaki Bay event. Tides in the Mediterranean are small; consequently,
harbors are not designed to accommodate large amplitude sea level changes asso
ciated with occasional meteotsunamis. Consequently, it is atmospherically generated
phenomena (not ordinary tsunamis) that are normally responsible for signiﬁcant
ﬂooding and damage in this region. However, the main reason for the damaging
nature of meteotsunamis is likely due to the strong currents in the harbor that
accompany the sea level oscillations. Seiches with a 10 min period give raise to cur
rents that are 70 times stronger than semidiurnal tides having the same amplitude.
Acknowledgments
This work was initiated by Professor Fred Raichlen (CalTech, Pasadena, CA); the
author sincerely appreciates his help and friendly support. He is also very grateful
to Professor Young Kim (California State University, Los Angeles), the Editor of
this Handbook, for his patience and useful comments and to Drs. Sebastian Mon
serrat (Universitat de les Illes Balears, Palma de Mallorca, Spain) and Ivica Vilibi´c
(Institute of Oceanography and Fisheries, Split, Croatia) for their assistance and
providing various observational data. Dr. Richard Thomson (Institute of Ocean Sci
ences, Sidney, BC, Canada) did tremendous work editing this chapter and encour
aging the author, Patricia Kimber (Sidney, BC) helped to draft the ﬁgures. Partial
ﬁnancial support was provided by the Russian Foundation on Basic Research, grants
060565210a, 080513582, 090500599 and 090501125a and by NATO Science for
Peace project SfP981382.
References
1. H. Akamatsu, On seiches in Nagasaki Bay, Pap. Meteor. Geophys. 33(2), 95–115
(1982).
2. A. Barberopoulou, A. Qamar, T. L. Pratt and W. P. Steele, Longperiod eﬀects of the
Denali Earthquake on water bodies in the Puget lowland: Observations and modeling,
Bull. Seism. Soc. Amer. 96(2), 519–535 (2006).
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
232 A. B. Rabinovich
3. J. A. Battjes, Surfzone dynamics, Ann. Rev. Fluid Mech. 20, 257–293 (1988).
4. W. A. M. Botes, K. S. Russel and P. Huizinga, Modeling harbor seiching compared
to prototype data, Proc. 19th Int. Coastal Eng. Conf., Houston (1984), pp. 846–857.
5. A. J. Bowen and D. A. Huntley, Waves, long waves and nearshore morphology, Marine
Geol. 60, 1–13 (1984).
6. E. C. Bowers, The modelling of waves and their eﬀects in harbors, in Hydraulic Mod
elling in Maritime Engineering (Thomas Telford, London, 1982), pp. 121–127.
7. J. Candela, S. Mazzola, C. Sammari, R. Limeburner, C. J. Lozano, B. Patti and
A. Bonnano, The “Mad Sea” phenomenon in the Strait of Sicily, J. Phys. Oceanogr.
29, 2210–2231 (1999).
8. D. C. Chapman and G. S. Giese, Seiches, Encyclopedia of Ocean Sciences, eds
J. H. Steele, K. K. Turekian and S. A. Thorpe, Vol. 5 (Academic Press, London, 2001)
pp. 2724–2731.
9. P. Colucci and A. Michelato, An approach to study of the ‘Marubbio’ phenomenon,
Boll. Geoﬁs. Theor. Appl. 13, 3–10 (1976).
10. A. Defant, Physical Oceanography, Vol. 2 (Pergamon Press, Oxford, UK, 1961), 598 pp.
11. M. P. C. de Jong, L. H. Holthuijsen and J. A. Battjes, Generation of seiches
by cold fronts over the southern North Sea, J. Geophys. Res. 108, 3117 (2003),
doi:10.1029/2002JC001422.
12. M. P. C. de Jong and J. A. Battjes, Lowfrequency sea waves generated by atmospheric
convection cells, J. Geophys. Res. 109, C01011 (2004), doi:10.1029/2003JC001931.
13. V. A. Djumagaliev, E. A. Kulikov and S. L. Soloviev, Analysis of ocean level oscilla
tions in Malokurilsk Bay caused by tsunami of 16 February 1991, Science Tsunami
Hazards, 11, 47–58 (1993).
14. V. A. Djumagaliev, A. B. Rabinovich and I. V. Fine, Theoretical and experimental esti
mation of transfer peculiarities of the Malokurilsk Bay coast, the Island of Shikotan,
Atmosph. Oceanic Phys. 30, 680–686 (1994).
15. W. L. Donn, Alaskan earthquake of 27 March 1964: Remote seiche simulation, Science
145, 261–262 (1964).
16. A. F. Drago, A study on the sea level variations and the ‘Milghuba’ phenomenon in the
coastal waters of the Maltese Islands, Ph.D. thesis, University of Southampton (1999).
17. E. Fontser´e, Les ‘seixes’ de la costa catalana. Servei Meteorol`ogic de Catalunya, Notes
d’Estudi, 58 (1934) (in Catalan).
18. M. Garcies, D. Gomis and S. Monserrat, Pressureforced seiches of large amplitude
in inlets of the Balearic Islands. Part II: Observational study, J. Geophys. Res. 101,
6453–6467 (1996).
19. G. S. Giese and D. C. Chapman, Coastal seiches, Oceanus 36, 38–46 (1993).
20. G. S. Giese, D. C. Chapman, P. G. Black and J. A. Fornshell, Causations of large
amplitude harbor seiches on the Caribbean coast of Puerto Rico, J. Phys. Oceanogr.
20, 1449–1458 (1990).
21. G. S. Giese and R. B. Hollander, The relation between coastal seiches at Palawan
Island and tidegenerated internal waves in the Sulu Sea, J. Geophys. 92, 5151–5156
(1987).
22. D. Gomis, S. Monserrat and J. Tintor´e, Pressureforced seiches of large amplitude in
inlets of the Balearic Islands, J. Geophys. Res. 98, 14437–14445 (1993).
23. D. G. Goring, Rissaga (longwave events) on New Zealand’s eastern seaboard: A
hazard for navigation, Proc. 17th Australasian Coastal Ocean Eng. Conf., 20–23
September 2005, Adelaide, Australia (2005), pp. 137–141.
24. E. E. Gossard and B. H. Hooke, Waves in Atmosphere (Elsevier, New York, 1975),
456 pp.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 233
25. H. P. Greenspan, The generation of edge waves by moving pressure disturbances,
J. Fluid Mech. 1, 574–592 (1956).
26. P. F. Hamblin, Seiches, circulation and storm surges of an icefree Lake Winnipeg,
J. Fisheries Res. Board Canada 33, 2377–2391 (1976).
27. D. L. Harris, The eﬀect of a moving pressure disturbance on the water level in a lake,
Meteorol. Monogr. 2, 46–57 (1957).
28. R. F. Henry and T. S. Murty, Tsunami ampliﬁcation due to resonance in Alberni Inlet:
Normal modes, inTsunami: Progress in Prediction, Disaster Prevention and Warning,
eds. Y. Tsuchiya and N. Shuto (Kluwer Acad. Publ., Dordrecht, The Netherlands,
1995), pp. 117–128.
29. T. Hibiya and K. Kajiura, Origin of ‘Abiki’ phenomenon (kind of seiches) in Nagasaki
Bay, J. Oceanogr. Soc. Japan 38, 172–182 (1982).
30. K. Honda, T. Terada, Y. Yoshida and D. Isitani, An investigation on the secondary
undulations of oceanic tides, J. College Sci., Imper. Univ. Tokyo (1908), 108 pp.
31. G. E. Hutchinson, A Treatise on Limnology. Volume 1. Geography, Physics and Chem
istry (J. Wiley, New York, 1957), 1015 pp.
32. G. H. Keulegan and L. H. Carpenter, Forces on cylinders and plates in an oscillating
ﬂuid, J. Res. Nat. Bureau Stand. 60, 423–440 (1958).
33. B. Kinsman, Wind Waves: Their Generation and Propagation on the Ocean Surface
(Prentice Hall, Englewood Cliﬀs, 1965), 676 pp.
34. B. J. Korgen, Seiches, Amer. Sci. 83, 330–341 (1995).
35. H. Lamb, Hydrodynamics, 6th Edn. (Dover, New York, 1945), 738 pp.
36. J. J. Lee, Wave induced oscillations in harbors of arbitrary geometry, J. Fluid Mech.
45, 375–394 (1971).
37. B. Le M´ehaut´e and B. W. Wilson, Harbor paradox (discussion), J. Waterways Harbor
Division, ASCE 88, 173–195 (1962).
38. P. L.F. Liu, S. Monserrat, M. Marcos and A. B. Rabinovich, Coupling between two
inlets: Observation and modeling, J. Geophys. Res. 108, 3069 (2003), doi:10.1029/
2002JC001478, 14(1–10).
39. M. S. LonguetHiggins and R. W. Stewart, Radiation stress and mass transport in
gravity waves, with application to ‘surfbeats’, J. Fluid Mech. 13, 481–504 (1962).
40. A. McGarr, Excitation of seiches in channels by seismic waves, J. Geophys. Res. 70,
847–854 (1965).
41. C. C. Mei, The Applied Dynamics of Ocean Surface Waves (World Scientiﬁc, London
1992), 740 pp.
42. M. A. Merriﬁeld et al., Tide gage observations of the Indian Ocean tsunami, December
26, 2004, Geophys. Res. Lett. 32, L09603 (2005), doi:10.1029/2005GL022610.
43. M. Metzner, M. Gade, I. Hennings and A. B. Rabinovich, The observation of seiches
in the Baltic Sea using a multi data set of water levels, J. Marine Systems 24, 67–84
(2000).
44. G. N. Mikishev and B. I. Rabinovich, Dynamics of a Solid Body with Cavities Partially
Filled with Liquid (Mashinostroyeniye Press, Moscow, 1968), 532 pp. (in Russian).
45. J. W. Miles, Ring damping of free surface oscillations in a circular tank, J. Appl.
Mech. 25, 26–32 (1958).
46. J. W. Miles, Harbor seiching, Ann. Rev. Fluid Mech. 6, 17–36 (1974).
47. J. Miles and W. Munk, Harbor paradox, J. Waterways Harbor Division, ASCE 87,
111–130 (1961).
48. G. R. Miller, Relative spectra of tsunamis, HIG72–8 Hawaii Inst. Geophys (1972),
7 pp.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
234 A. B. Rabinovich
49. S. Monserrat, A. Ibberson and A. J. Thorpe, Atmospheric gravity waves and the
“rissaga” phenomenon, Q. J. R. Meteorol. Soc. 117, 553–570 (1991).
50. S. Monserrat, A. B. Rabinovich and B. Casas, On the reconstruction of the transfer
function for atmospherically generated seiches, Geophys. Res. Lett. 25, 2197–2200
(1998).
51. S. Monserrat, I. Vilibi´c and A. B. Rabinovich, Meteotsunamis: Atmospherically
induced destructive ocean waves in the tsunami frequency band, Nat. Hazards Earth
Syst. Sci. 6, 1035–1051 (2006).
52. W. H. Munk, Surf beats, Eos, Trans. Amer. Geophys. Un. 30, 849–854 (1949).
53. W. H. Munk, Long waves, The Sea (J. Wiley, New York, 1962), pp. 647–663.
54. T. S. Murty, Seismic Sea Waves — Tsunamis (Bull. Fish. Res. Board Canada 198,
Ottawa, 1977), 337 pp.
55. T. S. Murty, Modiﬁcation of hydrographic characteristics, tides, and normal modes
by ice cover, Marine Geodesy 9, 451–468 (1985).
56. M. Nakano, Secondary undulations in bays forming a coupled system, Proc. Phys.
Math. Soc. Japan 3, 372–380 (1932).
57. M. Nakano, Possibility of excitation of secondary undulations in bays by tidal or
oceanic currents, Proc. Imp. Acad. Japan 9, 152–155 (1933).
58. M. Nakano and T. Abe, Standing oscillation of bay water induced by currents, Records
Oceanogr. Works in Japan, Spec. No. 3, 75–96 (1959).
59. M. Nakano and S. Unoki, On the seiches (secondary undulations of tides) along the
coast of Japan, Records Oceanogr. Works Japan, Spec. No. 6, 169–214 (1962).
60. M. Nakano and N. Fujimoto, Seiches in bays forming a coupled system, J. Oceanogr.
Soc. Japan 43, 124–134 (1987).
61. T. Nomitsu, A theory of tsunamis and seiches produced by wind and barometric
gradient, Mem. Coll. Sci. Imp. Univ. Kyoto A 18, 201–214 (1935).
62. M. Okihiro, R. T. Guza and R. J. Seymour, Excitation of seiche observed in a small
harbor, J. Geophys. Res. 98, 18201–18211 (1993).
63. J. OltmanShay and R. T. Guza, Infragravity edge wave observations on two California
beaches, J. Phys. Oceanor. 17, 644–663 (1987).
64. M. Orli´c, About a possible occurrence of the Proudman resonance in the Adriatic,
Thalassia Jugoslavica 16, 79–88 (1980).
65. J. Proudman, The eﬀects on the sea of changes in atmospheric pressure, Geophys.
Suppl. Mon. Notices R. Astr. Soc. 2, 197–209 (1929).
66. D. Prandle, The use of wave resonators for harbor protection, Dock Harbour Author
(1974), pp. 279–280.
67. A. B. Rabinovich, Possible vorticity eﬀect in longwave motions (surging) in harbors,
Trans. (Doklady) Russian Acad Scie, Earth Sci. Sec. 325, 224–228 (1992).
68. A. B. Rabinovich, Long Ocean Gravity Waves: Trapping, Resonance, and Leaking
(Gidrometeoizdat, St. Petersburg, 1993), 325 pp. (in Russian).
69. A. B. Rabinovich, Spectral analysis of tsunami waves: Separation of source and topog
raphy eﬀects, J. Geophys. Res. 102, 12663–12676 (1997).
70. A. B. Rabinovich and A. S. Levyant, Inﬂuence of seiche oscillations on the formation
of the longwave spectrum near the coast of the Southern Kuriles, Oceanology 32,
17–23 (1992).
71. A. B. Rabinovich and S. Monserrat, Meteorological tsunamis near the Balearic and
Kuril Islands: Descriptive and statistical analysis, Nat. Hazards 13, 55–90 (1996).
72. A. B. Rabinovich and S. Monserrat, Generation of meteorological tsunamis (large
amplitude seiches) near the Balearic and Kuril Islands, Nat. Hazards 18, 27–55 (1998).
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
Seiches and Harbor Oscillations 235
73. A. B. Rabinovich, S. Monserrat and I. V. Fine, Numerical modeling of extreme seiche
oscillations in the region of the Balearic Islands, Oceanology 39, 16–24 (1999).
74. A. B. Rabinovich, R. E. Thomson and F. E. Stephenson, The Sumatra tsunami of 26
December 2004 as observed in the North Paciﬁc and North Atlantic oceans, Surveys
Geophys. 27, 647–677 (2006).
75. A. B. Rabinovich and R. E. Thomson, The 26 December 2004 Sumatra tsunami:
Analysis of tide gauge data from the World Ocean Part 1. Indian Ocean and South
Africa, Pure Appl. Geophys. 164, 261–308 (2007).
76. A. B. Rabinovich, L. I. Lobkovsky, I. V. Fine, R. E. Thomson, T. N. Ivelskaya and
E. A. Kulikov, Nearsource observations and modeling of the Kuril Islands tsunamis
of 15 November 2006 and 13 January 2007, Adv. Geosci. 14, 105–116 (2008).
77. B. I. Rabinovich and Y. V. Tyurin, Numerical Conformal Mapping in Two
Dimensional Hydrodynamics (Space Research Institute RAS, Moscow, 2000), 312 pp.
78. F. Raichlen, Harbor resonance, Estuary and Coastline Hydrodynamics, ed. A. T. Ippen
(McGraw Hill Book Co., New York, 1966), pp. 281–340.
79. F. Raichlen, The eﬀect of shipharbor interactions on port operations, Maritime
Technol., 11–20 (2002).
80. F. Raichlen and J. J. Lee, Oscillation of bays, harbors and lakes, Handbook of Coastal
and Ocean Engineering, ed. J. B. Herbich (Gulf Publishing Company Houston, Texas,
1992), pp. 1073–1113.
81. C. Ramis and A. Jans`a, Condiciones meteorol´ogicas simult´aneas a la aparici´on de
oscilaciones del nivel del mar de amplitud extraordinaria en el Mediterr´ aneo occi
dental, Rev. Geof´ısica 39, 35–42 (1983) (in Spanish).
82. T. Sawaragi and M. Kubo, Longperiod motions of a moored ship induced by harbor
oscillations, Coast. Eng. Japan 25, 261–275 (1982).
83. D. J. Schwab and D. B. Rao, Gravitational oscillations of Lake Huron, Saginaw Bay,
Georgian Bay and the North Channel, J. Geophys. Res. 82, 2105–2116 (1977).
84. R. M. Sorensen and E. F. Thompson, Harbor hydrodynamics, Coastal Engineering
Manual, Part II (U.S. Army Corps. of Engineers, Washington, D.C., New York, 2002),
Chapter 7, pp. 1–92.
85. J. Tintor´e, D. Gomis, S. Alonso and D. P. Wang, A theoretical study of large sea level
oscillations in the Western Mediterranean, J. Geophys. Res. 93, 10797–10803 (1988).
86. V. V. Titov, A. B. Rabinovich, H. Mofjeld, R. E. Thomson and F. I. Gonz´alez, The
global reach of the 26 December 2004 Sumatra tsunami, Science 309, 2045–2048
(2005).
87. R. E. Thomson, P. W. Vachon and G. A. Borstad, Airborne synthetic aperture radar
imagery of atmospheric gravity waves, J. Geophys. Res. 97, 14249–14257 (1992).
88. R. E. Thomson, A. B. Rabinovich and M. V. Krassovski, Double jeopardy: Concurrent
arrival of the 2004 Sumatra tsunami and stormgenerated waves on the Atlantic
coast of the United States and Canada, Geophys. Res. Lett. 34, L15607 (2007),
doi:10.1029/2007GL030685.
89. M. J. Tucker, Surf beats: Sea waves of 1 to 5 minute period, Proc. Roy. Soc. London
Ser. A 202, 565–573 (1950).
90. W. G. Van Dorn, Some tsunami characteristics deducible from tide records, J. Phys.
Oceanogr. 14, 353–363 (1984).
91. I. Vilibi´c and H. Mihanovi´c, A study of resonant oscillations in the Split harbor
(Adriatic Sea), Estuar. Coastal Shelf Sci. 56, 861–867 (2003).
92. I. Vilibi´c, N. Domijan, M. Orli´c, N. Leder and M. Pasari´c, Resonant coupling of a
traveling airpressure disturbance with the east Adriatic coastal waters, J. Geophys.
Res. 109, C10001 (2004), doi:10.1029/2004JC002279.
July 31, 2009 8:18 9.75in x 6.5in b684ch09 FA
236 A. B. Rabinovich
93. X. Wang, K. Li, Z. Yu and J. Wu, Statistical characteristics of seiches in Longkou
harbor, J. Phys. Oceanogr. 17, 1063–1065 (1987).
94. R. L. Wiegel, Tsunamis, storm surges, and harbor oscillations, Oceanographical Engi
neering (PrenticeHall, Englewood Cliﬀs, NJ, 1964), Ch. 5, pp. 95–127.
95. B. Wilson, Seiches, Advances in Hydrosciences 8, 1–94 (1972).
96. J.K. Wu and P. L.F. Liu, Harbor excitations by incident wave groups, J. Fluid Mech.
217, 595–613 (1990).
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Chapter 10
Finite Diﬀerence Model for
Practical Simulation of Distant Tsunamis
Sung Bum Yoon
Department of Civil and Environmental Engineering
Hanyang University Ansan, 426791, Korea
sbyoon@hanyang.ac.kr
A practical numerical model based on the dispersioncorrection ﬁnite diﬀerence
scheme of Yoon et al. [Terres. Atmos. Oceanic Sci. 18(1), 31–53 (2007)] equipped
with the grid nesting scheme of Lim et al. [Natural Hazards 47(1), 95–118 (2008)]
is introduced. The model is applied to simulate the propagation of a historical
tsunami event that attacked the east coast of Korea. The calculated free surface
displacements are compared with the observations at two tidal stations along the
east coast of Korea. The comparison shows that the results agree well with the
observations. The analyses of the simulated results show that the underwater
topography such as submerged rises and ridges plays an important role in the
propagation of tsunamis in this region.
10.1. Introduction
Tsunamis are the ocean water surface waves generated by an undersea earth
quake, a landslide, a volcanic eruption, or even a meteoric impact on water surface.
As crust movement of the earth became active, the region along the rim of the
Paciﬁc Ocean experienced largescale tsunamis including the 1946 Alaskan tsunami,
the 1960 Chilean tsunami, and the 1964 Alaskan tsunami, for instance. These
largescale tsunamis propagated across the Paciﬁc Ocean, and attacked the coastal
area along the Paciﬁc rim. The damage including loss of human lives caused
by these catastrophic tsunamis was well documented by Lander and Lockridge.
3
More recent tsunamis in this area have been investigated by many scientists, and
reported by Satake et al.
4
for the 1992 Nicaragua tsunami, and Hokkaido Tsunami
Survey Group
5
for the 1993 Hokkaido tsunami. On the other hand, largescale
tsunamis including the 1992 and 1994 Indonesia tsunamis, 1995 Philippine tsunami,
1996 Indonesia tsunami, and 1998 Papua New Guinea tsunami were also generated
along the rim of the Indian Ocean. In particular, the tsunami of magnitude 9.3
237
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
238 S. B. Yoon
was generated oﬀ the west coast of Sumatra in the Indian Ocean on 26 December
2004. The west coast of northern Sumatra was most severely damaged by this
tsunami. The coastal areas along the rim of the Indian Ocean including Malaysia,
Thailand, Myanmar, Bangladesh, India, Sri Lanka, Arabian, and African countries
were attacked by this tsunami. More than 290,000 people were killed due to this
tsunami event. And the property damages and injuries cannot be countable. In
order to reduce such damages caused by tsunamis, a proper monitoring, forecasting,
and warning system should be established. For this we need to fully understand
the mechanism such as generation, propagation, and inundation of tsunamis, and
predict the possible quantities such as arrival times and runup heights of a tsunami
event. Thus, the development of an accurate and eﬃcient numerical model for com
puting all the aspects of tsunami propagation is indispensable.
When the tsunamis are generated by an earthquake, they have a shape of
solitary wave with a long crest. After they propagate a long distance from the
source area over a deep ocean, the tsunamis evolve into a train of waves due to
the dispersion eﬀect of waves. When the source area is narrow, the dispersion
eﬀect plays an important role in the deformation of tsunamis also in the rela
tively shallow water area. Boussinesq equations are one of the best choices for gov
erning equations to construct a numerical model for the farﬁeld tsunamis, because
they can take into account the dispersion eﬀects. The numerical models such as
FUNWAVE
6
and COULWAVE
7,8
were developed based on a fully nonlinear and
dispersive Boussinesq equations, and were applied to simulate the 26 December
2004 Indian Ocean tsunami.
9,10
However, these Boussinesq models require a small
mesh size to suppress the numerical dispersion error. The small mesh system con
sumes huge computer resources due to the implicit nature of the solution technique
to deal with dispersion terms. Thus, the Boussinesq model is not preferred for the
simulation of the farﬁeld tsunamis, and the shallowwater equations are generally
employed instead.
Considerable eﬀorts to construct a numerical model for the simulation of
tsunamis have been made. For example, ﬁnite diﬀerence models were developed by
Hwang and Divoky,
11
Goto and Shuto,
12
Kowalik and Murty,
13
and Mader and
Curtis.
14
Most of the existing numerical models are based on the shallowwater
equations and have been successfully applied to the nearﬁeld tsunamis which have
little dispersion eﬀect of waves. However, distant tsunamis generated far from the
region of interest are mainly transformed due to the accumulation of dispersion
eﬀects during the propagation. Thus, the numerical models based on the shallow
water equations will suﬀer from the lack of accuracy for the simulation of distant
tsunamis.
Imamura et al.
15
presented a ﬁnite diﬀerence model for the simulation of trans
oceanic propagation of tsunamis. The model solves the shallowwater equations
using a leapfrog scheme. The physical dispersion is compensated by the numerical
dispersion introduced by the truncation error of the numerical scheme. This can be
done only if the grid size is appropriately selected for the given water depth and
the time step satisfying the criterion proposed by Imamura et al.
15
Cho
16
improved
the numerical model of Imamura et al.
15
for tsunamis obliquely propagating to
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 239
the principal axes of computational grids. However, strictly speaking, those models
have no capability to consider the dispersion eﬀect of waves in the case of practical
application. The use of numerical models developed by Imamura et al.
15
and Cho
16
is limited to the case of constant water depth. Yoon
17
developed a ﬁnitediﬀerence
scheme that uses a uniform grid, but the actual computations are made on a hidden
grid of variable size determined from the condition proposed by Imamura et al.
15
This model satisﬁes local dispersion relationships of waves for a slowly varying
topography. However, the accuracy of the model becomes degraded when applied to
short waves due to numerical errors associated with the interpolation of quantities
at hidden grid points.
Yoon et al.
1
developed an eﬃcient and relatively accurate ﬁnitediﬀerence model
to simulate the propagation of distant tsunamis over varying topography. This
ﬁnitediﬀerence model solves a linear Boussinesqtype wave equation employing a
uniform ﬁnitediﬀerence grid. A considerable accuracy is achieved by eliminating the
numerical dispersion error and by providing the physical dispersion. The dispersion
correction is performed by controlling the parameters determined from the given
time step, grid size, and local water depth. As a result, an accurate dispersion
correction is possible although the grid size does not meet Imamura’s mesh
condition.
As the tsunamis approach the coastal area, they experience a shoaling due to
decreasing water depth over a continental shelf. The wavelength becomes short, and
the grid resolution becomes low near the coastline. Thus, the accuracy of the far
ﬁeld model proposed by Yoon et al.
1
is degraded. This diﬃculty can be solved by
nesting a ﬁner grid system to the coarse grid transoceanic model. As the tsunamis
approach the coastline, nonlinearity and bottom friction play a signiﬁcant role in
the transformation of tsunami waves. In this case, a nearﬁeld model such as full
Boussinesq or shallowwater equation model can be nested to the transoceanic
model as described in Yoon
17
and Lim et al.
2
Thus, various nesting schemes between
systems with diﬀerent grid sizes or governing equations are indispensable to extend
the applicability of the farﬁeld numerical model.
The objective of this chapter is to review a recent development of a prac
tical ﬁnitediﬀerence model to simulate tsunami propagation and runup for con
structing inundation maps. The contents in this chapter are not an original work
but a remake of research works presented by Yoon et al.
1
and Lim et al.
2,18
Section 10.2 describes governing equations and numerical scheme employed in the
dispersioncorrection ﬁnitediﬀerence model developed by Yoon et al.
1
and Lim
et al.
2,18
Section 10.3 presents the test of numerical model for uniform and variable
depth cases. Section 10.4 describes the nearﬁeld model. Section 10.5 introduces
the grid nesting scheme developed by Lim et al.
2,18
to extend the applicability of
the numerical model. In order to evaluate the applicability of the present model to
real tsunami events, a historical tsunami is simulated and the results are compared
with the observations at two harbors along the east coast of Korea in Sec. 10.6.
Then, the eﬀect of underwater topography on the propagation of tsunamis in this
sea is analyzed based on the simulated results. Finally, this study is summarized
in Sec. 10.7.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
240 S. B. Yoon
10.2. Governing Equations and Numerical Scheme
Tsunamis are generally regarded as long waves in comparison with the wind
generated waves. However, tsunamis can be short when the tsunami source is
narrow. In this case the dispersion eﬀect can play an important role in the trans
formation of tsunami waves. The dispersion eﬀect, rather than the nonlinear or
the bottom friction eﬀects, dominates the transformation of the farﬁeld tsunamis.
For the farﬁeld tsunami, the linear Bousinessq equations, which neglect the non
linear eﬀects but include the dispersion eﬀects, can be used as a governing equation.
On the other hand, the nonlinear eﬀect and the bottom friction eﬀect become
important for the transformation of tsunamis in the nearﬁeld where the water is
shallow near the coastal area. For the nearﬁeld tsunami, the nonlinear shallow
water equations are generally employed as a governing equation. In the present
numerical model, the linear Boussinesqtype wave equation proposed by Yoon et al.
1
is employed for the farﬁeld tsunami propagation, and the nonlinear shallowwater
equations with the bottom friction term are used for the nearﬁeld tsunamis.
For transoceanic propagation of tsunamis, the nonlinearity of waves can be
neglected because the free surface displacement is much smaller in comparison with
the water depth. However, the dispersion eﬀect of tsunami waves should be con
sidered properly for the farﬁeld tsunamis. Thus, the following linear Boussinesq
equations can be used as governing equations.
∂ζ
∂t
+
∂P
∂x
+
∂Q
∂y
= 0 , (10.1)
∂P
∂t
+ gh
∂ζ
∂x
=
h
2
2
∂
∂x
¸
∂
∂x
∂P
∂t
+
∂
∂y
∂Q
∂t
−
h
3
6
∂
∂x
¸
∂
2
∂t∂x
P
h
+
∂
2
∂t∂y
Q
h
, (10.2)
∂Q
∂t
+ gh
∂ζ
∂y
=
h
2
2
∂
∂y
¸
∂
∂x
∂P
∂t
+
∂
∂y
∂Q
∂t
−
h
3
6
∂
∂y
¸
∂
2
∂t∂x
P
h
+
∂
2
∂t∂y
Q
h
, (10.3)
where ζ represents the free surface elevation from still water level (m), and P and
Q are the depthintegrated volume ﬂuxes (m
2
/s) in the x and ydirections, respec
tively, g is the acceleration of gravity (m/s
2
), and h is the still water depth (m).
The equations of motion (10.2) and (10.3) include the dispersion terms in the
righthand side. These dispersion terms cause numerical diﬃculties in practice
because of the mixed form of diﬀerentiations with respect to both time and space.
Consequently, it calls for the use of implicit scheme, which solves a matrix system.
The implicit scheme requires very ﬁne grids to reduce the numerical dispersion
errors inherent in the numerical scheme such as ﬁnitediﬀerence method. A ﬁne grid
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 241
needs the tremendous computational eﬀort with a huge computer memory space
and excessive execution time.
To develop an eﬃcient and relatively accurate numerical model for the propa
gation of dispersive tsunamis over slowly varying topography, Yoon et al.
1
derived
the following linear Boussinesqtype wave equation (LBTWE) from the linear
Boussinesq equations (10.1)–(10.3) as
∂
2
ζ
∂t
2
−
¸
∂
∂x
gh
∂ζ
∂x
+
∂
∂y
gh
∂ζ
∂y
−
h
2
3
¸
∂
2
∂x
2
∂
∂x
gh
∂ζ
∂x
+
∂
∂y
gh
∂ζ
∂y
+
∂
2
∂y
2
∂
∂x
gh
∂ζ
∂x
+
∂
∂y
gh
∂ζ
∂y
= 0 , (10.4)
where the water depth is assumed to be slowly varying in the computational domain.
This equation includes the dispersion terms in the last part. Equation (10.4) can
be split into two ﬁrstorder partial diﬀerential equations in time as
∂v
∂t
= g
∂h
∂x
∂ζ
∂x
+
∂h
∂y
∂ζ
∂y
+ gh
∂
2
ζ
∂x
2
+
∂
2
ζ
∂y
2
, (10.5)
∂ζ
∂t
= v −γ∆x
2
∂
2
v
∂x
2
+
∂
2
v
∂y
2
, (10.6)
where v denotes an auxiliary variable introduced for computational convenience,
and γ is the dispersioncorrection parameter. The solution of (10.5) and (10.6) is
identical with that of (10.4) if γ is set to be −h
2
/3∆x
2
. Since the linear Boussinesq
type wave equations (10.5) and (10.6) can be solved by a pure explicit numerical
scheme, the dispersive tsunami propagation over a large domain can be simulated
with low computational cost.
Yoon et al.
