MULTICRITERIA ANALYSIS FOR LAND-USE MANAGEMENT

Environment & Management

VOLUME9

The titZes published in this series are listed at the end of this voZume.
Multicriteria Analysis for
Land-U se Management

edited by

EUROBEINAT

and

PETER NIJKAMP
Vrije Universiteit, Amsterdam, The Netherlands

SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-90-481-5077-9 ISBN 978-94-015-9058-7 (eBook)

DOI 10.1007/978-94-015-9058-7

Printed on acid-free paper

All Rights Reserved

© 1998 Springer Science+Business Media Dordrecht
OriginaIly published by Kluwer Academic Publishers in 1998
Softcover reprint ofthe hardcover 1st edition 1998

No part of the material protected by this copyright notice may be reproduced or

utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner
Table of contents

Preface ....................................................................................................................... vii

Euro Beinaf and Pefer Nijkamp

Land-use management and the path towards sustainability ........................................... 1

Part I: Problem structuring for land-use management

Theodor 1. Sfewart and Alison Joubert

Conflicts between conservation goals and land use for exotic forest plantations in South
Africa .......................................................................................................................... 17

Carlos A. Bana e Costa, Leonardo Ensslin and Alessandro P. Costa

Structuring the process of choosing rice varieties at the south ofBrazil ....................... 33

Part 11: The multiple dimensions of land-use management

Francesco Marangon and Tiziano Tempesta

Rurallandscape and economic results of the farm; a multi-objective approach ........... .49

Gerda Hermanides and Peler Nijkamp

Multicriteria evaluation of sustainable agriculturalland use: a case study of Lesvos .... 61

Giuseppe Munda, Massimo Parruccini and Giuseppe Rossi

Multicriteria evaluation methods in renewable resource management: integrated water
management under drought conditions ........................................................................ 79

Giulio Giannerini, Eliof Laniado and Giorgio Stagni

Assessment, evaluation and allocation of funds to infrastructure projects: the case of the
road network in Lombardy (Italy) ................................................................................ 95

Carlo Giupponi and Paolo Rosato

A farm multicriteria analysis model for the economic and environmental evaluation of
agricultural land use .................................................................................................. 115

Part 111: Multiple actors and multiple perspectives in land-use decisions

Euro Beinaf and Michiel van Drunen

Spatial conflicts in transport policies: an exploration ofthe perspectives ofregional and
local actors ................................................................................................................ 139

Diana S. Yakowifz and Mark Welfz

An algorithm for computing multiple attribute additive value measurement ranges under
a hierarchy of the criteria: application to farm or range land management decisions ... 163

v
Fred Wenstep and Arne J Carlsen
Using decision panels to evaluate hydropower development projects ......................... 179

Andrea Nardini
Improving decision-making for land-use management: key ideas for an integrated
approach based on MCA negotiation forums ............................................................. 199

Part IV: Spatial information in land-use management

J Ronald Eastman, Hong Jiang and James Toledano

Multi-criteria and multi-objective decision making for land allocation using GIS ...... 227

Ron Janssen and Marjan van Herwijnen

Map transformation and aggregation methods for spatial decision support ................. 253

Andrea Patrono
Multi-criteria analysis and Geographic Information Systems: analysis of natural areas
and ecological distributions ....................................................................................... 271

Mark A. Ridgley and Gerrit W. Heil

Multicriterion planning of protected-area buffer zones: an application to Mexico's Izta-
Popo national park ..................................................................................................... 293

Agro-ecological zones and Resource Management Domains (RMDs) in relation to land-

use planning .............................................................................................................. 311
Jaques Antoine, Frankjohn J Dent, Denis Sims and Robert Brinkman

Part V: Looking ahead

Henk J Scholten, Paul Misdorp and Martin van Beek

Land-use trends in Europe: simulation, presentation and scenario evaluation of future
options....................................................................................................................... 335

Hubert N. van Lier and Pa! D. Taylor

Long term comprehensive evaluation strategies for spatial planning, design and
management .............................................................................................................. 353

List of authors ........................................................................................................... 369

vi
Preface

The idea of this book started at approximately 33.000 feet, somewhere above the
Alps. On our way to a workshop in Venice we had the opportunity of appreciating
the different types of landscapes and the complex patchwork of urban areas,
agriculture, forests, rivers and lakes that can be seen from an aircraft. The
complexity of this puzzle, and the complex task of managing its evolution, became
the topic of conversation for the rest of the flight. It also became the topic of this
book.
Land-use management and multicriteria analysis offer countless opportunities
for mutual reinforcement. These two fields have developed largely independently,
but a trend towards the exploration of their synergies is now emerging. This is
clear from the recent literature on land-use management, spatial analysis and
spatial planning, which increasingly includes references to multicriteria
methodologies and decision analysis. At the same time, a growing share of
multicriteria applications now focus on environmental and land-use issues.
This book includes contributions from authors coming from a variety of
disciplines and backgrounds. All together they highlight current issues in
multicriteria analysis and land-use management from theoretical, methodological
and practical perspectives.
We would like to thank all authors for the time and effort they spent writing,
revising and adapting their papers to the book plan. All contributions have been
reviewed by two referees. Their help, critical comments and suggestions for
improving the quality of the book are gratefully acknowledged. We would also
like to thank the secretaries of the Institute for Environmental Sciences (Vrije
Universiteit of Amsterdam) for their editorial work. Finally, we take this
opportunity to thank the staff at Kluwer for revising the manuscripts and for
managing the project efficiently and smoothly.

Euro Beinat and Peter Nijkamp

Amsterdam, April 1998

vii
Land-use management and the path towards
sustainability

Euro Beinae.2 and Peter Nijkampl

Institute for Environmenta/ Studies and 2Facu/ty of Economics
J

Vrije Universiteit, Amsterdam, The Nether/ands

1. Introduction

Land-use changes have received major attention within the global environmental
change debate (see Meyer and Turner 11, 1994; Nijkamp, 1997; Ostrom, 1990; Parry,
1990). Major reasons for this renewed interest are the threats imposed by c\imate
change, deforestation, desertification and in general the loss of biodiversity. In this
context, sustainable land use has become an important analytical and policy issue (see
Finco and Nijkamp, 1997). Land use has a peculiar economic feature in that it has a
derived nature: human action (e.g., production, consumption, investment, recreation
etc.) requires for its operation the use of geographical space, which in a strict sense does
not have a value in itself (except as a capital asset). Economic activities are projected on
geographical space in various manifestations, depending on the economic functions
concerned (e.g. housing, facilities, infrastructure, agriculture, green space etc.). This
spatial mapping has immediate consequences for the environmental quality conditions
of an area, as there are in general spatially distinct, and hence conflicting, land-use
possibilities (see also Frederick and Rosenberg, 1994; Walker, 1993). Land use also
offers glaring examples of spatial environmental externalities, which in many cases may
be seen as distorted and unbalanced land use, biased in favour of specific
environmentally non-benign activities. These considerations justify the position of the
land-use management debate at the heart ofthe sustainability debate (see also Turner II
et al. , 1995).
Changes in land use have always accompanied economic development. The
historical trend shows a substantial and progressive transformation of natural areas into
areas which support agricultural, urban, or industrial functions. Table I illustrates this
trend by focusing on the changes in forest, grassland, and cropland between 1850 and
1980. Apart from Europe, where forests and grassland show a slight increase, the
overall trend is towards a substantial loss of natural land in favour of cropland. This
transformation affects as much as forty percent of the forests and grassland of some
areas, influenced by such factors as population growth, food production, income, wood
production and land tenure arrangements (Pearce, 1991).
The demand für space and natural resüurces will continue in the future, implying
further changes in land allocation and in land-use management. Table 2 shows aseries
ofprojeetions for the year 2010 (adapted from van Dieren, 1995). While irrigated land,
eropland, rangeland and pasture will inerease in absolute terms, the progressive
reduetion of forested areas eontinues along the trend illustrated in Table I. However,
the per eapita availability of all these resouree will deerease, implying a further pressure
on land. Without eountermeasures, this will neeessarily lead to an inereasing
environmentalload and to an impoverishment ofthe natural resourees eapital.

Table I. Percentage land- use changes in the period 1850-1980 (World Research Institute, 1987;
adaptedfrom Pearce, 1991).
Forests Grassland Cropland
Tropical Africa -20 +9 +288
Latin America -19 -23 +677
North America -3 -22 +309
China -39 -3 +79
South Asia -43 -1 +196
South East Asia -7 -25 +670
Europe +4 +8 -4
Former Soviet Union -12 -1 +147
All -15 -1 +179

Table 2. Availability olnaturat resources (adaptedfrom van Dieren, 1995).

1990 2010 Total Per capita
(million) (million) change change
(%) (%)
Population 5290 7030 +33
Irrigated land (hectares) 237 277 +17 -12
Cropland (hectares) 1444 1543 +5 -21
Rangeland and pasture (hectares) 3402 3540 +4 -22
Forest (hectares) 3413 3165 -7 -30

The negative effeets of land-use exploitation are manifested in soil erosion, loss of
habitats, inereased vulnerability of the soil, deerease in the earrying eapaeity of land,
landseape modifieation and loss ofnatural amenities. While their negative eonsequenees
are clear, land uses and land-use management in general are rather poorly understood,
given the multiple eontlieting funetions involved in spaee eonsumption. There are many
intrieate and eomplex linkages between the eeonomy, the social sphere and the environment,
in whieh land use and spaee usually aet as the vehicles for transmitting extemalities.
Signifieant improvement in understanding these extemalities has been aehieved, but major
gaps in our knowledge oftheir dynamie spatial eontext remain, as noted by The World Bank
Development Report (World Bank, 1992): "Degradation and destruetion of environmental
systems and natural resourees are now assuming massive proportions in some developing
eountries, threatening eontinued, sustainable development. It is now generally reeognised
that eeonomie development itself ean be an important eontributing faetor to growing

2
environmental problems in the absence of appropriate safeguards. A greatly improved
understandingof the natural resource base and environment systems that support national
economies is needed if patterns of development that are sustainable can be determined and
recommended to governments".
Understanding land-use manase ment and sustainable deve\opment has been partly
impeded by the limited attention given to natural resources within economics. The
Physiocrats recognised the productive capacity ofthe natural environment as a major source
of we\fare. This position, however, has been an exception rather than the rule in economic
thinking. Capital and labour, in addition to land. are regarded as the main we\fare generators
in classic economics, which also assigns a minor role to government and institutions that
frame actual decisions in the market.
Post WWII neo-classical thinking does not view nature as such as the generator of
welfare (e.g. income per capita) equivalent to labour, capital, technology and land. In this
view, "...there seemed no reason to accord land any special treatment that would suggest its
role is quite distinct from that of the other factors. Land could safely be subsumed under
broader aggregate of capital, ... " (Randall and Castle, 1985; p. 573). In general, the role of
environmental capital and goods in traditional neo-clässical economics is rather modest.
In the past few decades, the externalities and limits to growth (with regard to both
renewable and non-renewable resources) have become a new focaI point of economic
research. The major policy challenge is, in general, to avoid a "tragedy of the commons"
(Hardin, 1968) in view of the long-term threats exerted by the (seemingly) inevitable and
persistent changes in both local and global environmental conditions. Against this
background. land use and the spatial-environmental aspects of the economy deserve more
scientific attention from economists and science in general.

2. Conservation and development

Following the report of the Bruntland Commission (WCED, 1987), a wealth of research
has explored the meaning and consequences of sustainable development. Within this
context, land-use planning and management are activities which seek the assessment of
land potential and suitable land exploitation. Traditionally, this has been concerned with
the solution of a fundamental trade-off: conservation versus economic exploitation and
development. The view that conservation and development were conflicting objectives
dominated most of the twentieth century. Subsequent to the "Bruntland" report,
however, a complementary view has emerged. The conflict between conservation and
development is not necessarily an intrinsic feature of these two objectives, but rather a
feature of decisions and policies (and socio-economic patterns in general) which have
systematically disregarded one or the other.
In the recent past, a view has emerged that the win-win combination of conservation
and development should be attainable when sufficient economic resources are available.
The green Kuznets curve (Heintz and Verbruggen, 1997; de Bruyn and Opschoor, 1994;
Seiden and Song, 1994) synthesises this scenario by linking Gross National Product to

3
the total environmental load (Figure I). The decreasing trend at the right hand side of
the curve is the result of a shift towards the production and purchase of more
environmentally oriented goods and services, which is permitted by the availability of
higher wealth.

GNP per capita

Figure 1. The green Kuznets curve (adaptedfrom Heintz and Verbruggen. 1997).

There is mixed and partial evidence about the validity of this trend. The descending
path has been observed for environmental issues which have a direct relationship with
human health and which can be associated with high costs of environmental degradation
and remediation. O'Niell et al. (1996), however, point out that this is not a signal of a
general phenomenon: the evidence collected on "a subset of pollutants in a limited
number of pi aces cannot be accepted as surrogates for the complex interactions between
economic growth and the environment on which that growth takes place".
The validity of the green Kuznets curve would justify an emphasis on economic
growth in the medium term to reach a GOP-level commensurate with sustainability in
the long term. By implication, the need for environmental policy and land-use planning
is questioned. The assumptions involved, however, are based on conditions rarely met
in reality. This has led some authors to emphasise that, at least in the medium term,
"economic growth is no substitute for environmental policy" (Heintz and Verbruggen,
1997).
Besides the implications for the need and extent of environmental policy, the green
Kuznets curve does indicate that environmental policy cannot be static but must evolve
as societal and economic conditions change. The Outch Committee for Long-Term
Environmental Policy (1994) indicates five phases of the human-environment
relationship:
1. environmental degradation as a side effect;
2. environmental degradation as a cost factor;
3. the environment as a boundary condition;
4. the environment as a policy-determining factor;
5. the environment as an objective.

4
 Land-use planning and management emerge as effective policy instruments
especially in the fifth phase, because they are an indirect means of addressing
environmental quality by focusing on a derived concern: the spatial distribution of
human activities and natural resources.
Land use is the result of the interplay among a myriad of human and natural systems.
By addressing these systems in their spatial dimension (i.e., the distribution dimension
of location and density, and the interaction dimension of attraction, pressure and change
dynamics), ultimately the issues of sustainability and environmental resilience are
reached, as weil as those of social equity or economic competitiveness. Land-use
management is one means for comprehending the current state and for achieving
different states ofthe social, economic and environmental systems.

3. Issues in land-use management

It goes without saying that the multifunctionality and complexity of land use give rise to
much ambiguity in land-use management. There is no unidimensional denominator which
can be used to assess and evaluate land-use changes and policies. Consequently, there is the
need for a clear formulation of indicators encapsulating a wide diversity of attributes and
environmental assets in a spatial setting.
Policies on sustainable development and land-use management have increasingly
moved from a global level to a meso approach, such as areal level or a sector
intervention. The introduction of the spatial dimension has also permitted the
development of additional sustainability management concepts, such as strong and weak
sustainability (see Pearce and Turner, 1990; and van Pelt, 1993). Strong environmental
sustainability would imply that in all areas an improvement of environmental quality
conditions would take pI ace, whereas weak sustainability refers to a situation where in
some areas an environmental degradation has to be accepted, provided this is at least
compensated for by improvements elsewhere. If we extend the concept of
environmental sustainability towards the broader concepts of sustainable deve10pment
(including environmental, economic and social dimensions), the substitution
possibi1ities mayaIso be widened by a trade-off between environmental, economic and
social conditions. This can be visualised by the scheme in Figure 2. This scheme can be
used to identify the potential choice conflicts in land-use management, such as:
• whether environmental decay in a given area for a distinctive purpose (e.g. industrial
development) can be compensated for by enhancing the environmental quality of
another area (e.g. a tourist area), or;
• whether the benefits for a social group (e.g. citizens wishing to use rural areas for
recreation purposes) correspond to the interests of other groups (e.g. farmers
interested in reducing landscape complexity to increase production efficiency), or;
• whether the requirements placed on an economic sector (e.g. the ever increasing
mobility demand) are compatible with other requirements (such as the demand for a

5
 higher environmental quality at local and global scales).

L...--E_COD_omiC--,1 ~ I "',," - agriculture

- industry
- transport

1
Sodal
I~I Groups
- administrations
- citizens

I
~--------~ ~------~
- interest groups
- polical parties

f Environmental
I~I-.' I -
-
-
-
nations
regions
river basins
urban areas

Figure 2. Sustainable developmenf.

Some of these trade-offs are of a long range nature, thus adding a temporal
dimension to the graph above and leading to inter-temporal trade-offs. It is this multi-
faceted feature which attributes an integral economic value to land, such as for housing,
industry, infrastructure or agriculture. Consequently, the question whether some use of
land leads to a sustainable outcome depends not only on external sustainability criteria
ofland use (e.g. land degradation versus economic growth), but it is also determined by
internal sustainability criteria (e.g. agriculture versus tourism).

4. A scientific tool-box for land-use management

A proper answer to the above question can only be given if a satisfactory scientific tool
box is deployed for investigating the complexity and the solution strategies of the
management trade-offs. The literat ure and the practice of environmental and land-use
management offers a wide range of opportunities for developing such a tool box (see,
for instance, Giaoutzi and Nijkamp, 1994).
Impact analysis (cf. Therivel and Partidario, 1996; Wathern, 1988) serves to assess and
quantify the relationships between developments and the effects on environmental systems.
Impact analysis is widely used in environmental and land-use studies to assess the results of
policies or projects at national, regional or local levels. lt is a flexible tool as it permits us to
use several types of analytical methods like econometric models. input-output models,
simulation, prediction and scenario methods, goals achievement methods and qualitative
decision support models. Especially in studies concerning environmental and land-use
impacts which manifest themselves in the long run, a dynamic approach to spatial impact
analysis is necessary. In many cases dynamic models are used to assess the various effects in

6
an impact chain of a complex spatial system. In this respect, it is necessary to use plausible
parameter values (either statistically - econometrically estimated or otherwise calibrated) in
order to trace the multi-period consequences of changes in external conditions or policy
controls for the system at hand. In this context, the openness of spatial systems seen from the
land use perspective is worth emphasising.
Spatial analysis and efTective and accessible information systems are vital to economic
performance and strategic decision-making. The rapid development of digital and electronic
technologies ofTers new potential for information recording, handling, processing and
delivery. From a geographical viewpoint, the trend towards advanced information systems
has led to the design and use of GIS (cf. Scholten and StilweIl, 1990). A GIS enables a
coherent representation of a set of geographical units or objects which - besides their
10cation position - can be characterised by one or more attributes (feature, label or thematic
compound). GIS may be highly important für the planning of our scarce space, not only on a
global scale (e.g., monitoring of rain-forest development), but also on a local scale (e.g.,
physical planning). The past twenty years have witnessed the development of various
computer-based applications of information systems which have changed the activity
patterns and decision modes of spatial actors.
Evaluation methodologies, such as Cost Benefit Analysis (CBA; Brent, 1996), have long
been applied to assess the economic soundness of public decisions. The pros-and-cons of
these economic approaches have been discussed extensively (Munda, 1995; Nijkamp, 1977)
and their applicability has often been subject to strong criticisms, both from methodological
and ethical perspectives (cf. SagofT, 1988). In spite of this, CBA remains one of the most
known and used approaches, which can be extremely valuable, especially when it is
complemented by other evaluations which extend the scope beyond efficiency issues.
Methods for the analysis of the social component of complex decision environments have
been developed in fields like Stakeholder Analysis (Burgoyne, 1994) and Organisation
Theory (Simon, 1991; Hersey and Blanchard, 1988; Huigen ef al., 1993). These
approaches focus on the relationship and interaction between the multiple actors
involved in adecision process, including the role of procedures, information flows and
behavioural effects. A trend towards a specialisation of these approaches for land-use
management and spatial decisions can also be observed. An example is the appearance
of concepts like Collaborative Spatial Decision Support (NCGIA, 1996), which aim at
exploiting the potentiality of information technology to increase the scope and
effectiveness of the collaboration between actors in the search for solutions to complex
spatial decision problems.
Scenario analysis (cf. van der Heijden, 1996) tries to develop and judge a set of
hypothetical future development alternatives ("images") for a complex system, in order to
generate a rational frame of reference for evaluating different development alternatives. It
may play an important role as a leaming mechanism for decision-makers or physical
planners. By assessing all foreseeable and expectable impacts of development strategies,
decision makers may identify a strategy which fulfils the aim of an ecologically sustainable
economic system in combination with land use. It goes without saying that this idea is also

7
of utmost importance for the development of regional or local economic initiatives. Clearly,
one has to keep in mind that a scenario analysis often means the construction of hypothetical
development alternatives, which after solid empirical work may finally lead to the
construction of feasible and desired choice alternatives.
In spite of this wealth of approaches, land-use management and the relationship with
sustainability proves to be a complex issue for which a satisfactory scientific basis and a
methodological approach are still underdeveloped. More than anything, land use
requires an "intellectual family" of approaches (van Kooten, 1993), which combines the
experience and the strengths of many disciplines. However, aside from the search for a
more sophisticated approach to land-use management, scientific tools are only
instruments for understanding, explaining and achieving a more balanced and attractive
state ofthe environment through the land-use levy. The question ofwhat environmental
conditions we want to achieve, and what future we are willing to pursue, remains at the
core of the land-use debate. The importance of this fundamental social discussion will
remain valid in the future, and the degree to which we will be able to substantiate this
discussion will be the measurement of the success of land-use management approaches.

5. Multicriteria analysis and land-use management

The need for developing a platform on which different contributions can be integrated is one
of the main reasons for the growing interest in Multicriteria Analysis (MCA) for land-use
management (cf. Nijkamp et al., 1990; Beinat, 1997). MCA has evolved from a mechanism
for the selection of the best alternative from a set of competing options, to a range of
decision aid techniques. MCA now supports the structuring of adecision problem, the
exploration ofthe concerns of decision actors, the evaluation of alternatives under different
perspectives, and the analysis of their robustness against uncertainty. At present, MCA
comprises a wide set of tools, but MCA is especially a way of approaching complex
decision problems. The deve\opment of MCA for spatial problems (cf. Scholten and
StilweIl, 1990) and the integration of spatial concepts in multicriteria analysis imply that
MCA can now offer a substantial contribution to land-use management. There are many
ways in which the contribution of MCA can be exploited:
• MCA supports the analysis of multiple decision alternatives and land-use options and
aids the identification ofthe most suitable management solution for a given purpose;
• the effects of alternative options can be presented in a variety of forms, such as monetary
units, physical units, qualitative judgements, etc.
• it offers a framework for the exploration of the objectives and concerns of decision
actors, making it possible to understand and justify the main issues involved in a
decision;
• it makes it possible to analyse the trade-offs between different objectives and concerns,
thus supporting the analysis ofthe pros-and-cons of different options in a transparent and
effective way;

8
• it makes it possible to consider the different positions of different decision actors, thus
identifying and addressing potential conflicts at an early stage ofthe decision process;
• it offers the possibility of analysing the sensitivity and robustness of different choices
against the effects of uncertainty and on the basis of different future scenarios.

6. Objective and structure of the book

The objective of the book is to illustrate the state of the art and of the practice of
multicriteria methodologies applied to land-use management problems. The book
collects contributions from a wide range of disciplines, covering methodologieal, policy
and application issues. The contributions are divided into five parts.

Part I: Problem structuring for land-use management

The focus of papers in this section is on the process which starts from a cloud of
concepts, concerns, and ideas about adecision problem and leads to a formal problem
structure which is suitable for the analysis of management options.
The management of natural forests and commercial plantations of exotic trees in
South Africa is discussed by T. J. Stewart and A. Joubert. The focus of the paper is on
the development of policy scenarios and on the structure of the value tree for their
analysis. The paper discusses the use of multi attribute scoring systems using
"thermometer" scales as a means of communication of values between groups.
C. Bana e Costa, L. Ensslin and A.P. Costa illustrate an application of farm
management in Brazil. The paper shows the usefulness of cognitive maps and
multicriteria thinking to support the analysis ofthe points ofview ofthe decision actors.
The paper shows how this can be used to identify and structure the fundamental issues
which drive the choice of rice varieties in a Brazilian farm.

Part 11: The multiple dimensions of land-use management

This part emphasises the multiple perspectives which have to be addressed in land-use
management. The presence of multiple concerns and objectives makes it necessary to
analyse the trade-offs between (potentially) conflicting results, and the search for
compromise and acceptable solutions.
The trade-off between the maximisation of farm income and the preservation and
enhancement of rurallandscape is the focus of the paper by F. Marangon and T.
Tempesta. They propose a measure of landscape quality and use multiobjective analysis
to measure the trade-offs between landscape and farm income, also in situations when
subsidies to enhance landscape quality are introduced.
G. Hermanides and P. Nijkamp describe an application of qualitative multicriteria
analysis to sustainable agriculture in Greece. The paper i1lustrates the use of
sustainability indicators and a system to assess the situation in comparison to critical
threshold values. The use of a Regime approach then serves to explore the attractiveness
of different policy options.

9
 Soft multicriteria methods are also addressed by G. Munda, M. Parruccini and G.
Rossi, who focus on a fuzzy-sets approach to multicriteria analysis. An application to
integrated water management under drought conditions in southern Italy is iIIustrated,
together with the analysis of different management options for different decision
players.
G. Giannerini, E. Laniado and G. Stagni consider the problem of resource allocation
for infrastructure projects. The paper shows a system for allocation of public funds
based, on the one hand, on the intrinsic quality of the proposed options, and, on the
other hand, on the degree to which they contribute to the solution of critical problems in
the transport sector.
The use of multicriteria analysis for the economic and environmental evaluation of
agricultural land use is iIIustrated by C. Giupponi and P. Rosato. The authors first
discuss different farm management approaches based on income levels and risks. These
management options are then assessed in terms of their combined environmental and
economic effects, leading to the identification of measures to promote low-impact
agricultural practices.