1
developed a dispersioncorrection ﬁnitediﬀerence model which
solves the equations (10.5) and (10.6) using a relatively simple explicit numerical
scheme given by
v
n+1/2
i,j
−v
n−1/2
i,j
∆t
+ g
¸
h
n
i+1,j
−h
n
i−1,j
2∆x
ζ
n
i+1,j
−ζ
n
i−1,j
2∆x
+
h
n
i,j+1
−h
n
i,j−1
2∆y
ζ
n
i,j+1
−ζ
n
i,j−1
2∆y
− gh
¸
(1 −α)
ζ
n
i+1,j
−2ζ
n
i,j
+ ζ
n
i−1,j
∆x
2
+
ζ
n
i,j+1
−2ζ
n
i,j
+ ζ
n
i,j−1
∆y
2
+
α
2
ζ
n
i+1,j−1
−2ζ
n
i,j−1
+ ζ
n
i−1,j−1
∆x
2
+
ζ
n
i+1,j+1
−2ζ
n
i,j+1
+ ζ
n
i−1,j+1
∆x
2
+
α
2
ζ
n
i−1,j+1
−2ζ
n
i−1,j
+ ζ
n
i−1,j−1
∆y
2
+
ζ
n
i+1,j+1
−2ζ
n
i+1,j
+ ζ
n
i+1,j−1
∆y
2
= 0,
(10.7)
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
242 S. B. Yoon
and
ζ
n+1
i,j
−ζ
n
i,j
∆t
−v
n+1/2
i,j
+ γ∆x
2
×
v
n+1/2
i+1,j
−2v
n+1/2
i,j
+ v
n+1/2
i−1,j
∆x
2
+
v
n+1/2
i,j+1
−2v
n+1/2
i,j
+ v
n+1/2
i,j−1
∆y
2
=0, (10.8)
where indices i (and j) and n represent a spatial grid point and a time level, respec
tively, as shown in Fig. 10.1. ∆x(= ∆y) is the spatial grid size, and ∆t denotes the
time step. The weighting factor α to get orientationfree solution is given as 1/6,
and the dispersioncorrection parameter γ is given by
γ =
∆x
2
−(4h
2
+ gh∆t
2
)
12∆x
2
. (10.9)
If the dispersioncorrection parameter γ calculated by (10.9) for given local water
depth, spatial grid size, and time step are used, a considerably accurate solution
can be achieved for the linear Boussinesq wave equation (10.4). Thus, the selection
of grid size ∆xis free from the condition proposed by Imamura et al.
15
This means
that the present dispersioncorrection scheme is much more ﬂexible in the selection
of grid size than the conventional scheme.
The stability analysis of the dispersioncorrection numerical scheme shows that
the stability criterion is C
r
≤ 0.67 and −0.125 ≤ γ ≤ 0.083, where C
r
is the
Courant number. The lower limit of γ imposes a limitation on the ratio of grid
size to water depth as 1.27. This means that the uniform grid size greater than
approximately 1.27 times of a local water depth must be employed for the stability
of the present dispersioncorrection scheme. To satisfy the stability criterion of the
dispersioncorrection ﬁnitediﬀerence model, the range of applicable water depth is
limited if a uniform grid size is used for varying water depth. For practical purposes,
this stability criterion can be solved by imposing intentionally a limitation on the
Fig. 10.1. Sketch of the arrangement of variables and grid points; (a) grid system in space, (b) grid
system in time.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 243
γ value within the range of −0.125 ≤ γ ≤ 0.083. This means that the γ value is set
to be −0.125 if the γ calculated by (10.9) is less than −0.125. The γ value is always
less than 0.083. Thus, the upper limit of γ is automatically satisﬁed for very large
grid size.
10.3. Test of Numerical Scheme
10.3.1. Tsunami propagation over constant depth region
In order to test the applicability of the present model, the propagation of tsunamis
is ﬁrst simulated with an initial Gaussian shape of water surface for the case of
various constant water depths, and the computed free surface displacements are
compared with the analytical solutions of the linear Boussinesq equations.
19
The
initial free surface proﬁle and the velocity of free surface movement are described
by (10.10) and (10.11), respectively, as
ζ(r, θ) = 2e
−(r/a)
2
, (10.10)
∂ζ(r, θ)
∂t
= 0, (10.11)
where a is the characteristic radius of Gaussian function, r(= (x
2
+y
2
)
1/2
) denotes
the distance from the center of the Gaussian hump, and θ is the angle from the
xaxis, as shown in Fig. 10.2.
The analytical solution of the linear Boussinesq equations is given by Carrier
19
as
ζ(r, t) =
∞
0
a
2
e
−(ak)
2
/4
k cos
√
ghkt
1 +
(kh)
2
3
¸
¸
J
0
(kr)dk, (10.12)
where J
0
is the 0th order Bessel function of the ﬁrst kind and g = 9.81m/s
2
. The
parameters used for the solution are a = 7500m, ∆x = 2086m, and ∆t = 6 s. The
numerical simulations using the present scheme are performed for various constant
water depths of h = 500m, 1000m, and 1500m.
Figures 10.3 and 10.4 present the comparisons of analytical solutions and
numerical results calculated considering or neglecting the dispersioncorrection
Fig. 10.2. Coordinate system and initial free surface proﬁle to test the accuracy of the present
numerical scheme.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
244 S. B. Yoon
Fig. 10.3. Comparison of free surface calculated using FDM and analytical solution for the case
of h = 500 m (∆x
Im
= 1085 m, ∆x = 2086 m): (a) γ = 0.0, α = 0.0
15
; (b) γ = 0.0, α = 1/6
16
;
(c) γ = 0.061, α = 1/6 (present).
1
parameters γ and the weighting factor α. We remark here that the numerical scheme
of Imamura et al.
15
is recovered if we set intentionally γ = α = 0.0.
The orientationfree scheme of Cho
16
is obtained if γ = 0.0 and α = 1/6.
These ﬁgures show a time history of free surface displacements at the location of
150∆x(r = 312, 900m) far from the center of the initial Gaussian hump.
The free surface proﬁles calculated by the present ﬁnitediﬀerence model consid
ering or neglecting dispersioncorrection are compared with the analytical solution
of Carrier
19
for the case of 500 m water depth in Fig. 10.3.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 245
Fig. 10.4. Comparison of free surface calculated using FDM and analytical solution for the case
of h = 1500 m (∆x
Im
= 3087 m, ∆x = 2086 m): (a) γ = 0.0, α = 0.0
15
; (b) γ = 0.0, α = 1/6
16
;
(c) γ = −0.099, α = 1/6 (present).
1
Since the grid size ∆x(= 2086 m) employed in the computation is larger than
∆x
Im
(= 1085m) evaluated by Imamura condition, i.e., ∆x
Im
=
4h
2
+ gh∆t
2
,
the free surface proﬁles calculated without dispersioncorrection are more dispersive
than those of the analytical solutions as shown in Fig. 10.3(a). Especially, the
numerical model gives a less dispersive proﬁle along the diagonal direction (θ = 45
◦
)
than along principal axes (θ = 0
◦
, 90
◦
), while the numerical solution with α = 1/6
shows no directional dependency as shown in Fig. 10.3(b). The present model with
the dispersioncorrection parameter γ = 0.061 and the weighting factor α = 1/6
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
246 S. B. Yoon
gives a good agreement with the analytical solution in the surface proﬁles in all
directions as shown in Fig. 10.3(c). On the other hand, for the case of 1500 m water
depth, ∆x(= 2086 m) is smaller than ∆x
Im
(= 3087m). Thus, the numerical solu
tions calculated with γ = 0.0 show less dispersive nature in the surface proﬁles
than the analytical solutions based on the linear Boussinesq equations as shown
in Figs. 10.4(a) and 10.4(b). The present numerical model using γ = −0.099 and
α = 1/6, however, still gives a correct dispersion eﬀect in each direction as shown
in Fig. 10.4(c).
From the results discussed above, it is concluded that the present numerical
model is less sensitive to the choice of grid size than the models of Imamura et al.
15
and Cho.
16
As a result, the present model gives more ﬂexibility in selecting grid size.
10.3.2. Tsunami propagation over a submerged circular shoal
In this section, the validity of the present ﬁnitediﬀerence model (FDM) in the case
of slowly varying topography is tested using a submerged circular shoal. Figure 10.5
shows a computational domain with a submerged circular shoal with a conical
frustum shape. The center of the shoal is located at x
0
= 500 km and y
0
= 250 km.
The base radius R
1
and the top radius R
2
are 150 km and 86 km, respectively. The
bathymetry of the computational domain is given by
h(r) =
1500 r ≥ R
1
,
1500r
2
/R
2
1
R
2
< r < R
1
,
500 r ≤ R
2
,
(10.13)
where r =
(x −x
0
)
2
+ (y −y
0
)
2
is the distance from the center of the submerged
circular shoal.
Along x = 0, the initial water surface displacement in the form of a solitary
wave is prescribed as
ζ(x) = 2e
−(x/7500)
2
. (10.14)
Four wave gages are installed to measure the free surface proﬁle. Wave gages ➀
and ➂ are positioned at the front and back slopes of the shoal, respectively, where
Fig. 10.5. Schematic diagram of a submerged circular shoal, wave gages, and initial condition.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 247
the water depth is 1000 m. Gage ➁ is located at the center of the shoal where the
water depth is 500 m. Gage ➃ is placed behind the shoal. Sponge layers are placed
along both ends of the computational domain to absorb the energy of outgoing
waves.
Since no analytical solution is known for this case, numerical solutions using
the present dispersioncorrection ﬁnitediﬀerence model are compared with those
of the linearized Nwogu’s Boussinesq equations
20
implemented in the FUNWAVE
code.
6,21
FUNWAVE is a fully nonlinear Boussinesq wave model with improved dis
persion relationships for short waves. The accuracy of FUNWAVE has been veriﬁed
for various coastal problems such as shoaling, refraction, diﬀraction, and breaking
of waves. The numerical simulation using FUNWAVE is performed with a uniform
grid size of 500 m, the ﬁnest grid allowable on the personal computer employed in
this study, to minimize the numerical dispersion. An additional computation using
FUNWAVE is also conducted with 2000 m grid size to test the sensitivity of the
grid resolution. On the other hand, the numerical simulation employing the present
dispersioncorrection model is made with a uniform grid size of 2000m. The time
step ∆t is determined from the stability criteria for each wave model. The algorithms
to compute various physical processes, such as nonlinear advection, nonlinear dis
persion, and wave dissipation due to bottom friction and breaking, are eliminated
from the source code of the FUNWAVE model. Thus, the computational time for
FUNWAVE model to calculate only the propagation step of smallamplitude waves
can be measured for fair comparison with that of the present model.
The numerical simulation is conducted for 9000 s after the initial water surface
displacement imposed along x = 0 is released. The computational time elapsed
for diﬀerent models is presented in Table 10.1. The FUNWAVE model employing
a predictorcorrector scheme consumes a long computational time, while the
present fully explicit model takes only 1/10 of the computational time required for
FUNWAVE in the case of using the same grid size, i.e., ∆x = 2000m. The compu
tational eﬃciency of the present model can be realized even more dramatically if
the computational time is compared with that of the FUNWAVE using ﬁner grid
of ∆x = 500 m. The present model is approximately 2200 times faster than the
FUNWAVE model. If the accuracy of the present model is comparable to that of
FUNWAVE, it can be concluded that the present model is highly eﬃcient for prac
tical problems.
Figure 10.6 presents the comparison of the free surface time history at various
gage locations computed by FUNWAVE and the present FDM. Figure 10.6(a) shows
Table 10.1. Comparison of computational time for each simulation.
1
Model FUNWAVE FUNWAVE Present FDM
∆x (m) 500 2000 2000
Number of grids 3001×1001 751×251 751×251
∆t (s) 2 8 4
Number of time steps 4500 1125 2250
CPU time (s) 156,888 770 72
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
248 S. B. Yoon
Fig. 10.6. Comparison of time history computed by FUNWAVE and the present FDM:
(a) location ➀ (h = 1000 m); (b) location ➁ (h = 500 m); (c) location ➂ (h = 1000 m); (d) location
➃ (h = 1500 m).
1
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 249
the free surface time history recorded at the wave gage ➀ located on the front slope
of the shoal. This ﬁgure shows that the initial solitary wave evolves into a group of
waves due to dispersion of waves. Figure 10.6(b) presents the time history at the
location ➁ on the top of the shoal. The free surface calculated by the present model
shows a good agreement with that of the FUNWAVE with ﬁne grid of ∆x = 500 m.
The results calculated using FUNWAVE with 2000m grid size, however, show a
considerable phase lag in the tail of wave train. This phase lag is caused by the
accumulation of numerical dispersion errors due to poor resolution of computational
grid near the top of the shoal.
Although the same grid size of 2000m is employed, the present model still gives
more accurate results. This implies that the present model is capable of minimizing
the numerical dispersion eﬃciently through the dispersioncorrection algorithm.
Figure 10.6(c) shows the time history at location ➂ on the back slope of the shoal.
Two trains of waves arrive here. The ﬁrst train of waves propagating around the
shoulders of the shoal meets at this location at t = 6000s. The second train of waves
traveling over the top of the shoal arrives here at t = 6300 s. A good agreement
between the results calculated using the present model and the FUNWAVE with
ﬁne grids is achieved. The results calculated using the FUNWAVE with coarse
grids, however, suﬀers from the numerical dispersion. Figure 10.6(d) shows the time
history at location ➃ behind the shoal where the water depth h is 1500m. The
agreements between the numerical solutions are reasonable.
In summary, the present FDM is proven to be suﬃciently accurate in comparison
with the FUNWAVE model which can deal with a full coverage of dispersion eﬀects
for the varying water depth region. On top of this, the present model is shown to
be highly eﬃcient. Thus, the present model can be used as a practical numerical
model to simulate the propagation of transoceanic tsunamis over slowly varying
topography.
10.4. NearField Model
As the tsunamis propagate over a continental shelf and approach a coastal area, the
dispersion eﬀect of waves becomes weak, and the nonlinearity and bottom friction
of waves dominantly inﬂuence the transformation of the tsunamis. However, the
linear Boussinesqtype wave equation does not include the nonlinear and bottom
friction terms. Thus, the nonlinear shallowwater equations (NSWE) are employed
for the nearﬁeld transformation of tsunamis.
∂ζ
∂t
+
∂P
∂x
+
∂Q
∂y
= 0, (10.15)
∂P
∂t
+
∂
∂x
P
2
D
+
∂
∂y
PQ
D
+ gD
∂ζ
∂x
+
gn
2
D
7/3
P
P
2
+ Q
2
= 0, (10.16)
∂Q
∂t
+
∂
∂x
PQ
D
+
∂
∂y
Q
2
D
+ gD
∂ζ
∂y
+
gn
2
D
7/3
Q
P
2
+ Q
2
= 0, (10.17)
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
250 S. B. Yoon
where n denotes the Manning’s roughness coeﬃcient (sec/m
1/3
) for the bottom
friction, and D(= h + ζ) is the total water depth. A conventional leapfrog ﬁnite
diﬀerence scheme with wet–dry technique
17
can be employed.
10.5. Grid Nesting Scheme
In the central area of the East Sea of Korea, there exists a huge submerged shoal
called Yamato Rise. The depth diﬀerence between shallow and deep areas is signif
icant. To achieve the best performance from the dispersioncorrection model, the
grid size for deep area of 4 km water depth should be approximately 6 km. This grid
size causes a problem of poor resolution for the shallow water area near the top of
submerged shoal, because the wavelength becomes shorter due to shoaling. Thus,
the accuracy of the simulation will be seriously degraded. To overcome this diﬃ
culty, a nesting between coarser grid for deep area and ﬁner grid for shallow area is
indispensable.
Yoon
17
developed a dynamic grid nesting scheme for the ﬁnitediﬀerence model
to solve shallowwater equations. This grid nesting scheme uses a secondorder
interpolation in space and a ﬁrstorder interpolation in time. Numerical tests of
the nesting scheme developed by Yoon
17
show some reﬂection for short waves
along the nesting line. To improve the performance of the nesting scheme, a new
dynamic grid nesting scheme has been developed by Lim et al.
2,18
for the present
ﬁnitediﬀerence model based on the linear Boussinesqtype wave equations. They
employed a thirdorder interpolation in space and a secondorder interpolation
in time. The higher order interpolation helps short waves propagate with weak
reﬂection from the nesting line. Figures 10.7(a) and 10.7(b) show the layout of vari
ables in space and time, respectively, to nest the coarse and ﬁne grid regions. The
Fig. 10.7. Layout of variables for grid nesting; (a) grid nesting system in space, (b) grid nesting
system in time.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 251
grid size and the time step for the ﬁne grid region are 1/3 of the values of the coarse
grid region. Along the boundary between two regions, the information calculated
in one region should be transferred to another region for each time step. When the
information is transferred from the coarse grid region to the ﬁne grid region, both
spatial and temporal interpolations of the variables of the coarse grid region are
required.
Lim et al.
2
tested their grid nesting scheme for the propagation of a linear
solitary wave over a submerged circular shoal with a signiﬁcant depth diﬀerence
between the top and the base of the shoal. Numerical solutions calculated using
their nested model were compared with those of the linearized Nwogu’s Boussinesq
equations
20
implemented in the FUNWAVE code.
6,21
They showed that the present
model system is approximately 4000 times faster than the FUNWAVE model, while
the accuracy of the present model is comparable to that of FUNWAVE. Thus, it can
be concluded that the present model system is highly eﬃcient for practical problems.
To link the nearﬁeld model using nonlinear shallowwater equations to the far
ﬁeld model employing linear Boussinesqtype wave equations, special care should
be taken because the governing equations are diﬀerent between two regions. If the
dispersioncorrection is neglected along the nesting line, i.e., γ = 0.0, (10.6) can be
simpliﬁed to give
∂ζ
∂t
= v. (10.18)
Furthermore, the continuity equation (10.15) of the nonlinear shallowwater
equations can be rewritten as
∂ζ
∂t
= −
∂P
∂x
+
∂Q
∂y
. (10.19)
The righthand side of (10.18) and (10.19) should be identical. Thus, the aux
iliary variable v can be approximated as
v = −
∂P
∂x
+
∂Q
∂y
. (10.20)
The auxiliary variable v calculated in the farﬁeld propagation model using LBTWE
is transferred to the nearﬁeld model in the form of the divergence of ﬂow rate
as a boundary condition to solve the continuity equation of NSWE. The auxiliary
variable v required for the farﬁeld model can be computed by (10.20) using the
ﬂow rates P and Q calculated in the nearﬁeld model.
10.6. Numerical Simulations of Historical Tsunami Event
A historical tsunami event, i.e., the 1993 Hokkaido Tsunami, is selected to verify
the nested dispersioncorrection ﬁnitediﬀerence model (NDCFDM). The calculated
free surface displacements are compared with the tide gage records obtained at two
harbors: Mukho and Ulsan harbors, along the east coast of Korea.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
252 S. B. Yoon
10.6.1. Computational information
For the simulation of the historical tsunami event, the computational domain
bounded by four latitudinal and longitudinal lines, i.e., 127
◦
E, 142
◦
E, 33
◦
N, and
45
◦
N for west, east, south, and north boundaries, respectively, are selected as shown
in Fig. 10.8. This ﬁgure presents the bathymetry of the East Sea. For the ﬁnite
diﬀerence model, the computational domain is divided into 900 ×931 square grids,
for which the grid spacing is 1 min in the longitudinal direction. To make square
grids, the grid spacing in the latitudinal direction is adjusted to give the same grid
size of 1 min grid spacing in the longitudinal direction at each grid location.
18
This can be done if the following condition is satisﬁed.
∆φ = cos φ∆ψ, (10.21)
where φ and ψ represent the latitude and longitude of grid point, respectively. ∆φ
and ∆ψ are the radian grid size in latitudinal and longitudinal directions, respec
tively. Since ∆ψ is chosen as 1 min, ∆φ is obtained by multiplying ∆ψ by cos φ,
and ∆x is identical with ∆y everywhere. Thus, the actual grid size is given by
∆x = ∆y = Rcos φ∆ψ, (10.22)
where R(= 6.378 × 10
6
m) is the radius of the earth. Subsequently, the ﬁnite
diﬀerence mesh system taken into account the curvature of the earth can be regarded
Fig. 10.8. Computational domain and bathymetry of the East Sea (depth unit: m).
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 253
as locally uniform, and the actual grid size slowly changes in association with the
latitude of grid point. It is approximately 1.4 km around the central part of compu
tational domain.
The time step ∆t is chosen as 3 s, and the actual grid size ∆x slowly varies in the
range of 1312–1556m. The dispersioncorrection parameters γ calculated by given
conditions such as time step ∆t, actual spatial grid ∆x, and local depth h, give the
values between −2.93 and 0.08 in the whole computational domain. As mentioned
above, the minimum value of γ is restricted to −0.125 to maintain the stability for
the deep areas of the East Sea. Due to this limitation, the dispersion eﬀects are
underestimated in the numerical solutions. However, the present scheme gives more
accurate solutions than the conventional models which use a uniform γ value, i.e.,
γ = 0.0, everywhere.
As they approach the shallowwater region, the wavelength of tsunamis becomes
short due to shoaling, and the resolution of grids becomes lower. Thus, a set of ﬁner
grid system, B, C, D, and E, is nested dynamically to the coarse grid region A.
As shown in Fig. 10.8, two subregions, Mukho (C1) and Ulsan (C2) harbors, are
selected for computations with ﬁner grid system, because tide gage records are
available at these harbors. The computational information such as grid size and time
step employed for the nesting is listed in Table 10.2. The subregions are coupled by
the nesting scheme as proposed above. The linear Boussinesqtype wave equation
(LBTWE) is employed to compute the regions, A, B, C, and D. For the ﬁnest
grid regions, E1 and E2, the nonlinear shallowwater equations (NSWE) are solved
because the nonlinearity increases as the water depth decreases. Manningtype
energy dissipation for bottom friction is implemented in the nonlinear model. The
Manning’s roughness coeﬃcient is set to be equal to 0.05 s/m
1/3
.
To verify the applicability of the present model to the real topography, the
calculated tsunamifree surface elevation is compared with the elevation measured
by tide gage records at two harbors, Mukho and Ulsan harbors. The bathymetry
of computational subregions and the locations of tide gage at Ulsan harbor are
presented in Fig. 10.9.
Two kinds of boundary conditions, noﬂux boundary and absorbing boundary,
are mainly used in the tsunami propagation model. Noﬂux boundary condition is
employed to express the rigid impermeable wall, and an absorbing boundary con
dition proposed by Larsen and Darcy
22
is placed in front of the opensea boundary.
The fault parameters of the historical tsunami event, such as epicenter location
in latitude (
◦
N) and longitude (
◦
E), depth of upper rim of fault H(km), orientation
Table 10.2. Computational information for subregions of each harbor.
Region Grid spacing ∆ψ Time step ∆t (s) Remark
A 1 min 3.0 LBTWE
B 20 s 1.0 LBTWE
C1, C2 6.67 s 0.33 LBTWE
D1, D2 2.22 s 0.11 LBTWE
E1, E2 0.74 s 0.037 NSWE
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
254 S. B. Yoon
Fig. 10.9. Computational subregions and detailed bathymetry near Ulsan harbor (depth unit: m).
Fig. 10.10. Deﬁnition sketch for fault parameters of three historical tsunami events.
of fault with strike angle θ(
◦
), dip angle δ(
◦
), and slip angle λ(
◦
) from the north,
length of fault L(km), width of fault W(km), and dislocation of fault D(m), are
deﬁned in Fig. 10.10.
It is assumed that the water surface displacement created by the earthquake
is identical to the vertical displacement of sea bed induced by the ground motion.
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 255
The displacement ﬁeld can be determined by using the fault model proposed by
Mansinha and Smylie.
23
10.6.2. Simulation of the 1993 Hokkaido Tsunami
The magnitude of the 1993 Hokkaido Tsunami which originated from the deep sea
near the Okushiri Island of Hokkaido on 12 July 1993 was recorded as M7.8. The
source is composed of three faults. The south fault has the largest dislocation of
12 m. The fault parameters for the tsunami event proposed by Takahashi et al.
24
are listed in Table 10.3. Figure 10.11 presents the initial free surface proﬁle of the
1993 Tsunami.
The 1993 Tsunami is simulated. Figure 10.12 shows the propagation map with
5 min time interval obtained from the numerical simulation. Figure 10.13 shows
the distribution of the highest water levels obtained from the numerical simulation
Table 10.3. Fault parameters for the 1993 Tsunami.
24
Fault Lat. Long. H θ δ λ L W D
No. (
◦
N) (
◦
E) (km) (
◦
) (
◦
) (
◦
) (km) (km) (m)
1 42.10 139.30 5 163 60 105 24.5 25 12.00
2 42.34 139.25 5 175 60 105 30 25 2.50
3 43.13 139.40 10 188 35 80 90 25 5.71
Fig. 10.11. Initial free surface proﬁle of the 1993 Tsunami (unit: m).
2
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
256 S. B. Yoon
Fig. 10.12. Propagation map of the 1993 Tsunami (unit: min).
2
Fig. 10.13. Distribution of calculated highest water levels due to the 1993 Tsunami (unit: m).
2
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 257
Fig. 10.14. Comparison of calculated and measured time history of free surface displacements of
the 1993 Tsunami at (a) Mukho harbor and (b) Ulsan harbor.
2
of the 1993 Tsunami. The Shimane Peninsula (JF) of Japan and the Imwon area
(KA) of the east coast of Korea are shown to be vulnerable to the tsunami attacks
generated in the East Sea because of the large wave heights due to the wave trapping
over the Yamato Rise (YM) and the submerged ridges.
The calculated time histories of free surfaces at two harbors are compared with
the tide gage records of the 1993 Tsunami in Fig. 10.14. The calculated results
are generally in good agreement with measured data in many respects even though
there are slight phase and height diﬀerences.
10.6.3. Eﬀect of topography on propagation of tsunamis
As shown in Fig. 10.15, the underwater topography of the East Sea of Korea is
complicated. In the central part of the sea there exists a huge submerged shoal
called Yamato Rise (YM). This rise is connected to the Shimane Peninsula (JF)
of Japan by a submerged ridge. Two submerged ridges which protrude eastward
from the Imwon area (KA) on the east coast of Korea form a Kshaped underwater
topography. These underwater rises and ridges play a signiﬁcant role of a waveguide
for the tsunamis generated in this sea. Figure 10.15 shows a sketch of paths along
which the main tsunami energy is transmitted for the case of the 1993 Tsunami.
The tsunami waves are ﬁrst captured by Yamato Rise due to topographical lens
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
258 S. B. Yoon
Fig. 10.15. Schematic diagram of energy stream path for the 1993 Tsunami.
2
eﬀect, and then trapped again by the underwater ridge connecting the Yamato
Rise and the Shimane Peninsula of Japan. As a result, the main stream of tsunami
energy radiated toward the open sea turns back to the Shimane Peninsula (JF) of
Japan. On the other hand, the Kshaped submerged ridges catch the tsunami energy
propagating toward the east coast of Korea and carry it along the two branch ridges
toward the Imwon area (KA) where two ridges meet. Thus, the tsunami energy is
focused at this spot. Even though these regions, i.e., Shimane Peninsula (JF) of
Japan and Imwon (KA) of Korea, are located far from the tsunami source area,
these two regions are vulnerable to the attack of various tsunamis generated in the
East Sea of Korea.
10.7. Conclusions
A practical ﬁnitediﬀerence model developed by Yoon et al.
1
and Lim et al.
2,18
for the eﬃcient and relatively accurate simulation of tsunami propagation is intro
duced. This model solves a linear Boussinesqtype wave equation to achieve proper
dispersion eﬀect of waves. The present model is applied to simulate the propagation
of a historical tsunami event occurred in the East Sea of Korea. The calculated free
surface displacements for the 1993 Hokkaido Tsunami are compared with the obser
vations at two tidal stations along the east coast of Korea. The comparison shows
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
Finite Diﬀerence Model for Practical Simulation of Distant Tsunamis 259
that the results agree well with the observations. Based on the simulated results
of the historical tsunami that occurred in the East Sea of Korea, it is found that
the underwater topography, such as submerged rises and ridges, plays an important
role in the propagation of tsunamis in this region. The Yamato Rise located in the
central part of the sea behaves as a topographic lens to focus the tsunami energy,
and the submerged ridge connecting the rise to the Shimane Peninsula of Japan
behaves like a waveguide along which the focused energy is trapped and carried to
the Shimane Peninsula. The Kshaped submerged ridges exert a similar inﬂuence
on the tsunami incident on the Imwon area in the east coast of Korea.
Acknowledgment
This research was supported by a grant (No. NEMA06NH06) from the Natural
Hazard Mitigation Research Group, National Emergency Management Agency of
Korea.
References
1. S. B. Yoon, C. H. Lim and J. Choi, Dispersioncorrection ﬁnite diﬀerence model for
simulation of transoceanic tsunamis, Terres. Atmos. Oceanic Sci. CGU 18(1), 31–53
(2007).
2. C. H. Lim, J. S. Bae, J. I. Lee and S. B. Yoon, Propagation characteristics of historical
tsunamis that attacked the east coast of Korea, Natural Hazards 47(1), 95–118 (2008).
3. J. F. Lander and P. A. Lockridge, United States Tsunamis (including United States
Possessions) 1690–1988 (National Geophysical Data Center, Boulder, Colorado,
1989).
4. K. Satake, J. Bourgeois, K. Abe, Y. Tsuji, F. Imamura, Y. Iio, H. Katao, E. Noguera
and F. Estrada, Tsunami ﬁeld survey of the 1992 Nicaragua Earthquake, EOS Trans.
AGU 74, 156–157 (1993).
5. Hokkaido Tsunami Survey Group, Tsunami devastates Japanese coastal region, EOS
Trans. AGU 74, 417, 432 (1993).
6. J. T. Kirby, G. Wei, Q. Chen, A. B. Kennedy and R. A. Dalrymple, FUNWAVE 1.0
Fully nonlinear Boussinesq wave model, Documentation and User’s Manual (CACR
9806) Technical report, Center for Applied Coastal Research, Ocean Engineering
Laboratory, University of Delaware (1998).
7. P. Lynett and P. L.F. Liu, A twolayer approach to water wave modeling, Proc. Roy.
Soc. London A 460, 2637–2669 (2004).
8. S.C. Hsiao, P. Lynett, H.H Hwung and P. L.F. Liu, Numerical simulations of non
linear short waves using the multilayer model, J. Eng. Mech. ASCE 131(3), 231–243
(2005).
9. M. Ioualalen, J. Asavanant, N. Kaewbanjak, S. T. Grilli, J. T. Kirby and P. Watts,
Modeling the 26 December 2004 Indian Ocean tsunami: Case study of impact in
Thailand, J. Geophys. Res. 112, C07024 (2007), doi:10.1029/2006JC003850.
10. S. T. Grilli, M. Ioualalen, J. Asavanant, F. Shi, J. T. Kirby, P. Watts, Source
constraints and model simulation of the December 26, 2004, Indian Ocean tsunami,
J. Waterway, Port, Coastal Ocean Eng. ASCE 133(6), 414–428 (2007).
July 31, 2009 8:18 9.75in x 6.5in b684ch10 FA
260 S. B. Yoon
11. L. S. Hwang and D. Divoky, Tsunami generation, J. Geophys. Res. 75, 6802–6817
(1970).
12. C. Goto and N. Shuto, Numerical simulation of tsunami propagations and runup,
Tsunami — Their Science and Engineering, eds. K. Iida and T. Iwasaki (Terra Science
Publishing Company, Tokyo, 1983), pp. 439–451.
13. Z. Kowalik and T. S. Murty, Computation of tsunami amplitudes resulting from a
predicted major earthquake in the Shumagin seismic gap, Geophys. Res. Lett. 11,
1243–1246 (1984).
14. C. L. Mader and G. D. Curtis, Numerical modeling of tsunami inundation of Hilo
harbor, JIMAR Contribution No. 91–251, University of Hawaii, Honolulu (1991).
15. F. Imamura, N. Shuto and C. Goto, Numerical simulation of the transoceanic prop
agation of tsunamis, Proc. 6th Congress Asian and Paciﬁc Regional Division, IAHR,
Japan (1988), pp. 265–271.
16. Y. S. Cho, Numerical simulations of tsunami propagation and runup, PhD thesis,
School of Civil and Environmental Engineering, Cornell University, Ithaca, NY (1995).
17. S. B. Yoon, Propagation of distant tsunamis over slowly varying topography,
J. Geophys. Res. 107(C10), 3140 (2002), doi:10.1029/2001JC000791.
18. C. H. Lim, J. S. Bae, Y. J. Jeon and S. B. Yoon, Grid nesting for dispersioncorrection
ﬁnite diﬀerence model for tsunami simulation, Proc. 32nd Congress of the Interna
tional Association of Hydraulic Engineering and Research, Venice, Italy, Paper No.
D2.b050O (2007).
19. G. F. Carrier, Tsunami propagation from a ﬁnite source, Proc. 2nd UJNR Tsunami
Workshop, NGDC, Hawaii (1991), pp. 101–115.
20. O. Nwogu, Alternative form of Boussinesq equations for nearshore wave propagation,
J. Waterway, Port, Coastal Ocean Eng. 119(6), 618–638 (1993).
21. G. Wei and J. T. Kirby, A timedependent numerical code for the extended Boussinesq
equations, J. Waterway, Port, Coastal Ocean Eng. 121(5), 251–261 (1995).
22. J. Larsen and H. Dancy, Open boundaries in short wave simulations — A new
approach, Coastal Eng. 7, 285–297 (1983).
23. L. Mansinha and D. E. Smylie, The displacement ﬁelds of inclined faults, Bull. Seismol.
Soc. Am. 61, 1433–1440 (1971).
24. T. Takahash, N. Shuto, F. Iammura and M. Ortis, Fault model to describe Hokkaido
Nansei oﬀshore earthquake for tsunami, J. Coastal Eng. 41, 251–255 (1994) (in
Japanese).
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
Chapter 11
TsunamiInduced Forces on Structures
Ioan Nistor
∗
, Dan Palermo, Younes Nouri, Tad Murty
and Murat Saatcioglu
Department of Civil Engineering, University of Ottawa
161 Louis Pasteur, CBY, A115, Ottawa, Canada
∗
inistor@uottawa.ca
This chapter deals with the estimation of tsunamiinduced hydrodynamic forces
on infrastructure located in the vicinity of the shoreline. While extensive research
has been conducted on the impact of hydrodynamic forces on classical coastal
protection works (breakwaters, seawalls, reefs, etc.), there is limited research on
their impact on structures such as buildings and bridges located inland. The dev
astation brought by the 26 December 2004 Indian Ocean Tsunami on coastal com
munities in Indonesia, India, Sri Lanka, Thailand, and other countries outlined
the urgent need for research on the evaluation of structural resilience of infras
tructure located in tsunamiprone areas. This chapter summarizes the stateof
theart knowledge with respect to forces generated by tsunamiinduced hydraulic
bores, including debris impact. Further, sample calculations of tsunami loading
on a prototype structure are presented.
11.1. Introduction
Tsunami waves represent extreme, often catastrophic events, which signiﬁcantly and
adversely impact coastal areas. In spite of the lower frequency of occurrence com
paring to storms and storminduced surges, tsunamiinduced coastal ﬂooding often
leads to massive casualties and tremendous economic losses.
1–3
Hence, tsunamis are
rare events, highimpact natural disasters.
The devastating eﬀects of the 26 December 2004 Tsunami on many coun
tries bordering the Indian Ocean raised public concern and revealed existing deﬁ
ciencies within the current warning and defense systems against tsunamis. One of
the important elements that needs signiﬁcant improvement is the estimation of
forces generated by tsunamiinduced bores, as well as waterborne debris. Before
the 2004 Indian Ocean Tsunami, the design of structures against tsunamiinduced
forces was considered of minor importance when compared to the attention given
to tsunami warning systems. This was due to the assumption that tsunamis are
261
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
262 I. Nistor et al.
(a) (b)
(c) (d)
Fig. 11.1. Tsunami damage in Thailand and Indonesia (December 2004 Indian Ocean Tsunami):
(a) severe structural damage, Khao Lak, Thailand; (b) column failure of a reinforced con
crete frame, Phuket, Thailand; (c) column failure due to debris impact, Banda Aceh, Indonesia;
(d) punching failure of inﬁll walls, Banda Aceh, Indonesia.
4
rare events, with signiﬁcantly high return periods (sometimes more than 500 years).
Reconnaissance missions of the December 2004 Indian Ocean Tsunami disaster
revealed that tsunamiinduced forces can lead to severe damage or collapse of struc
tures as shown in Fig. 11.1.
3–11
Therefore, these forces should be properly accounted
for in the design of infrastructure built within a certain distance from the shoreline
in tsunamiprone areas.
The design of coastal structures such as breakwaters, jetties, and groins against
waves is typically based on considering the eﬀect of breaking waves and their asso
ciated forces, and is well established. Unlike coastal structures, the evaluation of
tsunamiinduced hydrodynamic forces on structures used for habitation and/or eco
nomic activity, received little attention by researchers and engineers.
Results of ﬁeld surveys conducted in the aftermath of the December 2004 Indian
Ocean Tsunami in Indonesia and Thailand showed that poorly detailed concrete
structures experienced severe damage.
3,4
This highlighted the fact that the current
structural design codes do not account for tsunamiinduced forces and the impact
of associated debris. Reinforced concrete structures have been observed to with
stand tsunamis with acceptable low levels of damage.
12
However, as shown in
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 263
0 1 2 3 4 5 6 7
Inundation Depth (m)
: The 2004 Indian Ocean Tsunami
Partial Damage
Lack of data
: Past Tsunamis
Damage
Withstand
20
Fig. 11.2. Relation between the inundation depth and degree of damage to reinforced concrete
buildings.
13
Fig. 11.2, inundation depths of more than 5 m can induce partial damage to concrete
structures.
Currently, there are no clearly established procedures to address the aforemen
tioned forces. Moreover, signiﬁcant disagreement on existing empirical formulae fos
tered new research interest in an eﬀort to properly address the inclusion of both
tsunamiinduced forces and the impact of debris into design codes. Aspects related
to these forces are discussed in this chapter. Some of the shortcomings and incon
sistencies of existing codes are also highlighted.
11.2. Literature Review
11.2.1. Tsunamiinduced hydraulic bores
As tsunami waves advance toward the shoreline and water depth decreases, their
wave height increases while celerity decreases. Tsunami waves may break oﬀshore or
at shoreline, inundating lowlying coastal areas in the form of a hydraulic bore. On
the other hand, tsunami inundation can also occur as a gradual rise and recession
of the sea level for the case of nonbreaking tsunami waves. The width of the conti
nental shelf, the initial tsunami wave shape, the beach slope, and the tsunami wave
length are the parameters which govern the breaking pattern of tsunami waves.
2,15
Figure 11.3 shows a tsunami wave approaching the Khao Lak Beach in Thailand
during the December 2004 Indian Ocean Tsunami.
Tsunami waves have a larger horizontal length scale compared to the vertical.
Consequently, implementing shallow water wave theory (i.e., depthintegrated equa
tions of momentum and mass conservation with the assumption of a hydrostatic
pressure ﬁeld) seems to be a reasonable method for representing tsunami wave prop
agation. Using these equations, it has been shown that the behavior and runup of
nonbreaking tsunami waves can be predicted with acceptable accuracy. However,
disagreement has been observed between experiment and prediction, in terms of
behavior and runup of breaking waves and the resulting bore.
16,17
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
264 I. Nistor et al.
(a) (b)
(c) (d)
Fig. 11.3. Tsunami wave in Khao Lak, Thailand (December 2004 Indian Ocean Tsunami):
(a) water recedes; (b) waves approach the shoreline; (c) tsunami waves break close to the shoreline;
(d) tsunami waves inundate the shoreline.
14
Signiﬁcant eﬀorts were directed toward the experimental investigation of the
mechanisms of tsunami bore runup.
2,18,19
Although a twodimensional dambreak
phenomenon was used in the experiments, the bore motion was observed to be fully
threedimensional and highly turbulent. This is in agreement with other observa
tions regarding the irregularity of the bore front in transverse direction and the
noticeable ﬂuctuation of the front propagation.
19
The onshore propagation of the tsunami wave is similar to the classical dam
break problem. Chanson
20
compared the instantaneous free surface ﬂow proﬁles
of a tsunamiinduced bore with ﬂoating bodies to a dambreak ﬂow on a hori
zontal bed. A framebyframe analysis of a video recording taken during the Indian
Ocean Tsunami in Banda Aceh, Indonesia was used to obtain the ﬂow proﬁle of
the tsunamiinduced bore. The agreement between the tsunami ﬁeld data and the
dambreak analytical formulation demonstrated the analogy between propagation
of tsunamiinduced bores and dambreak ﬂow.
Direct estimation of tsunami inundation bore velocities is limited. Tsutsumi
et al.
21
estimated the nearshore ﬂow velocity of the tsunami caused by the Southwest
Hokkaido Earthquake of 12 July 1993. Forces exerted on a bent iron handrail and
an iron guardrail were estimated using in situ strength tests. Then, the velocity of
the tsunami ﬂow was calculated from the forces using Morison’s equation.
22
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 265
Matsutomi et al.
13
and Chanson
20
estimated inundation ﬂow velocities in terms
of inundation depth. Framebyframe analysis of video recordings from the 2004
Indian Ocean Tsunami in Banda Aceh was conducted. The videos showed that
the inundation ﬂow carried numerous ﬂoating bodies with approximately the same
speed as that of the bore. This may have inﬂuenced the bore characteristics. Fur
thermore, Matsutomi et al.
13
developed an empirical model that predicts the bore
front velocity. Also, surveyed data were used to improve the criterion for estimating
the degree of structural damage.
11.2.2. Tsunamiinduced forces on structures
Snodgrass et al.
23
noticed that broken waves induced larger hydrodynamic hori
zontal forces on a test pile compared to waves breaking at the pile location. As
previously mentioned, broken tsunami waves inundate shoreline in the form of a
hydraulic bore, which is a fast moving body of water with an abrupt front. However,
mechanisms of impingement of broken tsunami waves on structures located inland
are not yet well understood.
Pioneering analytical and experimental attempts to quantify forces due to bores
date back to Stoker,
24
Cumberbatch,
25
Fukui,
26
Cross,
27
and Dames and Moore.
28
Comprehensive experimental investigation of the interaction of bores and dry
bed surges with a vertical wall was performed by Ramsden and Raichlen
29
and
Ramsden.
30,31
In these experiments, three ﬂow conditions were analyzed: (1) tur
bulent bores (initial still water downstream of the gate); (2) drybed surges (no
initial water depth downstream of the gate); and (3) solitary waves. Forces and over
turning moments due to bores and drybed surges were recorded and calculated,
respectively. The results of Ramsden’s studies are not applicable to the estimation
of impulsive forces.
31
It was observed that the pressure distribution during impact
is essentially nonhydrostatic. The experiment also demonstrated that the transition
from undular to turbulent bores led to a discontinuous increase in watersurface
slope, followed by an increase in measured runup, pressure head, and exerted forces
and moments. Figure 11.4 shows the diﬀerence between a strong turbulent bore and
a drybed surge with approximately the same celerity.
It was shown that recorded forces gradually increased to an approximately con
stant value for both the case of a surge and a bore. No impulsive (shock) force
exceeding the hydrodynamic force was observed. However, an initial impulsive
pressure equal to three times the pressure head, corresponding to the mea
sured runup, was recorded. Ramsden
31
further derived empirical formulae for the
maximum force and moment exerted on a vertical wall due to the bore impact
[Eqs. (11.1) and (11.2)].
F
F
l
= 1.325 + 0.347
H
h
+
1
58.5
H
h
2
+
1
7160
H
h
3
, (11.1)
M
M
l
= 1.923 + 0.454
H
h
+
1
8.21
H
h
2
+
1
808
H
h
3
, (11.2)
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
266 I. Nistor et al.
Fig. 11.4. Comparison of: (a) wave proﬁle; (b) runup; (c) pressure head; (d) force due to a strong
turbulent bore and a drybed surge.
31
where F is the force on the wall; F
l
is the force on the wall due to a runup equal to
twice the wave height, assuming hydrostatic pressure; H is the wave height at the
wall; h is still water depth; M is the moment on the wall; and M
l
is the moment
corresponding to F
l
.
Okada et al.
32
conducted a survey of previous studies on tsunami wave forces
and pressures, and identiﬁed ﬁve empirical formulae for tsunamiinduced forces or
pressures. It was found that calculation of tsunami load on structures using these
formulae would result in approximately the same magnitude of load. These formulae
are as follows:
• Tsunami wave pressure without soliton breakup,
33
• Tsunami wave pressure with soliton breakup,
33
• Tsunami wave pressure without soliton breakup,
34
• Tsunamiinduced wave forces on houses,
35
• Tsunami force exerted on houses.
36
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 267
Fig. 11.5. Measured and nondimensional force history for a square column.
37
Arnason
37
measured forces exerted on rectangular, rhomboidal, and circular
structures due to a hydraulic bore on a dry bed. It was observed that the surge
force overshot the hydrodynamic force in the case of a square column for small bore
heights (Fig. 11.5). However, no overshoot was recorded for the case of circular
and rhomboidal columns. This is in agreement with the results obtained by Nouri
et al.,
38
where similar experiments with larger bore heights, up to three times those
of Arnason,
37
were performed (Fig. 11.6).
Nouri et al.
38
conducted experiments with the objective of estimating bore
induced forces on freestanding structural components. The eﬀect of other param
eters such as upstream obstacles, ﬂow constrictions, and debris impact was also
investigated. The structural components were subjected to a dambreak ﬂow gen
erated by impoundment depths of 0.5, 0.75, and 0.85 m.
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
268 I. Nistor et al.
50
50
150
250
350
10 12 14 16 18 20 22
Time (s)
F
o
r
c
e
(
N
)
h0 = 1.00 m
h0 = 0.85 m
Fig. 11.6. Time history of exerted forces on a circular structure
38
; h0 is the impoundment depth.
0
10
20
30
40
50
60
2 3 8 13 18
P (kPa)
H
e
i
g
h
t
o
f
t
h
e
s
t
r
u
c
t
u
r
e
(
c
m
)
t=0.000 s
t=0.006 s
t=0.006 s
t=0.009 s
0
10
20
30
40
50
60
2 3 8 13 18
P (kPa)
H
e
i
g
h
t
o
f
t
h
e
s
t
r
u
c
t
u
r
e
(
c
m
)
t=0.160 s
t=0.170 s
t=0.180 s
t=0.190 s
Fig. 11.7. Variation of pressure distribution on the front face — circular structure
38
; t = 0.0 s is
the instant when the bore impacts the structure.
The variation of the vertical distribution of pressure was measured. Selected
snapshots from the variation of pressure distribution due to bore impact generated
by an impoundment depth of 1.0 m are shown in Fig. 11.7.
11.2.3. Debris impact force
Matsutomi
39
performed small and fullscale experiments on impact forces gen
erated by driftwood on rigid structures. Dambreak waves generated in a small
ﬂume carried pieces of lumber to the point of impact on a downstream wall. Also,
fullscale experiments in which wooden logs impacted a frame were conducted in
open air, and impact forces were measured. An empirical formula for estimating the
impact force, F, was derived using regression analysis of collected data [Eq. (11.3)].
F
γ
w
D
2
L
= 1.6C
M
u
√
gD
1.2
σ
f
γ
w
L
0.4
, (11.3)
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 269
0
40
80
120
0.0 1.0 2.0 3.0
V
A0
/(gD)
0.5
F
m
/γD
2
L
σ
f
/ γL = 2000
1500
1000
500
200
100
20
Fig. 11.8. Impact forces of wood logs for bores and surges.
39
where γ
w
is the speciﬁc weight of wood, D is the diameter of the log, L is the
length of the log, C
M
is a coeﬃcient which depends on the ﬂow passing around the
receiving wall (≈ 1.7 for bore or surge, and 1.9 for steady ﬂow), u is the velocity of
the log at impact, and σ
f
is the yield stress of the log. Figure 11.8 shows the design
chart based on Eq. (11.3).
Currently, three basic models are proposed for estimating the forces due to the
impact of debris on structures, which are used by a few design codes. In these
models, the impact force is calculated based on the mass and velocity of debris,
while ignoring the mass and rigidity of the structure. However, other than the mass
and velocity of debris, each model needs an additional parameter. The three models
and their corresponding additional parameters are
• Contact stiﬀness — stiﬀness between debris and structure,
• Impulse–momentum — stopping time of debris after impact and time history of
impact,
• Work energy — distance traveled from where initial contact occurs, to where
debris stops.
Haehnel et al.
40,41
used a singledegreeoffreedom model with eﬀective collision
stiﬀness as an additional parameter. They reviewed the current models discussed
above and demonstrated that none of the additional parameters are independent.
Hence, the three models are equivalent, provided that additional parameters are
appropriately selected. Further, small and largescale experiments were performed
in order to develop the singledegreeoffreedom model. Smallscale tests were per
formed in a ﬂume where wooden logs were released into the ﬂow and impacted a
load frame located further downstream. Largescale experiments were performed in
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
270 I. Nistor et al.
Fig. 11.9. Eﬀect of impact orientation on force.
40
a large basin where water was stationary and logs were placed on a movable car
riage. The eﬀect of parameters such as added mass of the water and the eccentricity
and obliqueness of the collision were also considered. Figure 11.9 shows the eﬀect of
impact orientation on the measured force. It was found that the maximum impact
force, F
i,max
, can be calculated using Eq. (11.4).
F
i,max
= Max
ˆ
kx = u
ˆ
km
1
, (11.4)
where u is the impact velocity of the log,
ˆ
k is the constant eﬀective stiﬀness between
the log and the structure, and m
1
is the mass of the log. Based on experiments,
Haehnel et al.
40
found the value of
ˆ
k = 2.4 MN/m to be the representative for the
upper envelope of the collected data.
11.3. Existing Design Codes
The design of structures in ﬂoodprone areas has previously been investigated.
However, few existing codes speciﬁcally address the design of onshore structures
built in tsunamiprone areas. Design codes that speciﬁcally address tsunamiinduced
forces were introduced in order to suggest provisions for designing infrastructure
in tsunamiprone areas. Posttsunami ﬁeld investigations of the December 2004
Indian Ocean Tsunami are indicative of the extreme loads generated by tsunami
induced ﬂoods,
3,4
and have outlined the need for developing new design guidelines.
Recent research work
42
indicated that tsunamiinduced loads are comparable or can
exceed earthquake loads. Tsunamiinduced forces and the impact of debris are not
properly accounted for in the current codes, and signiﬁcant improvement is needed.
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 271
At present, only four design codes and guidelines speciﬁcally account for tsunami
induced loads as listed below:
• FEMA 55: The code is adopted by the Federal Emergency Management Agency,
the United States, and recommends formulae for tsunamiinduced ﬂood and wave
loads.
43
• The City and County of Honolulu Building Code (CCH): The code, developed
by the Department of Planning and Permitting of Honolulu, Hawaii, United
States, makes provisions for regulations that apply to districts located in ﬂood
and tsunamirisk areas.
44
• Structural Design Method of Buildings for Tsunami Resistance (SMBTR): The
code is proposed by the Building Center of Japan
32
and outlines the structural
design for tsunami refuge buildings.
• Development of Guidelines for Structures that Serve as Tsunami Vertical Evacu
ation Sites: The guidelines were prepared by Yeh et al.
45
for estimating tsunami
induced forces on structures for the Washington State Department of Natural
Resources.
There are several other design codes (sometimes countryspeciﬁc) which contain
prescriptions and design guidelines for ﬂoodinduced loads. Examples of widely used
codes are indicated below:
• 1997 Uniform Building Code, Appendix 33, proposed by the International
Conference of Buildings Oﬃcials (ICBO),
47
• ASCE 705 of the American Society of Civil Engineers,
48
• 2006 International Building Code by the International Code Council.
49
However, none of the above codes address directly the tsunamiinduced forces,
which represent the focus of this chapter. The reader is advised to refer to these
codes when seeking guidance for the design of structures subjected to ﬂoodinduced
loads other than tsunamis: coastal ﬂooding due to storm surges, ﬂooding of river
banks above bankfull conditions, etc.
11.3.1. Tsunamiinduced forces
A broken tsunami wave running inland generates forces which aﬀect structures
located in its path. Three parameters are essential for deﬁning the magnitude and
application of these forces: (1) inundation depth, (2) ﬂow velocity, and (3) ﬂow
direction. The parameters mainly depend on: (a) tsunami wave height and wave
period; (b) coast topography; and (c) roughness of the coastal inland. The extent
of tsunamiinduced coastal ﬂooding, and therefore the inundation depth at a spe
ciﬁc location, can be estimated using various tsunami scenarios (magnitude and
direction) and modeling coastal inundation accordingly. However, the estimation of
ﬂow velocity and direction is generally more diﬃcult. Flow velocities can vary in
magnitude from zero to signiﬁcantly high values, while ﬂow direction can also vary
due to onshore local topographic features, as well as soil cover and obstacles. Forces
associated with tsunami bores consist of: (1) hydrostatic force, (2) hydrodynamic
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
272 I. Nistor et al.
(drag) force, (3) buoyant force, (4) surge force, and (5) impact of debris. A brief
description of these forces is further presented.
11.3.1.1. Hydrostatic force
The hydrostatic force is generated by still or slowmoving water acting perpen
dicular onto planar surfaces. The hydrostatic force per unit width, F
HS
, can be
calculated using Eq. (11.5), where ρ is the seawater density, g is the gravitational
acceleration, d
s
is the inundation depth, and u
p
is the normal component of ﬂow
velocity. Equation (11.5) is proposed by CCH and accounts for the velocity head.
Alternatively, FEMA 55 does not include the velocity head in its formulation since
it is assumed to be a negligible component of the hydrostatic force:
F
HS
=
1
2
ρg
d
S
+
u
2
p
2g
2
. (11.5)
The point of application of the resultant hydrostatic force is located at onethird
distance from the base of the triangular hydrostatic pressure distribution. In the
case of a broken tsunami wave, the hydrostatic force is signiﬁcantly smaller than the
drag and surge forces. Conversely, Dames and Moore
28
noted that the hydrostatic
force becomes important when tsunami is similar to a rapidlyrising tide.
11.3.1.2. Buoyant force
The buoyant force is the vertical force acting through the center of mass of a sub
merged body. Its magnitude is equal to the weight of the volume of water displaced
by the submerged body. The eﬀect of buoyant forces generated by tsunami ﬂooding
was clearly evident during posttsunami ﬁeld observations.
1,5,6
Buoyant forces can
generate signiﬁcant damage to structural elements, such as ﬂoor slabs, and are cal
culated as follows:
F
B
= ρgV, (11.6)
where V is the volume of water displaced by submerged structure.
11.3.1.3. Hydrodynamic (drag) force
As the tsunami bore moves inland with moderate to high velocity, structures are
subjected to hydrodynamic forces caused by drag. Currently, there are diﬀerences in
estimating the magnitude of the hydrodynamic force. The general expression for this
force is shown in Eq. (11.7). Existing codes use the same expression, but diﬀerent
drag coeﬃcient values (C
D
). For example, values of 1.0 and 1.2 are recommended
for circular piles by CCH and FEMA 55, respectively. For the case of rectangular
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 273
piles, the drag coeﬃcient recommended by FEMA 55 and CCH is 2.0:
F
D
=
ρC
D
Au
2
2
, (11.7)
where F
D
is the drag force acting in the direction of ﬂow, A is the projected area
of the body normal to the direction of ﬂow, and u is the tsunamibore velocity.
The ﬂow is assumed to be uniform, and therefore, the resultant force will act at
the centroid of the projected area. As indicated, the hydrodynamic force is directly
proportional to the square of the tsunamibore velocity. The estimation of the
bore velocity remains one of the critical elements on which there is signiﬁcant dis
agreement in literature. A brief discussion on the tsunamibore velocity is presented
below.
Tsunamibore Velocity. Previous research shows that signiﬁcant diﬀerences in esti
mating forces exerted on structures by tsunami bores, as well as impact of debris,
are due to diﬀerences in estimating bore velocity. Since the hydrodynamic force is
proportional to the square of the bore velocity, uncertainties in estimating veloc
ities induce large diﬀerences in the magnitude of the resulting hydrodynamic force.
Tsunamibore velocity and direction can vary signiﬁcantly during a major tsunami
inundation. Current estimates of the velocity are crude; a conservatively high ﬂow
velocity impacting the structure at a normal angle is usually assumed. Also, the
eﬀects of runup, backwash, and direction of velocity are not addressed in the current
design codes.
Although there is certain consensus in the general form of equation for the
hydrodynamic force, several researchers proposed diﬀerent empirical coeﬃcients.
The general form of the bore velocity is shown below [Eq. (11.8)]:
u = C
gd
s
, (11.8)
where u is the bore velocity, d
s
is the inundation depth, and C is a constant
coeﬃcient.
Various formulations were proposed by FEMA 55 (based on Dames and
Moore
28
), Iizuka and Matsutomi,
35
CCH,
44
Kirkoz,
50
Murty,
51
Bryant,
52
and
Camﬁeld
53
for estimating the velocity of a tsunami bore in terms of inundation
depth (Fig. 11.10). Velocities calculated using CCH and FEMA 55 represent a lower
and upper boundary, respectively.
11.3.1.4. Surge force
The surge force is generated by the impingement of the advancing water front of a
tsunami bore on a structure. Due to lack of detailed experiments speciﬁcally appli
cable to tsunami bores running up the shoreline, the calculation of the surge force
exerted on a structure is subject to substantial uncertainty. Accurate estimation of
the impact force in laboratory experiments is a challenging and diﬃcult task. CCH
recommends using Eq. (11.9), based on Dames and Moore
28
:
F
S
= 4.5ρgh
2
, (11.9)
where F
S
is the surge force per unit width of wall and h is the surge height.
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
274 I. Nistor et al.
0
4
8
12
16
0 2 4 6
d
S
(m)
V
(
m
/
s
)
CCH: u=h
FEMA 55: u=2(gh)^0.5
Iizuka: u=1.1(gh)^0.5
Kirkoz: u=(2gh)^0.5
Murty: u=1.83(gh)^0.5
Bryant: u=1.67h^(0.7)
Fig. 11.10. Comparison of various tsunamibore velocities as a function of inundation depth.
42
building
3h
Design inundation depth
z
h
3ρgh
qx
z
x
Fig. 11.11. Tsunami wave pressure for structural design recommended by SMBTR.
32
The point of application of the resultant surge force is located at a distance h
above the base of the wall. This equation is applicable to walls with heights equal
to, or greater than 3h. Structural walls with height less than 3h require surge forces
to be calculated using an appropriate combination of hydrostatic and drag force for
each speciﬁc situation.
SMBTR recommends using the equation for tsunami wave pressure without
soliton breakup derived by Asakura et al.
33
[Eq. (11.10)]. The equivalent static
pressure resulting from the tsunami impact is associated with a triangular dis
tribution where water depth equals three times the tsunami inundation depth
(Fig. 11.11):
qx = ρg(3h
max
−z), (11.10)
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 275
where qx is the tsunami wave pressure for structural design, z is the height of the
relevant portion from ground level (0 ≤ z ≤ 3h), ρ is the mass per unit volume of
water, and g is the gravitational acceleration.
Integration of the wave pressure formula for walls with heights equal to or greater
than 3h results in the same equation as the surge force formula recommended by
CCH [Eq. (11.9)]. The magnitude of the surge force calculated using Eqs. (11.9)
and (11.10) will generate a value equal to nine times the magnitude of the hydro
static force for the same ﬂow depth. However, a number of experiments
31,37
did not
capture such diﬀerences in magnitude. Yeh et al.
45
commented on the validity of
Eq. (11.9) and indicated that this equation gives “excessively overestimated values.”
On the other hand, Nakano and Paku
46
conducted extensive ﬁeld surveys in order
to examine the validity of the proposed tsunami wave pressure formula [Eq. (11.10)].
The coeﬃcient 3.0 in Eq. (11.10) was taken as a variable, α, and was calculated
such that it could represent the boundary between damage and no damage in the
surveyed data. A value of α equal to 3.0 and 2.0 was found for walls and columns,
respectively. The former is in agreement with the proposed formulae by both CCH
and SMBTR [Eqs. (11.9) and (11.10)].
The tsunami wave force may be composed of drag, inertia, impulse, and
hydraulic gradient components. However, SMBTR does not specify diﬀerent compo
nents for the tsunamiinduced force, and the proposed formula presumably accounts
for other components.
11.3.1.5. Debris impact force
A highspeed tsunami bore traveling inland carries debris such as ﬂoating automo
biles, ﬂoating pieces of buildings, drift wood, boats, and ships. The impact of ﬂoating
debris can induce signiﬁcant forces on a building, leading to structural damage or
collapse.
5,6
Both FEMA 55 and CCH codes account consistently for debris impact forces,
using the same approach and recommend using Eq. (11.11) for the estimation of
debris impact force:
F
i
= m
b
du
b
dt
= m
u
i
∆t
, (11.11)
where F
i
is the impact force, m
b
is the mass of the body impacting the structure,
u
b
is the velocity of the impacting body (assumed equal to the ﬂow velocity), u
i
is
approach velocity of the impacting body (assumed equal to the ﬂow velocity), and
∆t is the impact duration taken equal to the time between the initial contact of the
ﬂoating body with the building and the instant of maximum impact force.
The only diﬀerence between CCH and FEMA 55 resides in the recommended
values for the impact duration which has a noticeable eﬀect on the magnitude of
the force. For example, CCH recommends the use of impact duration of 0.1 s for
concrete structures, while FEMA 55 provides diﬀerent values for walls and piles for
various construction types as shown in Table 11.1.
According to FEMA 55, the impact force (a single concentrated load) acts hor
izontally at the ﬂow surface or at any point below it. Its magnitude is equal to
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
276 I. Nistor et al.
Table 11.1. Impact duration of ﬂoating
debris (FEMA 55).
Impact duration (s)
Type of construction Wall Pile
Wood 0.7–1.1 0.5–1.0
Steel N.A. 0.2–0.4
Reinforced concrete 0.2–0.4 0.3–0.6
Concrete masonry 0.3–0.6 0.3–0.6
the force generated by 455 kg (1000pound) of debris traveling with the bore and
acting on a 0.092m
2
(1 ft
2
) surface of the structural element. The impact force is
to be applied to the structural element at its most critical location, as determined
by the structural designer. It is assumed that the velocity of the ﬂoating body goes
from u
b
to zero over some small ﬁnite time interval (∆t). Finding the most critical
location of impact is a trial and error procedure that depends, to a large extent, on
the experience and intuition of the engineer.
11.3.1.6. Breaking wave force
Tsunami waves tend to break oﬀshore and approach shoreline as a broken hydraulic
bore or a soliton, depending on the wave characteristics and coastal bathymetry.
Consequently, classic breaking wave force formulae are not directly applicable to
the case of tsunami bores. Hence, this chapter does not discuss the estimation of
breaking wave forces.
11.3.2. Loading combinations for calculating
tsunamiinduced forces
Based on the location and type of structural elements, appropriate combinations
of tsunamiinduced force components (hydrostatic, hydrodynamic, surge, buoyant,
and debris impact force) should be used in calculating the total tsunami force.
This is due to the fact that a certain element may not be subjected to all of these
force components simultaneously. Loading combinations can signiﬁcantly inﬂuence
the total tsunami force and the subsequent structural design. Unlike the case of
tsunami waves, loading combinations for ﬂoodinduced surges are wellestablished
and have been included in design codes. The literature review revealed that proposed
tsunami loading combinations must be signiﬁcantly improved and incorporated into
new design codes. Tsunamiinduced loads are diﬀerent from ﬂoodinduced loads.
Therefore, load combinations based on ﬂood surges are not directly applicable to
tsunamis. Loading combinations proposed in the literature are as follows:
(i) FEMA 55 does not provide loading combinations speciﬁcally for calculation
of tsunami force. However, ﬂood load combinations can be used as guidance.
Flood load combinations for piles or open foundations, as well as solid walls
(foundation) in ﬂood hazard zones and coastal high hazard zones are presented
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 277
as follows:
Pile or open foundation:
F
brkp
(on all piles) + F
i
(on one corner or critical pile only), or
F
brkp
(on front row of piles only) + F
dyn
(on all piles but front row) +
F
i
(on one corner or critical pile only).
Solid (wall) foundation:
F
brkw
(on walls facing shoreline, including hydrostatic component) +
F
dyn
(assumes one corner is destroyed by debris),
where F
brkp
, F
i
, F
dyn
, and F
brkw
refer to breaking force on piles, impact force,
hydrodynamic force, and breaking force on walls, respectively. The reader can
refer to FEMA 55
43
for more details.
(ii) Yeh et al.
45
modiﬁed ﬂood load combinations provided by FEMA 55 and
adapted them for tsunami forces as follows:
Pile or open foundation:
F
brkp
(on column) + F
i
(on column), or
F
d
(on column) + F
i
(on column),
where F
d
is the drag force.
Solid (wall ) foundation (perpendicular to ﬂow direction):
F
brkw
(on walls facing shoreline) + F
i
(on one corner), or
F
s
(on walls facing shoreline) + F
i
(on one corner), or
F
d
(on walls facing shoreline) + F
i
(on one corner),
where F
s
is the surge force on walls.