Part III: Multiple actors and multiple perspectives in land-use decisions

Land-use management is a process in which multiple actors are involved. All actors
have a role and a stake in the decision, which implies the need for considering different
points of view, for facilitating interaction and co-operation, and for addressing
conflicts. The papers included in this section address the presence of multiple actors and
their different perspectives in the analysis of management options.
E. Beinat and M. van Drunen focus on spatial conflicts for large infrastructure
projects. The paper illustrates a methodological framework for the analysis of spatial
conflicts. For a case study in the Netherlands, the evidence collected during the public
participation phase and the results of a multicriteria approach serve to elucidate the
interests, perceptions and position of different policy players.
D. Yakowitz and M. Weltz address the problem of qualitative, hierarchical weights.
The paper shows an analytical method to calculate the minimum and maximum value
scores of the alternatives under evaluation. The method, which does not need a linear
programming sol ver, can be used to explore the effect of changing qualitative weights
at any tier of the hierarchy. An application to farm management in Iowa is illustrated.
Implications for use in multi-actor contexts are also discussed.
The environmental and economic effects of the construction of hydropower plants in
Norway is the topic of the paper by F. Wenst0p and A. Carlsen. Multicriteria analysis
and decision panels are used to trade-off between a set of evaluation criteria. The
process revealed systematic preference differences both within and across the panels,
consistent with known stakeholder interests and the sentiments of the population. The
analysis served to uncover and explain the roots of the existing controversy.
A. Nardini analyses the roles of Environmental Impact Assessment, Cost-Benefit
Analysis and Multicriteria Analysis to support land-use management decisions. The

10
unsatisfactory handling of different views and interest conflicts arising from the many
players involved is identified as a negative common structural feature. An integrated
approach is thus outlined and the role played by MCA is discussed. A case study on a
Chilean irrigation project is taken as an illustration.

Part IV: Spatial information in land-use management

This part of the book highlights the role of spatial information in land-use decisions and
the integration of spatial analysis and multicriteria analysis. This part includes
contributions which iIIustrate the issues of map evaluations, integration of MCA and
GIS, and the provision of spatial information for land-use decisions.
Recent developments in GIS software and in the conceptual basis for decision
making have led to dramatic improvements in the capabilities of GIS for resource
management. J.R. Eastman, H. Jiang and 1. Toledano discuss the use of GIS for land
allocation. The paper emphasises the problems of incorporating subjective expertise in
the context of participatory decision making, and the integration of MCA and GIS.
R. Janssen and M. van Herwijnen focus on the use of maps to represent and evaluate
environmental management alternatives. Aseries of spatial evaluation methods are
presented and applied to the comparison of mapped policy alternatives for the Green
Heart of the Netherlands. These methods do not provide the decision maker with a
single answer, but rather offer the possibility of exploring the features of different
spatial alternatives.
A. Patrono discusses the use of MCA and GIS in Environmental Impact Assessment.
The GlS-MCA approach is used in conjunction with animal distribution data for habitat
suitability assessment and connectivity analysis. This approach serves to analyse the
fragmentation of natural habitats and the environmental disturbance produced by a new
highway in northern Italy.
The design ofbuffer zones around protected areas is addressed by M. Ridgley and G.
Heil. Multiobjective optimisation is used both to formulate the managers' goals, and to
anticipate potential land-use changes resulting under different scenarios. Goal
programming and GIS are employed to help managers design spatial land-use
configurations attractive from landscape-ecological perspectives. An application to
Mexico's Izta-Popo national park is iIIustrated.
1. Antoine, F. Dent, D. Sims and R. Brinkman concentrate on land Resource
Management Domains (RMDs), which are homogeneous areas in terms of agro-
ecological and socio-economic conditions. RMDs are at the basis of interactive land-use
planning. The authors describe applications at the global, national, regional and local
scales.

Part V: Looking ahead

This final part of the book is devoted to two papers which offer the opportunity to think
about the future of land-use management. They offer two different views of the role of
quantitative and qualitative methods to support land-use evaluations and decisions.

11
 A Scenario Evaluation System is described by H. Scholten, P. Misdorp and M. van
Beek. The system is designed to support the prediction and analysis of future land-use
trends in Europe. The possibility of exploring the consequence of different scenarios,
and the effects of different policy actions, offers to policy makers a powerful way to
understand the long-term paths of land use and their desirability.
H.N. van Lier and P.D. Taylor offer an overview of the current issues in the
planning, design and management of the countryside. The paper reviews the status of the
debate between economic deveiopment and conservation illustrating the role played by
management strategies and evaluation concepts. This chapter assesses the efficacy of these
strategies and concepts, and urges a better balancing of human needs with those of the
physical environment.

References
Beinat, E. (1997) Valuefunctionsforenvironmental management, Kluwer, Dordrecht.
Brent, R.J. (1996) Applied cost-benefit analysis, Edward Eigar, Cheltenham.
Bruyn, S.M. de, J.B. Opschoor (1994) Is the economy ecologising?, Discussion paper Tl 94-65,
Tinbergen Institute, Amsterdam.
Burgoyne, 1.G. (1994) Stakeholder analysis, in C. Cassell, G. Symon (Eds.) Qualitative methods in
organisational research, Sage, London.
Dieren, W. van (1995, Ed.) Taking nature into account: areport to the club of Rome, Springer-
Verlag, New York.
Dutch Committee for Long-Term Environmental Policy (1994) The environment: towards a
sustainable future, Kluwer, Dordrecht.
Finco, A., P. Nijkamp (1997) Sustainable land use: methodology and application, Research
Memorandum, Dept. of Economics, Free University, Amsterdam.
Frederick, K.D., N.J. Rosenberg (1994, Eds.), Assessing the impacts of climate change on
natural resource systems, Kluwer, Dordrecht.
Giaoutzi, M., P. Nijkamp (1994) Decision support models for regional sustainable
development, Avebury, Aldershot.
Hardin, J. (1968) The tragedy ofthe commons, Science, 13: 1243-1248.
Heijden, K. van der (1996) Scenarios: the art ofstrategie conversation, Wiley, London.
Heintz, R.J., H. Verbruggen (1997) Meer groei en toch een schoner milieu? De groene Kumets-
curve (More growth and greener environment? The green Kumets-curve), Milieu, 12( I ):2-9.
Hersey, P., K.H. Blanchard (1988) Management of organisational behaviour, Prentice-Hall, New
York.
Huigen, 1., P.H.A. Frissen, P.W. Tops (1993) Het project Betuwelijn: spoorlijn of bestuurlijke co-
productie? PhD thesis, KUB, Tilburg.
Kooten, G.c. van (1993) Land resource economics and sustainable development: economic
policies and the common good, UBC Press, Vancouver.
Meyer, W.B., B.L. Turner II (1994, Eds.) Changes in land use and land cover: agiobai
perspective, Cambridge University Press, Cambridge.
Munda, G. (1995) Multicriteria evaluations in a fuzzy environment. Physica-Verlag,
Heidelberg.
NCGIA: National Center for Geographical Information and Analysis (1996)

12
 hup://www.ncgia.edulresearch/iI7/1-17_home.html
Nijkamp, P. (1997) Environmental security and sustainability in natural resource management:
adecision support framework, Research Memorandum, Dept. of Economics, Free
University, Amsterdam.
'Jijkamp, P. (1997) Theory and application of environmental economics, North-Holland,
Amsterdam.
Nijkamp, P., Rietveld, P., Voogd, H. (1990) Multicriteria evaluation in physical planning,
North Holland, Amsterdam.
O'Neill, R.V., lP. Kahn, lR. Duncan, S. Alliot, R. Efroymson, H. Caldwell, D.W. Jones
(1996) Economic growth and sustainability, a new challenge, Ecological Applications, 6(2):
23-24.
Ostrom, E. (1990) Governing the commons, Cambridge University Press, Cambridge
Parry, M.L. (1990) Climate Change and World Agriculture, Earthscan, London.
Pearce, D. (1991) Population growth, in Pearce, D. (Ed.) Blueprint 2: greening the world
economy, Earthscan, London
Pearce, D. W., K. Turner (1990) Economics of natural resources and the environment,
Harvester Weatsheaf, New York.
Pelt, MJ.F. van (1993) Ecological sustainability and project appraisal, Avebury, Aldershot.
RandalI, A., E.N. Castle (1985) Land Resources and land markets, in A.V. Kneese and lL. Sweeney
(Eds.) Handbook ofnatural resource and energy economics, North-Holland. Amsterdam.
Sagoff, M. (1988) The economy ofthe earth, Cambridge University Press, Cambridge.
Scholten, H., StilweIl, lC.H. (1990, Eds.) Geographical Information Systems and urban and
regional planning, Kluwer, Dordrecht.
Seiden, T.M., D. Song (1994) Environmental quality and development: is there a Kumets curve
for air pollution emissions? Journal of Environmental Economics and Management, 27:
147-162.
Simon, H.A. (1991) Bounded rationality and organisationalleaming, Organisation Science, 2: 125-
139.
Therivel, R., R.M. Partidärio (1996) The practice 0/ strategie environmental assessment, Earthscan,
London.
Turner 11, B.L., D. Skole, S. Sanderson, G. Fischer, L. Fresco, R. Leemans (1995) Land-Use
and Land-Cover Change, Science/Research Plan, IGBP Report No. 35/HDP report No. 7,
Stockholm and Geneva.
Walker, R. (1993) Deforestation and economic development, Canadian Journal of Regional
Science, 16(3): 482-497.
Wathern, P. (1988) Environmental Impact Assessment, Routledge, London.
WCED (1987) Our common future, World Commission on Environment and Development,
Oxford University Press, New York.
World Bank (1992) World development report: development and the environment, World Bank,
Washington D.C.
World Research Institute (1987) World reSOLlrces 1987, Basic Books, New York.

\3
 Part I
Problem structuring for
land-use management
Conflicts between conservation goals and land use for
exotic forest plantations in South Africa

Theodor J. Stewart and Alison Joubert

Department 0/ Statistical Sciences
University 0/ Cape Town, South A/rica

Abstract
South Africa has few natural forests and many commercial plantations of exotic trees have been
planted along mountain slopes in the east of the country. These plantations are of considerable
economic value, but are damaging large areas of the natural ecosystem, while also disturbing water
run-off patterns, and thus downstream river flows. This paper describes the application of
multicriteria decision analysis (involving key role players in workshop settings), to the problem of
obtaining an acceptable balance between socio-economic and conservation goals in the allocation of
land to forestry in one particular district. Over aseries of three workshops, a value tree of criteria
was established, and six detailed policy scenarios for the district were developed. The use of simple
multiattribute scoring systems using "thermometer scales" found broad acceptance amongst a wide
diversity of participants, and served as a means of communication of values between groups. The
study is still on-going, but a number of insights have already been achieved, holding promise of
eventually reaching a high level of consensus.

Keywords: multicriteria decision analysis; forestry.

1. Introduction

In much of the world, a serious environmental concem is that of deforestation, either as a

result of over-exploitation in the forest products industry, or as a result of deliberate
forest clearance for other land uses. In South Africa, an entirely opposite problem arises.
There are very few natural forests, and the demand for timber and wood pulp has led to
the establishment of extensive commercial forests in regions (largely on the eastern
escarpment at the edge of the inland plateau) which are naturally afro-montane
grasslands. There is thus areal danger that large portions of the grass land biome, which
is threatened throughout Africa, may be destroyed entirely.
The situation is further complicated by the socio-political situation in the country.
Demands on the state to rectify rapidly large economic inequities in the population,
coupled with the still developing world nature of the economy, has led in some quarters
at least to the perception that conservation iS an unaffordable luxury, unless it can
generate income through eco-tourism for example.
Furthermore, the recent political changes in the country have generated the
expectation that all regional development (i.e., land use) decisions will be taken only
after extensive consultation with all interested and affected parties, some of whom are

17
poorly educated and unsophisticated in their understanding of political. economic and
environmental processes.
The work reported in this paper arose from plans to extend commercial forests in the
northern part of the Eastern Cape province (the Maclear district. stretching
approximately 80 km along the foothilIs of the Drakensberg mountains, as iIIustrated by
the map in Figure 1), in an area which typifies the conflicts which often arise.

1.1. Conservation context: the afro-montane grassland biome

The Maclear distriet is at the southern end of the Eastern Mountain "hotspot" of plant
diversity, one of eight reeognised for southern Afriea. About 30 % of the plant species
are endemie and about 5 % of the "hotspot" is fonnally eonserved, almost exclusively at
the northern end (Cowling and Hilton-Taylor 1994). Afforestation, overgrazing and
inereased erop-farming are among the main threats to afro-montane grasslands
throughout Afriea, leading to its identification as one of the three most threatened
habitats in Afriea. In response to this, the World Wide Fund for Nature (WWF) funded a
survey of the little known fauna of the afro-montane grasslands in the area, whieh was
earried out by the Department of Nature Conservation of the University of Stellenboseh
(Annstrong et al., 1994). This study paid particular attention to the minimum needs for
ensuring survival of faunal eommunities, and its implications for the present study are
mentioned briefly in Seetion 3.

1.2. Socio-economic and political context

The ehange from predominantly eattle grazing to eommereial forestry ehanged the
economie and socia! strueture of the area eonsiderably. This needs to be seen in the
eontext of the following: eeonomic pressures had led many Maclear fanners to seIl, as
only larger farms seemed viable after a prolonged drought; the Eastern Cape has the
second highest unemployment figures in the eountry (around 45%); and the relatively
wealthy Maclear distriet is bordered on the east by the Transkei where population
pressures, overgrazing and erosion are extreme. Politieal ehanges in the region have seen
the Maclear distriet loeal eouncils pass from the eontrol of eommercial fanners to the
Afriean National Congress (ANC), who were seeking rapid upliftment of the previously
severely disadvantaged black eommunities.

18
 Transkd

Transkd
Eastem Cape

o 20 km 40 km

Figure I. Map of South Africa (inset ) and the Maclear magisterial district. showing areas presently
owned by the commercial forestry company (shaded areas). About 60% ofthis is actually afforested.

19
1.3. Forestry context and the permit system

One of the main forestry companies in South Africa (Mondi, operating as North East
Cape Forests) had by the time of this study (1995/1996) acquired a considerable amount
of land in the region, and had received govemment permission to afforest portions of this
land, and had planted mainly pine trees (Pinus radiata) on some 32000 hectares. The
area under afforestation was, however, deemed not to be economically sustainable, as
expected timber production would be insufficient to support the operation of a pulpmill
or a large sawmill, while transportation of logs to other regions was not economically
viable. North East Cape Forests was therefore seeking permission to extend
afforestation.
Because of the exotic nature of forestry, and the potential for commercial forests to
seriously restrict run-off into public streams and rivers, owners of land are required by
law to apply to the Department ofWater Affairs and Forestry for permits to plant forests
on their land. The functioning of the permit system had recently changed to allow for
representations from all affected parties which are reviewed by the multi-party
Afforestation Review Panel in each province. Although primarily concemed with affects
on run-off, a full impact assessment may be recommended for each application. This
approach has slowed down the operating of the permit system and also means that small
growers may potentially face the very high costs of funding the impact assessment. For
this reason, a govemment Green Paper identified the need for a more streamlined
approach to the issuing of permits, which, while still allowing for participation, does not
imply such large costs. Our study should be seen as contributing towards this aim.

1.4. Approach

In this paper we describe aseries of workshops (or "decision conferences") aimed at

facilitating the process of land-use decisions in the region, in such a way that the various
interested parties could be convinced that their concems had been adequately addressed.
To a large extent, the approach adopted was the use of multiple criteria decision making
("MCDM") tools, and in particular the use of additive preference models in the spirit of
"SMART' (Simple Multi-Attribute Rating Technique, as in von Winterfeldt and
Edwards, 1986, Chapter 8). This is not the place for substantial discussions on the merits
of different MCDM methods, but it has been our view that the simple additive model
approach has great merit in transparency to a wide range of participants in the decision
process, and (when used with care) in guiding the construction of preferences to be
consistent with axioms of "rationality" which appear to be widely accepted as
normatively desirable as a basis for objectivity in decision making. (Some further
discussion may be found in Stewart, 1992.) The paper follows for the most part the
sequence of events throughout the process.

20
2. Identifying values and appropriate scales for decision-
making

North East Cape Forests was concerned that the processing of these permit applications
was becoming unacceptably slow and costly, and was pushing for more streamlined
procedures. This led to the fIrst of the workshops, at which the aim was to develop a
value tree (hierarchy of criteria) on which decisions regarding the desirability of forestry
on any given farm could be based, as weil as measures of goal achievement for each
criterion. It was anticipated that assessment in terms of certain criteria might be
subjective, but the aim was to identify as many objectively measurable performance
measures as possible to act as surrogate criteria. For such measures, we would also seek
to assess value functions that would convert the natural measures into interval preference
scales (to justify additive aggregation across criteria) much along the lines of Keeney and
Raiffa (1976), Chapter 6, or von Winterfeldt and Edwards (1986), Chapters 8 and 9.
This fIrst workshop took place in November 1995, and involved representatives of
North East Cape Forests (Mondi), the provincial Department of Agriculture and Land-
Use Planning, the Department of Nature Conservation and Forestry of the University of
Stellenbosch (who had been responsible for the execution of the WWF study) and the
Division of Environmental Technology of the South Mrican Council for Scientific and
Industrial Research (CSIR). After various groups had presented position papers, a
relatively unstructured brainstorming session was used to develop an initial hierarchy of
criteria (or value tree), in which the primary interests and goals were sub-divided into
operationally meaningful criteria (but without at this stage associating any measures of
relative importance with each criterion). Once the basic hierarchy had been established,
ranges of potentialoutcomes for each criterion (i.e., taken across the range of farms
under consideration) were identified. The mid-value splitting technique was used to
estimate value functions (of a piecewise linear form) on a number of these criteria.
The process of establishing ranges and value functions for the criteria led to a number
of debates within the group which gradually revealed a serious practical problem in
setting a justifiable procedure for decisions on a farm by farm basis. This was the (in
hindsight obvious) problem of incrementalism. For some of the key criteria, such as
conservation of endangered species and preservation of traditional ways of life, the
impact of transforming agricultural, grazing or undeveloped land into forests was
minimal while substantial areas remained unforested, but at a certain point the impacts
become critical. It becomes impossible, therefore, to apply systematic decision support
methodologies at the level of single farms without consideration of what is to be done
with other farms in the region. In fact the real danger is that such procedures are applied
in an incremental manner (one permit application at a time): initially some criteria may be
found to playa minimal role, only to fmd at some later stage (when earlier decisions are
already irreversible) that their combination is highly undesirable relative to other
combinations of decisions that might have been made earlier.
The conclusion from the fIrst workshop, arrived at by the discipline of applying quite

21
standard multi-criteria decision analysis tools. was that decisions had to be made at what
came to be called the "meso" level; i.e .• considering land use and forestry development
across the entire Maclear districl. It was recognized that the results might eventually
have to be subjected to a review at the "micro" level (at which local water rights. for
example, might determine the precise locations at which forestry was to be permitted),
but it became c1ear that the "meso"-level decisions had to be taken fIrst, and that the
tools of decision analysis should be adapted to support such decisions in the fIrst
instance.

3. Scenario-based policy planning

The conclusion emerging from this fIrst workshop corresponded c10sely with fmdings
from our previous research on applying multiple criteria decision-making concepts in
water resources planning in South Africa. These ftndings, documented in Stewart and
Scott (1995), were that decision analysis at the regionalland-use planning level has to be
based on a policy scenario principle. By "policy scenario" we mean a description of a
potential set of policy actions in the region which is:

(a) sufficiently comprehensive in description, so that decisions between policy

scenarios can be made more-or-less independently of decisions elsewhere; and
(b) defmed at a just sufficient ("requisite" in the sense of Phillips, 1988) level of detail
so that interested and affected parties can differentiate between them in the sense
of being able to provide a preference ordering according to each point of view (or
criterion) which may be identifIed.

Policy scenario thinking is allied to conventional scenario planning concepts (cf.

Ascher and Overholt, 1983), but differs in that the emphasis is on different policy
directions rather than on exogenous future circumstances or environments. No attempt is
made to be fully comprehensive (enumerating all possible policy variations, or all
possible futures), the aim being rather to be representative of a broad range of
possibilities. In practice, the constraints of the "magic number 7 plus or minus 2" (MilIer,
1956) limit the number of scenarios that can be considered at a time to about 7. It is
nevertheless important at the outset to ensure that the coverage is sufficiently broad so
that no parties feel marginalized. In most cases, the process will be iterative in that the
initial set of policy scenarios may include a very wide range of options, but with
relatively little detail (since ordering of these options according to each point of view
may be quite straightforward). As one or two dominant scenarios begin to emerge, the
others will be discarded and these will be refIned and subdivided with greater levels of
detail. The assessment of the consequences of each scenario may involve the use of a
variety of models (e.g. hydrological models, economic models), use of expert groups,
and other traditional methods of project assessment, such as environmental and social
impact studies and cost benefIt analyses. Thereafter, the decision analysis phase of

22
scenario based policy planning (see below) permits value judgements to be incorporated
which recognize changing marginal values ("utilities") at different levels of goal
achievement, and differing relative importance of the criteria, in a manner which is
transparent and justifiable.
Thus, the final decision-analysis stage is to present the scenarios (now including the
assessed consequences) to decision conferences, at which the various parties can identify
the criteria according to which the scenarios ought to be compared, and carry out the
relevant comparisons. In the next section we shall describe in more detail the methods we
have used for this final stage, but for now we simply record that we have used the VISA
software (Visual Thinking International Ltd.) for this purpose. VISA implements in
effect something akin to the SMART method (e.g. von Winterfeldt and Edwards, 1986,
Chapter 8) for constructing additive preference models. Specific advantages of the VISA
software in this context are (i) that the scoring of policy scenarios in terms of each
criteria can be done by direct assessment on "thermometer scales", or by constructing
piecewise linear value functions on objectively assessed attributes; and (ü) that
aggregation across criteria is coupled with extensive sensitivity analysis. The
"thermometer scale" is obtained by arranging the scenarios along a 0-100 scale in order
of increasing preference (in terms of the specific criterion under consideration), in such a
way that the intervals, or "gaps" between scenarios are representative of relative
strengths of preferences, as illustrated in Figure 2. Piecewise linear value functions are
obtained in a similar fashion by mid-value splitting; i.e., by consideration of the relative
strengths of preference (or "gaps") for shifts between the lower end and mid point of a
range of values, and between the mid point and the upper end of the range respectively.

Best
100 .4--- II

80
"'''--0
..
60

40

20
...-----j@
o .4,--- III
Worst

Figure 2. Example of athermometer seale.

23
3.1. Scenario development

In the present context, the policy scenarios were defined primarily by the extent of
forestry permitted in the region, and by the associated primary processing (i.e., the
construction of saw mills or pulp mills in the region). Since it is not practical to return
land under forestry to its natural state (at least not within decades and without
considerable expenditure), the one extreme of policy scenarios had to be the status quo,
with approximately 32,000 ha under afforestation. At the other extreme, surveys by
North East Cape Forests had indicated that the maximum amount of land in the Maclear
district on which forestry is possible (taking into account soil types and depth, and
rainfall) is about 55,000 ha. The current level of afforestation could support a sawmill,
but only of a size considered by the company to be weH below the economic optimum.
The option preferred by the company was to build a pulp mill, requiring 60,000 to
100,000 ha of forests, which would require additional plantations in neighbouring areas
such as the Transkei. The introduction of this option did mean that choice between
scenarios would not be entirely independent of decisions elsewhere. Participants
accepted, however, that this would not be a serious hindrance to the implementation of
scenario-based policy planning, as these developments in the Transkei appeared likely to
take place whatever happened in the Maclear distriet
Before any evaluations or assessments can begin, the scenarios need to be amplified to
include as much information on the consequences of each set of policies as is available.
In the case of the afforestation scenarios indicated above, these consequences would
include the following.

(l) Economics: Information would include the value of the wood produced and its
value after primary processing, as weil as the volume and value of agricultural
products (an important competing land use), ensuring that comparable statistics are
used in each case. Information on the impacts on secondary industries is also
desirable, but at the time of this study was not readily available, and had to be
subjectively assessed during the workshop.
(2) Conservation: The WWF study identified 16 land types in the region, based on
considerations of altitude, geology and rainfall. A list of butterfly, grasshopper,
bird and small mammal species, and the total number of species and/or endemie
species was associated with each land type. It was considered that a combination
of 10 of these land types, each of a minimum area, would by complementarity
adequately conserve the faunal community at the district level. The areas of each of
the land types which remained untransformed formed a valuable measure of
conservation impact. Also important was the extent to which these land types
remained connected (by undeveloped "corridors"), but this had to be adjudged by
the relevant expert in the workshop.
(3) Hydrology: A study was commissioned by North East Cape Forests and the Water

24
 Research Commission to assess the effects of plantation forests on the quatemary
catchments in the district. This study produced figures showing reductions in mean
annual runoff and low flows with present levels of afforestation over a rotation,
and the effects of increased levels of afforestation (up to about 45,000 ha) over the
same time period. This information, together with the judgement of two of those
responsible for the study, formed the basis of water quantity impacts for the
scenarios.
(4) Social and employment: Information on changes in employment numbers and
access to services (such as housing, water, schools and clinics), was particularly
important given the rural upliftment, reconstruction and development objectives of
the new government. Some of this was available from Central Statistical Services
and from North East Cape Forests employment records, but information on
secondary and multiplier effects was unavailable, and this tumed out to be a crucial
problem (see below).

In this study, the relevant information was gathered from Central Statistical Services,
the Department of Water Affairs and Forestry, North East Cape Forests records, data
bases and maps, telephone interviews of a saw-mill and a pulp-mill elsewhere in the
country, and the WWF study.
Subsequent to the first workshop, inputs from North East Cape Forests and the
Department of Nature Conservation of the University of Stellenbosch were combined
into five policy scenarios, representing a practical series of increments in land area
allocated to forestry, from the status quo to maximum possible afforestation. The
provincial agricultural and land-use planning departments were unable to contribute to
the development of these scenarios at this stage, however, which Iimited the extent of
detail which could be incorporated. The scenarios included approximate information
regarding the geographic distribution of afforested areas and corridors available for
conservation, but here, too, the level of detail would turn out to be inadequate (see next
section).

4. Scenario and value tree construction

4.1. First phase

The second workshop was conducted in a formal decision (electronic meeting) room
established within the Centre for Information Systems Research at the University of Cape
Town, and included representatives of the national Department of Water Affairs and
Fores try, the provincial department of Nature Conservation, and the mayor of the
largest town in the district (Maclear), apart from those who had been represented at the
first workshop. The decision room environment provided the opportunity to link MCDM
concepts with electronic brainstorming methods, which were feIt apriori to offer the
opportunity for facilitating the idea generation phase of the discussion. Once the

25
scenarios had been presented to parttctpants, use was made of the GroupSystem~
software (Ventana Corporation. 1994) in order to generate a broad list of criteria which
needed to be taken into consideration when selecting between scenarios. This was
achieved by posing the following question to the group: "What points oj view. in te rests
or other issues need to be taken info consideration when selecting between jorestry
development scenarios?". The GroupSystems -software allows participants to enter their
ideas at their own terminals simultaneously and anonymously into a !ist of ideas
contributed by the rest of the participants. The complete list of ideas appears on all the
participants' screens, who can further comment on any of them. The GroupSystems
software was useful in that it allowed freedom of expression, and prevented the
domination of the early stages by more forceful personalities. However, as a result of the
widely divergent range of backgrounds of participants, those who lacked computer or
typing skills were placed at a serious disadvantage. (Later reflection suggested that this
problem rnight have been overcome to a large extent by abrief period of tuition prior to
the workshop.)
The brainstorming session produced a rich !ist of "points of view, interests and other
issues", which were then grouped by the workshop participants into the seven categories,
representing the interests of agriculture, water, people, tourism, forestry, conservation
and regional development. Some of the issues pertaining to each category are
summarized in Table 1. These issues contain the seeds of a set of criteria according to
which the scenarios can be compared and evaluated. Attempts were made at the
workshop to rank the scenarios on thermometer scales for a selection of the issues
represented in Table I, but two problems arose:

(1) The issues were insufficiently defmed, and lacked the necessary preferential
independence properties, to allow for a substantial level of consensus on the
rankings (and in this sense were not a useful set of "criteria" for evaluation of
scenarios); while
(2) it became apparent that the scenarios were not specified in enough detail to allow
the participants to discriminate between them for each criterion. For example, a
substantiallevel of controversy arose over how much of the land allocated to the
forestry industry would in fact remain untransformed, and thus still serve some
conservation purpose as corridors. To some extent this represented a lack of trust
between different interests represented, and it is clear that the scenario defmitions
need to cater for such dis trust.

In spite of the above problems in applying formal multicriteria decision analytical

tools, the participants were prepared to assess the scenarios holistically (from the point
of view of their own interests) on thermometer scales. The interesting observation from
this exercise was not so much that all participants found this a meaningful exercise (as
we had demonstrated the feasibility of the approach in Stewart and Scott, 1995, for
example), but rather that many participants clearly tried to consider other perspectives in
forming their scales, favouring compromises between extreme options. The result was an

26
identification of one of the scenarios as a potential starting point in the search for good
compromises. This scenario had only slightly less than the maximum level of
afforestation, but retained a fairly large corridor of untransformed grassland between the
western and eastern sides of the district. This scenario was rated highest by three of the
participants, second lowest by two, and somewhere in-between for the remaining five
participants. All the other scenarios were ra ted worst by at least one participant, and this
was of itself a useful insight for future deliberations.
Although the above outcomes were encouraging, it was clear that further work was
necessary in refining the scenarios and in turning the set of issues in Table 1 into a value
tree (or hierarchy of criteria).

Table 1. Summary ofthe categories defined in (he second workshop.

Agrieulture Water People Tourism Forestry Conservation Regional

------- ------,-_.- ._-------- __ __ c _

__ .~ ______ ~ __~~eJ?E'!.l~lIt__
Economie Effeeton Employment Economie Economic Untransforme Economie
value water supply and ineome value value d land area value of
(mean annual generated secondary
run-off; peak industries
and low flows)
Viability of Effeeton Stability of Numberof Thresholds Numbersof New wealth
small seale water d:mands income tourist sites necessary for a land types brought to
farming available viable i1dustry maintained in region
viable a-eas
Regional food Effects on Diversity of Numberof Sustainability Contiguity and Regional
produetion downstream employment tourist ofindustry geographie infrastruetural
users activities spread of land development
available types
Fire hazards to Effeet on Poverty I~els Stability of Degradation of Gross
farms water quality industry soil geographie
(sill and produet
ehemieal
loads)
Potential for Quality of life
good (services,
management personal
freedoms)

4.2. Second phase

In see king to refine the scenarios, one of the authors (AJ) visited the Maclear district, in
order to gather some of the relevant information first hand. Meetings were held with the
provincial agricuiture and land-use planning department, North East Cape Forests, and
the mayors and town clerks of the two towns in the area (Maclear and Ugie). A limited
number of farmers and individuals from the towns were also interviewed. The Forestry
company was able to supply details of the farms on which they had purchase options
pending, and where they would still like to expand further. The company also supplied
detailed maps of the actual arrangement of plantation blocks within their presently
afforested farms. It was at this stage that the possibility of developing forestry in adjacent

27
districts of the fonner Transkei area (which had been operated as an "independent"
homeland area under the previous South African government), as a community forestry
system in partnership with the company, was induded in some scenarios. This enabled
the forestry industry to achieve the scenario of around 100,000 hectares under
afforestation needed to justify the establishment of a large pulp-mil!. Initiatives were
already underway to negotiate with the tribai authorities and civic organizations in these
areas. Cognizance was taken of this in the relevant scenarios, although impacts outside
of the Maclear district were not directly assessed.
Although it was hoped to enter the relevant infonnation into a GIS fonnat, the
problem obviously lending itself to spatial analysis, this was not possible as insufficient
data was available in this fonn. As much infonnation as was possible was, therefore,
entered into a Microsoft Excel spreadsheet. Each farm and its surface area was linked to
an associated land type, based on altitude, rainfall and underlying geology. Each land
type was in turn linked to a !ist of associated grasshopper, butterfly and bird species. The
average pay and number of workers per hectare, on forestry and agricultural land in the
area, were determined from the State Central Statistical Services data for the district, and
from employment records of the forestry company, and linked to the areas of the fanns.
Tbe farms presently afforested were designated as such and the other scenarios were
entered into the Excel's "scenario" function. Thus, in order to change from one scenario
to another, it was necessary only to dick on the relevant scenario within this function.
Graphics and statistics were then appropriately updated. The link of each fann to aland
type and the associated species data, meant that the number of species represented at
least a certain number of times, in at least a minimum area could be calculated. The
number of times each land type occurred in an unafforested state was calculated, as was
the total afforested and unafforested area of each land type. Most of the unafforested
land remains as grassland and is used for grazing. Some land-use infonnation was,
however, available for the catchrnents in the area, which gave the percentage crop cover
and this could be used to modify the calculations to give an idea of untransfonned area
available for conservation. Tbe spreadsheet was available at the third workshop, so that
many questions could be addressed on-line.
The six scenarios emerging from this exercise cannot be described here in detail, but
can be summarized as folIows.

Scenario 1: Approximate Status Quo: Approximately 35 000 ha of

afforestation, with one timber saw mill to be constructed in the
region.
Scenario 2: Expansion of forestry to take up existing options on farms
(expanding forestry to 44 000 ha); saw mill to be constructed.
Scenario 3: Afforestation of all suitable land except for a small ecologically
sensitive area to the north of the region (giving 50 000 ha of
forests); saw mill to be constructed.
Scenario 4: Maximum afforestation (to 53 000 ha) plus the saw mill.

28
Scenario 5: Afforestation as in scenario 4, but with further afforestation in the
Transkei, with both forests feeding a pulp mill set up in the
Maclear district
Scenario 6: Afforestation as in scenario 4, but with timber transported to a
pulp mill in another region.

The issues raised in Table I, were converted into the value tree displayed in Table 2.
In constructing the value tree, consideration was given to both the operational
meaningfulness of the lower level criteria (Would participants be able to rank scenarios in
terms of these concepts?), and the likely preferential independence of the criteria (Would
participants be able to do the required rankings on one criterion independently of level of
achievement on the other criteria?). While it is difficult to ensure absolutely that these
two considerations are fully met, the value tree met with general approval at the third
workshop. Ranges (from ideal to worst cases) were identified for each of the lowest level
criteria, and the VISA software was used to score the scenarios in terms of a number of
these criteria, either directly on thermometer scales, or by assessing value functions in
cases where the criteria were associated with objectively determined measures.

Toble 2. Value tree (hierarchy %bjectives).

Regional Quality of Life Sodal Employment NPV of income
Number of jobs
Housing and services
Personal welI-being
Economic Agriculture NPV of agricultural production
Viability of small farms
Local food production
Forestry NPV of forestry output
Sustainability of forestry
Tourism potential
Secondary IndlStry NPV of secondary industry
Development of infrastructure
Environmental Conservation Number of land types preserved
Untransformed land area
Contiguity
Degradation
Water Quantity MAR
Lowflows
Peak flows
Quality Silt load
Chemicalload

At the time of our writing this article (November 1996), this work was still on-going.
The third workshop went a long way towards creating understanding between parties as
to their concems, and scores for the scenarios were obtained for a number of the criteria
in Table 2 which can serve to inform the later debate. In spite of this. and although

29
certain conclusions appeared to be emerging (for example that scenario 6 was to be
avoided, and that scenario I was an inefficient use of resources, probably unsustainable
in the long run), there remained unresolved issues. A key uncertainty related to the
employment that would be generated by the establishment of a pulp mill in the area.
Direct employment in the mill would be relatively smalI, but wide ranges of claims were
made for the multiplier effect of bringing relatively high technology industry into the
area. More detailed economic models are necessary to resolve this issue, which could
have a major effect on whether scenario 5 is to be preferred to scenarios 2, 3 or 4, and it
is this work which is continuing.

5. Conclusions
Although the project is at time of writing still incomplete, a number of lessons have been
learnt. Perhaps first and foremost of these is that even in issues as contentious as
conflicts between industrial and economic development and conservation, the use of the
formal tools of multi-criteria decision analysis can play a vital role in facilitating progress
towards understanding and consensus. Apart from this primary conclusion, it is useful to
document a few specific lessons which can assist in undertaking similar studies in the
future, as folIows.
(1) The use of formal decision analytic tools can reveal points at which the apparent
scope of the decisions to be made are too restrictive. It was precisely because we
sought to identify the relevant criteria and to score alternatives against these, that it
became evident that this would lead to incremental planning, and that a more
holistic systems view was needed.
(2) Relatively unstructured brainstorming sessions can generate useful sets of criteria
for evaluating alternative policy scenarios, but it requires a substantial amount of
time and knowledge of the underlying principles of value measurement (in
particular the understanding of and need for some form of preferential
independence) to structure these into a form which allows for meaningful
assessment by participants.
(3) The policy scenario approach (in which it is accepted that any policy
recommendation is no more than a "scenario", leaving many implementation details
unspecified), is an efficient manner in which views of divergent parties can be
brought into the planning process. It is, nevertheless, important to ensure that
crucial details, which can lead to controversy, are identified and the resulting
consequences evaluated in an objective manner before value judgements are
applied. This, too, can take considerable time, and may involve lengthy interviews
with participating parties outside of the formal workshop structure.
(4) Given the diversity of backgrounds of participating parties, any procedures used
for evaluating goal achievement according to specific criteria and aggregation of
these need to be simple, transparent and seen to be fair by all parties. The SMART
approach, as implemented in the VISA software, for example, linked to sensitivity

30
 analyses does seem to satisfy this requirement. In the studies reported here,
participants ranged from PhD level conservation scientists to representatives of
local communities with quite low levels of formal schooling, but all appeared able
to relate to the procedures being used (reinforcing the conclusions reached in
Stewart and Scott, 1995).

References
Armstrong, A.l, H.1. van Hensbergen, H. Geertsema (1994) Evaluation of afforestable montane
grasslands for wildlife conservation in the North-Eastern Cape, South Africa, South AJrican
ForestryJournal, 171: 7-20.
Ascher, W., W.H. Overholt (1983) Strategie Planning and Forecasting, lohn Wiley & Sons, New
York.
Cowling R. M., C. Hilton-Taylor (1994) Patterns of plant diversity and endemism in southern
Africa: an overview, Proceedings oJ a conJerence on the conservation and utilisation oJ
southern AJrican botanical diversity, Cape Town, September, 1993, Strelitzia, I: 31-52.
Keeney, R.L., H. Raiffa (1976'fJecisions with Multiple Objectives, Wiley, New York.
Miller, G.A. (1956) The magical Number Seven Plus or Minus Two: Some Limits on our Capacity
for Processing Information The Psychological Review, 63: 81-97.
Phillips, L.D. (1988) People-Centred Group Decision Support, in G. Doukidis, F. Land, G. Miller
(Eds.) Knowledge Based Decision Support Systems, Ellis Horwood, Chichester, 208-221.
Stewart, T.1. (1992) A critical survey on the status of multiple criteria decision making theory and
practic~ OMEGA The International Journal oJ Management Scienc~ 20, 569-586.
Stewart, T.1., L. Scott (1995) A scenario-based framework for multicriteria decision analysis in
water resources planning, Water Resources Research, 31: 2835-2843.
Ventana Corporation (1994) Group Systems Jor Windows: Quick ReJerence, Ventana Corporation,
Tueson, Arizona.
Von Winterfeldt, 0., W. Edwards (1986) Decision Analysis and Behavioral Research, Cambridge
University Press, Cambridge.

31
Structuring the process of choosing rice varieties at the
south ofBrazil'

Carlos A. Dana e Costa\ Leonardo Ensslin 2 and Alessandro P. Costa2

JIST-CESUR, Technical University 0/ Lisbon, Portugal
lFederal University o/Santa Catarina - EPS, Florianopolis, Brazil

Abstract
Today many varieties of rice seed can be found all over the world. This paper describes the
structuring phase of the decision-aid study conducted by the authors, in strict interaction with
rice producers in the south of Brazil, in order to help them solve the complex problem of
choosing the most adequate rice variety for seeding on a specific fann. The paper iIIustrates the
usefulness of integrating cognitive mapping and multicriteria structuring concepts when
practicing decision-aid. Cognitive mapping particularly facilitated the leaming process of
identifying and operationally describing the fundamental points of view for the evaluation of
alternative rice varieties.

Keywords: Multicriteria decision aid, structuring, cognitive mapping, land-use

management.

1. Introduction

The great majority of textbooks and articles reporting examples and cases of application
of multicriteria methodologies usually concentrate on evaluation. Structuring is ignored
or only superficially addressed, and options, criteria, and impacts are presented as pre-
specified. By contrast, this paper is almost totally confined to structuring (the process of
choosing rice varieties in the south of Brazil), in order to emphasize the complexity
involved in approaching ill-structured decision-making problems.
Sections 2 to 6 are devoted to the interactive and constructive structuring process.
We first describe the case and its decision environment, including the identification of
the decision-maker, his main objectives and the roles of the other actors. Section 3 is
concerned with the framing of the problem, including the screening phase that led to the
identification of the 14 alternative rice varieties considered in the study. Section 4

I This articIe is taken from a study (see Costa, 1996) carried out in the framework of an on-going joint

action-oriented research project between the Federal University of Santa Catarina - EPS and the Center
of Urban and Regional Systems - CESURIIST of the Technical University of Lisbon. Tbe project is
devoted to the integrated use in the decision-aiding of complementary contributions of various
multicriteria schools ofthought (see, von Winterfeldt and Edwards, 1986, Bana e Costa, 1990, Keeney,
1992, Roy, 1996, Larichev and Moshkovish, 1997) and the so-called Problem Structuring Methods (see
Rosenhead, 1989), under the methodological framework proposed by Bana e Costa (1992) and in line
with what has been advocated by a few authors (such as Belton et af., 1997).

33
addresses the cogmtlve mapping component of the leaming process of structuring.
closely linked to Section 5 in which the fundamental points of view for evaluating the
rice varieties are presented, and to Section 6 dedicated to the operational description of
those points of view. A1though the evaluation phase is not the main subject of this paper,
some highlights of the construction of the evaluation model are given in Section 7. Final
considerations and some conclusions are presented in Section 8.

2. The case and its decision environment

Brazil is the most important country of Latin America for the production of rice, with a
national production of 7.5 million tons (1990). The State of Rio Grande do Sul alone is
responsible for more than 40% of that total. Ricardo Gonralves da Si/va is an important
rice producer in Rio Grande do Sul. He is recognized as aleader by the other farmers in
the region and is considered a "benchmark" due to his remarkable performance in the
rice business. Ricardo is the owner of Säo Francisco, a big Farm of 3,500 hectares where
1,300 hectares of rice are cultivated each year. The Farm is located in the very south of
Brazil, at the boundary with Uruguay, near the city of Jaguaräo. Its limits are the Juncal
and Jaguaräo rivers and the Mirim lake. In addition, the farm includes an industrial unit
where rice is processed and prepared for commercialization.
The purpose of the study was to help Ricardo fmd an answer to the following
question: Among the many varieties of rice seed that can be found all over the world,
which of them is the most appropriate for Säo Francisco? This is a fundamental strategic
question'because a good choice of seed is crucial to high productivity. Moreover, rice
producers know from experience that such adecision is by no means an easy one.
Characteristics of the soil, weather conditions, time of seeding. availability of water, and
the morphological characteristics of each variety are but a few examples of the many
factors that can influence productivity. Market requirements, investments and human and
machinery resources are some other important business aspects that cannot be neglected
when selecting a rice variety to be seeded. Yet, traditionally, rice producers decide by
intuition.
Ricardo knows very well that this attitude is no longer compatible with the high
competitiveness of global economics that Brazilian agriculture now faces. As aleader, he
also knows that, ifhe embarks upon a new way of doing business, others will follow. For
these reasons, he is drawn to a new decision support methodology. Ricardo, as the
decision-maker, presented the problem as "to choose the variety that provides best profit
with small risks". Clearly, "maximize profit" and "minimize risks" are the strategic
objectives he wishes to attain. Two other farmers played. together with Ricardo, the role
of experts in the commercial area and in the industrial process. Moreover, a consultant in
agronomy provided bis expertise. Workers at Säo Francisco, commission-agents, rice
purchasers (as third parties) and ourselves (as the analysts or facilitators), complete the
cast of actors in this case. Last, but not least, it is important to point out that. although
constructed for the specific Säo Francisco conditions and under the value system of its

34
owner, the decision-aid model should be conceived to be easily transferable to other
farms, farmers, and future wider ranges of rice seed varieties.

3. Framing the problem: screening and identification of

options

In a frrst interview, the three farmers and the consultant were invited for an informal
discussion about the concerns intuitively taken into consideration by a rice producer
when analyzing a certain rice variety. They were frrst asked to think individually about all
the aspects that from their viewpoint most influence the achievement of the two broad
objectives "maximization of profit" and "minimization of risks". Afterwards, a group
debate gave rise to the elaboration by the facilitators of a large and unstructured list of
"primary evaluation elements" (Bana e Costa, 1992), many of them poorly defmed or
vaguely specified. The two concerns "profit" and "risk" topped the others.
The group was then confronted with the list, and asked to focus on those and other
possible elements in the list that enable the elimination of rice varieties obviously
inadequate for seeding in the region (for example, all those incompatible with Säo
Francisco soil or water conditions). Using such elements as screening criteria, the
agricultural expert easily identified a subset of 14 rice varieties deserving careful
comparison, hereafter designated by their usual technical acronyms or market labels:
EEA-406, IRGA-409, IRGA-41O, IRGA-412, IRGA-413, IRGA-414, IRGA-416,
IRGA-417, BLUEBELLE, EL PASO-I44, EMBRAPA-6, EMBRAPA-7, EMBRAPA-
38, and EMBRAPA-39. Ricardo accepted them as his options for seeding in Säo
Francisco, and therefore they constitute the "set of alternatives" in our decision-aid
study.
After this initial screening phase had taken place, the study had to determine whether
one of these varieties would appear significantly better than the other 13 options from
Ricardo's viewpoint. Thus, the technical decision-aid issue, from this stage onwards, was
to develop a selection procedure that could be used repeatedly to identify the best
options. As said before, the evaluation model should be as general and transferable as
possible - a good reason for adopting an approach in line with Keeney's (l992) "Value-
Focused Thinking".
Following this, all the primary elements that did not discriminate among the options
were deleted from the list, as weil as those considered redundant or not sufficiently
relevant for comparison. This resulted in a new "cloud" of 27 primary evaluation
elements (cloud in the sense that its boundaries were not stable, the value meaning of
each element was not yet weil captured by the involved actors, and possible relations
among elements were not formalized). Each participant was then invited to analyze the
list individually before meeting again.

35
4. Cognitive mapping': structuring the unstructured c10ud of
primary evaluation elements

A new session cornmenced by discussing more fully the elements in the list and checking
if their meaning and relevance were weil perceived by the four involved actors. This gave
rise to arevision of the list, for some of its elements were differently or more precisely
described, in order to avoid ambiguity, whereas others were abandoned, decomposed,
grouped, or replaced by new ones. Next, we questioned Ricardo as to whether in his
opinion the list could now be considered exhaustive. The question stimulated more
discussion, some obvious relations between elements naturally appeared and a few simple
clusters started to take form, giving rise to a fIrst family of 34 "points of view".
Following the ''points of view structuring approach" proposed by Bana e Costa
(1992), a point 0/ view (PV) is any element within a specillc decision context which at
least one actor considers relevant to the evaluation (ofrice varieties, in our case), and the
value meaning of which is weil perceived by all the involved actors. In this sense, stated
objectives and concerns of the actors, on the one hand, and active characteristics or
attributes of the options and possible consequences of potential actions, on the other, are
all points o/view (PsV) provided that theyare sufficiently weil defmed and do not give
rise. to ambiguity and miss-understanding. This is why it is important to distinguish
between an ill-defIned "primary evaluation element" and a "point of view". The notion of
a PV corresponds with the concept of a ''value dimension" for von Winterfeldt and
Edwards (1986), and with the broad notion of an "objective" for Keeney and Raiffa
(1976) - except that, unlike Keeney's (1992) defInition of an objective, there is not (yet)
necessarilya "direction of preference" associated with a PV. Moreover, we agree with
von Winterfeldt and Edwards (1986) when they state that "the word objective is
ambiguous". More fundamentally, although astated objective of an actor is a PV (such
as "minimize risk" and ''maximize profIt"), many types of PsV can exist that do not
emerge or are not actually stated as objectives by the actors - such as the productivity
constraints related to non-biological characteristics of the rice varieties such as
"tolerance to iron" (is it appropriate to say that "maximize rice tolerance to iron" is an
objective of Ricardo?).
At this stage in the leaming process, the information was introduced into the
computer by the facilitators using the Graphics Cope software (Banxia Software Ltd.,
1995), and a fust cognitive map (Eden, 1988) took form. In our cognitive maps (CMs) -
see Figure 1 - nodes represent PsV linked by arrows indicating that one PV influences,
leads or has implications for another. Relations between PsV were identified by
confronting the actors with questions such as why is this PV important? (Keeney, 1994).
For some PsV, such as (adequacy 0/ the rice variety to the) /arm's characteristics, the
answer was simply that the PV was important because it was viewed as essential for
''maximizing profIt" or ''minimizing risk". Such a PV was then considered as reflecting a
key-concem to be taken into consideration when comparing the rice varieties. Another
response was that a PV was relevant given its implications for another PV, such as the

36
relation between the initial strength of a plant and its tolerance to invaders (in Keeney's
tenninology (Keeney, 1992) the fonner can be considered a means to achieve the end
concern(s) captured by the latter). The question was then asked for this PV in turn. This
procedure also developed a perception of causeleffect relationships among PsV
previously identified and new ones that emerged in the meantime.
After a few workshops du ring which the initial CM had been interactively updated and
progressively redesigned, the CM of Figure I was designed, representing the system as
the four actors perceived it at that stage.

2 Phos o r · 17 Irrigalion 19 Drainage

6 QrgaJc su~c.. s1ruclure structure 22 Wa1er 1emperalure ~ 34lnttial slrength

3-Potassium 5 pH 7 Nitrogen ~ ~ / 35 Toierance 10

~
I
\, / 18 Aegularityol. 20 Wa1ersuppressioi k23SodIum invaders
4H1storv01th~8Fertilityofsoil thewa1er~,na ~
sOil L 21 Wa1erquality
16 Sansi1illity tok'" 37 MI~imiza inttlal

1
tmcic elemen1s and .. ____ nsles
9 Requi ents 01 Ihe immersion 38 Assure go<ilt "-
rice variety plants 36 Toierance 10 Iow
. . \ / 40 Tolerance 10 1emperalur..
11 StandardlZ81lOn ~ / 39 Specific _ illness..
~ 1 MI·· . Ies lI(d--- cheracteristics o f -_ __

:::-~r t"F.~~e~om thr~·

~c;;;e
7 _. /
15 Soil condttions /
24 Marl<et aspecls
\ ~
41 Toferance 10
Insects
43 Cycle period 44 Plant siza

~~a~~~ /' 26 Amylase

. 31 Aisk 01 chenging 42 Toxicologyto Fe
25 Produdivity in 27 Kind 01 gram variaty

indu.:l'ri"alizalion
process
/ \~ \ "'"-- 33 Practice 01 human
resources
29 Width 32 Knowledge. about
28 Appeerance 30 L.ength lhe new variatles

Figure 1. A cognitive map of the problem of choosing a rice variety.

The map was structured by the facilitators in such a way that the relations are oriented
towards its center, where the goal "minimize risks and improve profit from rice culture"
appear involved by several groups of closely related PsV.
Cognitive mapping helped the actors involved to expand their knowledge about the
problem and to work out their differences of opinion. It facilitated the mediation of
ditYerent perspectives and lead to agreement about the fundamental aspects of the
problem. The use ofGraphics-Cope's "cluster analysis" (Eden et al., 1992) proved to be
very helpful since it enabled extensive management of infonnation, namely by its
capability to fmd groups of closely linked PsV such that the number of links with other
clusters is minimized.

5. Family of Fundamental Points of View

The structuring learning-process should progressively lead to the identification of

fundamental points 0/ view (FPs V), often clustering several interrelated elementary
points of view (EPsV). The important distinction is that an FPV - not an EPV - is a PV
that the actors desire or accept to isolate from the other PsV, as an independent

37
evaluation axis (even if they do not agree on its relative "importance"). Consequently. it
must verify the necessary preference-independence conditions to enable eeteris paribus
value judgements about the attractiveness of the alternatives.
Within such a methodological framework, cognitive mapping is an intermediate step
in the process of identifying the FPs V. Indeed, some of the key concerns identified in the
previous sessions were too general in their defmition, in the sense of being "areas of
concern" (Keeney and Raiffa, 1976) reflecting broad values of interest not considered
sufficiently specific or precise for the evaluation of the rice varieties. The concept of
"specification" (Keeney, 1992) was used in the following sessions to decompose such
PsV into more than one FPV.
For example, the value meaning to the PV (Adequaey of the riee variety to the)
farm's eharaeteristies (node 10 in the CM of Figure l) became much more clear after it
was decomposed into Adequaey to soU eonditions and Adequacy to the irrigation
structure, considered the two FPsV for the comparison of the options in that area of
concern. As can he seen in the tree structure of Figure 2, each of these two FPsV was in
sequence specified by a few EPsV, and lead to a deeper and more accurate
understanding of the concerns captured by them.
The EPs V included in the tree are those that were later used when making the
respective FPsV operational for evaluation purposes (see next section), some of them
having been specified only in that phase. The entire tree is the result of the specification
process. The fmal "family" of the II FPsV is shaded in Figure 2 (family in the sense of
heing consensual, operational, non-redundant, concise, and exhaustive) and is hereafter
described. (The particular layout of the tree is intentional to avoid its mis-interpretation
as a hierarchical structure.)