(iii) Dias et al.
54
proposed two load combinations called “point of impact” and
“postsubmergence/submerged” (Fig. 11.12). These load combinations are
based on two conditions: (i) the instant that tsunami bore impacts the
structure, and (ii) when the whole structure is inundated.
Point of impact:
F
d
(on walls facing shoreline) + F
s
(on walls facing shoreline),
where F
s
is deﬁned as the hydrostatic force by Dias et al.
54
Postsubmergence/submerged:
F
d
(on walls facing shoreline) + F
b
(on submerged section of the structure).
The net hydrostatic force is zero and F
b
(γV) is the buoyant force.
F
d
F
s
W
F
s
F
s
F
d
W  V
(a) (b)
Fig. 11.12. Loading combinations: (a) point of impact/not submerged; and (b) postsubmergence/
submerged.
54
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
278 I. Nistor et al.
h
i
F
d
i
S
(a)
(b)
F
F
d
F
HS
F
HS
F
S
Fig. 11.13. Proposed loading conditions: (a) point of impact; and (b) postimpact.
42
(iv) Nouri et al.
42
proposed two new load combinations based on the two condi
tions considered by Dias et al.,
54
as shown in Fig. 11.13. The proposed load
combinations by Nouri et al.
42
are adapted to follow a consistent format as the
above combinations:
Columns:
F
s
(on front row of piles only) + F
i
(on one corner or critical column in the
front row only), or
F
d
(on all piles) + F
i
(on one corner or critical column only),
where F
s
is the surge force on walls.
Solid (wall) foundation:
F
s
(on walls facing shoreline) + F
i
(on walls facing shoreline), or
F
d
(on walls facing shoreline) + F
i
(on one critical wall facing shoreline) +
F
b
(on submerged section of the structure).
11.4. Design Example
Building codes provide guidance for the design of lateral force resisting systems
subjected to wind and seismic excitations. Tsunamiinduced loading is normally
not considered. The objective of this example is to demonstrate the levels of
lateral loading associated with tsunamis for a prototype reinforced concrete building
located in a tsunamiprone area. Speciﬁcally, the loads generated by a tsunami
bore are addressed. Other researchers have provided comparisons between tsunami
loading and other lateral loads (Okada et al.,
32
Pacheco and Robertson,
55
Nouri
et al.,
42
and Palermo et al.
56
).
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 279
6.0
6
.
0
6
.
0
6
.
0
1 2 5 4 3
A
B
D
C
6
6.0 6.0 m 6.0 6.0
Tsunami
Fig. 11.14. Plan view of structural layout of reinforced concrete momentresisting frame.
56
The following example consists of a momentresisting frame with simple
geometry, as shown in Fig. 11.14. The thickness of the slab is assumed to be
200 mm, and the beams are 450 mm deep (including the thickness of the slab) and
300 mm wide. The exterior and interior columns are 450 mm and 500 mm square sec
tions, respectively. The centertocenter storey heights are 3.65m, and a 10storey
structure is considered.
The components of the tsunamiinduced forces are calculated based on CCH,
FEMA 55, and SMBTR. The calculation of tsunamiinduced loads require a number
of assumptions, together with engineering judgment and lessons from reconnaissance
missions in regions aﬀected by tsunamis. The authors assume that the net hydro
static force exerted on the lateral system is zero in calculating the base shear. The
surge and drag forces require an eﬀective area for load transfer to the lateral force
resisting system. In this example, two scenarios are considered: (1) 100% breakaway
walls, which expose the structural elements; and (2) nonbreakaway walls. In the
latter, the exterior nonstructural elements remain intact. For breakaway walls, the
external elements will be damaged and all the columns of the structure will be
subjected to drag force. The forces are calculated for inundation depths of 1–5 m,
and the structure is oriented with its short side parallel to the shoreline. When
designing structures located in tsunamiprone areas, the inundation depth at a
speciﬁc location should be obtained from tsunami inundation hazard maps, when
available. Otherwise, numerical modeling based on various tsunami scenarios should
be conducted in order to estimate the tsunami inundation depth.
11.4.1. Hydrodynamic (drag) forces
For the case where 100% breakaway walls are assumed and the columns are exposed
to the hydraulic bore, the drag forces are based on a drag coeﬃcient of 2.0 for
square columns, as recommended by CCH and FEMA 55. For the second case,
nonbreakaway walls are assumed to remain intact and the hydraulic bore impacts
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
280 I. Nistor et al.
the entire surface of the building. In this situation, the drag coeﬃcients are taken
as 1.5 and 1.25 for CCH and FEMA 55, respectively.
11.4.2. Debris impact forces
To calculate debris impact, a mass of 455 kg is used to represent a ﬂoating object at
the water surface. The mass used is consistent with the recommendations of CCH
and FEMA 55. In this example, it is assumed that the debris will impact a single
reinforced concrete column. CCH assumes a duration of 0.1 s for concrete structures.
FEMA 55 recommends an impact duration ranging from 0.3 to 0.6 s for reinforced
concrete piles or columns. Hence, impact duration of 0.3 s is assumed in the design
example.
11.4.3. Surge forces
In this example, the surge force is applied over the full length of the building in the
direction of the tsunami for nonbreakaway walls. For the case of 100% breakaway
walls, it is assumed that the surge force will develop on the four exterior columns
which face the hydraulic bore. Note that the surge force is not applicable for FEMA
55 and that SMBTR assumes a diﬀerent surge force per unit width for columns and
walls, as mentioned in Sec. 11.3.1.4.
11.4.4. Sample calculation: Breakaway walls
The following is a sample calculation for breakaway walls for the given structure
subjected to a tsunami inundation depth of 5 m:
g = 9.81
m
s
2
; ρ = 1030
kg
m
3
; d
s
= 5 m.
Surge Force:
F
S
= 2.0ρgh
2
= 2.0
1030
kg
m
3
9.81
m
s
2
(5 m)
2
= 0.505 ×10
6
N
m
; SMBTR
F
S
×width =
0.505 ×10
6
N
m
(4(0.45 m)) = 909 ×10
3
N = 909 kN;
F
S
= 4.5 ρgh
2
= 4.5
1030
kg
m
3
9.81
m
s
2
(5 m)
2
= 1.14 ×10
6
N
m
; CCH
F
S
×width =
1.14 ×10
6
N
m
(4(0.45 m)) = 2046 ×10
3
N = 2046 kN.
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 281
Drag Force:
F
D
=
ρC
D
Au
2
2
u = C
gd
s
= 2
9.81
m
s
2
(5 m) = 14
m
s
; FEMA 55
u = d
s
= 5
m
s
; CCH
C
D
= 2.0 Rectangular columns;
A = (5 m)((16 ×0.45 m) + 8(0.50 m)) = 56 m
2
;
F
D
=
1030 kg/m
3
(2)(56 m
2
) (14m/s)
2
2
= 11,317 ×10
3
N = 11,317 kN; FEMA 55
F
D
=
1030 kg/m
3
(2)(56 m
2
) (5 m/s)
2
2
= 1442 ×10
3
N = 1442 kN. CCH
Debris Impact Force:
F
i
= m
u
i
∆t
= 455 kg
14m/s
0.3 s
= 21.2 ×10
3
N = 21 kN; FEMA 55
F
i
= 455 kg
5 m/s
0.1 s
= 22.8 ×10
3
N = 23 kN. CCH
11.4.5. Sample calculation: Nonbreakaway walls
The following is a sample calculation for nonbreakaway walls for the given structure
subjected to a tsunami inundation depth of 5 m.
g = 9.81
m
s
2
; ρ = 1030
kg
m
3
; d
s
= 5 m
Surge Force:
F
S
=
1.14 ×10
6
N
m
(18 m + 0.45 m)
= 20,973 ×10
3
N = 20,973 kN SMBTR/CCH
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
282 I. Nistor et al.
Drag Force:
F
D
=
ρC
D
Au
2
2
C
D
= 1.25 Walls FEMA 55
C
D
= 1.5 Walls CCH
A = (5 m)(18 m + 0.45 m) = 92 m
2
;
F
D
=
1030 kg/m
3
(1.25)(92 m
2
) (14 m/s)
2
2
= 11,652 ×10
3
N = 11,652 kN; FEMA 55
F
D
=
1030 kg/m
3
(1.5)
92 m
2
(5 m/s)
2
2
= 1782 ×10
3
N = 1782 kN. CCH
Debris Impact Force:
F
i
= m
u
i
∆t
= 455 kg
14 m/s
0.3 s
= 21.2 ×10
3
N = 21 kN; FEMA 55
F
i
= 455 kg
5 m/s
0.1 s
= 22.8 ×10
3
N = 23 kN. CCH
11.4.6. Results
Tables 11.2 through 11.7 provide the results for the calculation of the individual
force components for the structure considered using CCH, FEMA 55, and SMBTR,
respectively.
Given the force components, a loading combination must be speciﬁed in order
to evaluate the maximum tsunami load that would be used for either design or
analysis purposes. Yeh et al.
45
suggested loading combinations that are applicable
for tsunami loading. CCH does not speciﬁcally provide guidance to evaluate the
maximum tsunami load. Nouri et al.
42
proposed a twopart loading combination:
Initial impact and Postimpact ﬂow. For this example, these loading combinations
are similar to those of Nouri et al.
42
Table 11.8 provides the results of the tsunami
Table 11.2. Tsunamiinduced force components based on CCH for
breakaway walls.
Inundation depth Velocity Surge Drag Debris impact
Code (m) (m/s) (kN) (kN) (kN)
CCH 1 1 82 12 5
2 2 327 92 9
3 3 737 311 14
4 4 1310 738 18
5 5 2046 1442 23
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 283
Table 11.3. Tsunamiinduced force components based on CCH for non
breakaway walls.
Inundation depth Velocity Surge Drag Debris impact
Code (m) (m/s) (kN) (kN) (kN)
CCH 1 1 839 14 5
2 2 3356 114 9
3 3 7550 385 14
4 4 13,423 912 18
5 5 20,973 1782 23
Table 11.4. Tsunamiinduced force components based on FEMA 55
for breakaway walls.
Inundation depth Velocity Drag Debris impact
Code (m) (m/s) (kN) (kN)
FEMA 55 1 6 453 10
2 9 1811 13
3 11 4074 16
4 13 7243 19
5 14 11,317 21
Table 11.5. Tsunamiinduced force components based on FEMA 55
for nonbreakaway walls.
Inundation depth Velocity Drag Debris impact
Code (m) (m/s) (kN) (kN)
FEMA 55 1 6 466 10
2 9 1864 13
3 11 4195 16
4 13 7457 19
5 14 11,652 21
load calculation for CCH, FEMA 55 and SMBTR for an inundation depth of 5 m
based on loading combinations of Nouri et al.
42
For the prototype momentresisting frame structure with the short side per
pendicular to the advancing bore, it is apparent that nonbreakaway walls or rigid
exterior nonstructural components can lead to large design base shears. CCH and
SMBTR estimate signiﬁcantly larger base shears relative to FEMA 55 due to the
omission of a surge component in FEMA 55. It is evident that the width of exposed
surfaces aﬀects the magnitude of total forces exerted on a structure. Therefore, it
would be prudent to orient buildings such that the short side is placed parallel to the
shoreline. Furthermore, using breakaway or ﬂexible walls at the lower level would
reduce the lateral force that is transmitted to the lateral force resisting system. Note
that although the debris impact force is a negligible component in the base shear,
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
284 I. Nistor et al.
Table 11.6. Tsunamiinduced force components
based on SMBTR for breakaway walls.
Code Inundation depth (m) Surge (kN)
SMBTR 1 36
2 146
3 327
4 582
5 909
Table 11.7. Tsunamiinduced force components
based on SMBTR for nonbreakaway walls.
Code Inundation depth (m) Surge (kN)
SMBTR 1 839
2 3356
3 7550
4 13,423
5 20,973
Table 11.8. Tsunamiinduced load based on CCH, FEMA 55, and SMBTR for
5 m inundation depth.
Surge + Impact Drag + Impact Tsunami load
Standard Case (kN) (kN) (kN)
CCH Breakaway 2069 1465 2069
CCH Nonbreakaway 20,995 1804 20,995
FEMA 55 Breakaway N.A. 11,338 11,338
FEMA 55 Nonbreakaway N.A. 11,673 11,673
SMBTR Breakaway 909 N.A. 909
SMBTR Nonbreakaway 20,973 N.A. 20,973
it could be critical in the design of individual structural components subjected to
the debris impact.
Acknowledgment
Special thanks to Dr Andrew Cornett, Group Leader — Coastal, Ports and
Oﬀshore at the Canadian Hydraulics Centre, National Research Council of Canada
in Ottawa, Canada, for his valuable advice.
References
1. D. C. Cox and J. F. Mink, Bull. Seism. Soc. Am. 53, 1191–1209 (1963).
2. H. Yeh, Natural Hazards 4, 209–220 (1991).
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
TsunamiInduced Forces on Structures 285
3. A. Ghobarah, M. Saatcioglu and I. Nistor, Eng. Struct. 28, 312–326 (2006).
4. I. Nistor, M. Saatcioglu and A. Ghobarah, Ann. Conf. Canadian Society for Civil
Eng. (2006).
5. M. Saatcioglu, I. Nistor and A. Ghobarah, Earthquake Spectra (Earthquake Engi
neering Research Institute, 22, 2006), pp. 295–320.
6. M. Saatcioglu, I. Nistor and A. Ghobarah, Earthquake Spectra (Earthquake Engi
neering Research Institute, 22, 2006), pp. 355–375.
7. C. V. R. Murty, D. C. Rai, S. K. Jain, H. B. Kaushik, G. Mondal and S. R.
Dash, Earthquake Spectra (Earthquake Engineering Research Institute, 22, 2006),
pp. 321–354.
8. A. Ruangrassamee, H. Yanagisawa, P. Foytong, P. Lukkunaprasit, S. Koshimura and
F. Imamura, Earthquake Spectra (Earthquake Engineering Research Institute, 22,
2006), pp. 377–401.
9. B. K. Maheshwari, M. L. Sharma and J. P. Narayan, Earthquake Spectra (Earthquake
Engineering Research Institute, 22, 2006), pp. 475–493.
10. Y. Yamamoto, H. Takanashi, S. Hettiarachchi and S. Samarawickrama, Coastal Eng. J.
48(2), 117–145 (2006).
11. T. Tomita, F. Imamura, T. Arikawa, T. Yasuda and Y. Kawata, Coastal Eng. J. 48(2),
99–116 (2006).
12. N. Shuto, Tsunami Engineering Technical Report No.11, DCRC, Tohoku University
(in Japanese) (1994).
13. H. Matsutomi, T. Sakakiyama, S. Nugroho and M. Matsuyama, Coastal Eng. J. 48(2),
167–195 (2006).
14. CNN, http://www.cnn.com/interactive/world/0502/gallery.tsunami.photos/frameset.
exclude.html (2007).
15. H. Yeh, Int. Workshop on Fundamentals of Coastal Eﬀects of Tsunamis, Hilo, Hawaii
(2006).
16. S. Hibberd and D. H. Peregrine, J. Fluid Mech. 95, 323–345 (1979).
17. A. R. Packwood and D. H. Peregrine, Report AM8107, University of Bristol (1981).
18. H. Yeh, A. Ghazali and I. Marton, J. Fluid Mech. 206, 563–578 (1989).
19. H. Yeh and K. M. Mok, Phys. Fluids A 2, 821–828 (1990).
20. H. Chanson, Coastal Eng. J. 48(4), 355–370 (2006).
21. A. Tsutsumi, T. Shimamoto, E. Kawamoto and J. M. Logan, J. Waterway, Port,
Coastal, Ocean Eng., ASCE 126(3), 136–143 (2000).
22. J. R. Morison, M. P. O’Brien, J. W. Johnson and S. A. Schaaf, Petroleum Trans.,
AIME 189, 149–157 (1950).
23. F. E. Snodgrass, E. K. Rice and M. Hall, University of California, Berkeley, CA, Inst.
of Engineering Research, Technical Report Series 35, Issue 4, June (1951).
24. J. J. Stoker, Water Waves (Interscience Publishers, New York, 1957).
25. E. Cumberbatch, J. Fluid Mech. 7(3), 353–373 (1960).
26. Y. Fukui, M. Nakamura, H. Shiraishi and Y. Sasaki, Coastal Eng. Jpn. 6, 67–82 (1963).
27. R. H. Cross, PhD thesis, Dept. Civil Engineering, University of California, Berkeley,
CA (1966), p. 106.
28. Dames and Moore, in Design and Construction Standards for Residential Construction
in TsunamiProne Areas in Hawaii (Prepared for the Federal Emergency Management
Agency) (1980).
29. J. D. Ramsden and F. Raichlen, J. Waterway Port, Coastal and Ocean Eng., ASCE
116(5), 592–613 (1990).
30. J. D. Ramsden, Report No. KHR54, W. M. Keck Laboratory, California Institute of
Technology, Pasadena, California (1993), p. 251.
July 31, 2009 8:18 9.75in x 6.5in b684ch11 FA
286 I. Nistor et al.
31. J. D. Ramsden, J. Waterways, Port, Coastal and Ocean Eng. 122(3), 134–141 (1996).
32. T. Okada, T. Sugano, T. Ishikawa, S. Takai and T. Tateno, The Building Centre of
Japan (2005).
33. R. Asakura, K. Iwase and T. Iketani, Proc. Coastal Eng. JSCE 47, 911–915 (2000).
34. M. Ikeno, N. Mori and H. Tanaka, Proc. Coastal Eng. JSCE 48, 846–850 (2001).
35. H. Iizuka and H. Matsutomi, Proc. Conf. Coastal Eng., JSCE 47 (2000) (in Japanese).
36. M. Omori, N. Fujii and O. Kyotani, Proc. Coastal Eng. JSCE 47, 376–380 (2000).
37. H. Arnason, PhD thesis, University of Washington, Seattle (2005), p. 172.
38. Y. Nouri, I. Nistor, D. Palermo and A. Cornett, Coastal Structures 2007, Venice, Italy
(2007).
39. H. Matsutomi, J. Hyd. Coastal Environ. Eng. JSCE, No. 621/II47, 111–127 (1999)
(in Japanese, with English abs.).
40. R. B. Haehnel and S. F. Daly, Technical Report: ERDC/CRREL TR022, US Army
Corps of Engineers (2002), p. 40.
41. R. B. Haehnel and S. F. Daly, J. Hydraul. Eng. 130(2), 112–120 (2004).
42. Y. Nouri, I. Nistor, D. Palermo and M. Saatcioglu, 9th Canad. Conf. Earthquake
Engineering, Ottawa, Canada, June (2007).
43. Federal Emergency Management Agency, Coastal Construction Manual (3 Vols.),
3rd edn. (FEMA 55) (Jessup, MD, 2003).
44. Department of Planning and Permitting of Honolulu Hawai, Chapter 16, City and
County of Honolulu Building Code, Article 11 (2000).
45. H. Yeh, I. Robertson and J. Preuss, Report No 20054, Washington Dept. of Natural
Resources (2005).
46. Y. Nakano and C. Paku, Summaries of technical papers of Annual Meeting Architec
tural Institute of Japan (Kinki) (2005).
47. UBC, Int. Conf. Building Oﬃcials, 1997 Uniform Building Code, California (1997).
48. ASCE Standard, Minimum design loads for buildings and other structures, SEI/ASCE
705, 424 (2006).
49. International Code Council (INC), 2006 International Building Code 2006, Country
Club Hills, IL (2006), p. 675.
50. M. S. Kirkoz, 10th IUGG Int. Tsunami Symposium, Sendaishi/Miyagiken, Japan
(Terra Scientiﬁc Publishing, Tokyo, Japan, 1983).
51. T. S. Murty, Bull. Fisheries Res. Board of Canada, No. 198, Department of Fisheries
and the Environment, Fisheries and Marine Service, Scientiﬁc Information and Pub
lishing Branch, Ottawa, Canada (1977).
52. E. A. Bryant, Tsunami: The Underrated Hazard (Cambridge University Press, London,
UK, 2001), p. 320.
53. F. Camﬁeld, Tsunami Engineering (Coastal Engineering Research Center, US Army
Corps of Engineers, 1980), p. 222, Special Report (SR6).
54. P. Dias, L. Fernando, S. Wathurapatha and Y. De Silva, Int. Conf. Disaster Reduction
on Coasts, Melbourne (2005).
55. K. H. Pacheco and I. N. Robertson, Evaluation of Tsunami Loads and Their Eﬀect
on Reinforced Concrete Buildings. University of Hawaii Research Report (2005).
56. D. Palermo, I. Nistor, Y. Nouri and A. Cornett, PROTECT 2007, Whistler, Canada
(2007).
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Chapter 12
Nonconventional Wave Damping Structures
Hocine Oumeraci
LeichtweissInstitute for Hydraulic Engineering and Water Resources
Technical University Braunschweig
Beethovenstr. 51a, 38106 Braunschweig, Germany
h.oumeraci@tubs.de
After a brief discussion of the necessity (i) to consider in addition to the reduction
of wave heights in the sheltered area further aspects of the hydraulic performance
when developing and designing new coastal structures for the protection against
wave action and (ii) to develop a systematic “roadmap” of the existing concepts
and types of structures together with their hydraulic performance characteristics as
an important tool for practicing engineers and decisionmakers, ﬁve examples from
selected research studies performed in the last years by the LeichtweissInstitute
are presented in order to illustrate the development of nonconventional structures
with substantially improved performance as compared to their conventional coun
terparts. These examples include (i) a multichamber structure to overcome the
drawbacks of the perforatedJARLANtype breakwater concept, (ii) anartiﬁcial reef
made of successive submerged permeable screens to increase the wave damping per
formance and to better control both wave reﬂection and wave transmission, (iii) a
“High Mound Composite Breakwater” (HMCB) concept to decrease breaking wave
loads, wave reﬂection, overtopping, and spray generation at the structure, (iv) an
onshore wave damping barrier made of staggered walls, and (v) a rubble mound
breakwater with a core made of geotextile sand containers.
12.1. Introduction
In the past, the development of wave damping structures used in coastal protection
and harbor engineering was rather directed toward achieving a maximum reduction
of wave height in the sheltered area, while less attention was generally paid to
further performance aspects such as
(i) the reduction of wave loads, wave reﬂection, and wave overtopping,
(ii) the reduction of environmental impacts (degradation of the marine landscape
and of the water quality in the sheltered area, erosion of the foreshore, down
coast erosion, etc.),
287
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
288 H. Oumeraci
(iii) the control of wave periods in the sheltered area,
(iv) the reduction of spray generation at the structure,
(v) the adaptability of the wave damping structure for multipurpose use (recre
ational activities, wave power generation, etc.), and
(vi) the reversibility of built structures if not performing as expected.
In the future, the increasing interest in the practical implementation of the
sustainability principles in coastal engineering
18
will require that more eﬀort should
be put on the development of innovative low impact structures with better hydraulic
performance characteristics, lower capital and maintenance costs for the proper
sheltering of harbor and other facilities as well as for the proper protection of
threatened coasts against storm surges.
The main objective of this chapter is to illustrate, by means of three selected
example research studies performed in the last years by the LeichtweissInstitute
(LWI), the process of developing and testing nonconventional structures. This
process particularly consists of
(i) a better understanding of the hydraulic functioning of the existing concepts,
(ii) a clear identiﬁcation of their drawbacks with respect to the commonly accepted
and new emerging performance characteristics/requirements (see above), and
(iii) a better control of the physical processes and structure parameters which con
tribute to the improvement of the hydraulic performance and to the reduction
of wave loads.
However, before starting with these example studies, a ﬁrst attempt is made to
discuss, why and how a parameterization of the existing concepts and structures for
wave damping might help to develop a systematic “roadmap” of these concepts as
an important decision tool for practicing engineers and coastal managers.
12.2. Parameterization of Wave Damping Structures:
A First Attempt
A large variety of concepts and structures for wave damping is available. However,
the designer and coastal manager faced with the problem of ﬁnding the concept
mostly suited for the intended use have no proper tool to quickly make the right
choice before embarking into detailed investigations. For this purpose, a compact
overview of the most important concepts/types of structures is needed, including the
basic information on their hydraulic functioning (“roadmap”). The roadmap should
not only provide the required performance characteristics. Moreover, the unwanted
eﬀects and impacts may also be directly provided or easily derived for each type
of structure. The latter issue becomes increasingly important, because the choice
will not, as in the past, mainly or exclusively be dictated by the wave damping
eﬃciency, but also by further criteria associated with the sustainable protection of
coastal zones.
18
Unfortunately, such a practical decision tool, which is expected to substantially
help both engineers and coastal managers, is still missing. Due to the considerable
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 289
number of existing concepts/types of structures, it will certainly be very diﬃcult
or even impossible to elaborate any useful compact overview if a parameterization
of the existing types of structures is not undertaken. In order to illustrate this ﬁrst
requirement, a very tentative parameterization is proposed for a single structure as
a ﬁrst starting base in Fig. 12.1. Such a parameterization may then lead to the very
tentative compact overview in a parameterized form in Fig. 12.2.
SWL
E
i
E
r
E
t
H
L
ε
h
1
h
2
b
h
s h
Fig. 12.1. Tentative parameterization of single wave damping structure.
h
s
> h
h
s
< h h
s
= 0
h
1
= 0
h
2
= 0
h
1
> 0
h
2
= 0
h
1
= 0
h
2
> 0
h
1
> 0
h
2
> 0
h
1
+ h
2
= h
ε = 0
impermeable
ε ≠ 0
permeable
b = 0
thin
walls
b > 0
wide
structures
1.1 1.2 1.3 1.4
2.1 2.2 2.3 2.4
3.1
3.2
3.3 3.4
3.5
4.1 4.2 4.3 4.4 4.5
ε = 0
impermeable
ε ≠ 0
permeable
1.5
no structure
no structure
2.5
Fig. 12.2. Tentative compact overview of existing wave damping structures in parametrized form.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
290 H. Oumeraci
It is seen from Figs. 12.1 and 12.2 that only ﬁve parameters (structure width B,
porosity ε, structure height d
B
, submergence depth h
1
, and clearance h
2
) together
with the water depth h are suﬃcient to achieve a parametric description of most
types of existing wave damping structures. But this represents only a ﬁrst step. The
next important step is to provide the characteristic curves describing the hydraulic
performance for each of the parametrized structure types in Fig. 12.2. For this
purpose, the simplest procedure is to use the fact that the incident wave energy
E
i
is converted into a reﬂected (E
R
), a dissipated (E
D
), and a transmitted (E
T
)
energy component:
E
i
= E
R
+E
D
+E
T
. (12.1)
Equation (12.1) written in terms of the reﬂection (K
R
), dissipation (K
D
), and
transmission (K
T
) coeﬃcients yields
1 = K
2
R
+K
2
D
+K
2
T
. (12.2)
If only the wave damping performance
P
W
= 1 −K
2
T
(12.3)
is of interest, then the functioning curves P
W
= f(T) such as those qualitatively
described in Fig. 12.3 would be useful, because they clearly illustrate how each
type of structure will respond over the full range of incident wave periods T. For
instance, it is seen that the curtain wall founded on piles (Type 1.3) performs
well for shorter wave periods T, but has a very low wave damping performance
Wave Period T [s]
W
a
v
e
d
a
m
p
i
n
g
P
e
r
f
o
r
m
a
n
c
e
P
w
1.0
0.5
0.0
0
( )
= + = +
2 2 2
w R D T
P K K 1 K
K
R
= Reflection Coeff.; K
D
= Dissipation Coeff.; K
T
= Transmission Coeff.
Vertical Breakwater (3.1)
Horizontal Breakwater
on Piles (3.5)
Caisson Breakwater
on Piles (3.3)
CurtainWall on
Piles (1.3)
Submerged Breakwater
(3.2)
16
Fig. 12.3. Wave damping performance (principle sketch!).
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 291
Fig. 12.4. Hydraulic functioning of a single perforated wall under nonbreaking and nonover
topping wave conditions (adapted from Ref. 4).
for longer periods. Widening the wall to a caisson (Type 3.3) will considerably
improve the wave damping performance over a wider range of wave periods. Very
often, however, the wave damping performance alone is not suﬃcient and a “full”
description of the hydraulic functioning as a function of the most relevant parameter
governing wave reﬂection, wave transmission, and energy dissipation [see Eq. (12.1)]
is required. For instance, in the case of a single surface piercing perforated screen
(Type 2.1) and for nonbreaking and nonovertopping wave conditions, a kind of
“dynamic porosity” (here reﬂection number RN) which combines the inﬂuence of
incident wave height H
i
, water depth h, and wall porosity ε, is found to be the
most appropriate parameter to describe the reﬂected, transmitted, and dissipated
components of the total incident wave energy (Fig. 12.4).
For most of the types of structure shown in Fig. 12.2, curves describing the
hydraulic performance in terms of wave reﬂection, transmission, and damping
already exist which are derived empirically, analytically, or numerically.
2,11,19,21,22
12.3. Selected Example Studies for the Development
of Nonconventional Wave Damping Structures
In order to illustrate the processes associated with the development of noncon
ventional wave damping structures, the main results from three selected example
research studies performed in the last years by LWI are brieﬂy discussed below,
including
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
292 H. Oumeraci
(i) a multichamber caisson structure,
(ii) an artiﬁcial reef made of a submerged progressive ﬁlter,
(iii) a high mound composite breakwater (HMCBconcept),
(iv) an onshore wave damping barrier, and
(v) a rubble mound breakwater with a core made of geotextile sand containers
(GSC).
12.3.1. Multichamber caisson structure
One of the most useful concepts to cope with the high reﬂection induced by vertical
face breakwaters and seawalls is the perforated Jarlantype breakwater which was
introduced in 1960 in Canada.
9
It consists of a single dissipating chamber bounded
seaward by a perforated front wall (porosity ε ≈ 20%) and shoreward by an imper
meable back wall (One Chamber System).
The incident wave energy is partlyreﬂectedat the front wall andpartly transmitted
throughthe perforations into the wave chamber, where acertainamount of the incident
wave energy is reﬂected by the back wall while a large part is dissipated due to reso
nance phenomena, vortices, andfrictionlosses. The relative importance of the reﬂected
and dissipated part of the total incident wave energy, and thus the hydraulic perfor
mance, depends onthe porosityof the front wall, but is essentiallygovernedbythe ratio
of the chamber width B and the wave length Lof the incident waves (B/L).
Although the JARLANtype breakwater concept has been used more or less
successfully worldwide, it has a basic drawback (see Fig. 12.6) which requires a
further development of this concept. For this purpose, it was necessary to investigate
ﬁrst the key processes which contribute to the wave damping by friction (local losses
and vortices) and by destructive interference of the incident and reﬂected waves
over the full range of B/L ratios (i.e., over the full range of incident wave periods).
The experimental results in Fig. 12.5 well illustrate how a traditional JARLANtype
caisson works.
As shown by the upper curve in Fig. 12.6, the traditional JARLANtype caisson
(OCS) has, at its optimal working point (B/L ≈ 0.2), a much lower reﬂection
coeﬃcient (and thus a much larger energy dissipation) than a vertical impermeable
wall. However, the response is very selective with respect to the incident wave
periods; i.e., it performs satisfactorily only within a very narrow range of the B/L
ratios. In order to overcome this drawback, a new MultiChamber System (MCS)
was developed and tested in the Large Wave Flume of Hannover. As shown by the
lower curve in Fig. 12.6, the new MCS concept not only provides a lower reﬂection
coeﬃcient; moreover this reﬂection coeﬃcient is kept at its lowest level over the full
range of practical B/L ratios (i.e., for B/L > 0.25, where B is deﬁned as the overall
width of the MultiChamber System.
In addition to the substantial improvement of the hydraulic performance, which
is achieved by the new multichamber concept, the new concept has also the
advantage to strongly reduce the resulting horizontal wave forces F
+
total
(obtained
by superposition of the forces acting simultaneously on each wall). Because wave
forces are directly related to water surface elevation and thus to wave reﬂection,
3,6
the total force F
+
total
(related to force F
+
0%
on a single impermeable vertical wall
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 293
wave damping
by friction
wave damping by
destructive interference
total wave damping
relative chamber width B/L []
20%
60%
40%
80%
100%
0%
0 0.1 0.2 0.3 0.4 0.5
E
i
E
r
impermeable
back wall
wave chamber
perforated front
wall
B
reflected wave energy
incident wave
energy
Wave damping relative to incident wave energy E
i
Fig. 12.5. Wave damping friction and destructive interference (Adapted from Ref. 5).
with zeroporosity) exhibits a very similar behavior to the reﬂection coeﬃcient with
respect to the B/L ratio (compare Figs. 12.6 and 12.7).