RJee VII1ieIy

Figure 2. Tree 01 Ps V, in which the FPsVare shaded.

38
 Note that the two strategie objectives initially stated by Ricardo, maximize profit and
minimize risks, cross the whoJe family of FPsV:
FPVI - Adequacy to soil conditions - The extent to which the soil requirements of a
variety (in terms of "soil fertility" and "resistance to invaders") are consonant
with actual conditions on the farm, in order to reduce risk and improve
production with minimum land development costs.
FPV2 - Adequacy to the irrigation process - The extent to which the irrigation
requirements of a variety (in terms of "characteristics of the irrigation and
drainage structure", "saltiness of the irrigation water", and "availability of
water") are consonant with the existing irrigation structure, in order to improve
production at lower water costs.
FPV3 - Sensitivity to /ertilizers - The extent to which the return of a variety (grain
qualityand production/costs) responds to phosphate-based and potassium-based
fertilizers and nitrogen-based fertilizers, with "minimum lay down".
FPV4 - Resistance to illnesses - The extent to which a variety is likely to resist fungi.
Due to weather conditions in Säo Francisco, fungi attack is a serious problem.
This implies additional costs to combat fungi and the severe damage that they
cause to production. Two EPsV are of concern: "resistance to brusone" and
"resistance to other illnesses".
FPV5 - Tolerance to non-biological aspects - The extent to which a variety is likely to
tolerate iron in the soil and low temperatures. The culture of irrigated rice in the
region is usually on soils that have a high level of iron ("Fe"), and where the rice
is often exposed to "low temperatures" (lower than 11°C). These two non-
biological aspects may considerably reduce the productivity of some varieties
that are not tolerant to them. The farmers considered these aspects the most
constraining ones for rice culture in the region.
FPV6 - Duration 0/ the biological cycle - Each variety of rice has a biological cycle,
measured in days from seeding to cropping. The shorter the cycle the more
attractive the variety is in terms of production costs, since it will demand less
water for irrigation and land use.
FPV7 - Agricultural productivity - Each variety has an expected level of production of
rice grains, usually measured in tons per hectare. Of course, the higher its
expected production the hetter the variety. Variance of productivity was
considered later in a sensitivity analysis.
FPV8 - Loss 0/ grain - The extent to which a variety is resistant to grain loss or has a
constant maturation. Each variety presents a different biological capacity to
retain the grains when they are mature. Obviously, a higher capacity to avoid
this loss of grains in the crop period is preferred.
FPV9 - Industrial productivity - During the industrialization process many grains break
when husking and polishing the rice. This phenomenon is directly related to the
biological characteristics of the rice varieties. The hest varieties are those with
the highest expected percentage of entire grains after the industrialization

39
 process. This is a very important element in setting the commercial value of the
rice produced.
FPVIO - Grain quality - The quality of rice is evaluated in terms of its dimensions
(relation length / width) and its amylase content. The best varieties are those
with the highest relation length / width and with the lowest amylase content.
These aspects determine the price of the rice.
FPVll - Risks 0/ changing - The decision maker is averse to taking risks, so he prefers
the varieties that do not require different or new tasks ("variety" EPV) and that
have a well-known history on the farm or in the region ("knowledge" EPV).
Other risk components are already implicit in other FPsV, and are therefore not
of concern in this evaluation axis.

6. Making fundamental points of view operational for. the

evaluation

As said before, an FPV should be such that it is possible to appraise the (partial)
attractiveness of each rice variety in terms of the values made explicit by that FPV. The
more objectively that such values are made explicit, the better understood (less
ambiguous) and, therefore, the better accepted (less controversial) will be the evaluation
model. Operationally, this can be achieved by associating adescriptor to each FPV. A
descriptor is an ordered set of plausible (qualitative or quantitative) impact levels in
terms of an FPV, intended to serve as a basis to describe, as objectively as possible, the
impacts of options with respect to that FPV (Bana e Costa, 1992). 2 As said elsewhere
(Bana e Costa et al., forthcoming), the fact that adescriptor of an FPV is a set of levelS
that must be ordered in terms of their partial attractiveness, is necessary to ensure the
isolability of the respective FPV. Thus, if ordinal dependencies are detected in this
phase, feedback is necessary to re-structure the initial family of FPsV. On the other end,
the adjective "plausible" emphasizes the convenience, for better framing of the problem,
of restricting adescriptor to a plausibility domain (boundaries of which usually result
from screening), thus reflecting the context-dependent nature of an FPV. Adescriptor
can be direct or natural (its levels directiy reflect effects), indirect (an indicator of causes
more than effects), or constructed (a fmite scale of reference levels, often defined by
holistic ranking of combinations of levels on several EPsV, as exemplified hereafter).
Moreover, adescriptor can be quantitative or qualitative (or even pictorial), and
continuous or discrete.

2 This is similar to the classical notion of an "attribute" of Keeney and Raiffa (1976) and fits with the
broad notion of a "criterion" as "a tool allowing comparison of alternatives according to a particular PV"
(Bouyssou. 1990). An FPV is what Roy (1996) calls a "significance axis" of a criterion. The terms
"point of view" and "descriptor" were introduced in the decision-aid literature in the remarkable (and
unfortunately not very well-known) book of de Montgoltier and Bertier (1978). See also Bertier and de
Montgolfier (1974).

40
 Of course, an operational description of a FPV is better for evaluation purposes if it is
possible to associate it with a natural descriptor. This has been the case of the FPsV: 6 -
duration of the biological cyc\e of the rice described by the number of days from seeding
to cropping, ranging from 95 (best plausible impact level) to 145 days (worst plausible
impact level); 7 - agricultural productivity measured in tons of rice grains per hectare,
ranging from a best level of 8.5 ton/ha to 5.5 ton/ha; 8 - loss of grain described in
percentage units, between 2% and 12%; and, similarly, 9 - industrial productivity in
percentage of entire grains after the industrialization process (72% to 52%).
However, attributes direct1y reflecting effects are often difficult to quantify or describe
properly. In such cases, an indicator of facts or a natural descriptor of an EPV closely
related to the FPV in question can be used as an "indirect" (or "proxy") descriptor of the
impact levels in tenns of that FPV (see Beinat, 1997, for a discussion of the pros and
cons of indirect descriptors). Inspection of cognitive maps previously bullt can be very
useful to identify appropriate factual indicators. An example of an indirect descriptor is
the appraisal of resistance to invaders by an indicator of the initial strength of the rice
plants.
Other cases exist where it is rather convenient or even necessary to construct a
descriptor of an FPV by combining impact levels of several EPsV, in particular when the
FPV clusters interdependent EPsV. Such adescriptor is usually the result of a hard
leaming and time consuming process, requiring holistic value judgements over
combinations of levels of several EPsV. This was the case for all the other seven FPsV,
as we go on to show. Nevertheless, given the wide experience of the four actors involved
in the process, and their profound knowledge of the characteristics of Säo Francisco,
they developed a level of expertise which enabled them to easily assign to the technical
data a substantive meaning in tenns ofthe PsV in question. For this reason, the defmition
of a few qualitative reference impact levels, in tenns of each of the other seven FPsV,
was considered enough. This, of course, very much simplified the structuring process,
although at the expense of an accurate (although unnecessary in the actors' viewpoint)
description of impact levels. Note, however, that hereafter all reference levels defmed
correspond direct1y to the objective proftle characteristics of the various types of rice
seed existing on the market (this is, in some sense, a kind of pictorial description).
An example is the descriptor of the FPV5 (Tolerance to non-biological aspects). The
first phase of its construction was the defmition of reference levels for each of its two
EPsV (see Figure 2): "the extent to which a variety is likely to tolerate the level of iron
existing in Säo Francisco solls", simply distinguishing two reference levels, "tolerant"
and "not tolerant"; and "the extent to which a variety is likely to resist low
temperatures", for which three levels were seen as necessary, "resistant to temperatures
< 11°C", ''tolerant of low temperatures if the period of exposure is not too long", and
''not resistant to temperatures< 11°C". (Mitigation measures, to prevent effects from
rice intolerance either to iron or low temperature, are expensive and subject to a
significant risk of fallure.) Afterwards, the levels of the two EPsV were recombined in
six levels that could be appropriately ordered by (holistic) expert judgements in tenns of
the FPV5. This gave rise to the constructed qualitative scale ofTable I.

41
 Toble 1. Constructed descriptor of the FPV5 "1O[eranee 10 non-bio[ogica[ aspeets ".
Level Description
N6 The variety tolerates the iron existing in the soils of the farm and low temperature « 11°C).
N5 The variety tolerates the iron existing in the soils of the farm and resists low temperature «
11°C) only if the period of exposure is not long.
N4 The variety does not tolerate the iron existing in the soils of the farm but tolerates low
temperature « 11°C).
N3 The variety tolerates the iron existing in the soils of the farm but does not resist low
temperature « 11°C).
N2 The variety does not tolerate the iron existing in the soils of the farm and tolerates low
temperature « 11°C) only if the period of exposure is not long.
NI The variety neither tolerates the iron existing in the soils of the farm nor resists low
temperature « 11°C).

Note that whatever the level of tolerance to iron, more tolerance to low temperature
is always better than less, and vice-versa. This reveals that, for the actors, the two EPsV
are mutually ordinal preferential independent. However, a similar conclusion is false for
cardinal independence (see next section).

7. What came next: the evaluation phase

The assignment of an impact level of each descriptor to each alternative (thus defming
the "profIle of impacts" for each rice variety) has been very much facilitated not only by
the actors' expertise, but also by the fact that the characteristics of the various varieties
are well-known from studies developed by specialized research institutions; moreover,
full information is provided for the more relevant characteristics given the production
conditions in the region (namely by IR GA - the Rice Institute of Rio Grande, and
EMBRAPA - the Brazilian Agency for Agricultural and Veterinary Research). The
defmition of the impact-profIles of the 14 options marked a "turning point" in the
decision aid process, from its structuring phase - where objectivity was a key issue - to
evaluation - where the subjective notion of relative "attractiveness" (or "desirability") of
the alternatives played a more crucial role. Of course, subjectivity was always a factor
present.
In order to quantify the attractiveness of the rice varieties in terms of each FPV, fIrst a
cardinal value function was constructed for each descriptor, and second, scaling
constants were assessed in the operational framework of a simple additive value model.
Both 100 to 0 partial value sc ales and scaling constants were interactively defIned based
on qualitative value judgements expressed either by Ricardo or the agronomist, with the
support oftbe MACBETH approach (Bana e Costa and Vansnick, 1997a, 1997b).

42
 As mentioned in Section I, we will not describe in this article the phase of the
construction of the evaluation model.] However, it is worthwhile to emphasize that
judgemental dependencies between PsV were detected in this phase. For example,
concerning the FPV5 (see Table l) Ricardo judged as strong the difference of
attractiveness between level N6 and level N3, but he judged as very strong the difference
of attractiveness between level N4 and level NI. This reveals that, for Ricardo, the PV
"resistance to low temperature" is cardinaly dependent of the PV "resistance to iron",
due to some hidden synergetic effect between the levels of the two EPsV. In fact, the
measures necessary to mitigate the combined effects of intolerance to low temperature
together with intolerance to iron, involve a risk of failure perceived as higher than the
sum of the individual risks associated with the mitigating measures for each intolerance
phenomenon taken separately. Consequently, in the framework of the additive
aggregation model, it would be wrong to take the two PsV as isolated evaluation axes. 4
A feedback in the process might then be necessary to re-structure the family of FPsV.
Table 2 shows the partial values corresponding to the impacts of the rice varieties,
and the corresponding overall values resulting from the application of the additive model
with the scaling constants (''weights'') also indicated in the table. Finally, an extensive
sensitivity analysis was performed, to assist the actors in leaming with the model, namely
answering "what-if' questions, and to derive robust recommendations.

Table 2. Evaluation table: values and "weights" for the rice problem.
PARTIAL VALUES OVERALL
VARIETIES FPVI FPV2 FPV3 FPV4 FPVS FPV6 FPV7 FPV8 FPV9 FPVIO FPVII VALUE
EEA-406 70 45 20 100 15 40 29 0 38 0 50 32
IRGA-409 57 27 90 47 37 0 29 81 38 100 50 47
IRGA-410 57 27 75 63 37 0 57 23 0 100 0 36
IRGA-412 39 27 75 47 37 0 29 62 38 90 0 41
IRGA-413
IRGA-414
IRGA-416
57
39
39
27
45
45 100
75
75
63
32
63
54
54
40
10
42
86
86
° °92
55
62
90
78
38
100
100
90
86
100
100
54
61
55
IRGA-417 57 45 100 63 74 10 86 81 63 100 0 63
BLUEBELLE 0 45 90 32 0 10 71 46 78 90 50 47
EL PASO-l44 57 27 90 63 65 10 57 46 48 90 50 50
EMBRAPA-6 57 82 100 63 65 40 89 62 63 100 71 69
EMBRAPA-7 57 36 100 63 74 40 42 73 78 90 71 64
EMBRAPA-38 87 81 100 63 65 40 100 23 73 90 71 67
EMBRAPA-39

WEIGHTS
87 100

0,05 0,11
° 0,09
63 65 40 34 23 68 90

0,06 0,08 0,15 0,08 0,13 0,13

71 52

0,09 0,02

3 The interested reader can find in Bana e Costa et al. (forthcoming) a complete description of another
real-world application of this evaluation approach, in strategie management of the Santa Catarina
textile industry.
~ See other examples in Beinat, 1997, and Larichev and Moshkovich, 1997.

43
8. Final considerations and conciusions

In facilitating the structuring process of the rice varieties choice problem, a constructive
and interactive technical approach successfully led to progressively integrate initially
different perspectives about what were the relevant aspects of the problem into one
shared structure. The use of computational tools eased the interaction between
facilitators and actors and the management of information.
From the frrst meeting with Ricardo Gon~alves da Silva in March 1996, until the final
presentation of the study in July 1996, there were 14 meetings of one to two hours each.
The structuring phase of the problem using the cognitive mapping technique required
five sessions, including one for the approval of the fmal cognitive map and the
representation of the problem in the value tree format.
The initial structuring steps were the most difficult ones, mainly because of the ill-
defmed nature of the problem. Moreover, Ricardo had the strong conviction that the
problem should be stated simply as choosing the variety that provides the maximum
profit with minimum risk, and he was pessimistic about how the study could actually
contribute to the achievement of his objectives. But once the fmal cognitive map was
presented and the fundamental points of view started emerging, the decision-maker
changed his behaviour, and became completely and enthusiastically involved in the
process.
The construction of descriptors required four other sessions, and it was the stage that
consumed more time, almost three months. The evaluation phase took another five
meetings, including the model validation session. Afterwards, two long sessions were
devoted to working and leaming with the model and coming up with conclusions.
The fmal words of Ricardo were: "this new approach requires many hours of work,
but, once fmished, puts order in the process of choosing the rice variety". In general, the
actors directly involved in the study considered the leaming process particularly
interesting, because their understanding of the problem greatly increased. They all agreed
that the methodology generated better results than they were expecting at the beginning
of the study, and pointed out the possibility offered by the multicriteria analysis of
incorporating objective and subjective aspects, and qualitative and quantitative
information in the same model. These features, they said, increased their trust in the
model constructed, and their conviction that its adoption very much increases the quality
of decision-making and control of consequences when compared with intuitive decisions.

References
Bana e Costa, c.A. (ed.) (1990), Readings in Multiple Criteria Decision Aid, Springer, Berlin.
Bana e Costa, c.A. (1992), Structuration, construction et exploitation d'un modele multicritere
d' aide a la decision, Ph.D. Thesis, Universidade Tecnica de Lisboa.
Bana e Costa, c.A., L. Ensslin, E.C. Correa, J.C. Vansnick (forthcoming), Decision Support
Systems in action: integrated application in a multicriteria decision aid process, European
Journal of Operational Research.

44
Bana e Costa, e.A., J.e. Vansnick (l997a), Applications of the MACBETH approach in the
framework of an additive aggregation model, Journal oJ Multi-Criteria Decision Analysis,
6(2): 107-114.
Bana e Costa, e.A., le. Vansnick (I 997b), A theoretical framework for measuring
attractiveness by a categorical based evaluation technique (MACBETH), in J. Clfmaco (ed.)
Multicriteria Analysis, Springer-Verlag, Berlin, 15-24. .
Beinat, E. (1997), Value Functions Jor Environmental Management, Kluwer Acadernic
Publishers, Dordrecht.
Belton, V., F. Ackermann, I. Shepherd (1997), Integrated support from problem structuring
through to alternative evaluation using CaPE and VlSA, Journal oJ Multiple Criteria
Decision Analysis, 6(3): 115-130.
Bertier, P., J. de Montgolfier (1974), On multicriteria analysis: an application to forest
management problems, Metra, 13(1): 33-45.
Banxia Software Ud. (1995), Graphics COPE User Guide.
Bouyssou, D. (1990), Construction of Criteria: a Prerequisite for MCDA, in C.A. Bana e Costa
(ed.), Readings in Multiple Criteria Decision Aid, Springer-Verlag, Berlin, 58-80.
Costa, A.P. (1996), Metodologia multicriterio em apoio adecisäo para a selecräo de cultivares de
arroz para lavouras no sul do Estado do Rio Grande do Sul, Ms Thesis, Universidade Federal
de Santa Catarina.
Eden, C. (1988), Cognitive Mapping: a review, European Journal oJ Operational Research, 36:
1-13.
Eden, C., F. Ackermann, S. Cropper (1992), The analysis of cause maps, Journal oJ
Management Studies, 29(3): 309-324.
Keeney, R.L. (1992), Value-Focused Thinking: A Path to Creative Decisionmaking, Harvard
University Press, Cambridge, MA.
Keeney, R.L. (1994), Creativity in decision making with value-focused thinking, Sloan
Management Review, 33-41.
Keeney, R.L., H. Raiffa (1976), Decisions with Multiple Objectives: PreJerences and Value
Trade-offs, Wiley, New York.
Larichev, 0.1., H.M. Moshkovich (1997), Verbal Decision AnalysisJor Unstructured Problems,
Kluwer, Boston.
de Montgolfier, J., P. Bertier (1978), Approche Multicritere des Problemes de Decision,
Editions Hommes et Techniques, Paris.
Rosenhead, J. (ed.) (1989), Rational AnalysisJor a Problematic World, Wiley, Chi chester.
Roy, B. (1996), Multicriteria Methodology Jor Decision Aiding, Kluwer, Boston.
von Winterfeldt, D., W. Edwards (1986), Decision Analysis and Behavioral Research,
Cambridge University Press, Cambridge, MA.

45
 Part 11
The multiple dimensions
of land-use management
Rurallandscape and economic results of the farm; a
multi-objective approach 1

Francesco Marangon l and Tiziano Tempesta 2

I Dipartimento di Scienze Economiche, Universita di Udine, ltaly
2 Dipartimento Territorio e S.A.F., Universita di Padova, ltaly

Abstract
The protection and reorganisation of rurallandscape constitutes one of the main objectives in
European Union agro-environmental policy. Since 1985, numerous measures have been approved
with the aim of financing landscape improvement and farm reorganisation. These measures were
more c1early defined in 1992, when the timing and content of landscape measures were set in the
wider context of agro-environmental policy. European Union intervention was motivated by the
growing imbalance between the public demand for higher landscape quality and the land
transformation carried out by agriculture. On one hand, the societal and cultural changes which
have occurred in the last few decades have produced a rise in demand for green (rural) areas for
recreation. On the other hand, the spread of labour-saving technology in agriculture has led to an
increasing simplitication of the landscape and to a considerable reduction in rurallandscape
quality. This phenomenon can be attributed to the singular econornic characteristics of land'icape.
In many respects, land'icape represents a positive externality of agro-forest activity, which
assumes the role of a public good. This therefore necessitates corrective action by the public
policy maker who will have to use various means to increase the remuneration of those products
which generate positive externalities. Public intervention, however, runs the risk of being
transformed into welfare aid. In order to guarantee that it is used to remunerate a real service
undertaken by the farmer for the benefit of the whole community, the contribution must be
commensurate with the benefits produced. The objectives of this study are, first, to analyse which
elements of rural land'icape contribute to the aesthetic value of landscape and, second, to estimate
the trade-offs between economic results of the farm and landscape quality. An application in the
Venice Lagoon Basin Region allowed for the identification and quantification of compromise
solutions between the landscape .objective and the farm profits. These results can be used to
assess the economic consequences of improving land'icape quality and the suitability for this
purpose ofEuropean Union subsidies.

Keywords: rurallandscape, multiple objective decision making, agro-environmental

policies.

1 This study was carried out together by the two authors. Sections 1 and 2 must be attributed to Tiziano

Tempesta and Section 3 to Francesco Marangon. The Conclusions were made jointly. This research was
supported by National Research Council of haly. Special Project RAISA. Sub-project n.t.
49
1. Introduction

The Common Agriculture Policy (CAP) reform 2 aimed at progressively reducing the
price protections of agricultural products and, at the same time, at remunerating positive
externalities created by agriculture. In the latter case, the CAP suggested a system of
subsidies and incentives which may lead farmers to adopt cultivations that have a lower
environmental impact or can contribute to preserving or improving the landscape3 • The
system of subsidies, however, is unsatisfactory from many points of view. First of all,
the subsidies are very inelastic and sometimes scarcely operative because they cannot
take into account the organisational and environmental conditions in which farms
operate. Moreover, subsidies are determined in a way that hardly meet the public
demand of environment and landscape quality. Finally, the method used to determine
the subsidies is economically inefficient. If transaction costs are excluded, it can be
proved that subsidies based on surface unit are higher than subsidies based on the loss
of income caused by the introduction of crops that improve the environment and
landscape (Tempesta, 1997). However, it is inevitable that the implementation of a
system of subsidies to supplement lower incomes implies quite a number of operating
problems. In order to determine an effective subsidy system it is therefore necessary to
estimate both the environmental-Iandscape demand function and the farm production
function. Only in this way can alternative farm patterns be explored from the landscape-
agricultural point ofview, in which social benefits and private costs are known.
The aim of this study is to develop a methodology to determine the appropriate
compensation level for the loss of income which corresponds to higher levels of
landscape quality. First, a function of aesthetic appreciation of rurallandscape is
developed. This function relates the components ofrurallandscape to landscape quality.
After that, by means of a Multiple Objective Decision-Making (MODM) approach, the
decision strategies of different farms are explored. This serves to measure the income
losses corresponding to a higher landscape quality and, as a consequence, the amount to
be paid in order to allow these landscape improvements to be implemented. Finally, the
paper considers the introduction of the subsidy system and its effects on landscape
quality. The evidence provided in the paper is based on the structure of the agricultural
sector in the Venice Lagoon Basin4 (in Italy).

2 Regulations 2066/91, 1765/92,2078/92,2079/92 and 2080/92.

3 Regulations 2078/92 and 2080/92.
4 Venice Lagoon Basin has a surface of about 1,800 km 2 and covers a large part ofthe Veneto Lowland
(in the provinces ofTreviso, Padua and Venice) with highly diversified agricultural features. In the past,
many studies were carried out on this area, mainly to analyse the impact of agriculture and industry on
the water quality of the lagoon. The importance of positive extemalities created by agriculture was
rarely underlined, thus neglecting the fact that many actions aimed at the reduction of agriculture
pollution can have both positive and negative efTects on landscape.
50
2. Aesthetic appreciation and crop pattern: visual valuation

In order to identify the factors affecting the aesthetic appreciation of rurallandscape in

the Venice Lagoon Basin, preliminary research has been carried out on its main
landscape typologies. On the basis of previous studies (Franceschetti, 1988;
Franceschetti, 1990; Tempesta et al. 1990), seven main landscape types have been
identified (for instance, open views with trees in the background, or fields surrounded
by streams). Through a cartographic analysis, eighteen distinctive sites representative of
each landscape type were found in the area. They are considered as the best
representative points of the whole area. A photographic survey was then carried out in
these sites, together with the quantification of the different elements characterising the
rurallandscape. Finally, among the many slides taken, those most representative of the
reallandscape pattern of the site were chosen.
For each picture, all elements composing the image (such as ploughed land, trees,
hedges etc.) were listed. For some elements, such as ploughed land, a percentage of the
site coverage was computed. For some other elements, such as streams, only the
presence of the element was recorded.
The number of pictures selected was 126, seven for each site. During four meetings,
the pictures were projected and submitted to the aesthetic evaluation of 225 subjects.
Participants were asked to give a score between 1 (low appreciation of landscape) and
10 (high appreciation of landscape) to each picture. The average of individual
judgements was taken as the visual appreciation score of each picture.
A stepwise regression analysis was then used to compute the Visual-Aesthetic Index
(VAl). This index relates the elements which characterise the landscape to the visual
appreciation ofthe landscape obtained in the interviews. The result is the following (the
numbers in parenthesis show the statistical relevance ofthe coefficients):

VAl = -2.03 PLOUGHED -3.51 SETASIDE -4.24 GREENH -0.90 SEASON

(0.000) (0.000) (0.000) (0.000)

-1.67 ROADS -0.073 POLES +2.00 RIVERS +0.069 DENSlTY +3.09 ROWS
(0.052) (0.023) (0.003) (0.092) (0.00 I)

+ I.62 MEADOWS + 1.08 TREES I +0.74 TREES2 +2.16 HEDGES +5.80

(0.013) (0.024) (0.023) (0.069) (0.000)

Corrected R2 = 0.64

where:
VAL Visual Aesthetic Index;
PLOUGHED % of arable land;
SETASIDE % of set-aside surface;
51
GREENH presence of greenhouses;
SEASON dummy variable indicating the time of the year when the picture was
taken (O=Spring, 1=Summer);
ROADS dummy variable for asphalt road in the foreground;
POLES dummy variable for electric lights poles in the foreground;
RIVERS dummy variable for rivers and streams;
DENSITY dummy variable for hedge types with high density;
ROWS % surface covered with tree rows;
MEADOWS % surface covered with permanent meadows;
TREESI dummy variable for the presence of scattered trees in the foreground;
TREES2 dummy variable for the presence of trees in the background;
HEDGES % surface covered with hedges.

The form of this function shows a preference system which is very similar to those
obtained in other works carried out in other areas in the north-east of ltaly (Tempesta,
1997; Marangon and Tempesta, 1996a and 1996b). The factors that contribute most to
the pleasantness of landscape are hedges, rows of trees or scattered trees, as weil as
meadows and watercourses. On the contrary, ploughed land, set-aside and greenhouses,
as weil as paved roads and electricity poles, reduce landscape quality and appreciation.

3. The Multiple Objective Decision-Making approach

In the past few years a substantial amount of research has been carried out to develop
decision support tools for the management of agro-forestry resources. Among these
tools, the Multiple Criteria Decision-Making (MCDM) approach plays a prominent
role. Although there have been many applications ofthese techniques in water resources
management and forestry planning, their use in farm and land planning in agriculture is
relatively recent (cf. Romero and Rehman, 1989). The recognition of the multi-
functional character of this sector, in fact, occurred only very recently (Bernetti et al.,
1996; Marangon, 1997, Marangon and Rosato, 1995). Accordingly, evidence for the
potential of MCDM to solve problems of rurallandscape management is still limited
(Assfalg and Wemer, 1993; Marangon, 1997; Marangon and Tempesta, 1996a and
1996b; Werner, 1993). In this section, the applications of an MCDM technique to
particular farming cases will be described. This application focuses on the trade-off
between economic outputs and the generation of positive externalities (landscape).
Within the wide range of available MCDM techniques, Multiple Objective Decision
Making approaches (MODM) are more suitable for this application: they handle
continuous problems which involve several objectives which should be simultaneously
optimised. Multi-Objective Programming (MOP), in particular, allows both the
simultaneous optimisation of multiple objectives and the identification of several
efficient, or Pareto optimal, solutions (Marangon, 1994; Marangon and Rosato, 1995;
Romero and Rehman, 1989). Moreover, MOP makes it possible to construct

52
transformation curves between objectives, which allow the quantification of the trade-
off between the objectives and the degree to which they change when moving from one
alternative to another (Rehman and Romero, 1993).
The focus of the study was to investigate the effect of considering two conflicting
objectives simultaneously: the maximisation of the gross margin (GM) and the
maximisation of "landscape production" (L). These two objectives govern the farm
management choices. They can be used to simulate the behaviour of an entrepreneur
facing alternative choices which range from the production of marketable goods and the
production of services for the benefit of the whole community (landscape ). By
exploring the effects of different decision strategies it is possible to offer evidence of
the opportunity costs related to the production of positive externalities. Policy makers
should be aware of these costs, which are the basis for setting compensation
mechailisms in order to reach different levels oflandscape quality.
The formal structure of the MOP models is very similar to the typical structure of
monocriterion optimisation through LP. The two objective functions used in this
applications are:

Gross Margin (GM) (1)

Landscape (L) (2)

where:
GMj are the unitary Gross Margins (.000 Lira per hectare) of each cultivation type
obtained by subtracting production costs from the gross income;
lj are the hmdscape coefficients which correspond to each cultivation type;
x is the n-dimensional vector of decision variables (x; = surface area for crop i).

The objective function related to "landscape production" (2) requires the estimation
of the landscape coefficients lj (i=I,2, ... ,n) which correspond to the each type of
cultivation. These coefficients are based on the Visual-Aesthetic Index (VAl) described
earlier, and are calculated with the following formula:

lj = (VAI/SUR)'100 (3)

where VAl; are the levels of the aesthetic index for the landscape variables obtained
with the multiple regression (cf. Section 2), and SUR is the total arable land which is
shared among the different types of cultivation.
Unlike previous works (Marangon, 1997; Marangon and Tempesta, 1996a and
1996b), this MOP application allowed for the comparison of different organisational
farm models. More precisely, three farm sizes were selected as representatives of the
current situation and of the structure of the agricultural acti vities in the Venice Lagoon

53
 Basin, a situation characterised by a prevalence of cereals and oil seed production
(Table 1).
The first case (S) corresponds to a "smalI" farm (5 hectares) which relies on third
parties for the largest part of the farm activities. This type of farm inc1udes only a very
low amount of permanent work (10 hours per month). For this farm, most operation
costs are variable costs. The second model (M) is characteristic of a "medium" size
farm (10 hectares) which combines contracts to third parties with permanent labour
(160 hours per month). Finally, the last model (L) is characteristic of a "Iarge" farm
(100 hectares) with full-time workers (400 hours per month), not employing "third
parties". For this farm, most operation costs are fixed costs.
The range of possible crops is the same for all farm sizes: maize, soybeans, meadows
and poplars. For maize and soybean, however, a distinction has been made between
cultivation activities which make use of a large proportion of third parties (maize-c and
soybeans-c), and activities which are carried out in farms with full-time employees. The
rationale of this distinction is that the gross margin per hectare for soybeans and maize
depends on the proportion of fixed and variable costs, which are different for the three
farm sizes.
As far as the set-aside figures are concemed, the EC regulations make it compulsory
to set aside a given proportion (10%) of the surface allocated for maize and soybeans.
Finally, hedges and tree rows were introduced due to their high potential for improving
the landscape. On the basis of typical agronomic figures, the maximum surface
occupation for hedges was set at 20% of the total surface, while for tree rows the
maximum was set at 1%.
The parameters of the objective function for the MOP, together with the constraints
of the model, are shown in tabular form in Table 1. The three types of farms were
analysed under two different conditions. In the first case, the results of the farm are
based on standard gross margin levels corresponding to different crops; in the second
case the gross margin of poplar trees, meadows and hedges increase due to the effects
of subsidies made available by the Region Veneto to enforce EC Regulation 2078/92
(agri-environmental measures) and EC Regulation 2080/92 (forest enhancement
measures).
In Table I, the GMa and GMb rows show the gross margin per hectare for each
cultivation type (in .000 Italian lira). The GMb row inc1udes EC subsidies for poplars,
meadows and edges. The landscape row shows the VAl coefficients for different crops.
The constraints (part 1) indicate the number of man-hours per hectare of crop. Each row
limits the total number of hours per month to a lower level than the hours available in
the farm (the S, M and L columns). The second group of constraints (part 2) sets limits
for the surface occupied by hedges and tree rows (see above). Finally, the set aside and
the arable land rows simply state the total area for set aside and arable land for each
farm.

54
 Tahle I. Cocfficicllts ol the ohjcctive fUllctioll ami cOllstraillls for the MOl' model.

Ohjt.'Ctive MAlZE MAIZE- SOY SOY-C SETAS POPLAR MEAIJOW HEIlGES ROW
c'
function
GMa 3.200 2.500 2.200 1.500 1.000 2.300 1.200 0.750 0.750
GMh 3.200 2.500 2.200 1.500 1.000 3.320 1.520 2.0HO 0.750
Landscape -2.03 -2.()3 -2.03 -2.03 -3.51 1.60 2.16 3.0')

Constraints (part I) S M L
January 3.50 0.35 2.50 5.00 3.00 10 160 400
Fehruary 3.50 . 0.35 3.00 0.30 2.00 6.50 10 160 400
March 4.50 0.45 2.50 0.25 1.50 7.00 10 160 400
April 3.00 0.30 3.00 0.30 1.50 7.00 1.00 10 160 400
May 1.50 0.15 1.00 0.10 5.00 0.50 15.00 15.00 10 160 400
June 3.00 0.30 3.00 0.30 18.00 14.00 10 160 400
July 7.00 0.70 3.00 0.30 5.00 14.00 10 160 400
August 4.00 0.40 3.00 0.30 5.00 7.50 10 160 400
Septemher 4.00 0.40 7.50 10.00 10.00 10 160 400
Octoher 2.00 0.20 15.00 15.00 10 160 400
November 1.50 0.15 1.50 0.15 10 160 400
December 5.00 0.50 1.50 0.15 10 160 400

Constraints (part 2)
Hedges 1.00 I 2 20
Tree rows 1.00 0.05 0.10 1.00
Set-aside 0.10 0.10 0.10 0.10 -1.00 0 0 0
Arable land 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 5 10 100
• MAIZE-C and SOY -Co whlch Im ply large use of th Ird partles. were not consldered for the "'I arge" farm (L).

The starting point of the MODM analysis is the pay-off matrix (Romero and Rehman,
1989) in which the levels of objectives obtained by separate optimisations are shown
(Table 2).
As can be see in Table 2, the LP result obtained for the maximisation of the gross
margin objective shows higher levels per hectare in the case of the "medium" size farm,
with or without resorting to the subsidies of the CAP (the Mb and Ma models,
respectively). The optimisation of the economic objective can be obtained with
monocultural cultivations centred on maize. This holds for both the "smalI" and the
"medium" farms. In the "Iarge" farm, on the contrary, there are solutions from the Lb
model which include poplar trees and hedges in addition to soybeans.
The first se ries of MOP results (Sa, Ma and La) do not consider the effect of any
environmental and forest enhancement subsidies. Three efficient sets were determined for
these models. For the Sa and Ma models, the efticient set is formed by 7 solutions; while
for the La model the efficient set includes 6 solutions (Figure I).

55
 Table 2. Pay-offmatrices. The results oJmaximising the gross margin and the landscape quality are
shown Jor the "smaU", "medium" and "large" Jarms, both in the standard case (So, Ma and La
models) and in the case oJ subsidies (Sb, Mb. Lb).
"SMALL" FARM MODELSa MODEL Sb
OBJECTIVES MAXGMa MAXLAND MAXGMb MAXLAND
Gross Margin (,000 Lira) 12,576 6,784 12,746 7.768
Landscape (scores) -10.82 -6.00 -10.04 -6.00
Gross Marginlha. (,000 Lira) 2,515 1,357 2,549 1,554
Landscapelha. (scores) -2.16 -1.20 -2.01 -1-20

"MEDIUM" FARM MODELMa MODEL Mb

OBJECTIVES MAXGMa MAXLAND MAXGMb MAXLAND
Gross Margin (,000 Lira) 30,000 11,055 32,781 16,243
Landscape (scores) -21.65 17.27 -2.83 17.27
Gross Marginlha. (,000 Lira) 3,000 1,106 3,278 1,624
Landscapelha. (scores) -2.16 1.73 -0.28 1.73

"LARGE" FARM MODEL La MODELLb

OBJECfIVES MAXGMa MAXLAND MAXGMb MAXLAND
Gross Margin (,000 Lira) 240,909 169,873 260,067 198,523
Landscape (scores) -216.45 -74.94 -116.91 -74.94
Gross Marginlha. (,000 Lira) 2,409 1,699 2,601 1,985
Landscapelha. (scores) -2.16 .0.75 -1.17 .0.75

Landscape scorelHect.

2.0
-----_ ...... _... _........ _.......... _._ ..... +
1.5

1.0

0.5

0.0

-1.0

-1.5

-2.0

-2.5 + - - - - - t - - - - - - - - - t - - - - - + - - - - - + - - - - - - - - - - - l
1.000 1.500 2,000 2.500 3,000 3,500
Gross MarginlHect. (.000 lire)

__ smali ...... medium _large

Figure 1. Efficient solutions in the a-models (without subsidies) Jor the smalI. medium and {arge Jarms.

56
 Figure 1 c\early shows the difference between the efficient set of the "medium" farm
and of the other two farms. In the medium farm, there is such a structural composition
of land and workforce that a higher potential for landscape production emerges. The
same economic targets, in terms of gross margin per hectare, can be achieved together
with a systematically higher level of landscape quality. This result is mainly determined
by the relatively larger availability of labour in the "medium" farm. In this farm, there
is a constant amount of 16 hours per hectare per month (h/ha*m), whereas in the
"smalI" farm the amount is only 2 (h/ha*m) and in the "Iarge" farm is 4 (h/ha*m). The
key role played by permanent farm employment in relation to the landscape quality
performance of the farm is not new. Similar trends have been highlighted in previous
studies on Italian farms (Marangon, 1997; Marangon and Tempesta, 1996a and 1996b).
Concerning the structure of the solutions, it is worth noting that in four solutions of
the "medium" farm, where there are higher landscape scores, maize and soybeans (and
their set-aside) are totally excluded and replaced by meadows and poplars. Poplar trees
also predominate in high income solutions, but are replaced by a maize monoculture in
the hypothesis of the maximum gross income.
In the "smalI" farm the presence of ploughed land is constant, even in the solution
that maxi mi ses landscape quality, with soybeans being predominant (2/3 of arable land)
followed by hedges and meadows. In the other six solutions, soybean is replaced by
maize and, as income increases, the spread of poplars decreases together with the
gradual elimination of meadows, hedges and tree rows. The maximum gross margin in
this case also comes from maize monoculture.
Finally, in the "Iarge" farm, the best economic result comes from a maize-soybean
combination. Moving along the transformation curve towards the upper-left corner of
Figure 1, soybeans keep its share and maize is replaced by "landscape" productive
processes, such as tree rows, hedges and meadows. Poplars can be found only in two
low income solutions. It is also interesting to note that the maximum landscape solution
is the one in which 60% of arabie land is covered with soybeans, together with hedges,
meadows and tree rows.
Table 3 shows another type of comparison. Taking as a reference point the results of
the two objectives when the best economic result is achieved (cf. Table 2), Table 3
shows the trade-offs between economy and landscape for the three farm models.
Starting from the best solution in economic terms (the Max. GMa column), the table
shows the income losses and landscape gains which correspond to all other solutions of
the efficient frontier (cf. Figure 1). For each farm model, column 1 corresponds to the
best solution in landscape terms, column 2, to the second best etc. For instance, the
column 1 for the small farm indicates that the income loss going from the best income
solution (GMa=2515; landscape=-2.16) to the best landscape solution (GMa=1357;
landscape=-1.2) is equal to 1158 thousand lira, which produces alandscape increase of
44.6%.

57
 From this table it can be seen that the "medium" farm can achieve landscape
improvements around 85-90% with gross margin decreases per hectare slightly higher
than 600,000 Lira. The landscape score is doubled for economic losses of about
900,000 Lira per hectare.

Table 3. Trade-o.IJ between maximum gross margin (.000 liralha) and landscape. The GMa rows show
the income reductions (in thousands 0/ Lira per hectare) which corresponds to the landscape
improvement shown in the Landscape rows.
Small Max. GMa 2 3 4 5 6
GMa 2,515 1,158 428 319 172 55 40
·CäiidScape······ ..........................
-2.16
...................... ...................... ...................... ...................... ................... ...................
44.60% 40.39% 34.31% 24.80% 9.63% 7.21%
Medium Max.GMa I 2 3 4 5 6
GMa 3,000 1,895 1,805 1,216 872 632 608
Landscape -2.16 179.78% 174.61% 135.06% 111.52% 89.51% 86.90%
Large Max. GMa I 2 3 4 5
GMa 2,409 710 550 155 134 7
L'äiidsc·äp·e............·:2:T6........ '"6i38'%''' "'/54:38'%'" '"46:42%''' '"4'L02%''' '"f43%'''

In the case of the large farm model, the best landscape improvement is about 65%,
which corresponds to income reductions ranging between 550,000 and 710,000 Lira per
hectare. It must also be noted that significant landscape improvements (about 40-45%)
can be obtained with 140,000-150,000 Lira income losses per hectare. In the "smalI"
farm it seems more difficult to achieve large landscape improvements: a 34% increase
of the environmental objective would cost more than 300,000 Lira per hectare. If the
objective was set at 40%, the cost would reach 430,000 Lira per hectare.
Keeping these values as a reference point, and considering them with obvious
caution, the following step was to analyse the effect of the EC subsidies for meadows
(EC Regulation 2078/92), poplars and hedges (EC Regulation 2080/92) in the three
farm models.
Without going into details, the efficient frontier of Figure 1 moves upwards and
rightwards, as shown in Figure 2. While the ideal points for the landscape co-ordinate
do not change, the maximum level for the gross margin increases by 34,000, 278,000,
and 192,000 Lira per hectare in the "smalI", "medium" and "Iarge" farm", respectively.
Once again it is useful to analyse the results of the "medium" farm. Apart from the
solution which maximises landscape production, the remaining four solutions are all
characterised by higher income levels than 3 million Lira per hectare. This level is the
maximum which can be achieved by the medium farm when no subsidies are included.
When subsidies are taken into consideration, this result can be reached by devoting
about 80% of arable land to poplars.

58
 Landscape score/Hect.

2.0

1.5

1.0

0.5

0.0

\
.(l.5 \
\

)f--""' .........~"\
-1.0
"" " \

• - -----_-::!i.- "
\

\
-1.5 \
\
\
\
\
-2.0 \
~

-2.5 { - - - - - - - + - - - - - + - - - - - - + - - - - - - - - + - - - - - - - 4
1,000 1,500 2,000 2,500 3,000 3,500

Gross MarginIHect. (,000 Lire)

- . - small a -I; ~ medium a -,.. -Iarge a

_smallb ....... mediumb -Iargeb

Figure 2. Efficient solutions in a models (dashed fines) and b-models (thick. continuous fines).

As for the transformation curve of the "I arge" farm, although it achieves much lower
results than the "medium" farm, it shows a clear improvement in comparison with the
situation without subsidies. It is important to note that this is achieved without giving
up traditional crops, but rather by integrating them with cultivations subsidised for
environmental purposes.

4. Conclusions

This study analyses the trade-offs between farm income and landscape quality. On the
basis of a MODM approach, the analysis focuses on three farm models which are
characteristic of the current situation in the Veneto lowland in Italy. The solutions
presented by the Multiple-Objective model focus on production practices which
increase the landscape aesthetic values while minimising the loss of farm income.
The results obtained in this study shed light on the usefulness of EC subsidies to
improve landscape quality. In spite of the limits of the simplified assumptions, the
analysis demonstrates and confirms what has been suggested in previous studies
(Marangon, 1997; Marangon and Tempesta, 1996a and 1996b): the subsidies introduced
by EC Regulations 2078/92 and 2080/92 can reduce the conflict between the income

59
maximisation objective and the objective of landscape quality maximisation. MOP
solutions highlight that this effect is linked to the labour flexibility in the farm and to
the willingness to direct the production towards agriculture and forest activities which
complement the more traditional crops.
It should also be noticed that structural and organisational features can influence the
possibility of implementing these choices. With respect to this, the over-supply of
manpower, that partly still characterises some medium size farms in Italy, is of
fundamental importance.

References
Assfalg, W., R. Wemer (1993), Tbe optimal use of agricultural landscapes, in Applied
Geography anti Development - A Biannual Collection 0/ Recent German Conlributions,
Tübingen: Institut für Wissenschaftliche Zusammenarbeit, 60-97.
Bemetti, 1., F. Marangon, P. Rosato (Eds.) (1996), Metodi e applicazioni delI'analisi
multicriteriale nel settore agro-forestale e ambientale, CIRMOCOSAF, Quaderno n. 1.
Franceschetti, G. (1988), Il paesaggio rurale nella Provincia di Padova, Piano Territoriale
Provinciale: rapporto sullo stato dei Territorio, Padova.
Franceschetti, G. (1990), Indagine sui tipi prevalenti di paesaggio agrario deI veneziano, in
Provincia di Venezia, Speciale paesaggio agrario, n. 4.
Marangon, F. (1994), Nuovi metodi nelI'analisi di gestione delI'impresa agraria Tendenze
emergenti nel panorama italiano, in G.Trevisan (Ed.) L 'impresa agraria: attuali problemi di
organizzazione e di gestione, lNEA-Il Mulino, Bologna, 63-143.
Marangon, F. (1997), Redditivita aziendale e valorizzazione deI paesaggio rurale al vaglio
dell'Analisi a Molti Obiettivi, in T. Tempesta (Ed.) Paesaggio rurale e agro-tecnologie
innovative: una ricerca nella pianura Ira Tagliamento e Milano, AngeH, Milano.
Marangon, F., P. Rosato (1995), L'analisi multi criteri nella gestione delle risorse naturali. Il
caso delle risorse agroambientali, Economia delle/onti di energia e deli 'ambiente, 2: 45-89.
Marangon, F., T. Tempesta (1996a), La programmazione multiobiettivo nelI'azienda agricola:
paesaggio rurale e redditivita, in I. Benetti, F. Marangon, P. Rosato (Eds.) Metodi e
applicazioni dell'analisi multicriteriale nel settore agro-forestale e ambientale,
CIRMOCOSAF, Quademo n. I.
Marangon, F.,T. Tempesta (1996b), Farm Income versus Agricultural Positive and Negative
Lantiscape Externalities: a Multicriteria Approach, Proceedings of the 8th European
Association of Agricultural Economists (EAAE) Congress, Edinburgh, 3-7 September 1996.
Rehman, T., C. Romero (1993), Tbe application ofthe MCDM paradigm to the management of
agricultural systems: some basic considerations, Agricultural Systems, 3: 239-255.
Romero, C., T. Rehman (1989), Multiple Criteria Analysis/or Agricultural Decisions, Elsevier,
Amsterdam.
Tempesta, T., G. Marchetti, G. Milani, O. Santantonio (1990), I1 paesaggio agrario della
provincia di Treviso, Treviso.
Tempesta, T. (1997), Estemalitapositiveeproblemidiefficienzanell.alIocazione delle risorse,
in T. Tempesta (Ed.) Paesaggio rurale e agro-tecnologie innovative. Una ricerca nella
pianura Ira Tagliamento e Isonzo, Angeli, Milano.
Wemer, R. (1993), Ecologically and economically efficient and sustainable use of agricultural
landscape, Landscape and Urban Planning, 27: 237-248.

60
Multicriteria evaluation of sustainable agricultural land
use: a case study of Lesvos

Gerda Hermanides and Peter Nijkamp

Department of Economics. Vrije Universiteit
Amsterdam. The Netherlands

Abstract
Sustainable agriculture has become an important policy orientation in the area of land use. It serves
as a framework of analysis to reconcile conflicting interests. In this paper the notion of sustainable
agriculture will be operationalized by using the concept of environmental utilisation space. This
concept offers critical threshold values for various socio-economic and environmental indicators.
The paper allows for flexibility and uncertainty in expert opinion on sustainable agriculture by
distinguishing various ranges of carrying capacity. In order to map out the social costs of
environmental stress, a so-called 'flag model' is used for analysing the various trade-offs. Next, a
multicriteria method using the Regime method is introduced in order to identify compromise
solutions. The (stepwise) sustainability methodology is applied to the case of agriculture on the
island of Lesvos, Greece. Lesvian agriculture is faced with the threat of erosion and degradation
caused by various interlinked developments. Three poliey orientations on agriculture on Lesvos are
evaluated by means of the above mentioned methodology. According to the results of the flag
approach, a policy of structural support will likely lead to the most sustainable situation. The
scenario of environmental care in agriculture is the next most sustainable one, whereas the scenario
of the liberalisation of agricultural markets and agricultural trade appears to perform worst. This
outcome corresponds to the majority of the outcomes of the Regime evaluation. Thus, the
sustainability and continuity of agriculture seerns to be best served by a policy of public, socio-
economic support. Liberalisation of agricultural markets may certainly also be opted for, but this
would make the future of Lesvos less unambiguous.

Keywords: sustainability, critical threshold values, scenario analysis, flag approach,

Regime method.

1. Sustainable agriculture

The importance of sustainability as a key concept in environmentally-benign

policymaking has been increasing rapidly in the past years. 'Sustainability' is nowadays
accepted as a major signpost in many policy fields, although surprisingly the
interpretations of this concept are rather diverse. The basic elements of sustainability - or
more specifically, 'sustainable development' - can be found in the defmition of sustainable
development by the World Commission on Environment and Development (WCED):
sustainable development is economic development that meets the needs of the present
without compromising the ability of future generations to meet their own needs (WCED,
1987). In the view of many scientists and policy-makers, this definition implies that

61
policies or projects should explicitly be judged on their environmental impact. In other
words, a policy should be evaluated on the basis of the criterion of ecological
sustainability, in addition to the traditional criteria of efficiency and equity (see, for
instance, Fresco and Kroonenberg, 1992; Van Pelt, 1993).
Instead of confming the concept of sustainability exclusively to 'ecological
sustainability', several authors regard the broader notion of sustainable development as a
more proper policy objective, as it mirrors a threefold equilibrium situation: if a certain
development is sustainable, then economic as weIl as social and environmental interests
should be in balance, now and in the future. Sustainability is then characterised by three
prominent dimensions: the economic aspect, the social aspect and the environmental
aspect. The economic aspect is related to efficiency and welfare items like income,
production, investments, market developments, and price formation. The social aspect
concerns distributional and equity considerations, such as income distribution, access to
markets, and wealth and poverty positions of certain groups or regions. The
environmental dimension refers to pollution, landscape variety, quality of life, resource
scarcity, and related human variables.
The above mentioned three aspects of sustainability are strongly interlinked, but also
to a large extent mutually conflicting. Economic development, for example, can improve
social circumstances, but can also enlarge the gap between rich and poor. De Bruyn and
Opschoor (1994) studied the relationship between economic development and the
environment, and concluded that the environmental pressure varies for each stage of
economic development. On the other hand, poverty may have a detrimental influence on
the environment: poor societies may be unable to undertake long-term, environmentally
desirable investments, for instance, because of lack of fmancial resources or because they
may not have efficient (agricultural) production methods at their disposal. In addition,
environmental protection may simply not be in the direct (short-term) interest of poor
people (Reed, 1992).

Efficieney

Equity Environment

Figure 1. Möbius triangle ofthree key forces in agricultural evaluation.

The force field of the three dimensions of sustainability can be clarified by means of a
Möbius triangle (see Figure 1). Point A represents the present state of a socio-economic
and environmental system, while point B may denote a more desirable future state which
represents a sustainable state of affairs. Reaching this point is a matter of giving and

62
taking in respect to each of the three above mentioned sustainability dimensions.
In the agricultural sector, the interlinkage of efficiency, equity and environment is an
obvious one. The economic performance in agriculture depends, for instance, on the
availability, quality and efficient usage of natural resources. Environmental damage may
negatively affect agricultural productivity and income, which places pressure on the
economic and social structure of rural communities. If however, in the agricultural sector
a balance among efficiency, equity and environment is to be found, agricultural policy
will have to be guided by the concept of sustainability. The challenge is now to
operationalise the concept of sustainability in agriculture. This means that the notion of
sustainable agriculture should be converted into operational terms. The next section
discusses in more detail the question of how to make the notion of sustainable agriculture
operational in a practical context.