This ﬁrst example study has illustrated how a detailed insight into the physical
processes responsible for the hydraulic functioning of an existing concept may lead
to a clear identiﬁcation of the drawbacks of this concept by indicating how to
overcome them and how to achieve substantial improvements through the intro
duction of additional structure members. Given its potential to substantially reduce
and better control wave reﬂection (less risk to navigation and less sea bed scour),
wave loads, wave runup and overtopping, spray generation, etc., the new Multi
Chamber System represents an ideal alternative as a breakwater, jetty, and quay
wall as well as a seawall for the protection of reclaimed sea fronts and artiﬁcial
islands. Due to the ﬂexibility of caisson structures to allow any shape and size,
the seawalls can be adapted to incorporate promenades and any further facility for
recreation activities, etc.
12.3.2. Submerged wave absorber as artiﬁcial reef
for coastal protection
An interesting cost eﬀective and soft alternative to conventional seawalls for coastal
protection against erosion are artiﬁcial reefs which have the advantage
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
294 H. Oumeraci
Fig. 12.6. Reﬂection coeﬃcient versus relative chamber width for a traditional JARLANtype
caisson and for the new multichamber system (Adapted from Ref. 4).
Fig. 12.7. Resulting horizontal wave force versus relative chamber width for a traditional
JARLANtype caisson and for the new multichamber system (Adapted from Ref. 4).
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 295
(i) to attenuate the waves before they reach the shoreline,
(ii) to be invisible for viewers from the beach and therefore do not aﬀect the marine
landscape,
(iii) to reduce the morphological impact on the foreshore (erosion) and on the neigh
boring coast (downcoast erosion), and
(iv) to ensure a suﬃcient water exchange between the open sea and the sheltered
area.
However, the existing artiﬁcial reef concepts have serious drawbacks: the wave
damping performance is limited and sometimes not suﬃcient. Moreover, the overall
hydraulic performance is diﬃcult to control, due to the limitations associated with
the variation of the structure parameters, etc. Therefore, a reef concept made
of submerged permeable screens with predetermined porosity and spacing has
been experimentally tested in the Large Wave Flume of Hannover (GWK). As
schematically shown in Fig. 12.8 for a threeﬁlter system, this new reef concept
is particularly appropriate for the protection of such coastal areas which are
frequently used for recreation activities.
12−14
Before starting with the systematic study on the hydraulic performance of this
new reef concept, it was important to demonstrate ﬁrst its eﬃciency with respect
to the protection against beach erosion. For this purpose, a submerged twoﬁlter
system with porosities ε = 11% (front screen) and ε = 5% (back screen), spacing
B = 10.3 m and height d
B
= 4 m was installed in front of a beach proﬁle. The same
beach proﬁle was previously tested in the Large Wave Flume of Hannover
17
without
any protection under the same storm surge conditions (storm surge water level
h = 5.0 m, TMAwave spectrum with signiﬁcant wave height H
s
= 1.20 m and peak
Fig. 12.8. Submerged threeﬁlter system as an artiﬁcial reef for coastal protection.
23
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
296 H. Oumeraci
Fig. 12.9. Beach proﬁle development under storm wave attack with and without reef protection.
period T
P
= 6.6 s, storm duration t = 10 h). The comparison of the results related
to the development of the beach proﬁle at the diﬀerent stages of the tests with and
without the reef structure as shown in Fig. 12.9 provides a good demonstration of
the eﬃciency of the new reef concept as a soft protection alternative. In both cases
(i.e., with and without protection), one can indeed observe a beach erosion, a net
seaward transport, and a sand bar formation. However, the net seaward transport
rate is about twice larger for the unprotected beach than for the protected beach.
In addition, the reef causes the seaward transport to occur only within a limited
narrow zone, so that this sand bar does not extend further seaward as in the case
of an unprotected beach. Moreover, the eroded volume above the storm water level
(h = 5.0 m) for the protected beach is only half as much as the eroded volume
for the beach without any protection. As a result, the recession of the shoreline of
the protected beach is only half as much as the recession of the unprotected beach
(Fig. 12.10).
Regarding the hydraulic performance of submerged wave absorbers, a brief illus
tration is given in Fig. 12.11, showing how a twoﬁlter system damps the waves
over the full range of B/L ratios. The contribution of the seaward screen to the
total wave damping of the system varies between 30% for B/L ≈ 0.5 and 85% for
B/L ≈ 0.3. The maximum wave damping performance of the system also occurs at
B/L = 0.3, while the minimum value is at B/L = 0.5. These and further results
with a threeﬁlter system clearly show that for a given submergence depth (R
c
/H
i
)
the relative spacing B/L represents the most decisive parameter to describe the
hydraulic performance of wave absorbers.
23
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 297
Fig. 12.10. Beach proﬁle after storm with and without reef (Adapted from Ref. 11).
relative spacing B/L
r
e
l
.
d
a
m
p
i
n
g
[
%
]
r
e
l
a
t
i
v
e
w
a
v
e
h
e
i
g
h
t
wave damping at
filter 2
wave
damping at
filter 1
H
i
/ H
t
H
t1
/ H
t
H
i
/H
t
(measurement)
H
t1
/H
t
(measurement)
d
B
/h = 0.98
d
B
=
3 .9 4 m
h
11 % 5%
B
10m
H
i
H
r
H
t
H
t1
F i l ter 1 F i l ter 2
1 0.75 0.5 0.25 0
wave damping at filter 1/total wave damping
100
50
0
0
0.5
1
1.5
2.5
2
Fig. 12.11. Wave damping of a submerged twoﬁlter system (Adapted from Ref. 11).
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
298 H. Oumeraci
Fig. 12.12. Hydraulic performance of submerged single screens and ﬁlter systems.
From the comparative analysis of the hydraulic performance shown in Fig. 12.12
for a submerged single screen with diﬀerent porosities (ε = 0%, 5%, and 11%) and
submerged two or threeﬁlter systems it is seen that
(i) using a ﬁlter system instead of a single screen substantially increases the
amount of dissipated energy,
(ii) unlike a single screen, a ﬁlter system can substantially reduce and simultane
ously control both wave reﬂection and wave transmission,
(iii) using an optimized threeﬁlter system, more than 80% of the incident wave
energy can be dissipated,
(iv) the relative submergence depth R
c
/H
i
is an important parameter for the wave
damping performance of both single screens and ﬁlter systems, and
(v) for the range of practical submergence depths (R
c
/H
i
≈ −1), the highest
improvement of the wave damping performance is achieved when using a two
ﬁlter system instead of a single ﬁlter. A further increase of the number of
ﬁlters would lead to comparatively less improvement of the wave damping
performance.
Further results have indeed shown that the relative spacing B/L is much more
important than the number and porosity of the ﬁlters to control the reﬂected,
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 299
transmitted, and dissipated wave energy. A more detailed discussion on these issues
is provided by Oumeraci and Koether
23
and Koether.
11
As mentioned in the introduction, there are many cases where the reduction
of wave periods in the sheltered area may also be very important. Such situations
occur, for instance, when the reef is principally used to control sediment transport
or to reduce wave loads and overtopping at a seawall located behind the reef. An
example result to illustrate the eﬃciency of the new reef concept by reducing both
transmitted wave periods and heights is shown in Fig. 12.13 for two relative heights
of a threeﬁlter system.
This second example study has shown again that substantial improvements of the
existing concepts can only be achieved through a better insight into the mechanisms
and processes governing the hydraulic performance. In fact, using a single submerged
screen, only a very limited amount of incident wave energy can be dissipated, i.e.,
a decrease of wave transmission can only be achieved at the cost of the increase of
wave reﬂection (see Ref. 23 for more results).
Using a conventional reef made of rubble material would require a very wide
structure and a progressive decrease of the porosity in wave direction in order
to achieve a satisfactory wave damping performance. This is not only costly and
diﬃcult to construct and to maintain, but it is also very diﬃcult to control the
hydraulic performance by means of a variation of the structure parameters as it is
the case for this new artiﬁcial reef concept.
Fig. 12.13. Reduction of wave periods by a threeﬁlter system.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
300 H. Oumeraci
The understanding of the underlying mechanisms has shown that a new reef
concept made of two or three submerged thin ﬁlters is an elegant and cost
eﬀective alternative to overcome most of the drawbacks of the existing reef con
cepts, including a substantial reduction and a better control of the reﬂected and
transmitted components of the incident wave energy by means of the variation of
the structure parameters (submergence depth, porosity, number and spacing of sub
merged slit walls.) A theoretical model to optimize submerged wave absorbers has
been developed by Koether.
11,23
The model has been successfully validated by large
scale experimental data for regular and irregular waves as well as for submerged
singleslit wall and wave absorbers with two and three ﬁlters.
Preliminary tests in the Large Wave Flume of Hannover using comparatively a
singlewall as well as a twowall and a threewall submerged wave absorber subject
to 1 m high solitary waves have shown that this reef concept might also be applied
for the protection against tsunami. In fact, the part of wave energy of the incident
solitary waves dissipated is more than 75% and 85% for a twowall and a three
wall system, respectively. The experimental results showing the incident, reﬂected,
and transmitted waves for a single, two, and threewall systems are plotted and
discussed by Oumeraci and Koether.
23
12.3.3. High mound composite breakwater (HMCB)
Based on a historical concept which was ﬁrst used in 1830 in Cherbourg, France,
and in 1890 in Alderney, UK, a new HMCBconcept has extensively been inves
tigated within two joint research projects by LWI together with the Port and
Harbor Research Institute (PHRI), Yokosuka, Japan and with the Civil Engineering
Research Institute (CERI), Sapporo, Japan (Fig. 12.14).
Fig. 12.14. Old and new high mound breakwater concepts.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 301
The new HMCBconcept was intended to be used mainly as a seawall for the
protection of artiﬁcial islands (oﬀshore airport) and roads with heavy traﬃc along
the coast, but also as a breakwater [Fig. 12.14(c), 12.14(d)] in harbors. This concept
has recently been implemented at MORI Port in Hokkaido, Japan.
14
The governing
characteristic of the HMCBconcept is to cause the highest waves in the spectrum to
break before reaching the crest structure by means of a relatively ﬂat slope (about
1:3). This concept has the following advantages:
(i) the required amount of rubble material is much less than for a conventional
rubble mound breakwater,
(ii) the required armor units are smaller since they are all placed below still water
level, and
(iii) the required crest structure is much smaller than a conventional caisson
breakwater.
In order to further substantially reduce the breaking wave impact loads on
the superstructure and to overcome further drawbacks of the old HMCBconcept
(excessive wave reﬂection, overtopping, spray generation, etc.), a major innovation
was introduced to improve the performance of the concrete superstructure: a front
slit wall made of piles (porosity of about 30%) and a relatively short dissipation
chamber behind it. If a breaking wave reaches the structure, the total wave force is
split spatially and temporarily into the following force components (Fig. 12.15):
(i) a force component F
1
on the permeable front wall,
(ii) a force component F
2
on the impermeable back wall, and
(iii) a stabilizing downward force F
3
on the bottom slab of the dissipation chamber.
Fig. 12.15. New HMCBconcept.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
302 H. Oumeraci
In addition to the reduction of wave loads and the subsequent reduction of
the required size of the concrete superstructure, a substantial improvement of the
overall hydraulic performance characteristics is also achieved by using the new
HMCBconcept. A summary of the improvements in comparison to the older concept
(vertical impermeable superstructure) as obtained from extensive hydraulic model
investigations is given in Table 12.1.
With respect to the wave loads, it is seen that a substantial reduction of the
horizontal and uplift forces is achieved, which would result in a reduction of about
50% of the required weight of the superstructure to ensure sliding stability. With
respect to the hydraulic performance, it is seen that wave reﬂection is reduced by
about 25%, and as a result of the reduction of wave overtopping the required crest
level above still water level is reduced by about 40%. Even without using any splash
reducer at the front and back wall of the concrete superstructure, the splash/spray
heights are reduced by half. Further details on the results in Table 12.1 are given by
Oumeraci and Muttray,
26
Muttray et al.,
15
Oumeraci et al.,
25
Takahashi et al.,
33
Sch¨ uttrumpf et al.,
31
Muttray et al.,
16
and Oumeraci et al.
28
It might be important to stress that the reduction of spray generation at wave
damping structures is becoming an increasingly important design requirement, par
ticularly when the structures are used for the protection of oﬀshore airports, the pro
tection of roads with heavy traﬃc along the coastline, etc. The growing importance
of this relatively new emerging issue may be explained by the detrimental eﬀects that
spray might have on inland and nearshore infrastructures, operations behind the
structure as well as on vegetation. In fact, spray can be transported by wind up to
30 km inland and a ﬂux of spray salt up to 400µg/m
2
·s may result. Large quantities
of saltwater dispersed over wide coastal areas may result in the following eﬀects:
(i) shortterm detrimental eﬀects such as disturbance/stoppage of car traﬃc,
10
electric power supply, port, and airport operations and
Table 12.1. Main improvement by the new HMCBconcept.
Improved performance of new HMCB with slit front wall
(without splash reducers)
Hydraulic performance r
x
CoV [%]
Transmission coeﬃcient C
t
1.0 50
Reﬂection coeﬃcient C
r
0.75 30
Mean overtopping rate 0.5 30
Maximum overtopping rate q
max
0.6 25
Number of overtopping waves N
ow
0.7 25
Splash and spray height R
w
, R
s
0.5 60
Required freeboard R
c
0.6 30
Wave Forces and Stability
Horizontal forces F
h
0.4 50
Uplift forces F
u
0.6 40
Required weight for sliding stability W 0.5 50
CoV =
σ
x
¯ x
mean
(Coeﬀ. of vacation); ¯ τ
x
=
¯ x
new HMCB
¯ x
old HMCB
(mean
reduction factor)
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 303
(ii) longerterm impacts such as salt corrosion of buildings and other facilities,
damage to agriculture and inland vegetation, etc.
Therefore, one of the most challenging tasks to cope with salt spray consists of
the development of innovative shapes of the structure crest to substantially reduce
the splash/spray height induced by the breaking waves at the structure. For this
purpose, special tests were performed to analyze the eﬀectiveness of various alter
natives to reduce splash/spray height.
8,25,28
As shown in Fig. 12.16, the splash/spray height observed in the Large Wave
Flume tests for diﬀerent structure alternatives is related to the corresponding value
observed for the new HMCB without any splash reducer as a reference structure
(Type 1). Even without any splash reducer, the new HMCBconcept is capable of
reducing the splash/spray height by half as compared to a conventional impermeable
wall (Type 0). Adding a conical splash reducer on each pile of the front slit wall
and a splash reducer on the back wall will further reduce the splash/spray height by
about half (Type 3). The same improvement was achieved by using a continuous slab
on the crest of the front wall (Type 4), but this has the drawback to substantially
increase the uplift force.
28
Since the eﬃciency of Type 2, both with respect to the
reduction of splash heights and overtopping, was found very low, no further analysis
was performed.
This third example study has highlighted how a very old concept can basically
be improved to provide a new solution with a considerable improvement of the wave
Fig. 12.16. Eﬃciency of various alternatives for the reduction of splash and spray height.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
304 H. Oumeraci
loading conditions and of the hydraulic performance characteristics such as wave
reﬂection, wave overtopping and spray generation. The study also allowed to identify
the reduction of spray generation as a new emerging criterion for the development
and design of novel wave damping structures.
12.3.4. Onshore wave damping barrier (OWBD)
A nonconventional permeable barrier to damp storm waves running up a historical
promenade on the North Sea Island of Norderney was developed and tested in the
Large Wave Flume (GWK), Hannover.
29
The barrier consists of 1.30 m high straight
and curved staggered wall elements with a length of 5.5 m and 7.7 m, respectively
(Fig. 12.17).
The relatively low crest level and the discontinuous nature of the barrier resulted
from the requirements that, for the tourists on the promenade the view to the
sea and the direct access to the shoreline should not be obstructed. Moreover, the
barrier should architecturally and aesthetically ﬁt into the local landscape, so that
it will not necessarily be perceived as a coastal protection structure [Fig. 12.18(a)].
The eﬃciency of the OWBD concept in terms of wave overtopping reduction was
successfully tested in the Large Wave Flume (GWK) in Hannover [Figs. 12.18(b)
and 12.18(c)] and implemented in Norderney [Fig. 12.18(d)], where it withstood
without any damage several storm surges for more than ﬁve years.
Fig. 12.17. Onshore wave damping barrier for the North Sea Island Norderney (Adapted from
Ref. 29).
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 305
Fig. 12.18. Onshore wave damping barrier (OWBD).
Since broadening the crest of the embankment (Alternative 4) was practically
not feasible, the OWBD concept (Alternative 2) proved the most performant in
reducing wave overtopping (by a factor of 5) as compared to other conventional
alternatives (Fig. 12.19). More details on the results can be found in Refs. 29 and 32.
The fourth example study has shown that under some circumstances storm waves
can be damped eﬀectively by a discontinuous low barrier made of aesthetically and
functionally wellconceived staggered wall elements (Fig. 12.20).
This concept can be adapted for the protection against tsunami by using a
much wider barrier and more robust wall elements. In fact, the ﬁeld evidence expe
rienced with seawalls and breakwaters during the 2004 Indian Ocean Tsunami
clearly suggests that protective structures should not be designed to completely
stop the tsunami. Indeed, this is neither economically justiﬁable nor environmen
tally and socially acceptable. Therefore, protective structures would be preferable
which aim at progressively weakening the tsunami power without blocking com
pletely the inundation and which have the overall additional beneﬁt of broadly
blocking ﬂoating debris in a rather soft manner. Such a concept would par
ticularly be appropriate for urbanized and touristic coastal areas where forests
(bioshields) cannot be planted due to unfavorable local conditions and must
therefore be replaced by a manmade barrier ﬁtting architecturally in the local
marine landscape.
20
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
306 H. Oumeraci
Fig. 12.19. Reduction of wave overtopping by conventional and nonconventional alternatives
(Adapted from Refs. 29 and 32).
Fig. 12.20. Details of wall elements of OWBD in Norderney.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 307
12.3.5. Rubble mound breakwater with a core made
of geotextile sand containers
12.3.5.1. Motivations
There are several reasons which might lead the engineer and other decisionmakers
in practice to use sand instead of conventional quarry run for the core of rubble
mound breakwaters and structures, including among others:
(i) nonavailability of rock material in suﬃcient quantities and at aﬀordable costs;
(ii) sediment inﬁltration through rubble mound structures which may result in
the shoaling of navigation channels and harbor basins, and thus in higher
maintenance dredging costs; and
(iii) reduction of wave transmission through the structure which might particularly
be crucial in the case of long waves.
On the other hand, the use of sand as a quasiimpervious core instead of quarry
stone would result in an increase of
(i) wave setup and runup at the structure,
(ii) wave overtopping, and
(iii) wave reﬂection,
which might be detrimental to the stability of the structure, to the operations on
and behind the breakwater (due to excessive overtopping) as well as to navigation
and seabed stability (due to excessive wave reﬂection).
Moreover, serious diﬃculties arise in practice when trying to design and con
struct the ﬁlter to protect the sand core against washout by wave action. Applying
geometrically closed ﬁlter criteria would result in very complex, multiple, and rel
ative thin ﬁlter layers which will not only be very costly and very diﬃcult to build
in larger water depths, but also might certainly fail due to the almost unpredictable
very complex loading conditions of the sand core under cyclic pulsations by waves
and entrained air at the interface with the last ﬁlter layer. Such failures have indeed
been observed under both laboratory and ﬁeld conditions in the past. Laboratory
evidence has also shown that introducing the socalled “geometrically open ﬁlter”
criteria to design a “hydraulic sandtight ﬁlter” may indeed reduce the number of
ﬁlter layers. However, the main practical diﬃculties mentioned above will remain,
including those associated with the longterm stability of the sand core due to the
high complexity of the loading and its uncontrollability during the entire storm
duration and the life cycle of the structure.
Geotextile ﬁlters might present themselves as an alternative to the very complex,
costly, and uncertain ﬁlter made of multiple layers of granular material. However,
geotextile mats are not only diﬃcult to install under waves and currents, but also
may introduce a shear surface which might be detrimental to the stability of the
armor layer.
A more feasible alternative is to use a core made of geotextile sand containers
(geocore). This will not only allow to overcome the aforementioned core stability
problems, but also to provide (i) a better erosion stability of the core and (ii) an
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
308 H. Oumeraci
Fig. 12.21. Class of geocore structures in comparison to conventional rubble mound structures.
increased stability against seismic loads as compared to a core simply made of
loose sand. However, many of the drawbacks mentioned above remain with respect
to wave setup, runup, overtopping, reﬂection, and armor stability in comparison
to a conventional breakwater core. Therefore, an extensive research program has
been initiated at LWI to study both the hydraulic performance and armor stability,
including the processes involved and the development of prediction formulae for the
design of a class of rubble mound structures with a core made of geotextile sand
containers (Fig. 12.21).
In the following, only the case “rubble mound breakwater with GSCcore” in
Fig. 12.17(a) is brieﬂy addressed. Based on comprehensive hydraulic model tests
performed in the twinwave ﬂumes of LWI, a comparative analysis of the two break
water types was ﬁrst performed with respect to wave reﬂection, wave transmission,
wave runup, and wave overtopping performance. As a result, prediction formulae
were proposed for the geocore breakwater and compared with those for the conven
tional rubble mound breakwater alternative.
24
12.3.5.2. Wave reﬂection performance
Using the reﬂection model proposed by Oumeraci and Muttray,
27
which accounts for
the relative depth k
0
d as the primary inﬂuencing parameter, the following prediction
formula was derived for both breakwaters (Fig. 12.22):
K
r
= 0.43 ·
tanα
√
k
o
d
, (12.4)
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 309
Fig. 12.22. Reﬂection coeﬃcient K
r
versus relative water depth k
0
d.
with k
o
= 2π/L
0
(wave number in deepwater); tan α = steepness of the seaward
slope of the structure; d = water depth in front of the structure.
12.3.5.3. Wave transmission performance
Several models have been proposed in the literature
7,24
to predict wave transmission
through a rubble mound structure which, however, do not explicitly account for the
eﬀect of the core permeability. The parameter which mostly aﬀects the transmission
coeﬃcient K
T
has been found to be a function of the relative freeboard R
c
/H
s
and
the steepness s
m
= H
s
/L
0
of the incident waves
1
:
R
∗
=
R
c
H
s
s
m
2π
. (12.5)
In fact, the bestﬁt in analyzing the experimental results for the geocore and con
ventional breakwater was found by using the modiﬁed freeboard R
∗
according to
Eq. (12.5). As a result, the following formula was obtained for the transmission
coeﬃcient (Fig. 12.23):
K
T
= a
r
· (R
∗
c
)
b
r
, (12.6)
where a
r
and b
r
are constants, which depend on the permeability of the structure.
The structure parameters a
r
and b
r
are obtained for the conventional and the
geocore breakwater (Fig. 12.23).
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
310 H. Oumeraci
Fig. 12.23. Wave transmission coeﬃcient for conventional breakwater and geocore breakwater.
12.3.5.4. Wave runup performance
Based on the model proposed in CEM
7
for the conventional breakwater type, the
following runup formula was determined (CoV = 2.8%):
R
u2%
= 1.217 · ξ
0.274
om
H
mo
for ξ
om
= 3.3 −7.0, (12.7)
while for the geocore breakwater the following formula was obtained (CoV = 3.4%):
R
u2%
= 1.415 · ξ
0.274
om
H
mo
for ξ
om
= 4.1 −6.7, (12.8)
with ξ
om
= surf similarity parameter:
ξ
om
= tan α
H/
gT
2
m−1.0
2π
.
The diﬀerence between the two breakwater types is in the order of 20%; i.e., the
required crest level of a geocore breakwater should be increased accordingly if wave
runup or/and wave overtopping is an important issue (Fig. 12.24).
12.3.5.5. Wave overtopping performance
The bestﬁt of the experimental data for the conventional breakwater and the
geocore breakwater was obtained by using the model proposed by TAW (2002)
34
:
Q
∗
= a · exp
−
b · R
∗
γ
f
, (12.9)
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 311
Fig. 12.24. Runup for conventional breakwater versus runup for geocore breakwater.
where
Q
∗
=
q
(g · H
3
m0
)
0.5
,
q = average overtopping discharge [m
3
/s · m] and R
∗
= R
c
/H
m0
= relative free
board [].
Using a = 0.2 and b = 2.6 in Eq. (12.9) for both breakwater types, the correction
factor γ
f
, initially intended to account for the surface roughness eﬀect, is used to
distinguish between the eﬀect of permeability on the average overtopping rate q
(Fig. 12.25):
• γ
f
= 0.52 for the conventional breakwater (CoV = 12%),
• γ
f
= 0.60 for the geocore breakwater (CoV = 20%).
Based on the results in Fig. 12.25 and the results of further analysis,
30
the
geocore breakwater is associated with three to four times higher overtopping
discharges than the conventional breakwater.
12.3.5.6. Lessons learned from the study
The ﬁfth example study has shown that using sand encapsulated in geotextile con
tainers for the core of rubble mound breakwaters, which is about one order of mag
nitude less permeable than a conventional core made of quarry run, may provide
many advantages. This is particularly the case when rock material (quarry run) is
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
312 H. Oumeraci
Fig. 12.25. Average wave overtopping for conventional and geocore breakwater.
not available in suﬃcient quantities and at aﬀordable costs or when the reduction
of wave transmission and sediment inﬁltration through the rubble mound structure
is an important issue. However, one should keep in mind that using a GSCcore
instead of a conventional core will result in larger runup heights and higher wave
overtopping rates. Moreover, larger armor units are required to ensure hydraulic
stability. More results are given by Oumeraci et al.
24
The advantages of the nonconventional “geocore” concept are expected to
be particularly revealed when applied to a seawall protecting reclaimed land
(Fig. 12.17(d)), due to the diﬃculties to protect the latter from being washed out
through a conventional core made of quarry run.
12.4. Concluding Remarks and Perspectives
This chapter has attempted to show that one of the challenges of a sustainable pro
tection against wave attack is the development of innovative structures/concepts
which are not only capable to further reduce the wave heights in the sheltered area,
but also to achieve substantial improvements with respect to the reduction of wave
loads, wave reﬂection, and overtopping. Moreover, the novel types of structures
should also be able to cope with new emerging requirements such as (i) a better
control of the wave periods in the sheltered area, (ii) a reduction of spray gener
ation at the structure, and (iii) a better adaptability to multipurpose use. Since
the importance of these new emerging design criteria is expected to considerably
increase in the context of the sustainable development of coastal zones, they might
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 313
represent challenging candidate topics for future research. A further important can
didate issue for applied research will certainly consist of the elaboration of a kind
of “roadmap” of the existing wave damping concepts (possibly in a parametrized
form!), together with the associated characteristic curves which not only summarize
the associated hydraulic performance, but also the possible negative impacts and
further information required for a fast decisionmaking.
Acknowledgments
Most of the ideas, material, and illustrative examples used herein are based
on inputs taken from completed research projects supported by the German
Research Council (DFG, Bonn), the Federal Ministry for Education and Research
(BMBF, Bonn), the Science and Technology Agency of Japan (STA, Tokyo),
and the Fifth Port Construction Bureau of the Ministry of Transport (MOT,
Tokyo), the European Community within the EUMAST research program, NLWK
Norden and BBG Bauberatung GmbH & Co. KG, Germany. The author grate
fully acknowledges the ﬁnancial support. Since most of these projects were
joint projects with Port and Harbor Research Institute (PHRI, Yokosuka), Civil
Engineering Research Institute (CERI, Sapporo), the Institute of Naval Archi
tecture and Ocean Engineering, TU Berlin, and many further European Insti
tutes, the author also gratefully acknowledges the fruitful cooperation with these
partners.
References
1. N. W. H. Allsop, Lowcrest breakwaters, studies in random waves, Proc. Coastal
Structures, Specialty Conference on the Design, Construction, Maintenance and Per
formance of Coastal Structures, Arlington, Virginia, USA (1983).
2. H. Bergmann, Hydraulische Wirksamkeit und Seegangsbelastung senkrechter Wellen
schutzbauwerke mit durchl¨assiger Front, Dissertation, Mitteilungen des Leichtweiss
Instituts f¨ ur Wasserbau, Heft 147 (2001) (in German).
3. H. Bergmann and H. Oumeraci, Wave loads at perforated caisson structures,
Proc. 27th Int. Conf. Coastal Engineering (ICCE’2000), Sydney, Vol. 2 (2000),
pp. 1622–1635.
4. H. Bergmann and H. Oumeraci, Digue innovante en caissons multichambres — Fonc
tionnement et sollicitations hydrauliques, Revue fran¸ caise de g´enie civil. 5(7), 973–993
(2001) (Sp´ecial issue) (in French).
5. H. Bergmann and H. Oumeraci, Hydraulische Wirksamkeit und Wellenbelastung
senkrechter Wellenschutzbauwerke mit durchl¨assiger Front, Hansa (2002), pp. 1–8
(in German).
6. H. Bergmann and H. Oumeraci, Waveinduced water levels and pressure distribution
at perforated wall, Proc. COPEDEC VII, Dubai, UAE (2008), Paper No. B11
(in print).
7. CEM, Coastal Engineering Manual, Engineer Manual 111021100 (US Army Corps
of Engineers, Washington, DC, USA, 2003).
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
314 H. Oumeraci
8. T. Hayakawa, K. Kimura, S. Takahashi, M. Muttray, M. Kudella and H. Oumeraci,
Wave splash height on a highmound composite breakwater, Proc. 4th Conf. Hydro
dynamics (ICHD’2000), Yokohama, Japan, Vol. 2 (2000), pp. 671–676.
9. G. E. Jarlan, A perforated vertical wall breakwater, The Dock and Harbour Authority
(1961), pp. 394–398.
10. K. Kimura, T. Fujiike, K. Kamikubo, R. Abe and K. Ishimoto, Damages to vehicles
on a coastal highway by wave action, Proc. Conf. Coastal Structures’99, Santander,
Spain, Vol. 2 (2000), pp. 1009–1016.
11. G. Koether, Hydraulische Wirksamkeit getauchter Einzelﬁlter und Filtersysteme —
Prozessbeschreibung und Modellbildung f¨ ur ein innovatives Riﬀkonzept, Dissertation
an der TU Braunschweig, LeichtweissInstitut f¨ ur Wasserbau (2002), p. 163 (in
German).
12. G. Koether, H. Bergmann and H. Oumeraci, Wave attenuation by submerged ﬁlter
systems, Proc. 4th Int. Conf. Hydrodynamics (ICHD’2000), Yokosuka, Japan, Vol. 2
(2000), pp. 711–716.
13. G. Koether and H. Oumeraci, Schutzbauwerke f¨ ur Sandstr¨ ande mit touristischer
und ¨ okologischer Bedeutung — UnterwasserFiltersysteme, Hansa, Heft, 11 (2001),
pp. 93–98.
14. M. Mori, Y. Yamamoto and K. Kimura, Wave force and stability of upright section
of high mound composite seawall, Proc. Abstracts Int. Conf. Coastal Eng., Hamburg
(2008), p. 206.
15. M. Muttray, H. Oumeraci, K. Shimosako and S. Takahashi, Hydraulic performance
of high mound composite breakwater, Proc. 26th Int. Conf. Coastal Eng. (ICCE‘98)
(1998), pp. 2207–2220.
16. M. Muttray, H. Oumeraci, K. Shimosako and S. Takahashi, Wave load on an innovative
high mound composite breakwater: Results of large scale experiments and tentative
design formulae, Proc. Conf. Coastal Structures’99, Santander, Spain, Vol. 1 (2000),
pp. 353–362.
17. J. Newe, Methodik f¨ ur großmaßst¨abliche 2DExperimente zum Strandverhalten unter
Sturmﬂutbedingungen, Dissertation an der TU Braunschweig, LeichtweissInstitut f¨ ur
Wasserbau (2002) (in German).
18. H. Oumeraci, The sustainability challenge in coastal engineering, Keynote Lecture,
Proc. 4th Int. Conf. Hydrodynamics (ICHD), eds. Y. Goda et al., Yokohama, Japan,
Vol. 1 (2000), pp. 57–83.
19. H. Oumeraci, Breakwaters, Part 2, Planing and Design of Ports and Marine Ter
minals, ed. H. Agerschou (Thomas Telford, London, UK, 2004), pp. 155–262.
20. H. Oumeraci, Nearshore and onshore tsunami eﬀects, Background paper for
DFGRound Table Discussion, Hannover, http://www.fzk.unihannover.de/323.html
(2006).