2. Operationalising sustainability

Ideally, to judge the sustainability of a certain state of affairs or of a certain development,

we should be able to 'measure' sustainability. Measuring and judging sustainability
requires three building blocks: indicators, normative reference values (or standards), and
an impact assessment methodology. These three building blocks will be discussed
hereafter. Once these blocks are present in operational form, we may apply them in a
stepwise procedure (or sustainability methodology) in order to judge the sustainability of
a certain situation or policy choice. Following such a procedure would also allow the
evaluation of an already pursued policy, or of a policy still to be implemented. We will
first address the three building blocks of our analytical framework, while next the
sustainability procedure will be discussed in the second part of this section.
As mentioned above, measuring sustainability means the assessment of three
components. Indicators are numerical representations of variables that indicate (or
approximate ) the presence and/or state of phenomena that cannot be directly measured.
The state of education in a certain region, for example, can be indicated by the literacy
rate, or by the pupiVteacher ratio in primary education. In this context, sustainability
indicators should depict important numerical and observable aspects of sustainability.
They should be selected on account of their potential to measure relevant dimensions of
sustainable development. Clearly, the three relevant dimensions are environmental, social
and economic characteristics. The identifIcation and defmition of sustainability indicators
depend on the indicators' relevance regarding the region under consideration and its
specifIc socio-economic and land use conditions. In other words: which indicators should
be selected to 'measure' sustainability, depends on the region at hand, and its specifIc
sustainability problem. Another problem concerns the integration of various numerical
values reflecting only apart of a certain phenomenon into an aggregate indicator. This
requires proper aggregation procedures, based e.g. on non-statistical weightings or on
statistical multivariate analyses (see for an overview Coombes and Wong, 1994).
In our analysis, the policy-analytical concept of 'environmental utilisation space'

63
(EUS) is the basis of normative reference values. The concept of EUS expresses that at
any given point in time there are limits to the amount of environmental pressurethat the
earth's ecosystems can absorb without damage to these systems or the underlying life
support processes (Opschoor and Weterings, 1994). Environmental utilisation is what we
take (harvest) from the environment and what we return to the environment as waste.
Knowing how - and how fast - natural resources regenerate and to what extent the
environment can absorb waste, one can assess the degree of meaningful use we can make
of natural resources (Opschoor and Weterings, 1994). The notion prompts a search for
critical levels of environmental pressure beyond which actual environmental systems
rnight become damaged, either reversibly or irreversibly. These critical levels represent
the operational boundaries of the EUS, and can be used as the reference levels for
environmental indicators.
The application of the EUS concept can be broadened by including maUers of
sustainability in the wider sense of the sustainability concept. Reference levels can thus
also be assessed for economic as weIl as social indicators, since critical levels also exist
with regard to efficiency and equity. Thus, the boundaries of sustainability are
represented by a set of reference values for sustainability indicators. These normative
reference values, or critical threshold values, can be, for example, safe minimum or
maximum levels, 'natural' levels, and - perhaps the most popular though debatable point
of reference - present levels (Van Pelt, 1993). They may be assessed on the basis of
scientific research and/or expert opinion.
FinaIly, before any evaluation or comparative analysis of projected indicator values
can be made, a projection of anticipated effects - in the form of impact assessments -
should be made by means of an impact methodology. Impact assessment methods are
either ad hoc or structured (Blaas and Nij'kamp, 1992). Ad hoc impact assessment is a
way to analyse measurement problems for which no formal operational model can be
developed because of time constraints, non-repetitive situations, or lack of data. An
example of ad hoc analysis is an informal analysis using expert views. Ad hoc methods
are fairly fast: in a short time span relevant insights into the expected consequences of
any event can be generated, on the condition that sufficient expertise and experience is
available to enable realistic estimates of expected impacts. Ad hoc impact analyses
however, do not offer the same degree of precision, controllability and transferability as
structured impact analyses do. The main characteristic of structured impact analysis is
that it is systematic: the effects of (a set of) policy measures on relevant policy variables
are systematically traced under varying conditions. This me ans that projections are made
on the base of formal - usually quantitative (econometric or statistical) - techniques and
models.
After the assessment and design of indicators, normative reference values and an
impact methodology, the factual sustainability analysis has to be carried out, interpreted
here as a procedure for assessing the sustainability of an existing or future situation. This
procedure consists of four stages. First, the sustainability problem has to be identified.
Second, various policy scenarios (or strategies) for the future should be designed and
formulated. Third, each policy scenario (or strategy) should be evaluated, and finally, a

64
comparison of the various performances of policy options should be made.
First, the identification of the sustainability problem is an important initial step in
which the nature of the sustainability problem is assessed. A certain agricultural branch in
a given region can, for instance, be confronted with a poor or threatened environmental
quality, with social stress, or with substandard economic performance. The agricultural
sector can be distinguished according to main activities (e.g. crops), and the region can
be described by its size, its physical geography, and the like.
The use of scenario analysis (or strategie choice analysis) enables a systematic way
of scanning various uncertain future choice possibilities. A scenario is a possible image of
future events, such as an external policy development. Each scenario can be
characterised by a certain set of va lues of sustainability indicators, assessed by means of
an impact assessment method. In that way, each scenario can be judged on the basis of
its sustainability. Such a judgement however, requires an analysis of the projected
indicator values. This analysis will be carried out in the third step, policy evaluation.
To evaluate policy scenarios with regard to sustainability, the indicator values
characterising each scenario will have to be compared with the already discussed set of
normative reference values. These reference values, or critical threshold values (CTVs),
indicate the limits of sustainability; they cannot be exceeded without causing
unacceptably high damage and risk to the environment. The entire set of CTV s acts as a
reference system for judging actual states or future outcomes of scenario experiments, in
particular policy strategies.
The fact is that this reference system is not always unambiguous: CTVs may differ per
region and they may depend on local socio-econornic and natural conditions, and
different experts and decision-makers may have different views on the precise acceptance
level of a CTV. There is room, then, for an uncertainty analysis for both measurement
precision and expert knowledge. To avoid a high degree of ambiguity, a band width for
the CTV in question can be introduced by assessing a CTV min and a CTV max around this
median CTV. Assuming that the indicators are assessed as cost or damage variables,
CTV min indicates a conservative estimate of the maximum allowable threshold of the
sustainability indicator concerned. CTVmax refers to the maximum allowable value of the
sustainability indicator beyond which an alarming - i.e., environmentally devastating -
development will certainly start. To visualise the degree of sustainability of formulated
future scenarios, the set of indicator values characterising each actual scenario can be
compared with the pre-specified CTVs. This comparison results then in the so-called
'flag approach' (see Nijkamp, 1996). For example, we would denote a domain of
indicators of a given scenario with a green flag if the indicator value is below the most
conservative CTV; an orange flag, if the indicator value is below the median CTV, but
exceeds the CTV min; a red flag, if the indicator value exceeds the median CTV, but is still
below the CTV max; and a black flag, if the indicator value gives cause for substantial
concern since it exceeds the CTV rnax. This flag evaluation is clarified in Figure 2.

65
 CTVmin CTV CTVma,
green orange red black
flag flag flag flag
o JOO

Figure 2. Flag representation 0/ levels 0/ concern on indicator values.

Once more, it is assumed that the sustainability indicator is a so-called cost indicator
(ie., 'the lower, the better'), and that the minimum value of the indicator is 0, whilst the
median CTV is standardised at a value of 100. The colour of the flags should from a
sustainability perspective be interpreted as folIows:
• green flag: no reason for specific concern
• orange flag: be very alert
• red flag: reverse trends
• black flag: stop further growth or operation.

A further examination of the flag colours pertaining to the total set of indicators (or
evaluation criteria) may offer practical insight into the degree of sustainability of the
scenario in question.
The next step is the policy assessment; i.e., which scenario or policy strategy is more
desirable in light of the diverse characteristic judgement criteria. Here it is useful to
resort to multicriteria analysis. Given the uncertain degree of precision of the indicators
(including sometimes even qualitative information), we will use here the Regime method.
The Regime method, a multicriteria evaluation method which does not necessarily need a
cardinal assessment of CTVs, can be used as a complement to the flag approach. The
Regime method is a qualitative multiple criteria evaluation method that aims at providing
a rational basis for solving discrete choice problems characterised by multiple evaluation
criteria which are intangible and incommensurable (Hinloopen et al., 1982). The Regime
method is, like other multiple criteria analysis methods, based on two kinds of input data,
viz. an impact matrix and a set of political weights attached to the criterion effects. The
impact matrix represents the expected values of indicators (linked to policy criteria) for
each of the choice options or strategies. The weights indicate the importance of the
criterion in question in comparison to the other criteria. The result of the Regime method
is normally a ranking of scenarios, expressing which scenario is preferable.
The last step in the sustainability methodology, viz. the final assessment, concerns an
overall evaluation of the choice options. The performance levels of the various policy
options are compared, and the degree of sustainability of the fmal option is assessed and
communicated to policy-makers.
This stepwise sustainability methodology will be illustrated for a practical policy
choice problem, viz. the case of sustainable agriculture on the island of Lesvos, Greece.
This application and its results are described in the following sections.

66
3. The case of agriculture on the island of Lesvos

The issue of sustainable agricultural policy on the island of Lesvos is used to illustrate
the operational framework for the concept of sustainable agriculture. In our application,
the four steps described above are all passed through.

3.1. ldentification olthe sustainability problem

The very first step in operationalising sustainability is to identify the problem regarding
sustainability in the region under study. The area concemed is the island of Lesvos, one
of the north-eastem Aegean Islands in Greece, near to the Turkish co ast. It comprises
1630 square kilometres.
The main agricultural activity on the island is the cultivation of olives. The
approximately eleven million olive trees on the island render the island one of the most
important exporters of Greek olive oil. The next important branch of agricultural activity
is the breeding of goats and sheep. The primary sector accounts for about 25% of total
gross regional product, which means that the primary sector is a sector of substantial
importance. Employment on Lesvos is characterised by multi-activity; i.e., farmers do
not merely live from farming and farms are relatively smalI.
Since Lesvos is mainly a mountainous area and sometimes lacks sufficient water, the
Lesvian agriculture is characterised by terrace-cultivation. The mountainous conditions
form an impediment to the use of advanced agricultural implements. Another cause for
the traditional way of farming is the conventional attitude of Lesvian farmers. This
attitude can partly be explained by the fact that in the past few decades the island
population has been ageing. Young people left the rural areas and migrated to the urban
areas, especially to the mainland of Greece, where fast economic growth was stimulated
by the Greek govemment in the fifties (Loumou et al., 1995; Baaijens, 1996).
Centralisation of govemmental authorities in Athens, the breaking-up of the political
and economic relationships with Asia Minor after the liberation from the Turks, and the
fact that infrastructural facilities were less developed than on the mainland, also played a
role in the out-migration from the island of Lesvos in this century. Another factor
stimulating out-migration is the increasing importance of competitive substitute products
of olives and olive oil (Margaris, 1992; Loumou et al., 1995). The proliferation of such
products resulted in the economic decline of Lesvos, followed by an emigration of its
inhabitants to destinations in continental Greece and abroad. Both the out-migration and
the urbanisation in the mainland of Greece led to a decline in the rural population, a
development which threatens the continuity of the agricultural sector. The abandonment
and lack of maintenance of the olive groves is the most important symptom of the
declining importance of agriculture on the island of Lesvos.
The abandonment and neglect of agricultural areas, and of olive groves more
specifically, is the cause of the main environmental problem in the Lesvian agriculture,

67
viz. the threat of erosion and desertification. The abandonment of olive groves and of
plantations of other perennials has led to a neglect of the terraces, thus causing more
erosion. In addition, with the abandonment of terrace cultivation, animal husbandry
became more important on the island of Lesvos. The availability of feed for the animals
can be increased by setting frres to the vegetation of certain areas. Ouring the fIrst year
after such a frre, a lot of herbaceous plants appear, as a result of the activated
germination of seeds that lie in the soil seed bank. The increase in feed availability
however, is a short term increase, if intense grazing folIows. Then herbaceous plants are
either eaten or trampled by grazing animals, and the ecosystem has no time to regenerate.
The only plants which are able to survive are those which are resistant to grazing. Since
most of these plants are unpalatable, shepherds wish to get rid of them, and may set frres
more frequently, resulting in continuous degradation.
To combat erosion, new trees should be planted. The current market situation gives
no incentive to re-planting. In addition, reforestation is neither stimulated by any
governmental or European Community (EC) subsidy (cf. EC Regulation 2080/92). The
policy of the EC aims at maintaining current plantations, not at increasing the area and
production of perennials. The question is, whether alternative feasible policies will
change tbis environmental problem, and whether they will have an influence on economic
and social circurnstances in agriculture on Lesvos. The effects that a set of different
policies may have on sustainability will be discussed and evaluated hereafter.

3.2. Evaluation offuture options

The range of possible policies which affect agriculture on Lesvos is sizeable. In the
context of our paper, it is more useful and efficient to regard only a few contrasting
policy directions in order to get a clear idea of the differences and similarities among
alternative futures. We will discuss here three major policy scenarios. First, a
continuation of agriculture on Lesvos could be opted for. This means that an important
role is given to existing agriculture in the future of Lesvos, in the way it has always taken
place. Tbis scenario is called the 4S (FORCE) scenario, and will be discussed in more
detail hereafter. The second strategy for policy on Lesvos is towards environmental
priority. This means an explicit support for a preservation or improvement of the
environmental circumstances on the island. This scenario is called the GREEN scenario.
Third, one might choose not to intervene in the development of Lesvos and to rely on the
working of the market: the so-called 'invisible hand' will stabilise the markets and bring
the economy into equilibrium. In this case, the future of Lesvos will be determined by
economic forces. This scenario is the MARKET scenario. These three policy orientations
will now be discussed in more detail.
The 4S-scenario - StructuraL Support for Socio-economic Similarity - is a scenario of
structural policy which is based on EC Council Regulation 2019/93 (see Council of the
EC, 1993). This regulation is meant to support the smaller Aegean Islands in coping with
their specific socio-economic problems caused by their natural handicaps. The EC
recognises that the smallness of the islands thwarts an integrated development or

68
realisation of advantages of scale; that the smalI, ageing population in decline is cause for
concern; that the remote location of the production centres from mainland of Greece
hardly makes the exploitation of the islands' small amount of raw materials a viable
enterprise; that the dry climate and mountainous, infertile soil are negative factors; and
that the natural environment as a whole is very vulnerable. The EC also recognises that
the islands are scattered, which hinders commercial traffic and causes high transportation
costs. All these natural handicaps are to a large extent reflected on the island of Lesvos
and cause significant economic deprivation in agricultural income with respect to other
regions in the European Comrnunity. They also cause differences in the agricultural
structure in comparison to other Comrnunity regions, a fact demonstrated by the
different amounts of land available to a farmer, the different contributions of agriculture
to gross domestic product, and the different degrees of labour productivity (Slot, 1988).
This scenario of structural policy takes traditional agriculture as a starting point, but aims
at alleviating differences in income and socio-economic conditions. This means that the
production of several arable crops and fruits is supported, that various traditional
agricultural activities are promoted, and that a certain degree of policy control and a
public management of undesired movements in the agricultural markets, is strived for.
This has clear implications for land-use policy on the island. To summarise, then, 4S is a
scenario that aims to maintain, restore, or improve agricultural activities on Lesvos, so
that the socio-economic arrears will be overtaken.
The GREEN scenario - Gains to Ruralization and Environment ENlinked - is a
scenario of environmental care in agriculture. The starting-point of this scenario is the
idea that a farmer has a dual role as a producer of food and guardian of the countryside.
Managing the soil and the countryside is aprerequisite for the viability of agricultural
production in the long term. Too much emphasis on production, however, can lead to
over-intensification and thus to over-exploitation and degradation of the natural
resources on which agriculture itself depends. The basis of the GREEN scenario is an
EC-regulation which is also relevant for agriculture on Lesvos, viz. the Council
Regulation 2078/92 (see Council of the EC, 1992). This regulation exemplifies the
integration of environmental and agricultural policy, and stresses the dual role of farmers.
Therefore, the use of farming practices which reduce the polluting effects of agriculture
is promoted; an environmentally favourable extensification of agricultural practices is
furthered; the upkeep of abandoned farmland and woodlands is promoted; long-term set-
aside of agriculturalland is favoured; land use management for public access is furthered;
education and training for farmers is supported; and ways of using agricultural land
which are compatible with the protection and improvement of the environment, the
countryside, the landscape, natural resources, the soil and with genetic diversity are
advanced. Apart from fmancial incentives, much emphasis would have to be placed here
on information campaigns.
The third scenario is the MARKET scenario: Minimising Agroproduction Relief and
Knocking down Established Tariffs. This is a scenario of liberalisation of agricultural
markets and trade. The ideas in this scenario are in conformity with the sector
Agreement on Agriculture, a sector-specific agreement in the framework of the General

69
 Agreement on Tariffs and Trade (GATI). In the Agreement on Agriculture, concluded
du ring the Uruguay Round, it was agreed that a gradual liberalisation of agriculture
should be initiated, after agriculture had come to be effectively excluded from the reach
of GATI disciplines in the course of time. Roughly, the Agreement on Agriculture deals
with three issues: market access, domestic support, and export competition (Commission
of the EC, 1994; Hoekman, 1995; Trebilcock and Howse, 1995). Regarding market
access it was agreed that existing nontariff barriers (NTBs) should be converted into
tariffs and that new measures of this kind should be prohibited. The reason for this is that
NTBs interfere with efficiency and produce inequities, and that tariffs are more
transparent than NTBs (Kenen, 1994). It was also agreed that all agricultural tariffs be
bound, which means that the rates cannot be raised, unless specific conditions set out in
other parts ofthe GATI are satisfied. In addition, average tariffs should be reduced, and
minimum market opening criteria should be established through minimum import levels.
Secondly, domestic production support to agriculture is to decline, and fmally, export
subsidies are to be reduced. The export subsidy levels are bound, and the use of new
export subsidies is prohibited.
As a consequence of the provisions agreed upon in the Agreement on Agriculture,
two courses of actions and events are conceivable for the case of Lesvos island. First, we
may think of a future where smaU scale agriculture will disappear and where the Lesvian
agriculture will be dominated by a few large landowners. This would mean that fmancial
sources per farm would be larger, which could offer the possibility to use more advanced
farming methods, thus increasing production efficiency. This scenario is called
'MARKET I: Scaling Up'. The second possibility one could think of is that small
landowners will co-operate by sharing common agricultural equipment, purchased
together, by putting their products on the market commonly, etc .. This could give
farmers the opportunity to use more advanced farming methods and to lower some
overhead costs. In this version, caUed 'MARKET 11: Co-operation', each farmer's feeling
of responsibility for his own land and livestock would remain, since his income would
direcdy depend on these resources.
Having now concisely discussed the relevant scenarios, we have to assess the
sustainability consequences of each of these scenarios. Expectations regarding the
sustainability effects of each aforementioned scenario for Lesvos were expressed by
severallocal experts. Here, the impact methodology used is obviously an ad hoc impact
assessment method, using expert views. The experts were asked to express, what,
according to them, the influence of each scenario would be on each of the twenty
evaluation criteria or indicators distinguished in our study. In this way, the range of
qualitative foreseeable effects of each scenario can be assessed in a systematic way. The
expectations are summarised in Table 1. Although there is some variation in expert
opinion, it turned out that most experts had more or less the same view on the
foreseeable consequences of these scenarios. It should be noted that the list of indicators
is by no means ideal, but is based on the available information from the island. This may
mean that certain relevant data are excluded. Furthermore, there are certain variables
(like erosion, biodiversity, or water supply) which are not included either. However,

70
these are supposed to be a result of the stimulus variables in Table land can in principle
be taken into consideration, if an impact assessment model existed.
The results shown in Table I demonstrate that the effects of the 4S-scenario are
expected to be relatively positive, while the MARKET scenarios appeared to perform
worst, and GREEN had an intermediate position. However, on the base of the plain data
obtained mainly from interviews, the sustainability impacts of each scenario cannot be
accurately assessed. For a precise analysis of the scenarios, one needs evaluation
methods, thereby enabling a proper interpretation of the data. The flfst evaluation
method applied to the case of agriculture on Lesvos, is the flag evaluation method, as
described earlier. Clearly, this approach requires establishing a set of CTVs.
Acknowledging that CTV-levels may vary per region, local expertise has again been
consulted in establishing numerical values for the CTVs. The opinion of local experts on
CTVs is represented in Table 2, which again shows a reasonable degree of consensus.

Table I. Impact matrix.

45 GREEN MARKET MARKET
I 11
ECONOMIC
general & structural
I. GDP of the primary sector as a percentage of total + +/- -- --
GDP
2. average income out of farming activities as a + +/- +/- -
percentage of annual household expenditure
3. number of farms + +/- - +/-
4. farm size +/- +/- + +/-
livestock numbers
5. number of goats +/- - -- --
6. number of sheep +/- - -- --
7. number of cattle + - -- --
production figures
8. production of olives + + +/- -
9. production of meat + - -- --
10. production of milk + - -- --
land use
11. total agriculture area in use +/- - -- -
12. total area in fields and pasture +/- - -- -
13. surface area planted with olive trees
+/- +/- -- -
SOCIAL
14. total population +/- +/- - -
15. economically active in primary sector + +/- - -
16. employment in the primary sector as a percentage + +/- -- -
of total employment
ENVIRONMENTAL
17. number of olive trees +/- +/- -- --
18. olive yield per hectare +/- - + +/-
19. area of abandoned olive groves as a percentage of - - ++ +
total olive groves
20. number of sheep and goats per hectare of pasture +/- -- --
land
++ =suhstanllal mcrease; + = sltghl mcrease
+/- =neither increase. nor dccrease
- = slighl decrease; -- = suhstantial decrease

71
 TOOfe 2 C7Vs regarding Lesvian agriculture
lowest CTV highest CTV
ECONOMIC
general & structural •
1. GDP ofthe primary sector as a percentage oftotal GDP 25%
2. average income out offarming activities as a 20%
percentage of annual household expenditures'
3. number offarms 20,000

..
4. farm size 1 Ha
livestock numbers
5. number of goats 40,000
6. number of sheep 200,000
7. number of cattle' 9,000
production figures
8. production of olives 90,000 t./yr
9. production of meat 4,300 t./yr
10. production ofmilk 30,000 t./yr
land use
11. total agricultural area in use 60,000 ha
12. total area in field and pasture 60,000 ha
13. surface area planted with olive trees' 46,500 ha
SOCIAL
14. total population 90,000
15. economically active in primary sector' 8,000 9,000
16. employment in the primary sector as a percentage oftotal 30%
employment'
ENVIRONMENTAL
17. number of olive trees 11,000,000
18. olive yield per ha' 1,500 kg/ha 2,000 kg/ha
19. area ofabandoned olive groves as a percentage oftotal area 20%
of olive groves
20. number of sheep and goats per ha of oasture land 1.4 headlha
* refers to the nomos (department) ofLesvos

For most indicators present values are defined as critical threshold values. The idea
behind this is, that the current state of the environment is still acceptable even though the
environment of the island is indeed under pressure. This means that the sustainability
concem is particularly relevant in case of future changes in the current system.
Combining the CTVs and the projections for each of the three scenarios enables the
application of the flag approach. In light of the scarcely available qualitative expert
information, for none of the indicators numerical data related to each of the precise four
flags could be defmed; the information available on CTV s appeared to enable only either
a two-flag representation (red and green), or a three-flag representation (red, orange, and
green). Since the 4S scenario shows mostly green flags (based on a frequency counting),
the 4S-scenario appears to lead to the most sustainable situation. It also shows a lot of
mixed green/red colours, suggesting that many indicators fluctuate around the borders of
sustainability. The 4S-scenario performs very weil with regard to the economic and
social dimensions of sustainability, but its performance with regard to the environmental

72
aspect is ambiguous. Its performance regarding abandonment of olive groves is, for
example, sustainable, but the olive yield and the number of sheep and goats per hectare
of pasture land lead to a non-sustainable result.
The least sustainable strategies are found in versions of the MARKET scenario. Only
for some general economic and structural indicators more or less sustainable scores can
be identified.
The GREEN scenario appears to perform on an intermediate level. The economic
performance of this scenario is for the most part unsustainable, and in regard to the
social dimension, it balances on the edge of sustainability. Remarkable is the outcome of
the GREEN scenario with regard to the environment: the number of flags for the
environment do not show a better performance in this field than the other scenarios do,
although one would expect that the GREEN scenario would perform better.
As a complement to the flag evaluation, the (computerised) Regime method is used to
evaluate the scenarios. The Regime method is based on an impact matrix and a set of
political weights attached to the criterion effects. By considering all indicators - twenty,
in the case of Lesvos - more than 160 million different weight sets are, in principle,
possible. We will now examine one of the relevant possibilities in more detail.
The selection of weights is obviously a political matter: the importance attached to
each criterion, class of criteria or aspect of sustainability depends on political choices.
These political choices are inter alia based on certain assumptions regarding the present
state of affairs. One can assume, for example, that a wealthy society can afford to pay
more attention to the environment than a poor society can. This would mean that
stimulating economic growth would result in an improvement of environmental
circumstances. Economic stimulation would then be important, whereas environmental
policy as such would be of little importance: environmental improvement would then
follow economic growth as a matter of course. The weight set with regard to the three
aspects of sustainability would then be:

weighteconomic > weightsocial > weightenvironmentai.

The economic aspect of sustainability comprises four sub-aspects: general and

structural characteristics, livestock numbers, production Jigures, and land use. The
importance of these sub-aspects should be weighed against each other as weIl. One
might, for example, be of the opinion, that the general and structural indicators give the
best picture of the economic state of a sector in a region. Consequently, the category of
structural and general indicators would get the highest weight. Income, which is an
essential variable in the structural and general category, is generated by production
activities. Production figures would then be the second most important indicators of the
economic state. The capital of a farmer consists of - among other things - land and
livestock. Since olives are a more important agricultural product than meat or dairy
products on the island of Lesvos, land use will then be the third important category, and
livestock the fourth. The weight set regarding the economic sub-aspects will then be as
folIows:

73
 weightgeneral & Jlruclural > weightl'mduclion figur" > weightland UJe > weightlil'eJlock.

Finally, weights should he attached to each individual indicator. Assuming that the
economic indicators are quite homogeneous within each category, the weight sets for the
economic indicators are as follows (the numbers correspond to those given in Tables 1
and 2):

weightl = weight2 = weighlJ = weighl!,

weights =weighlö =weighb,
weightg = weight9 = weightlo,
weightll =weightl2 =weight l3.

With regard to the social aspect, a rational argumentation might he as folIows: the
'survival' of the island does not just depend on agriculture. Agriculture is indeed an
important sector, but the equilibrium on the island requires that people work in other
sectors as weIl. The most important social indicator thus is the total population. No
population, or too few people on the island, would me an no future for the island. To give
an indication of the significance of agriculture on the island, employment in the primary
sec tor as a percentage of total employment is a hetter indicator than the absolute
number of the economically active in the primary sector. As a result, the weight set
regarding the social aspect would be:

weightl4 > weightl6 > weightls.

The environmental aspect, in conclusion, can best be measured by the numher of

sheep and goats per hectare of pasture land, since one may argue that overgrazing is the
most important cause of the major environmental decay in Lesvian agriculture. The area
of abandoned olive groves as a percentage of the total area of olive groves gives a hetter
indication of the most important environmental problem on the island than olive yield per
hectare does. Olive yield per hectare is influenced by many other factors than the
environmental state of an area, like climate and way of harvesting. The least important
environmental indicator is the number of olive trees. The plantation of trees indeed
prevents the soil from eroding, but the mere number of olive trees gives no indication of
the quality of the plantation and the soil, and hence is not a proper identifier. The
resulting weight set would then be:

weight20 > weightl9 > weightl8 > weight l7.

74
 In summary, the weight set is as follows (in order of decreasing weight):

1. GDP of the primary sec tor as a percentage of total GDP

average income out of farming activities as a percentage of household expenditure
number of farms
fann size
2. production of olives
production of meat
production of milk
3. total agricultural area in use
total area in field and pasture
surface area plan ted with olive trees
4. number of goats
number of sheep
number of cattle
5. total population
6. employment in the primary sector as a percentage of total employment
7. economically active primary sector
8. number of sheep and goats per hectare of pasture land
9. area of abandoned olive groves as apercentage of total area of olive groves
10. olive yield per hectare
11. number of olive trees.

Given the ranking of the policy weights, a multiple criteria Regime method can be
applied. We will not give here the details of the method and its results (see Nijkamp et al.
1991), but mainly focus on the general fmdings. Applying then the Regime method to the
weight set leads to the following ranking of policy scenarios:

I.4S
2. GREEN
3.MARKETI
4. MARKET 11.

We see that, in this case, the 4S scenario is regarded the most sustainable scenario.
This means that socio-economic support should be preferred to a policy of environmental
care, and to a policy of liberalisation. Liberalisation might increase net economic welfare
in a certain region, but, as a result of welfare redistribution, the Lesvian agriculture is
worse off if agricultural markets are liberalised. Therefore, sustainability is not ensured in
case of the MARKET scenario.
Clearly, this outcome is just one of the many possible outcomes. In 99.8 percent of an
weight sets, however, 4S appears to rank first, and in 99.7 percent of an weight sets,
both MARKET scenarios are the worst performing scenarios. just as in the case
described above.

75
 If we look at the results in more detail, the Regime method designates the 4S-scenario
as the most sustainable scenario for all weight sets, if just the economic aspect is
considered. The GREEN scenario is the second most sustainable in that field, and the
MARKET scenarios are the least sustainable.
The same holds for the social aspect: all weight sets make GREEN rank second,
before the MARKET scenarios, and after the 4S scenario. The outcome of the Regime
method regarding the environmental aspect however, is not so unambiguous. More than
40 percent of the weight sets leads to the highest ranking of 4S, while about 30 percent
of the weight sets regarding the environmental indicators results in the highest ranking
for the GREEN scenario. Another 30 percent of the weight sets designates the
MARKET I scenario as the most environmentally sustainable, whereas the chance that
MARKET II is ever ranked fIrst, is almost nil. Although one cannot say that 4S is
conclusively perfonning the most sustainably in regard to the environmental state of
affairs (followed by GREEN, MARKET I, and, fmaIly, MARKET II), it appears that the
main share of the weights indeed leads to a number one ranking of the 4S scenario.
Clearly, it is not certain whether all of these weight sets are equally plausible (for a
complete overview of the possible weight sets and the accompanying scenario rankings,
see Hennanides, 1996).
If we compare the results of the flag evaluation with those of the Regime evaluation,
one may conclude that the final outcome of the flag evaluation corresponds to the
majority of the outcomes of the Regime evaluation. The outcomes of both evaluation
methods per aspect of sustainability are largely similar as weIl, although a scenario
ranking regarding the environmental aspect is somewhat complex. The fact that the
Regime method allows for policy priorities expressed by weights, clearly makes this
method more subtle than the flag method.

4. Final assessment
The fmdings of our research are not just of theoretical value, but of practical value as
weIl because we can learn from them in several ways. First, of course, the evaluation of
policies suggests a direction in which a sustainable policy regarding the Lesvian
agriculture should be developed. It appears from the policy evaluation that the
sustainability and consequently the continuity of agriculture on Lesvos is best served by
public, socio-economic support. If the Lesvian agriculture is left to fend for itself, it is
doomed to unsustainability, and in the long run agriculture might disappear from the
island altogether. Nevertheless, one may still choose to liberalise agricultural markets,
since it might yield a net econornic benefit. In this case, the future of Lesvos would be
more uncertain. The question is, then, what economic activity will replace agricultural
activity. Lesvos' natural handicaps may not just fonn obstacles for t~e agricultural sector,
but they might also cause disadvantages to other sectors. FinaIly, dramatic emigration
might result in the worst case.
The second lesson to be leamt from the findings of the research concems the

76
shortcomings encountered in the research methodology and the flaws in the application
of the methodology. A serious problem was faced in identifying the experts' estimates
for the future. They were asked to express how indicators would change with respect to
the present state of affairs. To use these estimates properly in the flag model however,
one should not ask how things would change in respect to the present indicator values,
but in respect to the critical threshold values. This problem obfuscated fIfteen percent of
the picture resulting from the flag approach.
In addition, a more balanced view on the future might have been obtained if more
experts from divergent disciplines could have been consulted. The experts consulted in
our case had more or less the same background: they were mainly ecologists and
agricultural engineers. This may have resulted in a more coloured picture of the future
expectations.
Finally, some further research on feedback effects may be useful, in order to refme the
sustainability methodology. Each of the four steps of the sustainability methodology
follows the previous one in a logical, almost natural order. However, the result of each
step might influence one or more of the previous steps, as a kind of feedback. For
example, the results of the policy evaluation can defmitely change one's view on the
sustainability problem, which is identified in the fIrst step of the methodology. If the
evaluation has given insight into the impact of a certain policy on sustainability, and if the
policy is about to be pursued, then this may change the sustainability problem. As a
result, other indicators will have to be selected in successive steps, and the expected
indicator values may then be different. This makes the methodology a repeating process,
in wh ich each and every step can be adjusted and readjusted.
A second feedback-loop may also be distinguished: the recommendation concerning
the future of Lesvos, resulting from the policy evaluation, should be re-considered if new
policy plans are envisaged and formulated. The new policy plans should then pass
through the entire procedure of the methodology, and areal 100p' with a feedback
structure would consequently have to be created. There is sufficient scope for further
experimental research in the fIeid of sustainable agriculture.

References
Baaijens, S.R. (1996), Raming van Toeristische Inkomensmultipliers met Meta-analyse: Een
casusstudie voor Lesbos,M.A. Thesis, Amsterdam, Vrije Universiteit.
Bruyn, S.M. de, J.B. Opschoor (1994), Is the Economy Ecologising?: De- or Relinking Economic
Development with Environmental Pressure, Tinbergen Institute Discussion Paper TI 94-65,
Amsterdam, Tinbergen InstituteIVrije Universiteit.
Blaas, E., P. Nijkamp (1992), Methods of Regional Impact Assessment: A New Approach to the
Evaluation of the Performance of the European Regional Development Fund, Research
memorandum 1992-70, Amsterdam, Vrije Universiteit.
Coombes, K, C. Wong (1994), Methodological Steps in the Development of Multivariate Indexes
for Urban and Regional Policy Anal ysisßnvironment and PLanning A, 26, pp. 1297-1316.
Commission of the European Communities (1994), Uruguay-ronde: MondiaLe OvereenkomH -
MondiaLe VoordeLen, Luxemburg, Bureau voor ofticiele publikaties der Europese

77
 Gemeenschappen.
Council of the European Communities (1992), Official Journal. of the European Communities No L
215, pp. 85-92. Brussels.
Council of the European Communities (1993), Official Journal of the European Communities No L
184, pp. 1-7. Brussels.
Fresco, L.O., S.B. Kroonenberg (1992), Time and Spatial Scales in Ecological Sustainability. Land
Use Policy. 9, pp. 155-168.
Hermanides, G.c. (1996), Sustainability Analysis in Agriculture: A Multicriteria Evaluation of
Policy Scenarios for Lesvos,M.A. Thesis, Amsterdam, Vrije Universiteit.
Hinloopen, E., P. Nijkamp, P. Rietveld (1982), Qualitative Discrete Multiple Criteria Choice
Models in Regional Planning,Research memorandum 1982-23, Amsterdam, Vrije Universiteit.
Hoekman, B.M. (1995), Trade Laws and Institutions: Good Practices and the World Trade
Organisation, World Bank discussion papers No. 282, Washington DC: The World Bank.
Keren, P.B. (1994),The international economy, Cambridge University Press, Cambridge.
Loumou, A., S. Koukoulas, Ch. Giourga, P. Dimitrakopoulos, N.S. Margaris, 1. Hatzopoulos
(1995), The Contribution of the Tourism in the Conservation of the Agro-ecosysterns: The Case
ofLesvos Island, Greece, Unpublished manuscript, Mytilini: University ofthe Aegean.
Margaris, N.S. (1992), Primary Sector and Environment in the Aegean Islands, Greece.
Environmental Management. 16, pp. 569-574.
Nijkamp, P. (1996), Sustainability analysis in agriculture: a Decision Support System framework,
Research Memorandum, Dept. ofEconomics, Free University, Amsterdam.
Nijkamp, P., P. Rietveld, H. Voogd (1991), Multicriteria Analysisfor Physical Planning, Elsevier,
Amsterdam.
Opschoor, J.B., R. Weterings (1994), Environmental Utilisation Space: an Introduction, Milieu. 9,
pp. 198-205.
Pelt, MJ.F. van (1993), Ecologically Sustainable Development and Project Appraisal in
Developing Countries, Ecological Economics. 7, pp. 19-42
Reed, D. (1992), Structural Adjustment and the Environment. London, Earthscan Publications
Ltd.
Slot, N. (1988), Structurele en Regionale Problemen in de EG-Iandbouw, In: EG-Landbouwpolitiek
van Binnen en van Buiten (J. de Hoogh and HJ. Silvis, eds.), Wageningen, Pudoc.
Trebilcock, MJ., R. Howse (1995), The Regulation of International Trade: Political Economy
and Legal Order, pp. 205-210, LondonINew York: Routledge.
WCED, Our Common Future, Oxford Community Press, New York, 1987.

78
Multicriteria evaluation methods in renewable resource
management: integrated water management under
drought conditions

IGiuseppe Munda, 2Massimo Parruccini and 3Giuseppe Rossi

JUniversitat Autonoma de Barcelona, Spain
2European Commission, Joint Research Centre, lspra, Italy
3Universita di Catania, Catania, Italy

Abstract
The limits inherent in conventional decision theory methodologies and the necessity to analyse
conflicts between policy objectives have led to many calls for more appropriate analytical tools
for strategie evaluation. As such multicriteria evaluation does not itself provide a unique criterion
for choice; rather it helps to frame the problem of arriving at a political compromise. This paper,
as a first step, deals with the role of multicriteria evaluation methods in the framework of
renewable natural resource management. As a second step, the possibility is studied of using
multicriteria evaluation methods to tackle problems of integrated water management under
drought conditions. We take water management under drought conditions into account because
this is an important issue in many southern countries. The term "integrated water management" is
used because it is evident that the problems underlying water management can be dealt with only
if all the conflicting activities and uses that affect the resource are taken into account. A real-
world example in the area of the city of Palermo (western Sicily) is also considered' From this
case study, which is apart of a larger project that has been commissioned by the Sicily region, a
number of useful lessons can be learned for comparing alternative strategies for the management
of a water system under drought conditions.

Keywords: environmental sustainability, water management, drought conditions,

multicriteria evaluation, fuzzy uncertainty, NAIADE method.

1. Multicriteria evaluation in integrated natural resource

management

The growth of the world's population and the rapid growth of economic activities have
caused environmental stress in all socio-economic systems. There is a wide scientitic
consensus that problems such as the greenhouse effect, ozone depletion, acid rain, loss of
biodiversity, toxic pollution and renewable and non-renewable resource depletion are
clear symptoms of environmental un-sustainability. It is evident that fabricated
environments (cities, industrial areas, airports, etc.) are neither self-supporting nor self-
maintaining. They are sustained by their dependence on the solar-powered natural and
domesticated environments (life-supporting ecosystems). Stress, caused by the disposal

79
of wastes and pollutants. negatively affects recycling. feedback loops and control
mechanisms in the life-supporting ecosystem and thereby the production and
maintenance of environmental goods and services.
In the eighties. the awareness of actual and potential conflicts between economic
growth and the environment has led to the concept of "sustainable development". Such
a concept is based on the assumption that certilin sorts of natural capital are deemed
critical. and not readily substitutable by man-made capital (see e.g. Costanza et al.. 1996;
Munda. 1997; and for applications in water management, Simonovic. 1996). As a
consequence. a sustainable economy is an economy that maintains the exploitation of
renewable resources (fish. trees. water) within the limits of their rates of renewal; it is an
economy that uses non-renewable resources (e.g. petroleum) in line with achievement of
a transition (through substitution. technological change. etc.) to a base of renewable
resources. It is an economy that keeps the production of waste and contamination within
the assimilative capacity of the environment. A sustainable economy is an economy that
preserves biodiversity in all its express ions.
Unfortunately, human beings are currently witnessing the emergence of serious
problems in the ecological, economic and social dimensions. These problems are
interconnected and urgent, thus there is a need to develop synthetic conceptual
framework and analytical approaches which begin to explore the interconnections among
these three dimensions and the prospects for solutions that simultaneously address
problems in all those domains. This is the aim of integrated environmental assessments
(IEA) (Bailey et al., 1995). Most IEA applications are based on a cost-benefit analysis
approach.
Cost-benefit analysis (CBA) is based on the neo-cIassical maximisation premise on
behaviour, stating that rational decisions coincide with utility maximisation. This
approach is based on the assumption that different objectives can be expressed in terms
of a common denominator by means of trade-offs (complete commensurability), so that
the loss in one objective can be evaluated against the gain in another. From a theoretical
point of view, the optimising principle is elegant since it provides an unambiguous tool to
evaluate alternative strategies on the basis of their contributions to community welfare.
From an operational point of view, the value of the optimising approach is rather limited,
because the specification of a community welfare function requires complete information
about all possible combinations of actions, about the relative trade-offs between all
actions and about all constraints prevailing in the decision making process. Such
information is generally not available in the context of environmental decision making,
and in any case the validity of the proposed trade-offs is likely to be contes ted by
affected groups (Munda. 1996).
During the last two decades, more support has emerged for the view that welfare is a
multidimensional concept (Bana e Costa, 1990; Munda. 1995; Nijkamp et al .• 1990;
Paruccini. 1994). One should note that incommensurability (i.e. taking into account of all
the different dimensions of adecision problem without any monetary reductionism) does
not imply incomparability. It is possible to distinguish between the concepts of strong
commensurability (common measure of the different consequences of an action based on

80
a cardinal scale of measurement), weak commensurability (common measure based on an
ordinal scale of measurement), strong comparability (there exists a single comparative
term by which all different actions can be ranked), and weak comparability (one has to
accept the existence of conflicts between all different consequences of an action)
(Martinez-Alier et al., 1996). Weak comparability can be considered to be the
philosophical basis of multicriteria evaluation.
A great number of multicriteria methods have been developed and applied for
different policy purposes in different contexts. As a tool for conflict management,
multicriteria evaluation has demonstrated its usefulness in many environmental
management problems (Munda et al., 1994). From an operational point of view, the
major strength of multicriteria methods is their ability to address problems marked by
various conflicting evaluations. Real-world problems are generally not direct win-Iose
situations and a certain degree of compromise is needed. Multicriteria evaluation
techniques cannot solve all conflicts, but they can help to provide more insight into the
nature of conflicts and into ways to arrive at political compromises in case of divergent
preferences in a multi-group or committee system so increasing the transparency of the
choice process.
In general, a multicriteria model presents the foJlowing aspects:
• there is no solution that optimises all the criteria at the same time and therefore the
decision maker has to find compromise solutions;
• the relations of preference and indifference are not enough in this approach, hecause
when an action is hetter than another one for some criteria, it is usually worse for
others, so that many pairs of actions remain incomparable with respect to a dominance
relation.

The main advantage of multicriteria models is that they make it possible to consider a
large number of data, relations and objectives which are generally present in a specific
real-world decision problem, so that the decision problem at hand can be studied in a:
multidimensional fashion. On the other side, an action a may be better than an action b
according to one criterion and worse according to another. Thus when different
conflicting evaluation criteria are taken into consideration, a multicriteria problem is
mathematically ill-defmed. The consequence is that a complete axiomatization of
multicriteria decision theory is quite difficult (Arrow and Raynaud, 1986).
According to Simon (1983), a distinction must be made between the general notion of
rationality as an adaptation of available means to ends, and the various theories and
models based on a rationality which is either substantive or procedural. This terminology
can be used to distinguish between the rationality of adecision considered independently
of the manner in which it is made (in the case of substantive rationality, thc rationality of
evaluation refers exc1usively to the results of the choice) and the rationality of adecision
in terms of the manner in which it is made (in the case of procedural rationality, the
rationality of evaluation refers to the decision making process itself). Other authors
quest ion the usefulness of choice as a pervasive metaphor for describing and interpreting
human behaviour in the traditional paradigm of decision making, c1aiming that thc issue

81
is not that of choosing but of generating (March. 1978; Winograd and Flores. 1988).
Thus. we can conclude that the validity of a given multicriteria method depends on two
main factors (Roy, 1985):
• mathematical and descriptive properties which make it conform to given
requirements;
• the way it is used and integrated in adecision process.

An important point is the way in which a problem is structured. The results of any
decision model depend on the available information. Since this information may assume
different forms, it is useful that decision models can take this differentiation into account.
Another problem related to the available information concerns the uncertainty associated
with this information. Ideally, in the Laplacian perspective on science and decision
making, the information should be precise, certain, exhaustive and unequivocal. But in
reality, it is often necessary to use information which is known not have those
characteristics.
If it is impossible to establish exactly the future state of the problem faced, a
stochastic uncertainty may be postulated; this type of uncertainty is weil known and it
has been thoroughly studied in probability theory and statistics. According to a well-
established tradition originating with by the c1assical contributions of Knight (1921) and
Keynes (1921), a distinction is made between two kinds of uncertainty: a weak variety
called "risk" and a strong one called "uncertainty". In particular, risk refers to probability
distributions based on reliable c1assification of possible events and uncertainty refers to
events whose probability distribution is not fully definable for lack of reliabIe
classification criteria. Existing decision theories, based on substantive rationa1ity, are
lirnited in dealing with environmental decision situations characterised by a high degree
of uncertainty about nature, incidence and/or timing of possible environmental costs,
especially situations where the effects may be catastrophic for future generations.
Another framing of uncertainty, called fuzzy uncertainty, focuses on the ambiguity of
information, in the sense that the uncertainty does not concern the occurrence of an
event but the event itself, which cannot be described unambiguously (Zadeh, 1965). This
sort of situation is readily identifiabIe in complex systems. Spatial-environmental systems
in particular, are emergent complex systems characterised by subjectivity, incompleteness
and imprecision (e.g., ecological processes are quite uncertain and little is known about
their sensitivity to stress factors such as various types of pollution). Fuzzy set theory is a
useful mathematical theory for modelling situations of such a type, i.e. it aims to portray
in terms of fuzzy uncertainty, some of the indeterminacy of the socio-ecological system
under study.

2. The NAIADE method

In designing models for environmental and resource policy making the following three
main types of policy objectives may be distinguished:

82
• socio-economic objectives, e.g. "maximum production of goods and services";
• nature conservation objectives, e.g. "minimum exploitation of natural systems",
"optimum yield";
• mixed objectives, e.g. "maximum sustainable use of resources and environmental
services at minimum (private and social) cost".

It is dear that in policy-relevant economic-environmental evaluation models, socio-

economic and nature conservation objectives are to be considered simultaneously.
Consequently, multicriteria methods are in principle an appropriate modelling tool for
environmental decision making issues: a compromise solution that takes account of
different conflicting values can in principle be identified.
Given the problem of the differences in the measurement levels of the variables used
for economic-ecological modelling, multicriteria methods able to deal with mixed
infonnation (both qualitative and quantitative measurements) can be considered
particularly useful. A new multicriteria method, based on some aspects of the partial
comparability axiom, called NAIADE' (Novel Approach to Imprecise Assessment and
Decision Environments) has been developed in Munda (1995). It is a discrete
multicriteria method whose impact (or evaluation) matrix may include either crisp,
stochastic or fuzzy measurements of the performance of an alternative an with respect to
a judgement criterion gm, thus it is very flexible for real-world applications. From an
empirical point of view, this model is particularly suitable for economic-ecological
modelling incorporating various degrees of precision of the variables taken into
consideration. From a methodological point of view, two main issues are then faced:

• the problem of equivalence of the procedures used in order to standardise the various
evaluations (of a mixed type) of the perfonnance of alternatives according to
different criteria;
• the problem of comparison of fuzzy numbers typical of all fuzzy multicriteria
methods.

Both problems are tackled by means of a new semantic distance presenting relevant
theoretical properties (Munda et al., 1995). Since in a fuzzy context, any attempt to
reach a high degree of precision on the results tends to be somewhat artificial. a pairwise
linguistic evaluation of alternatives is used. This is done by means of the notion of fuzzy
relations and linguistic quantifiers. In the aggregation process, particular attention is paid
to the problem of diversity of the single evaluations, while the entropy concept is used as
a measure of the associated "fuzziness". Such linguistic evaluations can be used in
different ways according to the decision environment at hand.
In short, the main properties of the NAIADE method can be synthesised as follows
(for more details see Munda, 1995):

Naiade is the Italian name for the Greek nymphs of rivers.

83
• communication with the decision maker is required to elicitate different relevant
parameters, thus a constructive decision aid framework is implied;
• the method is based on some aspects of the partial comparability axiom, in particular,
a pairwise comparison between alternatives is carried out, and incomparability
relations are allowed;
• intensity of preference is taken into account; this implies that a certain degree of
compensation between criteria is allowed, given the characteristics of the method it
may be c1assified among partial compensatory methods;
• for the indifference relation no transitivity is implied, the preference relation is max-
min transitive;
• a partial (or total) order of feasible alternatives is supplied. It has to be noted that the
fmal ranking is a function of all the alternatives considered, this implies that if a
dominated or a dominating action is introduced, the ranking may change; moreover if
the best action is eliminated, the ranking of the other alternatives mayaIso change,
thus NAJADE does not respect the independence of irrelevant alternatives axiom.

Since in environmental and resource management and policy aiming at an ecologically

sustainable development many conflicting issues and interests emerge, particular
attention has to be given to the problem of different values and goals of different groups
in society. Equity and conflicting values in multicriteria decision aid are traditionally
introduced in two different ways:
(I) by weighting the different criteria (unfortunately, in public decision making a single
point-value solution (e.g. weights) tends to lead to deadlocks in adecision process
because it imposes too rigid conditions to reach a compromise);
(2) by taking into consideration a set of ethical evaluation criteria (a weak point of this
approach is that it could lead to an excessive number of evaluation criteria.
Furthermore, the .identification of ethical criteria may not be an easy task).

A third possibility is the use of conflict analysis procedures to be integrated with

multicriteria evaluation in order to allow policy-makers to seek for "defendable"
decisions that could reduce the degree of conflict (in order to reach a certain degree of
consensus) or that could have a higher degree of equity on different income groups.
NAIADE uses a fuzzy conflict resolution procedure. Starting with a matrix showing the
impacts of different courses of action on each different interestlincome group, a fuzzy
c1ustering procedure indicating the groups whose interests are c10ser in comparison with
the other ones is used. Therefore, fmally a compromise solution that takes into account
the three conflicting values of efficiency, equity and sustainability can in principle be
identified.

84
3. Multicriteria analysis of integrated water management
under drought conditions: the case of Palermo

3.1. Decision making in water system management under drought

conditions

Water resource management is characterised by the presence of a strong competition

among different categories of consumptive water uses and, as a consequence, among
various interest groups. Such a competition also exists between consumptive uses as a
whole and "ecological uses" (strongly supported by environmental pressure groups)
which aim at limiting water diversion for off-stream uses in order to preserve the
ecological equilibrium of ecosystems (in-stream flow uses). It is evident that the issues
underlying water management can be dealt with only if all the competitive uses and
activities, that effect the resources, are taken into account according to the concept of
"integrated water management" (OECD, 1989).
This permanent condition of competition may become areal conflict under drought
conditions, i.e. when there is a temporary and casual reduction of available water
resources due to a long and severe decrease of rainfall (compared to mean or median
natural values). The problem of water shortages due to drought is particularly relevant in
Southern Europe for the following main reasons:
• the high variability of the annual precipitation,
• the increase of the used percentage of the mean available water resources (this leads
to an increased risk of not satisfying water demands during the years with the biggest
deviations from normal conditions).
As a consequence, correct water management rules have to take into account the need
for various measures that can reduce the risk of water shortages and/or minimise drought
impacts. Water shortages depend not only on hydrological drought which in turn follows
from meteorological drought, but also on water supply system characteristics and
demand levels, which are both affected by different drought mitigation measures (see
Figure 1).
When a complex water supply system is considered, i.e. a water system which has to
satisfy multiple purposes (e.g. municipal, irrigation and industrial water demands) by
using various water sources, water management under drought conditions implies
decisions on (Rossi et al., 1995):
• the sharing of water shortages among different users,
• the inter-temporal distribution of release from each source,
• the priority of release from various sources.
In the following sections, we will present a multicriteria approach to this kind of
decision making situation in the case of the water supply system of the city of Palermo
(western Sicily). This case study is part of a project which has been commissioned from
the Sicily region and executed in the context of the European Commission DGXVI

85
structural funds (Paruccini, 1996). This case study was developed over two years of
interaction mainly between the authors of this article and the management body of the
water supply system of the city of Palermo (plus some social actors involved in the final
step of the study).

MEASURESFOR
/ DEMAND REDUCTION
~------'

Metereological
drought

1 Economical losses
and intangible
impacts

MEASURESFORSUPPLY MEASURES FOR DROUGHT

INCREASE OR DEFICIT IMPACT REDUCTION
RlSK REDUCTION

Figure 1. Drought Effects on Water Supply System and Socio-Economic System.