21. H. Oumeraci, G. F. Clauss, R. Habel and G. Koether, Unterwasser Filtersysteme zur
Wellend¨ampfung, Final Report of BMBFResearch Project (2001), p. 311 (in German).
22. H. Oumeraci, A. Kortenhaus, N. W. H. Allsop, M. B. De Groot, R. S. Crouch, J. K.
Vrijling and H. G. Voortman, Probabilistic Design Tools for Vertical Breakwaters
(Balkema, Rotterdam, The Netherlands, 2001), pp. 392.
23. H. Oumeraci and G. Koether, Hydraulic performance of a submerged wave absorber
for coastal protection, Advances in Coastal Ocean Engineering (2008), pp. 36 (in print).
24. H. Oumeraci, A. Kortenhaus, H. Breustedt and P. Schley, Hydraulic model investiga
tions on breakwaters with a core made of geotextile sand containers, Research Report
No. 933, Report LeichtweißInstitut f¨ ur Wasserbau, TU Braunschweig (2007), pp. 71
and Annexes.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
Nonconventional Wave Damping Structures 315
25. H. Oumeraci, M. Kudella, M. Muttray, K. Kimura and T. Hayakawa, Wave runup
and wave overtopping on a high mound composite breakwater, Final Report No. 831,
LeichtweissInstitut, TU Braunschweig (1998), pp. 88.
26. H. Oumeraci and M. Muttray, Large scale model tests on a highmound composite
type breakwater, Research Report No. 818, LeichtweissInstitute for Hydraulics, TU
Braunschweig (1997), pp. 115.
27. H. Oumeraci and M. Muttray, Bemessungswellenparameter vor Strukturen mit ver
schiedenen Reﬂexionseigenschaften, Abschlussbericht DFGProjekt, OU 1/33, Braun
schweig, Germany (2001), 93 S (in German).
28. H. Oumeraci, M. M. Muttray, K. Kudella, K. Kimura and S. Takahashi, Wave loading
of a high mound composite breakwater (HMCB) with splash reducers, Proc. 4th Int.
Conf. Hydrodynamics (ICHD’2000), Yokohama, Japan, Vol. 2 (2000), pp. 659–670.
29. H. Oumeraci, H. Sch¨ uttrumpf, A. Kortenhaus, M. Kudella, J. M¨oller and M. Muttray,
Untersuchungen zur Erweiterung bzw. zum Umbau des Deckwerks am Nordstrand
von Norderney, Research Report No. 853, LeichtweissInstitute for Hydraulics, TU
Braunschweig (2000b) (in German).
30. P. Schley, Hydraulische Wirksamkeit von geschuetteten Wellenbrechern mit
herkoemmlichen Kern und einem Kern aus geotextilen Sandcontainern — eine experi
mentelle Vergleichsstudie, Diplomarbeit am LeichtweissInstitut f¨ ur Wasserbau, Fach
bereich Bauingenieurwesen, TU Braunschweig, Germany (2006) 80 S, 8 Anlagen
(in German).
31. H. Sch¨ uttrumpf, H. Oumeraci, K. Kimura, T. Hayakawa and J. Moeller, Wave over
topping on a high mound composite breakwater, Proc. Conf. Coastal Structures’99,
Santander, Spain, Vol. 1 (2000), pp. 397–404.
32. H. Sch¨ uttrumpf, H. Oumeraci, F. Thorenz and J. M¨ oller, Reconstruction and rehabili
tation of a historical seawall at Norderney, ICE, Proc. Breakwaters, Coastal Structures
Coastlines, ed. N. W. H. Allsop (Thomas Telford, London, 2002), pp. 257–268.
33. S. Takahashi, K. Shimosako, H. Oumeraci, M. Muttray and K. Kimura, Reduced wave
overtopping characteristics of a new high mound composite breakwater: Results of
smallscale experiments, Proc. Conf. Coastal Structures’99, Santander, Spain, Vol. 1
(2000), pp. 389–396.
34. TAW: Technical Report on wave runup and wave overtopping at dikes, Techn.
Advisory Committee on Flood Defence, Delft, The Netherlands.
July 31, 2009 8:18 9.75in x 6.5in b684ch12 FA
This page intentionally left blank This page intentionally left blank
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Chapter 13
Wave Interaction with Breakwaters
Including Perforated Walls
KyungDuck Suh
Department of Civil and Environmental Engineering
Seoul National University, San 561, Shillimdong, Gwanakgu
Seoul 151744, Republic of Korea
kdsuh@snu.ac.kr
During the past several decades, breakwaters including perforated walls have been
introduced to resolve various problems associated with gravitytype breakwaters.
In this chapter, ﬁrst, the mathematical models are described that predict various
hydrodynamic characteristics of single or multiplerow curtainwallpile break
waters, the upper part of which is a curtain wall and the lower part consisting of an
array of vertical piles. Their extension to irregular waves is also described. These
models can be used for curtainwall breakwaters by just removing the piles. They
can also be used for pile breakwaters by removing the curtainwall and extending
the piles to the water surface. Second, the mathematical model to predict wave
reﬂection from a fullyperforatedwall caisson mounted on a rubble foundation is
described, and its applicability to a partiallyperforatedwall caisson and irregular
waves is described. Third, a discussion is given for the calculation of the socalled
permeability parameter, which represents the energy dissipation and phaseshift
of ﬂows passing across a perforated wall.
13.1. Introduction
Gravitytype breakwaters using rubble mound or vertical caissons have been widely
used to provide a calm basin for ships and to protect harbor facilities from rough
seas. In general, the width of these breakwaters increases with water depth, leaving
a large footprint and requiring a great amount of construction material, especially
when built in deeper water. Often, they block littoral drift and cause severe erosion
or accretion in neighboring beaches. In addition, they prevent the circulation of
water, thus deteriorating the water quality within the harbor. In some places, they
obstruct the passage of ﬁshes and bottomdwelling organisms. A solid soil foun
dation is also needed to support such heavy structures.
317
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
318 K.D. Suh
In order to resolve the abovementioned problems, perforated wall structures
have been introduced especially in small craft harbors. The simplest perforated wall
structure may be a curtainwall breakwater (sometimes called wave screen or skirt
breakwater), which consists of a vertical wall extending from the water surface to
some distance above the seabed.
1−3
Recently, Isaacson et al.
4
proposed a slotted
curtainwall breakwater. Another simple perforated wall structure may be an array
of vertical piles, which is called a pile breakwater in this chapter. The closely spaced
piles induce energy dissipation due to viscous eddies formed by the ﬂow through the
gaps. To examine the wave scattering by vertical piles, hydraulic model tests have
been used.
5−8
Eﬀorts toward developing analytical models to calculate the reﬂection
and transmission coeﬃcients have also been made.
8−11
Recently, Suh et al.
12
intro
duced a curtainwallpile breakwater, the upper part of which is a curtain wall and
the lower part consisting of an array of vertical piles. They developed a mathe
matical model that predicts various hydrodynamic characteristics of a curtainwall
pile breakwater. More recently, Suh and Ji
13
extended the model to a multiplerow
breakwater.
Another type of breakwaters including perforated walls is a perforatedwall
caisson breakwater, even though it is still a gravitytype breakwater. It reduces not
only wave reﬂection, but also wave transmission due to overtopping. It also reduces
wave forces, especially impulsive wave forces, acting on the caisson.
14,15
In order
to examine the reﬂection characteristics of a perforatedwall caisson breakwater,
hydraulic model tests have been used.
16−19
Mathematical models for predicting the
reﬂection coeﬃcient have also been developed.
10,20,21
On the other hand, Fugazza
and Natale
22
proposed a closedform solution for wave reﬂection from a perforated
wall caisson. The aforementioned mathematical approaches dealt with the case in
which the waves are normally incident to the perforatedwall caisson lying on a
ﬂat sea bottom. To resolve these limitations, Suh and Park
23
developed a math
ematical model that can predict the wave reﬂection from a fullyperforatedwall
caisson mounted on a rubble mound foundation when regular waves are obliquely
incident on the breakwater at an arbitrary angle. Recently, Suh et al.
24
described
how to apply this model to a partiallyperforatedwall caisson and irregular waves.
In this chapter, ﬁrst, the mathematical models of Suh et al.
12
and Suh and Ji
13
are described. These models were developed to predict the hydrodynamic character
istics of curtainwallpile breakwaters. However, they can also be used for curtain
wall breakwaters or pile breakwaters by just removing the piles or by removing
the curtain wall and extending the piles to the water surface, respectively. Second,
the mathematical model of Suh and Park
23
and its applicability to a partially
perforatedwall caisson and irregular waves is described. Third, a discussion is given
for the calculation of the socalled permeability parameter, which represents the
energy dissipation and phaseshift of ﬂows passing across a perforated wall.
13.2. CurtainWallPile Breakwaters
13.2.1. Singlerow curtainwallpile breakwater
A curtainwallpile breakwater is sketched in Fig. 13.1, in which h = constant water
depth in still water; d = height of the curtain wall below the still water level;
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 319
Fig. 13.1. A deﬁnition sketch of a curtainwallpile breakwater: (a) side view; and (b) front view.
b = thickness of the wall. A Cartesian coordinate system (x, z) is deﬁned with the
positive x directing downwave from the crest line of the breakwater and the vertical
coordinate z being measured vertically upward from the still water line. The distance
between the centers of two neighboring piles is denoted as 2A, and the width of an
opening is 2a so that the porosity of the lower part of the breakwater at x = 0 is
deﬁned as r
0
= a/A. A regular wave train with wave height H
i
is incident on the
positive xdirection. The ﬂuid domain is divided into region 1 (x ≤ 0) and region 2
(x ≥ 0).
Assuming incompressible ﬂuid and irrotational ﬂow motion, the velocity
potential exists, which satisﬁes the Laplace equation. Linearizing the freesurface
boundary conditions, the following boundary value problem for the velocity
potential Φ(x, z, t) is obtained:
∂
2
Φ
∂x
2
+
∂
2
Φ
∂z
2
= 0 , (13.1)
∂Φ
∂z
−
ω
2
g
Φ = 0 at z = 0 , (13.2)
∂Φ
∂z
= 0 at z = −h, (13.3)
where ω = wave angular frequency, and g = gravitational acceleration. Assuming
periodic motion in time t, we can assume the solution to the above problem as
Φ(x, z, t) = Re
−
igH
i
2ω
1
cosh(kh)
φ(x, z) exp(−iωt)
, (13.4)
where i =
√
−1; and the symbol Re represents the real part of a complex value.
The wave number k must satisfy the dispersion relationship, ω
2
= gk tanh(kh).
The spatial variation of the velocity potential φ(x, z) should be determined in each
region.
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
320 K.D. Suh
We assume that the length scale of the ﬂow near the breakwater is of the order of
the wall thickness, which is much smaller than the farﬁeld length scale of O(k
−1
),
so that the wall has no thickness mathematically and the 3D feature near the
breakwater does not signiﬁcantly aﬀect the 2D farﬁeld solutions. Then, φ
1
(x, z)
and φ
2
(x, z) must satisfy the following matching conditions at x = 0:
∂φ
1
∂x
=
∂φ
2
∂x
= 0 for −d ≤ z ≤ 0, x = 0 , (13.5)
∂φ
1
∂x
=
∂φ
2
∂x
= iG(φ
1
−φ
2
) for −h ≤ z ≤ −d, x = 0 , (13.6)
where the subscripts indicate the regions of the ﬂuid domain. The ﬁrst matching
condition describes that the horizontal velocities vanish on both sides of the curtain
wall of the breakwater. The second one for the lower part of the breakwater describes
that the horizontal mass ﬂuxes (or indirectly horizontal velocities) in the two regions
must be the same at the breakwater and that the horizontal velocity at the opening is
proportional to the diﬀerence of velocity potentials, or the pressure diﬀerence, across
the breakwater. The proportional constant G, often called permeability parameter,
is in general complex, which will be discussed in detail later.
To solve the boundary value problem [Eqs. (13.1) to (13.3)], we use the eigen
function expansion method of Isaacson et al.
4
The velocity potential is expressed in
a series of inﬁnite number of solutions:
φ
1
= φ
i
−
∞
¸
m=0
A
m
cos[µ
m
(h +z)] exp(µ
m
x) , (13.7)
φ
2
= φ
i
+
∞
¸
m=0
A
m
cos[µ
m
(h +z)] exp(−µ
m
x) , (13.8)
where φ
i
= cosh[k(h+z)] exp(ikx) is the incident wave potential. The wave numbers
µ
m
are the solution of the dispersion relation, ω
2
= −gµ
m
tan(µ
m
h), which has an
inﬁnite discrete set of real roots ±µ
m
(m ≥ 1) for nonpropagating evanescent waves
and a pair of imaginary roots µ
0
= ±ik for propagating waves. We take µ
0
= −ik
so that the propagating waves in Eqs. (13.7) and (13.8) correspond to the reﬂected
and transmitted waves, respectively. We also take the positive roots for m ≥ 1 so
that the nonpropagating waves die out exponentially with the distance from the
breakwater.
Now, Eqs. (13.7) and (13.8) satisfy the freesurface boundary condition in
Eq. (13.2) and the bottom boundary condition in Eq. (13.3). Also, they automat
ically satisfy the requirement that the horizontal velocities must be matched at
the breakwater. In order to solve for the unknown coeﬃcients, A
m
s, we use the
matching conditions at the breakwater. First, Eqs. (13.7) and (13.8) are substi
tuted into Eqs. (13.5) and (13.6), respectively. Multiplying each resulting equation
by cos[µ
n
(h + z)], integrating with respect to z over the appropriate domain of z
(i.e., z = −d to 0, or z = −h to −d), and ﬁnally adding them, we obtain a matrix
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 321
equation for A
m
:
∞
¸
m=0
C
mn
A
m
= b
n
for n = 0, 1, 2, . . . , ∞, (13.9)
where
C
mn
= µ
m
f
mn
(−d, 0) + (µ
m
−2iG)f
mn
(−h, −d) , (13.10)
b
n
= −µ
0
[f
0n
(−d, 0) +f
0n
(−h, −d)] , (13.11)
f
mn
(p, q) =
q
p
cos[µ
m
(h +z)] cos[µ
n
(h +z)]dz
=
1
2
¸
sin[(µ
m
+µ
n
)(h +z)]
µ
m
+µ
n
+
sin[(µ
m
−µ
n
)(h +z)]
µ
m
−µ
n
q
p
for m = n,
1
4µ
m
[2µ
m
(h +z) + sin[2µ
m
(h +z)]]
q
p
for m = n.
(13.12)
Note that the mathematical model developed for a curtainwallpile breakwater
can be used for a pile breakwater just by setting d = 0. It can also be used for
a curtainwall breakwater by using the permeability parameter G = 1/b, which
was derived from the energy dissipation formula for a curtainwall breakwater of
Kriebel.
25
Once the wave potentials are calculated, we can obtain various engineering wave
properties. The reﬂection and transmission coeﬃcients are given by
C
r
= A
0
 (13.13)
and
C
t
= 1 +A
0
 , (13.14)
respectively. The wave runup on the upwave face of the breakwater is given by
R
u
=
H
i
2
1 −
1
cosh(kh)
∞
¸
m=0
A
m
cos(µ
m
h)
. (13.15)
In the limiting case of a fulldepth impermeable vertical wall (d → h or r
0
→ 0),
A
m
= 0 for all m ≥ 1 and A
0
= −1, so that C
r
= 1.0, C
t
= 0.0, and R
u
= H
i
as
expected.
Since the vertical distributions of wave pressure on both upwave and downwave
sides of the breakwater are known, the wave force can also be calculated. The
maximum horizontal wave force F
max
per unit width of the breakwater is given by
F
max
= ρgH
i
1
cosh(kh)
∞
¸
m=0
A
m
µ
m
{sin(µ
m
h) −r
0
sin[µ
m
(h −d)]}
+
ρg
8
H
2
i
¸
1 −
1
cosh(kh)
∞
¸
m=0
A
m
cos(µ
m
h)
¸
2
−
¸
1 +
1
cosh(kh)
∞
¸
m=0
A
m
cos(µ
m
h)
¸
2
¸
, (13.16)
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
322 K.D. Suh
where ρ = density of ﬂuid. The second term on the righthand side represents
the secondorder force contribution of the wave crest regions on the upwave and
downwave sides of the breakwater.
26
Without this term, in the limiting case of a
fulldepth impermeable vertical wall (d → h or r
0
→0), A
m
= 0 for all m ≥ 1 and
A
0
= −1 so that the preceding equation becomes
F
s
max
= ρgH
i
1
k
sinh(kh)
cosh(kh)
, (13.17)
where the superscript s stands for the standing wave in front of an impermeable
vertical wall. In another limiting case of no breakwater (d → 0 and r
0
→ 1), F
max
becomes zero, as expected.
13.2.2. Extension to multiplerow breakwaters
The singlerow curtainwallpile breakwater still gives large transmission for long
period waves. To reduce the wave transmission, the draft of the curtain wall must
be increased, or the porosity between the piles must be decreased. Then, however,
the wave reﬂection and the wave forces and moments acting on the breakwater
may increase. Therefore, it is diﬃcult to increase the draft or decrease the porosity
beyond certain limits. A multiplerow breakwater may be a solution for these
problems. It may also be advantageous for providing a space on top of the break
water for ﬁshing, walking, and maintenance of the breakwater, and so on.
Suh and Ji
13
extended the mathematical model to a multiplerow breakwater,
whose deﬁnition sketch is shown in Fig. 13.2, in which d
j
= height of the jth curtain
wall below the still water level; and b
j
= thickness of the jth wall. The center of the
jth wall is located at x = x
j
. The distance between the centers of two neighboring
piles is denoted as 2A
j
, and the width of the gap between the piles is 2a
j
so that
the porosity of the lower part of the jth wall is deﬁned as r
j
= a
j
/A
j
. The ﬂuid
domain is divided into J + 1 regions by the J walls. The upwave and downwave
regions of the jth wall are deﬁned as Ω
j−1
and Ω
j
, respectively.
Using the same assumptions as those made for a singlerow breakwater, the
same boundary value problem as that given by Eqs. (13.1)–(13.3) is obtained for
the velocity potential Φ
j
(x, z, t) in each region, which can be assumed as
Φ
j
(x, z, t) = Re
−
igH
i
2ω
1
cosh(kh)
φ
j
(x, z) exp(−iωt)
. (13.18)
The spatial variation of the velocity potential φ
j
(x, z) should be determined in each
region.
The solutions in the regions of Ω
0
and Ω
J
are given by
φ
0
= A
00
cos[µ
0
(h +z)] exp[−µ
0
(x −x
1
)]
+
∞
¸
m=0
B
0m
cos[µ
m
(h +z)] exp[µ
m
(x −x
1
)] , (13.19)
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 323
x
1
x
2
x
j
x
J
j1 0
1 2 j J1 J
z
h
H
Ω
Ω
Ω Ω
Ω Ω Ω
i
x
d
1
b
1
r
1
d
2
b
2
r
2
d
j
b
j
r
j
d
J
b
J
r
J
d
j
b
j
2a
j
2A
j
(a) (b)
Fig. 13.2. A deﬁnition sketch of a multiplerow curtainwallpile breakwater: (a) side view; and
(b) bird’seye view of jth row.
φ
J
=
∞
¸
m=0
A
Jm
cos[µ
m
(h +z)] exp[−µ
m
(x −x
J
)] , (13.20)
respectively. Here, A
jm
and B
jm
are the coeﬃcients of the component waves prop
agating forward and backward, respectively. The ﬁrst subscript (j) indicates the
row of the wall, while the second one (m) indicates the wave component. The
wave numbers µ
m
are the same as those used for a singlerow breakwater. Taking
A
00
= 1, the ﬁrst term on the righthand side of Eq. (13.19) represents the incident
wave potential, and B
00
 and A
J0
 are the reﬂection and transmission coeﬃcients,
respectively. On the other hand, the solutions in the regions from Ω
1
to Ω
J−1
are
given by
φ
j
=
∞
¸
m=0
{A
jm
exp[−µ
m
(x −x
j
)] +B
jm
exp[µ
m
(x −x
j+1
)]} cos[µ
m
(h +z)] ,
j = 1, 2, . . . , J −1. (13.21)
The matching conditions at each row of the breakwater are the same as
Eqs. (13.5) and (13.6), i.e.,
∂φ
j−1
∂x
=
∂φ
j
∂x
= 0 for −d
j
≤ z ≤ 0, x = x
j
, (13.22)
∂φ
j−1
∂x
=
∂φ
j
∂x
= iG
j
(φ
j−1
−φ
j
) for −h ≤ z ≤ −d
j
, x = x
j
. (13.23)
These matching conditions are used to solve for the unknown coeﬃcients A
jm
s and
B
jm
s. First, for x = x
1
, Eqs. (13.19) and (13.21) are substituted into Eqs. (13.22)
and (13.23). Multiplying each resulting equation by cos[µ
n
(h+z)], integrating with
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
324 K.D. Suh
respect to z over the appropriate domain of z (i.e., z = −d
1
to 0, or z = −h to
−d
1
), and ﬁnally adding them, we obtain
∞
¸
m=0
[{µ
m
f
mn
(−d
1
, 0) + (µ
m
−iG
1
)f
mn
(−h, −d
1
)}B
0m
+iG
1
f
mn
(−h, −d
1
)A
1m
+iG
1
exp(−µ
m
∆x
1
)f
mn
(−h, −d
1
)}B
1m
]
= µ
0
A
00
{f
0n
(−d
1
, 0) +f
0n
(−h, −d
1
)} +iG
1
A
00
f
0n
(−h, −d
1
)
for n = 0, 1, 2, . . . , ∞ (13.24)
and
∞
¸
m=0
[µ
m
{f
mn
(−d
1
, 0) +f
mn
(−h, −d
1
)}B
0m
+µ
m
{f
mn
(−d
1
, 0) +f
mn
(−h, −d
1
)}A
1m
−µ
m
exp(−µ
m
∆x
1
){f
mn
(−d
1
, 0) +f
mn
(−h, −d
1
)}B
1m
]
= µ
0
A
00
{f
0n
(−d
1
, 0) +f
0n
(−h, −d
1
)} for n = 0, 1, 2, . . . , ∞, (13.25)
where
∆x
j
= x
j+1
−x
j
, (13.26)
and f
mn
(p, q) are the same as given in Eq. (13.12). For x = x
j
(j = 2 to J −1),
Eq. (13.21) is substituted into Eqs. (13.22) and (13.23). Again, multiplying each
resulting equation by cos[µ
n
(h + z)], integrating with respect to z over the appro
priate domain of z, and ﬁnally adding them, we obtain
∞
¸
m=0
[exp(−µ
m
∆x
j−1
){−µ
m
f
mn
(−d
j
, 0)+µ
m
f
mn
(−h, −d
1
)
+iG
j
f
mn
(−h, −d
j
)}A
j−1,m
+{µ
m
f
mn
(−d
j
, 0) −µ
m
f
mn
(−h, −d
j
)
+iG
j
f
mn
(−h, −d
j
)}B
j−1,m
−iG
j
f
mn
(−h, −d
j
)A
jm
−iG
j
exp(−µ
m
∆x
j
)f
mn
(−h, −d
j
)B
jm
] = 0
for n = 0, 1, 2, . . . , ∞ (13.27)
and
∞
¸
m=0
[µ
m
exp(−µ
m
∆x
j−1
){f
mn
(−d
j
, 0)+f
mn
(−h, −d
j
)}A
j−1,m
−µ
m
{f
mn
(−d
j
, 0) +f
mn
(−h, −d
j
)}B
j−1,m
−µ
m
{f
mn
(−d
j
, 0) +f
mn
(−h, −d
j
)}A
jm
+µ
m
exp(−µ
m
∆x
j
){f
mn
(−d
j
, 0) +f
mn
(−h, −d
j
)}B
jm
] = 0
for n = 0, 1, 2, . . . , ∞. (13.28)
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 325
Finally, for x = x
J
, Eqs. (13.20) and (13.21) are substituted into Eqs. (13.22) and
(13.23). Using the similar procedure as above, we obtain
∞
¸
m=0
[exp(−µ
m
∆x
J−1
){−µ
m
f
mn
(−d
J
, 0) +µ
m
f
mn
(−h, −d
J
)
+iG
J
f
mn
(−h, −d
J
)}A
J−1,m
+{µ
m
f
mn
(−d
J
, 0) −µ
m
f
mn
(−h, −d
J
) +iG
J
f
mn
(−h, −d
J
)}B
J−1,m
−iG
J
f
mn
(−h, −d
J
)A
Jm
] = 0 for n = 0, 1, 2, . . . , ∞ (13.29)
and
∞
¸
m=0
[µ
m
exp(−µ
m
∆x
J−1
){f
mn
(−d
J
, 0)+f
mn
(−h, −d
J
)}A
J−1,m
−µ
m
{f
mn
(−d
J
, 0) +f
mn
(−h, −d
J
)}B
J−1,m
−µ
m
{f
mn
(−d
J
, 0)
+f
mn
(−h, −d
J
)}A
Jm
] = 0 for n = 0, 1, 2, . . . , ∞. (13.30)
If we take M wave modes (i.e., one progressive and M −1 evanescent modes), the
number of unknown coeﬃcients is 2JM. Equations (13.24), (13.25), and (13.27)–
(13.30) give 2JM equations, which can be solved for the unknown coeﬃcients.
For a multiplerow breakwater, the reﬂection and transmission coeﬃcients are
given by
C
r
= B
00
 , (13.31)
and
C
t
= A
J0
 , (13.32)
respectively. The wave runup on the upwave side of the ﬁrst row of the breakwater
is given by
R
u
=
H
i
2
1
cosh(kh)
A
00
cos(µ
0
h) +
∞
¸
m=0
B
0m
cos(µ
m
h)
. (13.33)
The wave force on each wall can be calculated by integrating the wave pressure
acting on upwave and downwave sides of the wall. The maximum horizontal wave
force F
max
per unit width of the ﬁrst wall of breakwater is given by
F
max
=
ρgH
i
2
1
cosh(kh)
A
00
µ
0
sin(µ
0
h)
+
∞
¸
m=0
1
µ
m
{B
0m
−A
1m
−B
1m
exp[µ
m
(x
1
−x
2
)]} sin(µ
m
h)
¸
−r
1
A
00
µ
0
sin[µ
0
(h −d
1
)]
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
326 K.D. Suh
+
∞
¸
m=0
1
µ
m
{B
0m
−A
1m
−B
1m
exp[µ
m
(x
1
−x
2
)]} sin[µ
m
(h −d
1
)]
¸
+
ρgH
2
i
8
1
cosh
2
(kh)
A
00
cos(µ
0
h) +
∞
¸
m=0
B
0m
cos(µ
m
h)
2
−
∞
¸
m=0
{A
1m
+B
1m
exp[µ
m
(x
1
−x
2
)]} cos(µ
m
h)
2
.
(13.34)
Again, the second term on the righthand side represents the secondorder force
contribution of the wave crest regions on the upwave and downwave sides of the
wall. Without this term, in the limiting case of a fulldepth impermeable vertical
wall, B
0m
= 0 for all m ≥ 1, B
00
= A
00
= 1, and A
1m
= B
1m
= 0 for all m ≥ 0, so
that the preceding equation becomes Eq. (13.17).
13.2.3. Extension to irregular waves
Using the above regular wave models, the reﬂection and transmission coeﬃcients
can be calculated diﬀerently for each frequency component, i.e., C
r
(f) and C
t
(f),
where f = wave frequency. In the computation of the permeability parameter G in
Eqs. (13.6) and (13.23), the rootmeansquared (rms) wave height is used in place
of the incident wave height H
i
, because the ﬂow separation due to irregular waves
and the resulting energy dissipation are induced not by the individual component
waves but by the superposition of the component waves. The rms wave height is
calculated by H
rms
= H
s
/
√
2, where H
s
is the signiﬁcant wave height. The spectral
densities of the reﬂected and transmitted waves, respectively, are calculated for a
particular frequency component by
S
η,r
(f) = C
r
(f)
2
S
η,i
(f) , (13.35)
S
η,t
(f) = C
t
(f)
2
S
η,i
(f) , (13.36)
where S
η,i
(f) = incident wave energy density.
The frequencyaveraged reﬂection and transmission coeﬃcients are then calcu
lated as
27
C
r
=
m
0,r
m
0,i
, (13.37)
C
t
=
m
0,t
m
0,i
, (13.38)
where m
0,i
, m
0,r
, and m
0,t
= zeroth moments of the incident, reﬂected, and trans
mitted wave spectra, respectively. Note that the reﬂection and transmission coef
ﬁcients calculated by the conventional method, as the ratio of the reﬂected and
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 327
transmitted rms wave heights to the incident one, are equivalent to the frequency
averaged coeﬃcients, because the rms wave height is proportional to
√
m
0
.
The force spectrum, S
F
(f), is calculated by
S
F
(f) = T
F
(f)
2
S
η,i
(f) , (13.39)
where T
F
(f) is the frequencydependent transfer function between wave amplitude
and force amplitude, which can be computed by using the linear term in
Eq. (13.16) as
T
F
= 2ρg
1
cosh(kh)
∞
¸
m=0
A
m
µ
m
{sin(µ
m
h) −r
0
sin[µ
m
(h −d)]} . (13.40)
The zeromoment force, F
m0
, can then be determined as
F
m0
= 2
¸
∞
0
S
F
(f)df
1/2
. (13.41)
Note that a factor of 2 instead of 4 was used in order to represent the force amplitude.
13.3. PerforatedWall Caisson Breakwaters
13.3.1. Fullyperforatedwall caisson breakwater
Let us consider the perforatedwall caisson breakwater sketched in Fig. 13.3, in
which θ
1
= incident wave angle; and B = wave chamber width. The xaxis and
yaxis are taken to be normal and parallel, respectively, to the breakwater crest
line, and the water depth is assumed to be constant in the ydirection. The vertical
Fig. 13.3. A schematic diagram and coordinate system for calculation of wave reﬂection form a
perforatedwall caisson breakwater.
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
328 K.D. Suh
coordinate z is measured vertically upward from the still water line. In region 2
(−b ≤ x ≤ 0), the water depth h(x) is a varying function of x. For x ≤ −b (region 1)
and 0 ≤ x ≤ B (region 3), the water depth is constant and equal to h
1
and h
3
,
respectively. The rubble mound foundation is assumed to be impervious.
Assuming incompressible ﬂuid and irrotational ﬂow motion, the velocity
potential Φ(x, y, z, t) for the monochromatic wave propagating over the water depth
h(x) with the angular frequency ω and wave height H can be expressed as
Φ(x, y, z, t) = Re
−
igH
2ω
φ(x, y, z) exp(−iωt)
. (13.42)
Linearizing the freesurface boundary conditions, the following boundary value
problem for the potential φ(x, y, z) is obtained:
∇
2
φ +
∂
2
φ
∂z
2
= 0 , (13.43)
∂φ
∂z
−λφ = 0 at z = 0 , (13.44)
∂φ
∂z
+
∂φ
∂x
dh
dx
= 0 at z = −h(x) , (13.45)
where ∇represents the horizontal gradient operator, and λ = ω
2
/g is a wave number
in deepwater.
The Galerkineigenfunction method is used to formulate the problem. The
Galerkin approach assumes that the function φ(x, y, z) can be expanded in terms
of M + 1 depthdependent functions Z
m
(x, z):
φ(x, y, z) =
M
¸
m=0
ϕ
m
(x, y)Z
m
(x, z) . (13.46)
The functions Z
m
(x, z) are taken as
Z
m
(x, z) =
cos[µ
m
(z +h)]
cos(µ
m
h)
, (13.47)
so as to form a complete orthogonal set of eigenfunctions in the domain
[−h(x), 0]. The wave numbers µ
m
are the solution of the dispersion relation,
λ +µ
m
tan(µ
m
h) = 0. The function Z
0
(x, z) represents the free propagating wave
mode, while the functions Z
m
(x, z) (m ≥ 1) correspond to the nonpropagating
evanescent modes. The functions Z
m
(x, z) satisfy the free surface boundary con
dition in Eq. (13.44) and do not satisfy the bottom boundary condition in Eq.
(13.45) individually. However, the global set of orthogonal functions should satisfy
this condition. This is known as a tau method. In the tau method, a suﬃcient
number of functions ϕ
m
(x, y) in the approximated solution of Eq. (13.46) is chosen
to ensure exact satisfaction of the bottom boundary condition.
Turning back to the breakwater problem sketched in Fig. 13.3, the solution of
the boundary value problem given by Eqs. (13.43)–(13.45) can be constructed from
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 329
the particular solutions in each region of the ﬂuid domain:
φ
1
(x, y, z) = {exp[ik
1
(x +b) cos θ
1
] −exp[−ik
1
(x +b) cos θ
1
]}
× exp(iχy)
coshk
1
(z +h
1
)
coshk
1
h
1
+
¸
µ
1,m
R
m
exp[ν
1,m
(x +b)] exp(iχy)Z
1,m
(h
1
, z), (13.48)
φ
2
(x, y, z) =
¸
µ
2,m
˜ ϕ
m
(x) exp(iχy)Z
2,m
(h
2
, z) , (13.49)
φ
3
(x, y, z) =
¸
µ
3,m
T
m
exp(−ν
3,m
x) exp(iχy)Z
3,m
(h
3
, z) , (13.50)
where
ν
j,m
=
µ
2
j,m
+χ
2
, χ = k
j
sinθ
j
= k
1
sin θ
1
= constant (j = 1, 2, 3) . (13.51)
If m = 0 (propagating mode), the following relationships are obtained:
λ = k
j
tanh(k
j
h
j
), ν
j,0
= ±ik
j
cos θ
j
. (13.52)
For ν
j,0
, we take the negative sign for the reﬂected wave in region 1, while we need both
positive and negative signs for the waves inside the wave chamber (i.e., region 3).
The potential φ
j
(x, y, z) must satisfy the matching conditions which provide
continuity of pressure and horizontal velocity, normal to the vertical planes sepa
rating the ﬂuid regions and noﬂux condition at the wall on the downwave side of
the wave chamber, i.e.,
φ
1
= φ
2
,
∂φ
1
∂x
=
∂φ
2
∂x
(x = −b, −h
1
≤ z ≤ 0) , (13.53)
φ
3
= φ
2
+
i
G
∂φ
2
∂x
,
∂φ
2
∂x
=
∂φ
3
∂x
(x = 0, −h
3
≤ z ≤ 0) , (13.54)
∂φ
3
∂x
= 0 (x = B, −h
3
≤ z ≤ 0) . (13.55)
As shown in Eq. (13.46), the Galerkin solution consists of a free propagating
wave mode and nonpropagating evanescent modes. The evanescent wave modes
would be of importance in the region near the breakwater. However, since we are
interested in the reﬂected wave far from the breakwater, we focus on the solution
for the propagating mode. Massel
28
showed that in region 2, the function ˜ ϕ
0
(x)
satisﬁes the following ordinary diﬀerential equation:
d
2
˜ ϕ
0
dx
2
+D(x)
d˜ ϕ
0
dx
+E(x) ˜ ϕ
0
= 0 , (13.56)
where
D(x) =
k
τ +kh(1 −τ
2
)
¸
1 −3τ
2
+
2τ
τ + kh(1 −τ
2
)
dh
dx
, (13.57)
E(x) = k
2
¸
1 +
u
1
k
2
u
0
dh
dx
2
+
u
2
k
2
u
0
d
2
h
dx
2
¸
−χ
2
, (13.58)
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
330 K.D. Suh
where τ = tanh(kh), and u
0
, u
1
, and u
2
are given by
u
0
=
1
2k
tanh(kh)
¸
1 +
K
sinh K
, (13.59)
u
1
=
sech
2
(kh)
4(K + sinh K)
(sinh K −K coshK) , (13.60)
u
2
=
ksech
2
(kh)
12(K + sinh K)
3
[K
4
+ 4K
3
sinh K −9 sinhK sinh(2K)
+3K(K + 2 sinhK)(cosh
2
K −2 coshK + 3)] , (13.61)
where the abbreviation K = 2kh was used.
In order to solve Eq. (13.56) in region 2, we need the boundary conditions at
x = −b and x = 0. Suh and Park
23
derived these boundary conditions as
d˜ ϕ
0
(−b)
dx
= i[2 − ˜ ϕ
0
(−b)]k
1
cos θ
1
, (13.62)
˜ ϕ
0
(0) =
¸
1
ν
3,0
exp(−ν
3,0
B) + exp(ν
3,0
B)
exp(−ν
3,0
B) −exp(ν
3,0
B)
−
i
G
d˜ ϕ
0
(0)
dx
. (13.63)
The diﬀerential equation (13.56) with the preceding two boundary conditions can
be solved using the ﬁnitediﬀerence method. Using the forwarddiﬀerencing for
d˜ ϕ
0
(−b)/dx, backwarddiﬀerencing for d˜ ϕ
0
(0)/dx, and centraldiﬀerencing for the
derivatives in Eq. (13.56), the boundary value problem of Eqs. (13.56), (13.62), and
(13.63) can be approximated by a system of linear equations, AY = B, where A
is a tridiagonalbandtype matrix, Y is a column vector, and B is also a column
vector. The subroutines given in the book of Press et al.
29
can be used to solve
this matrix equation. In particular, we are interested in ˜ ϕ
0
(−b), from which the
reﬂection coeﬃcient C
r
is calculated as follows:
C
r
= ˜ ϕ
0
(−b) −1 . (13.64)
As stated in the introduction, the advantage of this model is that it can be used
for a perforatedwall caisson mounted on a rubble mound foundation when waves
are incident at an arbitrary angle. In order to examine the eﬀect of rubble mound
foundation upon wave reﬂection, Suh et al.
30
calculated the reﬂection coeﬃcients
by changing the height and slope of the rubble mound. Total water depth in front
of the breakwater h
1
= 10 m, wave period T = 8 s, incident wave height H = 2 m,
thickness of the perforated wall l = 0.6 m, porosity of the wall r = 0.3, and wave
chamber width B = 10 m were used. Figure 13.4 shows the ratio of the reﬂection
coeﬃcient of the breakwater with a mound to that of the breakwater without a
mound (i.e., lying on a ﬂat sea bottom). When the mound height is approximately
less than 0.3 times the total water depth, the reﬂection coeﬃcient of the breakwater
with a mound is less than that of the breakwater without a mound. For the mound
height greater than this value, however, the wave reﬂection from the mound slope
becomes signiﬁcant so that the reﬂection coeﬃcient becomes greater than that of the
breakwater without a mound. In addition, the reﬂection coeﬃcient increases with
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 331
0.0 0.1 0.2 0.3 0.4 0.5
(h1h3)/h1
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
M
o
u
n
d
S
l
o
p
e
Fig. 13.4. Contours of the ratio of the reﬂection coeﬃcient of the breakwater with a mound to
that of the breakwater without a mound as a function of mound height and slope.
the height and slope of the mound as expected. Within the range of the practical
design situations, however, the rubble mound eﬀect on wave reﬂection is not so large.
However, it should be noted that the eﬀect of permeability of the rubble mound was
not take into account because the rubble mound was assumed to be impervious.
Suh et al.
30
also examined the eﬀect of angle of incidence upon wave reﬂection.
The reﬂection coeﬃcients were calculated by changing the wave angle from 0
◦
to
80
◦
and Bcos θ/L from 0.1 to 0.5. The caisson was situated in a constant depth
without a mound so that h
1
= h
2
= h
3
= 30 cm. Wave height H = 3 cm, wave
period T = 1.554 s, porosity of the wall r = 0.25 were used, while the perforated
wall thickness was taken to be zero (i.e., the inertia eﬀect of the perforated wall was
neglected). The calculated results are shown in Fig. 13.5. For normal incidence of
waves, the wave reﬂection is minimum at Bcos θ/L = 0.25 and increases as it gets
away from this value. The same trend is kept for oblique incidence of waves, but
the solution exhibits singular behavior (i.e., the reﬂection coeﬃcient shows a sharp
peak) at B/L = 0.5, 1.0, 1.5, and so on, as shown in Fig. 13.5. In the practical design
of a perforatedwall caisson, the value of B/L is usually less than 0.5. Thus, when
the waves are obliquely incident to the breakwater, to minimize the wave reﬂection,
the value of Bcos θ/L should be around 0.25, but the condition B/L
∼
= 0.5 should
be avoided.
13.3.2. Application to partiallyperforatedwall caisson breakwater
A conventional perforatedwall caisson consists of a front wave chamber and a back
wall as shown in Fig. 13.6(a), and the water depth inside the wave chamber is
the same as that on the rubble foundation. The weight of the caisson is less than
that of a vertical solid caisson with the same width, and moreover, most of this
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
332 K.D. Suh
0 20 40 60 80
Wave Angle (degree)
0.1
0.2
0.3
0.4
0.5
B
c
o
s
θ
/
L
B/L = 0.5
1.0
1.5
Fig. 13.5. Contours of the reﬂection coeﬃcient as a function of wave incident angle and B cos θ/L;
the lines of B/L = 0.5, 1.0, and 1.5 are also included.
weight is concentrated on the rear side of the caisson. Therefore, diﬃculties are
sometimes met in the design of a perforatedwall caisson to satisfy the design criteria
against sliding and overturning. In addition, particularly in the case where the
bearing capacity of the seabed is not large enough, the excessive weight on the rear
side of the caisson may have an adverse eﬀect. In order to solve these problems,
a partiallyperforatedwall caisson as shown in Fig. 13.6(b) is often used, which
provides an additional weight to the front side of the caisson. In this case, however,
other hydraulic characteristics of the caisson such as wave reﬂection and overtopping
may become worse than a fullyperforatedwall caisson.
The mathematical model described in Sec. 13.3.1 assumes that the water depth
inside the wave chamber is the same as that on the mound berm as in a fully
perforatedwall caisson breakwater shown in Fig. 13.6(a). However, for a partially
perforatedwall caisson [see Fig. 13.6(b)], these water depths are diﬀerent from each
other, having depth discontinuity at the location of the perforated wall. In order
to apply the model to the case of a partiallyperforatedwall caisson, it can be
assumed that the lower part of the front face of the caisson (below the perforated
wall) is not vertical but has a very steep slope. As can be seen in Eq. (13.58), the
model includes the terms proportional to the square of the bottom slope and to the
bottom curvature, so that it can be applied over a bed having substantial variation
of water depth. In order to examine the eﬀect of the slope of the lower part of the
caisson (which is inﬁnity in reality), Suh et al.
24
calculated the reﬂection coeﬃcient
by changing the slope from 0.1 to 10. They found that the reﬂection coeﬃcient
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 333
(a)
(b)
Fig. 13.6. Bird’seye views of (a) a fullyperforatedwall caisson breakwater and (b) a partially
perforatedwall caisson breakwater.
virtually does not change for slopes greater than 2.0, and they used the slope of 4.0
in their comparison between model prediction and experimental results.
13.3.3. Extension to irregular waves
As for irregular waves, as explained in Sec. 13.2.3, the reﬂection coeﬃcient is cal
culated diﬀerently for each frequency component. The wave period is determined
according to the frequency of the component wave, while the rms wave height is
used for all the component waves to compute the energy dissipation at the perfo
rated wall. The spectral density of the reﬂected waves is then calculated for a par
ticular frequency component as in Eq. (13.35), and the frequencyaveraged reﬂection
coeﬃcient is calculated as in Eq. (13.37).
13.4. Calculation of Permeability Parameter
Equations (13.6) and (13.23) state that the ﬂuid velocity normal to the perfo
rated wall is proportional to the pressure diﬀerence across the wall with a complex
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
334 K.D. Suh
constant of proportionality. To elucidate the physical meaning of the permeability
parameter G, however, it is more convenient to express the equations in the form
of Eq. (13.54). Then, the real part of i/G is associated with the phase diﬀerence
between the velocity and the pressure due to inertial eﬀects, and the imaginary part
of i/G corresponds to the resistance of the wall. There are several approaches to
calculate the permeability parameter G. Mei et al.
31
expressed the resistance and
the inertial eﬀects in terms of a headloss coeﬃcient and an eﬀective oriﬁce length,
respectively. On the other hand, Sollitt and Cross
32
expressed them in terms of a
friction coeﬃcient and an added mass coeﬃcient, respectively.
13.4.1. Mei et al.’s method
The permeability parameter based on Mei et al.
31
can be expressed as
G =
1
β
ω
−i
, (13.65)
where β = energy dissipation coeﬃcient derived by linearizing the nonlinear con
vective acceleration term in the equation of motion; and = eﬀective oriﬁce length,
which is the length of the jet ﬂowing through the perforated wall.
The linearized dissipation coeﬃcient β for a pile breakwater is given by Kim
33
as
β =
8α
9π
H
w
ω
1
(R + 2)
2
+P
2
5 + cosh(2kh)
2kh + sinh(2kh)
, (13.66)
where H
w
= incident wave height at the perforated wall; P = k; R = βk/ω; and
α = headloss coeﬃcient. The preceding equation was derived for a pile breakwater.
However, it could be used for a curtainwallpile breakwater as well, because the
mechanism of energy dissipation between piles must be similar for these two break
waters. A pile breakwater has energy dissipation associated with the free surface,
but it may be small when compared with the energy dissipation between piles. On
the other hand, for a perforatedwall caisson (see Fig. 13.3), Fugazza and Natale
22
gave β as
β =
8α
9π
H
w
ω
W
W
2
(R + 1)
2
+F
2
5 + cosh(2k
3
h
3
)
2k
3
h
3
+ sinh(2k
3
h
3
)
, (13.67)
where W = tan(k
3
B), R = βk
3
/ω, F = 1 − PW, and P = k
3
. Rearrangement of
Eq. (13.66) or (13.67) gives a quartic polynomial of β, which can be solved by the
eigenvalue method (e.g., Press et al.
29
).
Suh et al.
34
showed that the eﬀective oriﬁce length is related to the blockage
coeﬃcient C by
= 2C . (13.68)
For rectangular piles, Flagg and Newman
35
proposed the blockage coeﬃcient as
C =
b
2
1
r
0
−1
+
2A
π
¸
1 −log(4r
0
) +
1
3
r
2
0
+
281
180
r
4
0
. (13.69)
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 335
The headloss coeﬃcient α could be given by the plate oriﬁce formula
36
:
α =
1
r
0
cos θC
c
−1
2
, (13.70)
where C
c
= empirical contraction coeﬃcient, for which Mei et al.
31
suggested using
the formula:
C
c
= 0.6 + 0.4r
2
0
, (13.71)
and r
0
cos θ denotes the eﬀective ratio of the opening of the perforated wall taking
into account the oblique incidence of the waves to the wall. Very recently, on the
other hand, Yoon et al.
37
proposed the headloss coeﬃcient taking into account the
eﬀect of wall thickness as
α = (0.6 + 0.4e
−2.5(b/2a)
)
1
r
0
C
c
−1
2
, (13.72)
for normally incident waves.
For circular piles, Kakuno and Oda
38
proposed the blockage coeﬃcient as
C =
π
4
A(1 −r
0
)
2
1
1 −ξ
; ξ =
π
2
12
(1 −r
0
)
2
. (13.73)
Park et al.
11
proposed a formula for the headloss coeﬃcient for circular piles as
α =
1
rC
c
−1
2
, (13.74)
where C
c
is the contraction coeﬃcient given by Eq. (13.71), and the ad hoc porosity
r is given by
1
r
2
=
1
D
D/2
−D/2
dx
r(x)
2
, (13.75)
with the spatially varying porosity
r(x) = 1 −
D
2
2
−x
2
D
2
+a
, (13.76)
where D = diameter of piles.
In the case of a perforatedwall caisson mounted on a rubble foundation as in
Fig. 13.3, the incident wave height at the perforated wall H
w
is a priori unknown.
In the case where the caisson does not exist and the water depth is constant as h
3
for x ≥ 0, Massel
28
has shown that the transmitting boundary condition at x = 0
is given by
i ˜ ϕ(0)k
3
cos θ
3
=
d˜ ϕ(0)
dx
. (13.77)
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
336 K.D. Suh
The governing equation (13.56) and the upwave boundary condition in Eq. (13.62)
do not change. After solving this problem, the transmission coeﬃcient C
t
is given by
C
t
= Re{ ˜ ϕ(0)}, from which H
w
is calculated as C
t
times the incident wave height
on the ﬂat bottom.
The above method to compute the permeability parameter is advantageous com
pared with other methods in that all the related parameters are known, i.e., incident
wave height and period and the geometrical parameters of the barrier. However,
Suh et al.
12
showed that this method gives a wrong result for very long waves. As
can be seen in Eq. (13.66), as kh goes to zero, β goes to inﬁnity, indicating complete
dissipation of the long waves. This leads to zero transmission for long waves, which
is deﬁnitely wrong because in the limit of long waves the breakwater is invisible to
the waves so that complete transmission of the waves should occur.
This method can be used for irregular waves as well. Using the regular wave
model, the reﬂection and transmission coeﬃcients are calculated diﬀerently for each
frequency component. In the computation of the energy dissipation coeﬃcient β in
Eq. (13.66), however, the rms wave height should be used instead of the height of
the component wave, because the ﬂow separation due to irregular waves and the
resulting energy dissipation are induced not by the individual component waves but
by the superposition of the component waves.
13.4.2. Sollitt and Cross’s method
The permeability parameter based on Sollitt and Cross
32
can be expressed as
G =
r
0
b (f
c
−is)
, (13.78)
where f
c
= friction coeﬃcient, and s = inertia coeﬃcient given by
s = 1 +C
m
1 −r
0
r
0
, (13.79)
where C
m
= added mass coeﬃcient.
In the method of Sollitt and Cross,
32
the values of friction and added mass coef
ﬁcients are not known a priori, and they are estimated on the basis of a best ﬁt
between measurement and prediction. Many researchers (e.g., Sollitt and Cross
32
;
Losada et al.
39
; Yu
40
; Isaacson et al.
4,41
; Zhu and Chwang
42
; Hossain et al.
43
)
suggested to use the value of C
m
= 0, i.e., s = 1 by comparing their numerical
predictions of reﬂection and transmission coeﬃcients of perforated structures with
experimental results. As for the friction coeﬃcient, recently, Li et al.
44
proposed the
following empirical equation for estimating f
c
in terms of b/h, along with the use
of C
m
= 0:
f
c
= −3338.7
b
h
2
+ 82.769
b
h
+ 8.711 for 0.0094 ≤
b
h
≤ 0.05 . (13.80)
They also recommended using f
c
= 2.0 if b/h ≥ 0.1.
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 337
13.4.3. Inertia eﬀect on wave reﬂection
from perforatedwall caisson
It is well known that wave reﬂection from a perforatedwall caisson breakwater
depends on the width of the wave chamber relative to the wavelength. For regular
waves, neglecting the inﬂuence of the inertia term, Fugazza and Natale
22
showed
analytically that the resonance inside the wave chamber is important so that the
reﬂection is minimal when B/L = (2n + 1)/4 (n = 0, 1, 2, . . .) in which L is the
wavelength. However, the resonant mode of practical interest is the fundamental
mode (n = 0), i.e., B/L = 0.25 because of the width limit of the breakwater. This
is also shown in Fig. 13.5. In reality, however, the reﬂection coeﬃcient becomes
minimal at the value of B/L somewhat smaller than 0.25 because of the inﬂuence
of the inertia term.
24,27
In front of a perforatedwall caisson breakwater, a partial
standing wave is formed due to the wave reﬂection from the breakwater. If there were
no perforated wall, the node would occur at a distance of about L/4 from the back
wall of the wave chamber, and hence the largest energy loss may occur at this point
because the horizontal velocity becomes a maximum at a node. But, in reality, there
exists inertia resistance at the perforated wall, which interrupts the smooth pass of
the wave, thus creating pressure (or free surface) diﬀerence between the front and
back of the perforated wall. Consequently, the location of the maximum horizontal
velocity will move toward the breakwater, and the point where the maximum energy
loss is gained becomes smaller than L/4. Thus, the minimum reﬂection occurs at
a value of B/L smaller than 0.25. In the methods of Mei et al.
31
and Sollitt and
Cross,
32
the eﬀective oriﬁce length and inertia coeﬃcient s, respectively, represents
the inertia eﬀect.
Acknowledgments
This chapter was prepared while the author was supported by the Korea Sea Grant
Program and the Project for Development of ReliabilityBased Design Method for
Port and Harbor Structures of Korea Ministry of Marine Aﬀairs and Fisheries.
References
1. R. L. Wiegel, J. Waterways Harbors Coastal Eng. Div., Am. Soc. Civ. Eng. 86, 1
(1960).
2. D. L. Kriebel and C. A. Bollmann, Proc. 25th Int. Conf. Coastal Eng., ASCE (1997),
p. 2470.
3. D. L. Kriebel, C. Sollitt and W. Gerken, Proc. 26th Int. Conf. Coastal Eng., ASCE
(1999), p. 2069.
4. M. Isaacson, S. Premasiri and G. Yang, J. Waterways, Port, Coastal, Ocean Eng. 124,
118 (1998).
5. T. Hayashi, M. Hattori, T. Kano and M. Shirai, Proc. 10th Int. Conf. Coastal Eng.,
ASCE (1966), p. 873.
6. H. Kojima, M. Utsunomiya, T. Ijima, A. Yoshida and T. Kihara, Proc. 35th Jpn Conf.
Coastal Eng., ASCE (1988), p. 542.
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
338 K.D. Suh
7. T. Uda, A. Omata and T. Kawamura, Rep. Public Works Res. Inst., 2891, Japan
Ministry of Construction (1990) (in Japanese).
8. S. Kakuno and P. L.F. Liu, J. Waterways, Port, Coastal, Ocean Eng. 119, 302 (1993).
9. K. Hagiwara, Proc. 19th Int. Conf. Coastal Eng., ASCE (1985), p. 2810.
10. G. S. Bennett, P. McIver and J. V. Smallman, Coastal Eng. 18, 231 (1992).
11. W. S. Park, B. H. Kim, K. D. Suh and K. S. Lee, Coastal Eng. J. 42, 253 (2000).
12. K.D. Suh, S. Shin and D. T. Cox, J. Waterways, Port, Coastal, Ocean Eng. 132, 83
(2006).
13. K.D. Suh and C.H. Ji, Proc. 30th Int. Conf. Coastal Eng., ASCE (2006), p. 4303.
14. S. Takahashi and K. Shimosako, Proc. HydroPort ’94, Port and Harbour Res. Inst.,
Yokosuka, Japan (1994), p. 747.
15. S. Takahashi, K. Tanimoto and K. Shimosako, Proc. HydroPort ’94, Port and Harbour
Res. Inst., Yokosuka, Japan (1994), p. 489.
16. G. E. Jarlan, The Dock and Harbour Authority, Vol. XII(486) (1961), p. 394.
17. M. Marks and G. E. Jarlan, Proc. 11th Int. Conf. Coastal Eng., ASCE (1968), p. 1121.
18. F. L. Terret, J. D. C. Osorio and G. H. Lean, Proc. 11th Int. Conf. Coastal Eng.,
ASCE (1968), p. 1104.
19. K. Tanimoto, S. Haranaka, S. Takahashi, K. Komatsu, M. Todoroki and M. Osato,
Technical Report, Port and Harbour Res. Inst., Min. of Transport, Japan, No. 246
(1976) (in Japanese).
20. H. Kondo, Proc. Coastal Structures ’79, ASCE (1979), p. 962.
21. S. Kakuno, K. Oda and P. L.F. Liu, Proc. 23rd Int. Conf. Coastal Eng., ASCE (1992),
p. 1258.
22. M. Fugazza and L. Natale, J. Waterways, Port, Coastal, Ocean Eng. 118, 1 (1992).
23. K. D. Suh and W. S. Park, Coastal Eng. 26, 117 (1995).
24. K.D. Suh, J. K. Park and W. S. Park, Ocean Eng. 33, 264 (2006).
25. D. L. Kriebel, Proc. Coastal Structures ’99, Balkema, Rotterdam (2000), p. 525.
26. R. G. Dean and R. A. Dalrymple, Water Wave Mechanics for Engineers and Scientists
(World Scientiﬁc, River Edge, New Jersey, 1991).
27. Y. Goda, Random Seas and Design of Maritime Structures (World Scientiﬁc,
Singapore, 2000).
28. S. R. Massel, Coastal Eng. 19, 97 (1993).
29. W. H. Press, S. A. Teukolsky, W. T. Vetterling and B. P. Flannery, Numerical Recipes
in FORTRAN: The Art of Scientiﬁc Computing, 2nd edn. (Cambridge University
Press, Cambridge, 1992).
30. K. D. Suh, W. S. Park and K. D. Yum, Proc. Int. Conf. Coastal and Port Eng.
Developing Countries (1995), p. 2311.
31. C. C. Mei, P. L.F. Liu and A. T. Ippen, J. Waterways, Harbors Coastal Eng. Div.,
Vol. WW3, ASCE (1974), p. 217.
32. C. K. Sollitt and R. H. Cross, Proc. 13th Int. Conf. Coastal Eng., ASCE (1972),
p. 1827.
33. B. H. Kim, Interactions of waves, seabed and structures, PhD. dissertation, Seoul
National University, Seoul, Korea (1998).
34. K. D. Suh, S. Y. Son, J. I. Lee and T. H. Lee, Proc. 28th Int. Conf. Coastal Eng.,
ASCE (2002), p. 1709.
35. C. N. Flagg and J. N. Newman, J. Ship Res. 15, 257 (1971).
36. C. C. Mei, The Applied Dynamics of Ocean Surface Waves (Wiley, New York, 1983).
37. S. B. Yoon, J. I. Lee, D. H. Nam and S. H. Kim, J. Korean Soc. Coastal Ocean Engrs.
18, 321 (2006) (in Korean).
38. S. Kakuno and K. Oda, J. Jpn. Soc. Civil Engrs. 369, 213 (1986) (in Japanese).
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
Wave Interaction with Breakwaters Including Perforated Walls 339
39. I. J. Losada, M. A. Losada and A. Baquerize, Appl. Ocean Res. 15, 207 (1993).
40. X. Yu, J. Waterway, Port, Coastal, Ocean Eng. 121, 275 (1995).
41. M. Isaacson, J. Baldwin, S. Premasiri and G. Yang, Appl. Ocean Res. 21, 81 (1999).
42. S. Zhu and A. T. Chwang, J. Eng. Mech. 127, 326 (2001).
43. A. Hossain, W. Kioka and T. Kitano, Coastal Eng. J. 43, 83 (2001).
44. Y. Li, Y. Liu and B. Teng, Coastal Eng. J. 48, 309 (2006).
July 31, 2009 8:18 9.75in x 6.5in b684ch13 FA
This page intentionally left blank This page intentionally left blank
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Chapter 14
Prediction of Overtopping
Jentsje van der Meer
Van der Meer Consulting
P. O. Box 423, 8440 AK Heerenveen, The Netherlands
jm@vandermeerconsulting.nl
Tim Pullen
HR Wallingford
Howbery Park, Wallingford, Oxon, OX10 8BA, UK
tap@hrwallingford.co.uk
William Allsop
HR Wallingford
Howbery Park, Wallingford, Oxon, OX10 8BA, UK
w.allsop@hrwallingford.co.uk
Tom Bruce
School of Engineering, University of Edinburgh
King’s Buildings, Edinburgh, EH9 3JL, UK
tom.bruce@ed.ac.uk
Holger Sch¨ uttrumpf
RWTHAachen University
Institute of Hydraulic Engineering and Water Resources Management
MiesvanderRoheStr. 1, 52065 Aachen, Germany
schuettrumpf@iww.rwthaachen.de
Andreas Kortenhaus
LeichtweissInstitute for Hydraulics, Technical University of Braunschweig
Beethovenstr, 51a, 38106 Braunschweig, Germany
kortenhaus@tubs.de
341
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
342 J. W. van der Meer et al.
This chapter describes the processes of wave overtopping at sea defense and
related coastal or shoreline structures. It introduces a range of methods to
calculate mean overtopping discharges, individual and maximum overtopping
volumes, and the proportion of waves overtopping a seawall. It describes the
principal hazards from wave overtopping and will help engineers by suggesting
limiting tolerable discharges for frequent, design, and extreme wave conditions.
This chapter is supported by more detailed material in Chaps. 15 and 16 which
focus on the methods to predict overtopping for rubble mound structures (with
partly sloping embankments), and on vertical structures and battered walls. All
of these three chapters have been based closely on the new EurOtop Overtopping
Manual.
5
14.1. Introduction
14.1.1. Wave overtopping
Wave overtopping has always been of principal concern for coastal structures con
structed to defend against ﬂooding: often termed sea defenses. Similar structures
may also be used to provide protection against coastal erosion: sometimes termed
coast protection. Other structures may be built to protect areas of water for ship
navigation or mooring within ports, harbours, or marinas: often formed by break
waters or moles. Within harbours, or along shorelines, reclaimed areas must be
defended against both erosion and ﬂooding. Some structures may be detached from
the shoreline, often termed oﬀshore, nearshore, or detached, but most structures
used for sea defense or similar function form a part of the shoreline.
Sloping dikes have been widely used for sea defenses along the coasts of the
Netherlands, Denmark, Germany, and the United Kingdom. Dikes or embankment
seawalls are also used to defend lowlying areas in the Far East, including China,
Korea, and Vietnam. Historically, dikes or embankment seawalls were built along
many North Sea coastlines, sometimes subsuming an original sand dune line, pro
tecting the land behind from ﬂooding, and sometimes providing additional amenity
value. Similar structures have been formed by clay materials or even from a veg
etated shingle ridge, in both instances allowing the side slopes to be steeper. All
such embankments need some degree of protection against direct wave erosion, often
using a revetment facing on the seaward side (Fig. 14.1). Revetment facing may
take many forms, but may commonly include closelyﬁtted concrete blockwork, cast
in situ concrete slabs, or asphaltic materials. Embankment or dike structures are
generally most common along rural frontages.
A second type of coastal structure consists of a mound or layers of quarried rock
ﬁll, protected by rock or concrete armour units (Fig. 14.2). The outer armour layer
is designed to resist wave action without signiﬁcant displacement of armour units.
Underlayers of quarry or crushed rock support the armour and separate it from ﬁner
material in the embankment or mound. These porous and sloping layers dissipate a
proportion of the incident wave energy in breaking and friction. Simpliﬁed forms of
rubble mounds may be used for rubble seawalls or as protection to vertical walls or
revetments. Rubble mound revetments may also be used to protect embankments
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 343
Fig. 14.1. Wave overtopping on a revetment seawall.
Fig. 14.2. Wave overtopping on a rubble mound breakwater.
formed from relict sand dunes or shingle ridges. Rubble mound structures tend to
be more common in areas where harder rock is available.
Along urban frontages, especially close to ports, erosion or ﬂooding defense
structures may include vertical (or battered/steep) walls (Fig. 14.3). Such walls
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
344 J. W. van der Meer et al.
Fig. 14.3. Wave overtopping on a battered/vertical seawall.
may be composed of stone or concrete blocks, mass concrete, or sheet steel piles.
Typical vertical seawall structures may also act as retaining walls to material behind.
Shaped and recurved wave return walls may be formed as walls in their own right,
or smaller versions may be included in sloping structures. Some coastal structures
are relatively impermeable to wave action. These include seawalls formed from
blockwork or mass concrete, with vertical, nearvertical, or steeply sloping faces.
Such structures may be liable to intense local wave impact pressures, may overtop
suddenly and severely, and will reﬂect much of the incident wave energy. Reﬂected
waves cause additional wave disturbance and/or may initiate or accelerate local bed
scour.
It is worth noting that developments along waterfronts are highly valued with
purchase or rental prices substantially above those for properties not on the water
front. Yet, direct (or indirect) eﬀects of wave overtopping have the potential to
generate signiﬁcant hazards to such developments and their users. Residential and
commercial properties along a waterfront will often be used by people who may
be unaware of the possibility, of the severity, or of the eﬀects of wave overtopping
in storm conditions. Regulatory authorities may therefore wish to impose onerous
ﬂood defense requirements on new developments. For instance, protection against
ﬂooding (including wave overtopping) for any new developments in the United
Kingdom is now required to be 0.5% annual probability, equivalent to 1:200 year
return. Exposure to overtopping of many coastal sites will, however, be inﬂuenced
by climate change, probably increasing wave heights and periods as well as sea
level rise.
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 345
14.1.2. Predicting wave overtopping
A number of diﬀerent methods may be available to predict overtopping of particular
structures (usually simpliﬁed sections) under given wave conditions and water levels.
Each method will have strengths or weaknesses in diﬀerent circumstances. In theory,
an analytical method can be used to relate the driving process (waves) and the
structure to the response through equations based directly on the knowledge of
the physics of the process. It is, however, extremely rare for the structure, the
waves, and the overtopping process to all be so simple and wellcontrolled that an
analytical method on its own can give reliable predictions. Analytical methods are
not therefore discussed further in this chapter.
The primary prediction methods are therefore based on empirical methods
that relate the overtopping response to the main wave and structure parameters.
These are by far the most commonly used methods to predict overtopping. Two
other methods have been derived during the CLASH European project
3
based on
the use of measured overtopping from model tests and ﬁeld measurements. The
ﬁrst of these techniques uses the CLASH database of structures, waves, and over
topping discharges, with each test described by 31 parameters. Using the database
is, however, potentially complicated, requiring some familiarity with these type of
data. A simpler approach, and much more rapid, is to use the Neural Network tool
that has been trained using the test results in the database. The Neural Network
tool can be run automatically on a computer as a standalone device, or embedded
within other simulation methods.
For situations for which empirical test data do not already exist, or where the
methods above do not give reliable enough results, then two alternative methods
may be used, but both are more complicated than the methods above. A range of
numerical models can be used to simulate the process of overtopping. All such
models involve some simpliﬁcation of the overtopping process and are therefore
limited to particular types of structure or types of wave exposure. They may,
however, run sequences of waves giving overtopping (or not) on a wavebywave
basis. Generally, numerical models require more skill and familiarity to run success
fully. They will not be described in this chapter.
The ﬁnal method to be mentioned is physical modeling in which a scale model
is tested with correctly scaled wave conditions. Typically, such models may be built
to a geometric scale in the range 1:10 to 1:60. Waves will be generated as random
wave trains each conforming to a particular energy spectrum. The model may rep
resent a structure cross section in a 2D model tested in a wave ﬂume. Structures
with more complex plan shapes, junctions, transitions, etc., may be tested in a 3D
model in a wave basin. Physical models can be used to measure many diﬀerent
aspects of overtopping such as wavebywave volumes, overtopping velocities and
depths, as well as other responses.
14.1.3. Performance requirements
Most sea defense structures are constructed primarily to limit overtopping volumes
that might cause ﬂooding. For defenses that protect people living, working, or
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
346 J. W. van der Meer et al.
enjoying themselves, designers and owners of these defenses must, however, also
deal with potential direct hazards from overtopping. This requires that the level of
hazard and its probability of occurrence be assessed, allowing appropriate action
plans to be devised to ameliorate risks arising from overtopping. Section 14.6 deals
with tolerable wave overtopping.
14.2. Empirical Models, Including Comparison of Structures
14.2.1. Mean overtopping discharge
Empirical methods use a simpliﬁed representation of the physics of the process pre
sented in (usually dimensionless) equations to relate the main response (overtopping
discharge, etc.) to key wave and structure parameters. The form and coeﬃcients of
the equations are adjusted to reproduce results from physical model (or ﬁeld) mea
surements of waves and overtopping. Empirical equations may be solved explicitly,
or may occasionally require iterative methods to solve. Historically, some empirical
methods have been presented graphically, although this is now very rare.
The mean overtopping discharge, q, is the main parameter in the overtopping
process. It is, of course, not the only measure of overtopping, but it is relatively easy
to measure in a laboratory wave ﬂume or basin (or even in the ﬁeld), and most other
parameters are related in some way to this overtopping discharge. The overtopping
discharge is generally calculated in m
3
/s per m width, but in practical applications
it may be quoted as liter/s per m width. Although it is given as a discharge, it is
usually very far from a steady discharge as the actual processes of wave overtopping
are much more dynamic. For most defenses, only large waves will reach the crest
of the structure and will overtop, but they may do so with a lot of water in a few
seconds. The individual volumes in wavebywave overtopping are more diﬃcult to
measure in a laboratory than the mean discharge; so data on wavebywave volumes
are much rarer.
As mean overtopping discharges are relatively easy to measure, many physical
model tests have been performed all over the world, both for idealized structures
and real applications or designs. The European CLASH project
3
collected a large
database worldwide with more than 10,000 wave overtopping test results on all kinds
of structures (see Sec. 14.5). Some series of tests have been used to develop empirical
methods for the prediction of overtopping. Such empirical methods or formulae are,
however, only directly applicable to idealized structures, like smooth slopes (dikes,
sloping seawalls), simple rubble mound structures or vertical structures (caissons)
or walls, and may require extrapolation when applied to many existing structures.
14.2.2. Comparing overtopping performance
Chapters 15 and 16 will describe overtopping formulae for diﬀerent kinds of struc
tures, based on the EurOtop Overtopping Manual.
5
In this section, an overall view
is given to compare the performance of diﬀerent structure types and to give insight
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 347
into how wave overtopping behaves for diﬀerent structures. Those structures con
sidered here are: smooth sloping structures (dikes, seawalls); rubble mound struc
tures (breakwaters, rock armored slopes); and vertical structures (caissons, sheet
pile walls).
The principal prediction formula for many types of wave overtopping is
q
gH
3
m0
= a exp(−bR
c
/H
m0
). (14.1)
It is an exponential function with the dimensionless overtopping discharge
q/(gH
3
m0
)
1/2
and the relative crest freeboard R
c
/H
m0
. This type of equation shows
a straight line on a loglinear graph, which makes it easy to compare formulae for
diﬀerent structures. Speciﬁc equations are given in Chaps. 15 and 16.
Two equations are considered for pulsating waves on a vertical structure. Allsop
et al.
1
consider relatively shallow water and Franco et al.
6
more deep water
(caissons). Vertical structures in shallow water, and often with a sloping foreshore in
front, may become subject to impulsive forces, i.e., high impacts and water splashing
high up into the air. Speciﬁc formulae have been developed for these kinds of
situations.
For easy comparison of diﬀerent structures, like smooth and rubble mound
sloping structures and vertical structures for pulsating and impulsive waves, some
simpliﬁcations will be assumed.
In order to simplify the smooth structure, no berm is considered (γ
b
= 1), only a
normal wave attack is considered (γ
β
= 1), and the sloping seawall does not feature
any wavewall on top (γ
v
= 1). As the slope is smooth and impermeable, γ
f
= 1.
This limits the structure to a smooth and straight slope with normal wave attack.
The slope angles considered for smooth slopes are cotα = 1−8, which means from
very steep to very gentle. If relevant, a wave steepness of s
m−1,0
= 0.04 (steep storm
waves) and 0.01 (long waves due to swell or wave breaking) will be considered.
Detailed deﬁnitions are given in Chap. 15.
The same equation as for smooth slopes is applicable for rubble slopes, but now
with a roughness factor of γ
f
= 0.5, simulating a rock armoured structure. Rubble
mound structures are often steep, but rock armoured slopes may also be gentle.
Therefore, slope angles with cotα = 1.5 and 4.0 are considered.
For vertical structures under pulsating waves, both formulae of Allsop et al.
1
and Franco et al.
6
will be compared, together with the formula for impulsive waves.
Impulsive waves can only be reached with a relatively steep foreshore in front of
the vertical wall. For comparison, values of the breaker ratio (wave height/water
depth) of H
m0
/h
s
= 0.5, 0.7, and 0.9 will be used. These will be discussed further
in Chap. 16.
Overtopping on smooth slopes can be compared with rubble mound slopes and
with vertical structures under pulsating or impulsive conditions. First, the tradi
tional graph is given in Fig. 14.4 with the relative freeboard R
c
/H
m0
versus the
logarithmic dimensionless overtopping q/(gH
3
m0
)
1/2
.
In most cases the steep smooth slope gives the largest overtopping. Steep means
cotα < 2, but also a little gentler if long waves (less steepness) are considered. Under
these conditions, waves surge up the steep slope. For gentler slopes, waves break as
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
348 J. W. van der Meer et al.
1.E06
1.E05
1.E04
1.E03
1.E02
1.E01
1.E+00
0 0.5 1 1.5 2 2.5 3 3.5 4
Relative freeboard R
c
/H
m0
D
i
m
e
n
s
i
o
n
l
e
s
s
o
v
e
r
t
o
p
p
i
n
g
q
/
(
g
H
m
0
3
)
0
.
5
steep smooth slopes, cota<2
gentle smooth slope, cota=4, so=0.04
steep rubble mound slope, gf =0.5
gentle rubble mound slope, gf=0.5
vertical structure Allsop (1995)
vertical structure Franco et al. (1994)
impulsive vertical Hm0/hs=0.5
impulsive vertical Hm0/hs=0.7
impulsive vertical Hm0/hs=0.9
Fig. 14.4. Comparison of wave overtopping formulae for various structure types.
plunging waves and this reduces wave overtopping. The gentle slope with cotα = 4
gives much lower overtopping than the steep smooth slopes. Both slope angle and
wave period have inﬂuence on overtopping for gentle slopes.
The high roughness and permeability of a rubble mound can reduce over
topping substantially (Fig. 14.4). A roughness factor of γ
f
= 0.5 was used here
although γ
f
= 0.4 (two layers of rock on a permeable under layer) would reduce
the overtopping further. A gentle rubble mound slope with cotα = 4 gives very low
overtopping.
Vertical structures under pulsating waves
1,6
give lower overtopping than steep
smooth slopes, but more than a rough rubble mound slope. The impulsive conditions
give a diﬀerent trend. First of all, the inﬂuence of the relative water depth is fairly
small as all curves with diﬀerent H
m0
/h
s
are quite close. For low vertical structures
(R
c
/H
m0
< 1.5), there is hardly any diﬀerence between pulsating and impulsive
conditions. The large diﬀerence is present for higher vertical structures and certainly
for the very high structures. With impulsive conditions, water can be thrown high
into the air, which means that overtopping occurs even for very high structures.
The vertical distance that the discharge travels is more or less independent of the
actual height of the structure. For R
c
/H
m0
> 3 the curves are almost horizontal.
Another way of comparing the eﬀectiveness of structure types is to show the
inﬂuence of slope angle on wave overtopping, as in Fig. 14.5. A vertical structure
means cotα = 0. Battered walls have 0 < cotα < 1. Steep slopes are generally
described by 1 ≤ cotα ≤ 3. Gentle slopes have roughly cotα ≥ 2 or 3. Overtopping
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 349
1.E07
1.E06
1.E05
1.E04
1.E03
1.E02
1.E01
1.E+00
0 1 2 3 4 5 6 7 8
Slope angle cot α
D
i
m
e
n
s
i
o
n
l
e
s
s
o
v
e
r
t
o
p
p
i
n
g
q
/
(
g
H
m
0
3
)
0
.
5
vertical, Allsop (1995)
vertical, Franco et al. (1994)
impulsive vertical, Hm0/hs=0.9, so=0.04
smooth slope, so=0.04
smooth slope, so=0.01
R
c
/H
m0
=1.5
R
c
/H
m0
=3.0
Fig. 14.5. Comparison of wave overtopping as a function of slope angle.
prediction curves for two relative freeboards: R
c
/H
m0
= 1.5 and 3.0 are shown in
Fig. 14.5.
Of course, similar conclusions can be drawn as for the previous comparison.
Steep slopes give the largest overtopping, which reduces for gentler slopes, for a
given wave condition and water level. Vertical slopes give less overtopping than steep
smooth slopes, except for a high vertical structure under impulsive conditions. This
graph gives also the method to calculate for a battered wall: interpolate between a
vertical wall and a slope 1:1 with cotα as the parameter to interpolate.
Details of all equations used here are described in more detail in Chaps. 15 and
16 (sloping smooth structures, rubble mound structures, and vertical structures).
14.2.3. Overtopping volumes and V
max
Wave overtopping is a dynamic and irregular process and the mean overtopping
discharge, q, does not cover this aspect. But by knowing the storm duration, t, and
the number of overtopping waves in that period, N
ow
, it is possible to give some
description of this irregular and dynamic overtopping, if the overtopping discharge,
q, is known. Each overtopping wave gives a certain volume of water, V , and these
can be described by a distribution.
The twoparameter Weibull distribution can be ﬁtted to many distributions. This
equation has a shape parameter, b, and a scale parameter, a. The shape parameter
gives a lot of information on the type of distribution. Figure 14.6 gives an overall
view of some wellknown distributions. The horizontal axis gives the probability of
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
350 J. W. van der Meer et al.
0
2
4
6
8
10
12
14
16
probability of exceedance (%)
o
v
e
r
t
o
p
p
i
n
g
v
o
l
u
m
e
(
m
3
/
m
) b=3
b=2
b=1
b=0.75
b=0.65
b=0.85
100 90 70 50 30 20 10 5 2 1 0.5 0.1
Fig. 14.6. Various distributions on a Rayleigh scale graph. A straight line (b = 2) is a Rayleigh
distribution.
exceedance and has been plotted according to the Rayleigh distribution. It is known
that wave heights in deep water generally conform to a Rayleigh distribution; so,
responses governed by deep water wave conditions will plot on or close to a straight
line, whilst shallow water eﬀects will show deviations from the Rayleigh distribution.
When waves enter shallow water and the highest waves break, wave heights more
closely match a Weibull distribution with b > 2. An example with b = 3 is shown
in Fig. 14.6, and this indicates that there are more large waves of similar height.
The exponential distribution (often found for extreme wave climates) has b = 1,
and shows that extremes become larger compared to most of the data. Such an
exponential distribution would give a straight line in a loglinear graph.
The distribution of overtopping volumes for all kinds of structures has average
values even smaller than b = 1. Such a distribution is even steeper than an expo
nential distribution. It means that the wave overtopping process can be described
by many fairly small overtopping volumes and a few very large volumes. The EA
manual
2
gives values for b and a, based on limited data sets. The bvalues are mostly
within the range 0.6 < b < 0.9. For comparison, curves with b = 0.65 and b = 0.85
are given in Fig. 14.6. The curves are very similar, except that the extremes diﬀer
a little. It is for this reason that an average value of b = 0.75 was chosen for smooth
slopes and not diﬀerent values for various subsets. The same average value has been
used for rubble mound structures, which makes smooth and rubble mound struc
tures easily comparable. The exceedance probability, P
V
, of an overtopping volume
per wave is then similar to
P
V
= P (V ≤ V ) = 1 −exp
−
V
a
0.75
, (14.2)
with
a = 0.84 · T
m
·
q
P
ov
= 0.84 · T
m
· q · N
w
/N
ow
= 0.84 · q · t/N
ow
. (14.3)
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 351
Equation (14.3) shows that the scale parameter a depends not only on the over
topping discharge, q, but also on the mean period, T
m
, and the probability of over
topping, N
ow
/N
w
, or (similarly) on storm duration, t, and the actual number of
overtopping waves N
w
.
The maximum overtopping during a certain event is fairly uncertain as most
maxima, but depends on the duration of the event. In a 6h period, individual peak
volume is likely to be larger than that in 15 min. The maximum overtopping volume
by a single wave during an event depends on the actual number of overtopping
waves, N
ow
, and can be calculated by
V
max
= a · [ln(N
ow
)]
4/3
. (14.4)
A comparison can be made between the mean overtopping discharge, q, and the
maximum overtopping volume in the largest wave. Note that the mean overtopping
is given in l/s per m width and that the maximum overtopping volume is given in
liters per m width.
Again, three example structures are considered: a smooth 1:4 slope; a rubble
slope at 1:1.5; and a vertical wall, and again speciﬁc equations are given in Chaps. 15
and 16. The storm duration has been assumed as 2 h (the peak of the tide), and
a ﬁxed wave steepness of s
m−1,0
= 0.04 has been considered. Figure 14.7 gives the
q −V
max
lines for the three structures and for relatively small waves of H
m0
= 1 m
and for fairly large waves of H
m0
= 2.5 m.
100
1000
10000
100000
0.1 1 10 100 1000
Overtopping discharge q (l/s per m)
M
a
x
i
m
u
m
v
o
l
u
m
e
i
n
o
v
e
r
t
o
p
p
i
n
g
w
a
v
e
(
l
p
e
r
m
)
Smooth slope Hs=2.5m
Rubble mound Hs=2.5m
Vertical Hs=2.5m
Smooth slope Hs=1m
Rubble mound Hs=1m
Vertical Hs=1m
Fig. 14.7. Relationship between mean discharge and maximum overtopping volume in one wave
for smooth, rubble mound, and vertical structures for wave heights of 1 m and 2.5 m.
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
352 J. W. van der Meer et al.
A few conclusions can be drawn from Fig. 14.7. First of all, the ratio q/V
max
is about 1000 for small q (roughly around 1 l/s per m), and about 100 for large q
(roughly around 100 l/s per m). So, the maximum volume (in liters per m width) in
the largest wave is about 100–1000 times larger than the mean overtopping discharge
(in l/s per m).
Secondly, the lines for a 1 m wave height are lower than for a larger 2.5 m wave
height, which means that for lower wave heights, but similar mean discharge, q, the
maximum overtopping volume is also smaller. For example, a vertical structure with
a mean discharge of 10 l/s per m gives a maximum volume of 1000 l/m for a 1 m
wave height and a volume of 4000 l/m for a 2.5 m wave height.
Finally, the three diﬀerent structures give diﬀerent relationships, depending on
the equations to calculate q and the equations to calculate the number of over
topping waves.
14.2.4. Wave transmission by wave overtopping
For structures like nearshore breakwaters, large overtopping can be allowed as this
overtopping simply falls into the water behind the structure causing new waves
behind the structure. This is termed wave transmission and is most easily deﬁned
by the wave transmission coeﬃcient K
t
= H
m0,t
/H
m0,i
, with H
m0,t
= transmitted
signiﬁcant wave height, and H
m0,i
= incident signiﬁcant wave height. The limits
of wave transmission are K
t
= 0 (no transmission) and 1 (no reduction in wave
height). If a structure has its crest above water, the transmission coeﬃcient will
never be larger than about K
t
= 0.4–0.5.
Wave transmission was studied in the European DELOS project,
4
and the fol
lowing prediction formulae were derived for smooth sloping structures:
K
t
=
−0.3 ·
R
c
H
m0,i
+ 0.75 · (1 −exp (−0.5 · ξ
op
))
· (cos β)
2/3
, (14.5)
with a minimum K
t
= 0.075 and maximum K
t
= 0.8; and limitations 1 < ξ
op
< 3;
0
◦
≤ β ≤ 70, and 1 < B/H
m0,i
< 4, where β is the angle of wave attack and B is
the crest width (and not berm width).
The transmission coeﬃcient K
t
is shown in Fig. 14.8 as a function of the relative
freeboard R
c
/H
m0
and for a smooth structure with slope angle cotα = 4 (a gentle
smooth lowcrested structure). Three wave steepnesses have been used: s
op
= 0.01
(long waves), 0.03 and 0.05 (short wind waves). Again, normal wave attack has been
assumed. Wave transmission decreases for increasing crest height, but longer waves
give more transmission. Wave overtopping can be calculated for the same structure
and wave conditions (see Chap. 15 and Fig. 14.9).
Wave overtopping and transmission can be related if Figs. 14.8 and 14.9 are
combined, and Fig. 14.10 shows this relationship. For convenience, the graphs are
not dimensionless, but for a wave height of H
m0,i
= 3 m.
A small transmitted wave height of H
m0,t
= 0.1 m is found if overtopping exceeds
q = 30−50 l/s per m. In order to reach a transmitted wave height of H
m0,t
= 1 m
(K
t
≈ 0.33), the overtopping discharge should exceed q = 500−2500 l/s per m or
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 353
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.5 1 1.5 2
R
c
/H
m0
K
t
so=0.01
so=0.03
so=0.05
Fig. 14.8. Wave transmission for a gentle smooth structure of 1:4 and for diﬀerent wave steepneses.
0.001
0.010
0.100
1.000
0 0.5 1 1.5 2
R
c
/H
m0
q
/
(
g
H
m
0
3
)
0
.
5
so=0.01
so=0.03
so=0.05
Fig. 14.9. Wave overtopping for a gentle smooth structure of 1:4 and for diﬀerent wave
steepnesses.
0.5–2.5 m
3
/s per m. One may conclude that any signiﬁcant wave transmission is
always associated with (very) large wave overtopping.
Wave transmission for rubble mound structures has also been investigated in
the European DELOS project,
4
and the following prediction formulae were derived
for wave transmission:
K
t
= −0.4R
c
/H
m0
+ 0.64B/H
m0
−0.31(1 −exp(−0.5ξ
op
))
for 0.075 ≤ K
t
≤ 0.8. (14.6)
Wave overtopping for a simple trapezoidal rubble mound can be calculated by
methods presented in Chap. 15. A typical rubble mound structure has been used as
an example, with cotα = 1.5, armour rock of 6–10 ton (D
n50
= 1.5 m), and crest
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
354 J. W. van der Meer et al.
0
500
1000
1500
2000
2500
3000
3500
0 0.5 1 1.5 2
H
m0 transmitted
(m)
q
(
l
/
s
p
e
r
m
)
so=0.01
so=0.03
so=0.05
Fig. 14.10. Wave transmission versus wave overtopping for a smooth 1:4 slope and a wave height
of H
m0
= 3 m.
0
500
1000
1500
2000
2500
3000
3500
0 0.5 1 1.5 2
H
m0 transmitted
(m)
q
(
l
/
s
p
e
r
m
)
so=0.01
so=0.03
so=0.05
Fig. 14.11. Wave transmission versus wave overtopping discharge for a rubble mound structure,
cotα = 1.5, 6–10 ton rock, B = 4.5 m, and H
m0
= 3 m.
width of B
w
= 4.5 m (3D
n50
). A wave height of H
m0,i
= 3 m has been assumed
with the following wave steepness: s
m−1,0
= 0.01 (long waves), 0.03 and 0.05 (short
wind waves). In the calculations, the crest height has been changed to calculate
wave transmission as well as wave overtopping.
These are compared in Fig. 14.11 which shows that longer waves (s
m−1,0
= 0.01)
give more wave transmission, even for similar overtopping discharge. The reason is
probably the eﬀect of wave action penetrating through the permeable upper layers,
easier for long waves, thus contributing to the waves behind the structure.
So, for permeable mounds, there may still be signiﬁcant wave transmission
through the structure even without substantial overtopping discharge. In this
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 355
example, transmitted wave heights between H
m0,t
= 0.5 m and 1 m are found for
overtopping discharges smaller than q = 100−200 l/s per m.
A simple equation for wave transmission at vertical structures has been given
by Goda,
7
although again more complete methods are given in Chap. 16:
K
t
= 0.45 −0.3R
c
/H
m0
for 0 < R
c
/H
m0
< 1.25. (14.7)
For vertical walls, transmission is primarily governed by the relative crest height
with little or no eﬀect of wave period or steepness. A simple vertical structure has
been used as an example with H
m0,i
= 3 m. Overtopping and wave transmission are
compared in Fig. 14.12, where in the calculations the crest height has been changed
to calculate wave transmission as well as wave overtopping.
For comparison, the same rubble mound structure has been used as the example
in Fig. 14.11, with cotα = 1.5, 6–10 ton rock (D
n50
= 1.5 m) as armour, a crest
width of 4.5 m (3D
n50
), and a wave steepness s
op
= 0.03. The curve for a smooth
structure (Fig. 14.10) and for s
op
= 0.03 has been given too in Fig. 14.12.
A rubble mound structure gives more wave transmission than a smooth structure
for similar overtopping discharge, but a vertical structure can give even more trans
mission. The reason may be that overtopping water over the crest of a vertical
breakwater always falls in a mass from a distance into the water, rather than ﬂowing
relatively smoothly over or through a sloping structure.
One may conclude that even without considerable wave overtopping discharge at
the crest of a vertical structure, there still might be signiﬁcant wave transmission.
In this example of a vertical structure, transmitted wave heights between 0.5 m and
1 m are found for overtopping discharges smaller than 100–200 l/s per m.
Finally, an example is shown of both wave overtopping and wave transmission
on a rubble mound breakwater in Fig. 14.13.
0
500
1000
1500
2000
2500
3000
3500
0 0.5 1 1.5 2
H
m0 transmitted
(m)
q
(
l
/
s
p
e
r
m
)
vertical
rubble mound (so=0.03)
smooth (so=0.03)
Fig. 14.12. Comparison of wave overtopping and transmission for a vertical, rubble mound, and
smooth structure.
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
356 J. W. van der Meer et al.
Fig. 14.13. Wave overtopping and transmission at breakwater IJmuiden, the Netherlands.
14.3. PCOVERTOPPING
The computer programpcovertopping was created using the results of the TAW
8
Report “Wave runup and wave overtopping at dikes” and is used for the ﬁveyearly
safety assessment of all water defenses in the Netherlands. The TAW
8
Report has
now been replaced by Chap. 5 (dikes and embankments) in the EurOtop Over
topping Manual
5
and extended for rubble mound and vertical structures in Chaps. 6
and 7 of that manual. pcovertopping has been translated into English, and is
available from the EurOtop Overtopping Manual
5
web site.
The program was based on diketype structures. The structure should be sloping,
although a small vertical wall on top of the dike may be included. Some eﬀects of
roughness and/or permeability can be included, but not a crest with permeable and
rough rock or armour units. In such a case, the structure should be modeled up to
the transition to the crest, and other formulae should be used to take into account
the eﬀect of the crest.
pcovertoppingwas set up so that almost every sloping structure can be modeled
by an unlimited number of sections. Eachsection is given by x–y coordinates, and each
section can have its own roughness factor. The programcalculates most relevant over
topping parameters (except ﬂowvelocities and ﬂowdepths), such as:
• 2% runup level;
• mean overtopping discharge;
• percentage of overtopping waves;
• overtopping volumes per wave (maximum and for every percentage deﬁned by
the user);
• required crest height for given mean overtopping discharges (deﬁned by the user).
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 357
The main uses of pcovertopping are:
• modeling of overtopping on any sloping structure, including diﬀerent roughnesses
along the slope;
• calculation of most overtopping parameters, not only the mean discharge.
A disadvantage of pcovertopping is that it does not calculate overtopping
for vertical structures or for rubble structures with rough/permeable crest.
An example illustrates some of its capabilities. An example dike cross section
with the design water level at +1 m CD is shown in Fig. 14.14. Diﬀerent materials
are used on the slope: rock, basalt, concrete asphalt, open concrete system, and
grass on the upper part of the structure. The structure has been schematized in
Fig. 14.15 by x–y coordinates and selection of the material of the top layer. The
program selects the right roughness factor.
The input parameters are wave height, wave period (either spectral period
T
m−1,0
or peak period T
p
), wave obliquity, water level (with respect to the same
level as used for the structure geometry), and ﬁnally, number of waves (derived from
the storm duration and mean period) for the calculation of overtopping volumes,
etc. Fig. 14.16 gives the input ﬁle.
Fig. 14.14. Example cross section of a dike.
Fig. 14.15. Input of geometry by x–y coordinates and choice of top material.
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
358 J. W. van der Meer et al.
Fig. 14.16. Input ﬁle.
Fig. 14.17. Output of pcovertopping.
The output is given in three columns (Fig. 14.17). The left column gives the
2% runup level, the mean overtopping discharge, and the percentage of overtopping
waves. If the 2% runup level is higher than the actual dike crest, this level is cal
culated by extending the highest section in the cross section. The middle column
gives the required dike height for the given mean overtopping discharges. Again,
the highest section is extended if required. Finally, the righthand column gives
the number of overtopping waves in the given storm duration, together with the
maximum overtopping volume, and volumes for speciﬁed overtopping percentages,
given as a percentage of the total number of overtopping waves.
pcovertopping also provides a check on whether the results for the 2% runup
level and mean overtopping discharge fall within the measured ranges. Results on
which the formulae were based are shown in the runup or overtopping graphs (see
Figs. 14.18 and 14.19). These show the measured runup or overtopping, including
eﬀects of reductions due to roughness, berms, etc. The curve gives the maximum
for smooth straight slopes with normal wave attack. The program then plots the
calculated point in these graphs (the point within the circle).
14.4. Neural Network Tools
Artiﬁcial neural networks are tools that allow meaning to be extracted from very
large quantities of data. Neural networks (NN) are organized in the form of layers,
within which there are one or more processing elements called “neurons.” The ﬁrst
layer is the input layer, and the number of neurons in this layer is equal to the
number of input parameters. The last layer is the output layer and the number of
neurons in this layer is equal to the number of output parameters to be predicted.
The layers between the input and output layers are the hidden layers, consisting of a
number of neurons to be deﬁned in conﬁguring the NN. Neurons in each layer receive
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 359
Fig. 14.18. Check on 2% runup level.
Fig. 14.19. Check on mean overtopping discharge.
information from the preceding layer, carry out a standard operation, and produce
an output. Each connection has a weight factor assigned from the calibration of the
NN. The input of a neuron consists of a weighted sum of the outputs of the preceding
layer. This procedure is followed for each neuron; the output neuron generates the
ﬁnal prediction of the NN.
Artiﬁcial NNs have applications in many ﬁelds including coastal engineering
where examples have been applied to predicting armour stability, forces on walls,
wave transmission, and wave overtopping. The development of an artiﬁcial NN is
useful, where:
• the process to be described is complicated with many parameters involved,
• there is a large amount of data.
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
360 J. W. van der Meer et al.
INPUT
LAYER
HIDDEN
LAYER
OUTPUT
LAYER
q (m
3
/s/m)
h H
m0,toe
T
m1,0,toe
h
t
B
t f
cot
d
cot
u
R
c
B h
b
tan
B
A
c
G
c
INPUT
LAYER
HIDDEN
LAYER
OUTPUT
LAYER
q (m
3
/s/m)
h H
m0,toe
T
m1,0,toe
h
t
B
t f
cot
d
cot
u
R
c
B h
b
tan
B
A
c
G
c
INPUT
LAYER
HIDDEN
LAYER
OUTPUT
LAYER
q (m
3
/s/m)
h H β
m0,toe
T
m1,0,toe
h
t
B
t f
cot
d
cot
u
R
c
B h
b
tan
B
A
c
G
c
Fig. 14.20. Conﬁguration of the NN for wave overtopping.
It has already been seen that overtopping cannot be predicted by a single
formula, but requires a number of diﬀerent formulae. A single NN can, however,
cover the full range of structures, provided that suﬃcient data are available to
“train” the NN. If too few data are available, predictions in the lesswell populated
regions will be unreliable, particularly where the prediction is trying to extrapolate
out of range. Providentially, international cooperation supported by the European
CLASH research project
3
collected many test results on wave overtopping for all
kinds of coastal structures and embankments.
Within the CLASH project, two NNs were developed, one within the main
project, and one alongside a PhD project.
9
In both cases, the NN conﬁguration was
based on Fig. 14.20, where the input layer has 15 input parameters (β, h, H
m0toe
,
T
m−1,0toe
, h
t
, B
t
, γ
f
, cotα
d
, cotα
u
, R
c
, B, h
b
, tanα
b
, A
c
, G
c
) and one output
parameter in the output layer (i.e., mean overtopping discharge, q). CLASH was
focused on a threelayered NN, where a conﬁguration with one single hidden layer
was chosen.
Development of an artiﬁcial NN requires that all data be checked thoroughly
(rubbish in = rubbish out), and that training be done by those with appropriate
skills. Using NN as a prediction tool, however, is easy for most practical engineers!
It is for this reason that the CLASH NN was adopted as part of the EurOtop
Overtopping Manual.
5
Applying NN requires an Excel or ASCII input ﬁle with parameters, run the
program (push a button), and get a result ﬁle with mean overtopping discharge(s).
It is therefore as easy as using a formula programmed in Excel and does not need
knowledge about NNs. The advantages of the neural network are:
• it works for almost every structure conﬁguration;
• it is easy to calculate trends instead of just one calculation with one answer.
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
Prediction of Overtopping 361
The input exists of 10 structural parameters and four hydraulic parameters.
The hydraulic parameters are wave height, wave period, and wave angle and water
depth just in front of the structure. The structural parameters describe almost
every possible structure conﬁguration by a toe (two parameters), two structure
slopes (including vertical and wave return walls), a berm (two parameters), and
a crest conﬁguration (three parameters). The tenth structural parameter is the
roughness factor for the structure (γ
f
) and describes the average roughness of the
whole structure. Although guidance is given, estimation is not easy if the structure
has diﬀerent roughness on various parts of the structure. An overall view of possible
structure conﬁgurations is shown in Fig. 14.21. It clearly shows that the NN is able
to cope with most structure types.
Fig. 14.21. Overall view of possible structure conﬁgurations for the NN.
August 13, 2009 14:43 9.75in x 6.5in b684ch14 FA
362 J. W. van der Meer et al.
H
s
= 5 m; T
p
= 10 s β = 0
h = 12 m h
t
= 9 m
B
t
= 4 m
Xbloc 1:1.5
A
c
= 5 m
G
c
= 5 m
R
c
= 4 m
Fig. 14.22. Example cross section with parameters for application of NN.
Very often one is not only interested in one calculation, but in more. As the
input ﬁle has no limitations in the number of rows (= number of calculations), it
is easy to incrementally inc