3.2. The Palermo Water Supply System

In general terms, a water system is composed of a set of natural elements (water bodies),
a set of artificial elements (e.g. reservoir, conveyance conduit, distribution network) and
a set of institutional-management structures which interact to achieve one or more
objectives (Rossi et al.. 1995). A water supply system includes all facilities which allow
the supply of surface water, ground water and non-conventional resources to different
types of users (municipalities, irrigation districts, industrial districts and so on).
The water system of Palermo has to supply water to municipal, agriculture and
industrial users by using surface water and groundwater. The water of the "Piana degli
Albanesi" reservoir is also used for energy production. In order to evaluate the
vulnerability of the water supply system to future severe droughts, here a few facilities
that are not yet in operation, such as the "Rosamarina" reservoir and the "Acque dei
Corsari" treatment plant, have been introduced in the scheme of the system.
The Palermo water system under consideration includes (see Figure 2):
• The reservoir of Scanzano, on the river Eleuterio, which provides mainly drinking
water (only a very sm all percentage is used for agriculture);
• The reservoir of Piana degli Albanesi on the river Belice is mainly used for energy
production. Only the volume above 11 Mm 3 is available for agriculture and drinking

86
 water. The percentages are 90% for drinking water and 10% for agriculture. The
upper annual constraints are 9 Mm 3 for drinking water and I Mm 3 for agriculture;
• The reservoir Poma, on the river Iato. is used for agriculture (60%) and drinking
water (40%). Sometimes part of the water is used for municipalities close to
Palermo. The total annual available volume is 30 Mm 3 ;
• The spring-water Scillato mixed with the waters of "Galleria Presidiana" (in the
ratio of 3: I) supplies drinking water to Palermo and some municipalities in the east
ofthe city;
• Other springs (Risalaimi, Gabriele) which provide water to Palermo and several
wells supplying municipal and agricultural uses;
• A few surface water diversions (Imera, Monte Tesoro, S.Caterina) devoted to supply
drinking water to Palermo;
• The reservoir Rosamarina, on the river San Leonardo is used for agriculture (65%)
and drinking water (35%). The total annual available volume is 55 Mm 3 ;
• The waste waters of Palermo, treated by the plant of "Acqua dei Corsari" are used
for agriculture only in the volume of 5 Mm 3 (the annual possible maximum volume
being 13 Mm\

3.3. Drought mitigation alternatives and evaluation criteria

The alternative management options under drought conditions can be divided ioto two
main groups:
• alternatives that try to satisfy 100% of the water demand (group A),
• alternatives that do not satisfy completely the water demand (group B).

Group A:
al: Reduction of the hydroelectric use of water in Piana degli Albanesi (i.e., it is possible
to use the water volume under ll Mm 3 for municipal use). In this case, a penalty to the
company managing the hydroelectric production (ENEL) has to be paid.
82: Use of the waste water treatment plant of "Acqua dei Corsari" up to its maximum
capacity (13 Mm 3) for irrigation.
a3: Relaxation of the ecological constraint of maintaining minimum io-stream flows on
the whole set of rivers and use of these resources for water supply.

87
 .bll
O.. t.ioR 8oIiaoil

SCANZANO PlANA ~G.I ALBANESI PQMA

Reservoir Reservoir Reservoir

R ~N I
GftlRIElE
NOIfIE TESORO
ItSUIl G61EU
IM[ RA

PN.ERMO
MunicicolilY
CICIIA

"'OOA 08
~

A Reoervoir D> Purifloalion Plan! 0 Infiltrolioo Gallery

tra O""';on ~ West. Wot... Trmtmenl Su~1'iY conduR

@ Weil EI H\'drooowerPlanl
Il<mond Zone
IX> Sprin~ 0 PurnD Sla!ion D
Figure Z. Scheme afthe Palermo Water Supply System.
Group B:
The demands are satisfied only in the following percentages:
bl: Drinking water 100%; agriculture 85%; and industry 85%.
b2: Drinking water 100%; agriculture 70%; and industry 85%.
b3: Drinking water 90%; agriculture 70%; and industry 85%.

These alternatives were evaluated for the longest drought experienced in the past (4
years) according to the following set of criteria.

Economic criteria
These criteria measure the monetary impact of the different alternatives on the different
economic activities of the region. Costs are computed for the 4 years of the drought
period. 80th costs and returns have been considered for the organisation in charge of the
water system (AMAP):
(1) The costs for AMAP include penalties due to ENEL, total costs connected to water
pumping, and incremental costs for the operation of the distribution network under
drought conditions.
(2) Returns for AMAP include all categories of users and are referred to the consumed
volumes.
(3) Industrial costs measure the additional costs of searching for alternative water
sources du ring drought events. The industrial areas of Carini and Termini Imerese
are considered.
(4) Agriculture is very sensitive to water supply. The criterion used measures the total
return of each alternative.

For AMAP and industry the values are crisp. For agriculture the uncertainty is quite
high. so symmetric fuzzy numbers of a Gaussian form are used. All the values are in
Billion Italian Lire (approximately 1US Dollar is 1600 ltalian Lire).

Social criteria
Hygienic risk and social discomfort are considered.
(5) Hygienic risk refers to the risk of water contamination Iinked to low pressure in the
water distribution network. This phenomenon is connected to a number of different
factors, e.g. maintenance of the network. sewer conditions and distribution, etc ..
Here, a good proxy has been considered the number of days in four years for which
the threshold of unsatisfied demand is exceeded.
(6) Social discomfort is considered as a consequence of three hydrologic indicators
obtained by means of simulation models. The indicators are average quantity of
water not supplied (deficit), frequency of water deficit, and maximum quantity not
supplied in a month. The aggregation of these indicators is not easy. After technical
simulations, a linguistic score has been selected based on the ranking obtained with a
multicriteria application.

89
Environmental criteria
(7) The in-stream flow requirement (IFR), defined as the discharge which maintains a
stream ecosystem or aquatic habitat, has been considered because it is easy to
quantify and gives a good picture of the ecological conditions of the ecosystems
affected. This criterion measures the water deficit to satisfy the minimum in-stream
flow requirement constraint. The measurement unit is 106 m3 per year.

The multicriteria evaluation matrix shown in Table I shows all the performance scores
of the alternatives under evaluation.

Table I. Multicriteria evaluation matrix.

business al a2 a3 bl b2 b3
as usual
l.costs AMAP
(l031T. Lire) 106,968 101,988 103,106 99,832 106,121 106,831 100,721

2.retums AMAP
(103 1T. Lire) 523,321 549,240 575,514 559,747 527,193 541,411 515,270

3.Incr.Ind.par costs
(109 1T. Lire) 0 0 0 0 6437 6437 6437

4.Agricul. product. about about about about about about about

(109 1T. Lire) 864 865 1143 864 737 618 711

5.Hygienic risk
(No. of days) 89 54 36 34 83 54 33

6.Social Discom. very high more or moderate moderate very high more or high
(Linguis.) less high less high

7.Missing waterfor 48.4 48.4 43.1 97.3 45.8 44.7 39.8

IFR(l(f m3)

3.4. Application 0/ an aggregation procedure: the NAJADE method

Given that our problem is characterised by mixed information and various distributional
issues, the NAJADE method has been considered particularly suitable. By applying
NAJADE to the impact matrix described in Table 1, a quite stable ranking of the
alternatives is obtained. The subset {al, a2, a3 and b3} contains the best alternatives. b3
is always worse than the others. Thus, the decision is among al, a2 and a3. In particular,
a2 seems to be slightly beUer than the others.

90
 The interest groups taken into account are the following: city inhabitants, farmers,
industrial organisations, ENEL, AMAP and environmentalists. The impacts have been
evaluated by means of interactions with relevant local people (various interviews with
representatives of the interest groups). The equity impact matrix is shown in Table 2.

TOOfe 1. Equity Impact Matrix.

business al a2 a3 bl b2 b3
as usual
city bad moderate good good bad moderate more or
inhabit. less bad

Farmers more or more or very more or moderate bad bad

less good less good good less good
Industr. good good good good bad bad bad

ENEL good bad good good good good good

AMAP more or more or good good moderate moderate more or

less bad less good less bad
Environ. moderate moderate good very bad good good good

By applying NAIADE, the dendrogram shown in Figure 3 is obtained. The results are
quite weIl in agreement with prior expectations about the attitudes and behaviour of the
interest groups. The interests of the city inhabitants and AMAP seem to run fairly
parallel. The same applies to the industrial organisations and the farmers. Anyway, the
group composed by farmers, industrial organisations, city inhabitants and AMAP may
have some common compromise solutions. ENEL and environmentalists presents a more
individualistic character. One should note that a2 has a good impact on all the interest
groups considered. On the contrary al and a3 could be vetoed by ENEL and
environmentalists, thus by considering the minority principle a2 would be strongly
recommended.

4. Conclusions

Some lessons were leamed from the experience made in this case study. First of all, the
use of a multicriteria framework is a very efficient tool to implement a multidisciplinary
approach necessary in natural resource management. The technical actors involved had
various backgrounds (mainly in engineering, economics and mathematics). At the
beginning the communication process was very difficult. However, it was astonishing to
realise that a common language emerged as soon as it was decided to structure the
problem in a multicriteria fashion. Since some of the criterion scores were characterised
by approximate evaluations, the use of fuzzy sets was very useful. It was also evident

91
that the possibility of taking into account distribution issues explicitly increased the
transparency of the study. It also made the interaction with and between various social
actors possible and effective. We think that these characteristics of the approach make it
suitable for applications in other spatial and cultural contexts.

City Inhab. AMAP Form. Industrial. ENEL Env.

0.707
0.706

0.638

0.532

0.506

Figure 3. Dendrogram ofthe coalitionformation process.

The results of this study were considered relevant for the real-world integrated
management of Palermo water supply system. However, one has to admit that time and
resource constraints limited the possibility of a better structuring of the decision problem
at hand (e.g. a more complete set of criteria, more technical simulations on the
hydrological behaviour of the system, and so on). Attempts to improve these aspects of
the approach will be the aim of future researches.

Acknowledgements
We wish to thank Prof. R. Mazzola and engineers A. Ancarani. A. Cancelliere. V. Di
Leonardo and L. Menegolo for the useful information and suggestions provided. This
study has been developed in the framework of POP Sicily. contract No. 10122-94-03
TIPC ISP I.

92
References
Arrow KJ., Raynaud H. (1986) Social choice and multicriteria decision making, M.I.T. Press,
Boston.
Bailey P.D., Gough c.A., Chadwick MJ. McGranahan G. (1995) Methods for integrated
environmental assessment: research directions for the European Union, Discussion paper
for the European Commission Directorate General for Science, Research and Development
XIIID-5 Workshop on Integrated Environmental Assessment, Brusse1s.
Bana e Costa C.A. (Ed.) (1990) Readings in multiple criteria decision aid, Springer, Berlin.
Costanza R., Martinez-Alier J., O.Segura (Eds.) (1996) Getting down to earth: practical
applications of ecological economics, Island Press/lSEE, Washington D.C.
Keynes J.M. (1921) A Treatise on Probability, St. Martin's Press, New York.
Knight, F.H. (1921) Risk, Uncertainty and Profit, Houghton & Mifflin, Boston.
March J. (1978) Bounded rationality, ambiguity and the engineering of choice, The Bell Journal
of Economics, 9(2): 587-608.
Martinez-Alier J., Munda G., O'Neill J. (1996) Incommensurability of values in ecological
economics, in proceedings of "Inaugural Conference of the European branch of the
International society for Ecological Economics", Paris.
Munda G., Nijkamp P. and Rietveld P. (1994) Qualitative multicriteria evaluation for
environmental management, Ecological Economics, 10(2): 97-112.
Munda G., Nijkarnp P. and Rietveld P. (1995) Qualitative multicriteria methods for fuzzy
evaluation problems, European Journal of Operational Research, 82: 79-97.
Munda G. (1995) Multicriteria evaluation in afuzzy environment, Theory and applications in
ecological economics, Physica-Verlag, Heidelberg.
Munda G. (1996) Cost-benefit analysis in integrated environmental assessment: some
methodological issues, Ecological Economics, 19(2): 157-168.
Munda G. (1997) Environmental economies, ecological economies and the concept of sustainable
development, Environmental Values, 6(2): 213-233.
Nijkamp, P., Rietveld P. and Voogd H. (1990) Multicriteria evaluation in physical planning,
North-Holland, Amsterdam.
OECD (1989) Water resources management: integrated policies, Paris.
Paruccini M. (Ed.) (1994) Applying multiple criteria aid for decision to environmental
management, Kluwer, Dordrecht.
Paruccini M. (1996) Project A - Water resource management, Final Report of the POP Sicily,
contract No. 10122-94-03 TIPC ISP I, Ispra, Italy.
Rossi G., Ancarani A., Cancelliere A. (1995) Gestione dei sistemi idrici durante i periodi di
siccita: il ruolo dei modelli, Universita di Catania, Istituto di Idraulica, Idrologia e Gestione
delle acque, Catania.
Roy B. (1985 ) Methodologie multicritered' aide a la decision, Economica, Paris.
Si mon H. A. (1983) Reason in Human Affairs, Stanford University Press, Stanford.
Simonovic S. P. (1996) Decision support systems for sustainable management of water
resources: l. General principles, Water International, 21: 223-232.
Winograd T., Flores F. (1988) Understanding computers and cognit;on, Addison-Wesley. New
York.
Zadeh L.A. (1965) Fuzzy sets, Information and Control8: 338-353.

93
Assessment, evaluation and allocation of funds to
infrastructure projects: the case of the road network in
Lombardy (Italy)

Giulio Giannerini, Eliot Laniado and Giorgio Stagni

Centro di Teoria dei sistemi - CNR, Dipartimento di Elettronica e Informazione
Politecnico di Milano, Milano, Italy

Abstract
Many Italian regions have recently introduced new methodologies to allocate regional funds to
projects of regional interest. In 1992 the region Lombardy introduced a tool called Frisl which is
used to assign loans in different sectors involving land use. These loans regard projects for road
and public transport networks, town planning and so on. This research shows an assessment
system for selecting the projects to be financed which aims at overcoming the limits of a simple
weighted-sum method, a system used so far within Fris!. In the new assessment a weighted sum is
used only after the projects are divided into "priority dasses", which represent degrees of
importance of the problems to be solved by the project. The thresholds, which determine priority
dasses, do not require weighting schemes. A project which belongs to a high priority dass is
always considered as more urgent than all those which belong to a lower dass. This mechanism
reduces the importance of weights in the evaluation. The assessment system has been developed
in detail for a sampie sector: the road network in Lombardy. It has been tested on 40 projects that
had submitted applications for loans to Frisl in 1992.

Keywords: allocation of funds, weighted sum, threshold system, transports.

1. Introduction

In the recent past, many Italian regions have faced the problem of managing the
distribution of fmancial resources among aseries of different projects proposed by local
bodies (Provinces, Municipalities). In 1992 the region Lombardy introduced a tool,
called Frisl (Reconstitution Fundfor Sodal Facilities in Lombardy), to manage loans in
different sectors involving land use and social facilities. Public organisations, and in some
cases private individuals, have the possibility of obtaining loans to be repaid in ten years
without interest.
The aim of this paper is to show a new assessment system for the selection of projects
to be fmanced in sectors usually within the Regional Administrations' competence for
which funds are generally insufficient to cover all requests. The current system uses a
weighted sum to rank the proposals and finance the highest ranked up to depletion of the
funds available. The new system described in the paper is an evolution of the weighted
sum approach which overcomes some of its shortcomings.

95
 The methodology combines a threshold classification and an utility ranking within
classes. A simple utility ranking, such as the result of a weighted sum, may be
inappropriate due to the compensation mechanism: projects with average performances
on several criteria may emerge even if they are not able to solve critical problems on on~
or few criteria. For this reason, a combined class-utility ranking system is used. The
projects are first classified into three priority classes and then ranked within a class. The
system works in four stages:
1. Select a group of indicators. These are measurable quantities which provide evidence
as to the degree to which fund allocation objectives are met.
2. For each indicator, select two thresholds to classify the projects into three classes:
high, medium and low priority for this indicator. This serves to rate the context, rather
than the intrinsic qualities of the projects, so that it is possible for the Public
Administration to fmd the critical situations or emergencies, and to direct intervention
- i.e., funding - to them.
3. Assess the intrinsic quality of the projects, rated on the basis of a traditional weighted
sum (utility ranking).
4. Obtain the final rank order by combining threshold classification and utility ranking.

The structure of the paper is the following. In Section 2 the Frisl is first analysed, with
particular emphasis on transport sector applications (Section 3). The methodology is
then described in Section 4. Section 5 shows the evaluation scheme and its application to
the road network system. The method is tested against the results of the old Frisl scheme
for 40 projects financed in 1992 (Section 6).

2. The Frisl tool

Up to now, Lombardy Administrators have applied simple methodologies for the
selection of projects to be fmanced. Among these methods, a weighted-sum system
appeared since 1992 in Fris!. The main aim of using the weighted sum was to increase
transparency and objectivity in assigning funds. Since its implementation (Reg.
Lombardia, 1991), Frisl has distributed three sets of loans in 1992, 1993 and 1994. For
each year and for each sector identified by the Regional Administration, the analysis of
the project proposals is made byan Evaluation Staff (Reg. Lombardia, 1994). The initial
screening identifies non-admissible projects which do not satisfy legal specifications. The
remaining projects are then ranked with a weighted sumo Projects are financed starting
from the frrst in the rank and then going downwards until funds are available. Table 1
gives an overview of projects presented and fmanced in 1992, 1993 and 1994, divided
into sectors. The loan has to be repaid in ten years, without interest, starting from the
second year after allocation.

96
 Table 1. Admissible andjinanced projects (billions 0/ Liras).
Years 1992 1993 1994
Sectors Projects Loans Projects Loans Projects Loans
Adm. Fin. Adm. Fin. Adm. Fin.
A) Elderly people 116 46 106.5 51 51 136.8 108 60 145.0
B) Arts and environment 114 30 40.0 93 73 75.4 145 37 80.0
C) Reception facility 18 18 13.1 7 7 3.7 10 10 6.3
D) Public transportation - - - - - - 7 3 50.0
E) Road network 198 41 89.7 50 26 60.0 33 13 59.0
F)Mountain 17 8 20.0 4 4 7.1 8 4 19.0
G) Waste disposal 80 80 63.0 71 71 58.5 77 65 54.6
H) Suburban areas - - - 13 10 47.9 10 7 30.0
I) Local road network - - - - - - 102 71 39.8
Totals 543 223 332.2 289 242 389.3 500 270 483.8
Adm. = admissible appllcatlOns
Fin. =projectsjinanced

The criteria used for the assessment can be classified into nine fundamental groups
(Reg. Lombardia, 1995):
• regional priority (project consistency with regional planning);
• advice on priority expressed by Provincial Administration;
• quality/effectiveness of the project;
• feasibility;
• economic efficiency (cost per user);
• synergy between a project and its context;
• completeness of a project;
• secondary investments.

The specific criteria used within a group depend on the application sector. After their
selection and after rating each project on each criterion, the weighted sum provides the
evaluation scheme for financing the projects. Many criteria are used for more than one
sector - or even for all sectors, as it happens for the last three criteria. This fact allows a
certain degree of standardisation in the assessment of different sectors, but prevents the
use of quantitative indicators specific for each single sector. In general, qualitative
criteria are used more frequently than quantitative criteria. For instance, the
quality/effectiveness of a project is expressed only on the basis of a descriptive indicator
"highlmediumllow" and not by using specific sector indicators, such as road capacity of
the road network, average traffic, seats available in public transports etc. Quantitative
criteria, such as economic efficiency, appear only in the assessment offew sectors.
The application of the weighted sum, which assumes a compensation mechanism, has
shown some disadvantages. The regional administration faced some difficulties in
ensuring a correct response to real priorities; that is, to fmance projects which solve a
specific critical situation and cannot be deferred. The current assessment scheme seems

97
to be more suitable for a general evaluation of a generic project, in terms of economical
and administrative feasibility, but it is less suitable for a technical assessment of projects
in each specific sector. The regional officers interviewed during this research conftrmed
this opinion and added that they had often used their own experience, rather than the
assessment criteria, to determine whether a project was really necessary and coherent
with regional planning. The new assessment system tries to correct these deficiencies.

3. Analysis of the road network sector

The main characteristics required for a project which applies for Frisl are established by
regionallaws. They are:

Objects of intervention: improvement of road networks at a provincial or local

Minimum investment:
Maximum investment:
Non-repayable amount:
Interested subjects:
Particular limitations:
level;
200 million liras;
8,000 million liras;
max 30% , only under particular conditions;
municipalities, provinces, mountain communities;
works must be in conformity with the general town-
planning scheme.

The analysis of the Road Network sector, summarised in Table 2, shows the great
importance of the Frislloan, with a number of applications which is always higher than
the funds available.

Table 2. Total projects for each year.

Year Presented Admissible Financed
1992 305 64.9% 41 13.4%
1993 98 51.0% 26 26.5%
1994 226 59.8% 84 37.2%
Total 629 60.9% 151 24.0%

On the basis of 151 projects financed during the past three years:
• 76 projects consist of completely new works (20 of which represent works on
specific points such as crossings or junctions);
• 75 projects consist of widening or restructuring pre-existent roads.
In total, 153 km of roads, 60 of which are new roads, are affected by Frisl loans. In
addition to this, 16 two-Ievel crossings, 6 level crossing, 10 bridges and 41 rotary
intersections were also made. These projects have been c1assified into 9 c1asses as shown
in Table 3. From this c1assification, the Frisl loans seems to particularly address the

98
improvement of short, pre-existing local roads, distributed between normal routes and
mountain connections (specifically towards isolated villages).

Table 3. Typological subdivision offinanced projectsfor /992. /993 and /994 (totals).
Projects total km average loan per
project (million liras)
Important route in dense urban areas 8 6.7 4,868
Main by-pass 7 12.1 4,293
Ordinary by-pass 17 18.4 2,353
Main town link 16 21.5 1,940
Minor town link 23 35.8 1,561
Unique connection to isolated village 15 11.7 567
Local connections 20 24.8 513
Road serving specific installations 10 10.3 1,784
Urban road 35 12.3 954
Totals 151 153.7 1,622

Table 4 shows the list of assessment criteria used by Frisl in the 1992 edition. This list
confrrms the comments made before; i.e., remarks made about the use of qualitative
criteria and a predominance of administrative factors rather than technical ones.

Table 4. Assessment criteria and related scoresfor Frisl, Road Network sector (/992 edition).
A) Provincial priority (scores 0; 5; /0; /5)
Subdivision into four levels of priority, stated by each Provincial Administration.
B) Technical qualitie s of the project (scores 0; 5; /0; /5)
Qualitative function of some typological characteristics (e.g. presence of structures, crossings
or 2-level junctions).
Cl Regional priority (scores 0; 5; /0)
Qualitative function of project typology: routes according to Regional planning. by-passes. less
important roads, urban roads.
D) Completeness of the project (scores /-/0)
Scores are given to projects that complete a previously started batch of work, that are consistent
with town planning and that do not involve urban constraints.
E) Applicant's self-financing availability (scores /-/0)
Linear function of the availability of other financing means (applicant's own resources andlor
other non-regionalloans).

4. The assessment of infrastructure projects

4.1. The mixed threshold-score system

The Frisl assessment system, at the time of its introduction. was surely innovative for the
Lombardy Administration. However it has a main drawback: it is very common that a
project solving a very critical situation with respect to a single criterion is ranked lower
than another project which has average performances along many criteria.

99
 Tbe new assessment bas been specifically conceived for combining the simplicity and
immediate applicability of the weighted-sum method with a mechanism adapted to
emphasise emergencies. Tbis assessment system will be referred to as "mixed threshold-
score system", and can be summarised in five steps (see Figure I):

Step 1) Test administrative admissibility of each application.

Step 2) Define a hierarchy of the projects into priority classes.
Step 3) Calculate the utility score of the projects.
Step 4) Detennine the rank order.
Step 5) Distribute funds.

4.1.1. Admissibility
Tbe overall conditions required for making an application admissible are first checked.
This step involves the adherence to law parameters and can be easily managed by the
Administrative Offices offering the loans.

4.1.2. Classification 0/ projects

This is the innovative part of the mixed threshold-score system. On the basis of
thresholds, the projects are classified into three intervention classes. The assessment
requires, for each criterion, the selection of two thresholds called attention and alarm. In
this way three classes are generated: high priority (the project cannot be postponed),
medium priority (intervention is needed) , low priority (the project can be deferred, if
more urg~nt cases are present). At this_ stage~ the project is not evaluated on the basis of
its intrinsic perfonnances, but mainly on the degree to which it solves specific critical
problems (the so-called emergencies). Emergencies are defmed on the basis of context
information, such as an area with an accessibility level much lower than the average, or a
road with an accident ratio much higher than average.
This means, for instance, that if someone proposes a project for increasing
accessibility in an area where accessibility is already sufficient, then the project goes into
the low priority class (it can not be considered urgent, according to regional policy). If a
similar project is proposed for an area in which accessibility is a critical issue, then the
project goes into the high priority class as it has a great chance of increasing accessibility
and can be considered coherent with regional policy.
Tbe context may take into consideration average spatial values (referred, for example,
to a province or a region). In some other cases, reference values can be based on
planning needs which are not currently fulfilled (e.g., a bad ratio between the traffic
volume and the road capacity). They can also result from the regulations in force, for
instance as concerns noise and air pollution.
Tbe following is an example of threshold classification for the indicator accident rate.
Thresholds for the accident rate can be fixed at 50% and 200% of the regional average
value. If a project aims at decreasing the accident rate, it will be classified as folIows:

100
 current accident rate< 50% low priority (class C)
50% < current accident rate< 200% medium priority (cla~s B)
current accident rate> 200% high priority (class A).

With this approach, the state of the context in which the project is included is the
driving force for the classification. Therefore, a project which improves a critical
situation is considered as more important than one which may have overall better scores,
but is unable to impinge on a critical problem.
If only one criterion has been considered for a threshold classification, the procedure
is concluded. When, on the contrary, the critical state is described by more than one
criterion threshold, then the fmal classification depends on all thresholds. For road
network projects, for instance, this means that a project goes into the high priority class
either if it is a priority in terms of congestion or if it is a priority in terms of accident rate
or both.
The formulation of several thresholds does not require direct comparisons between
criteria. Using the example of congestion and accident rate again, this means that it is not
necessary to state which is the relative weight between safety and congestion. It is also
not necessary to normalise the accident rate and the volume/capacity ratio, because there
is no need to sum them up. The only need is to fix the numeric values of the two
thresholds, which gives rise only to an indirect comparison across criteria.
If classification results in three sets with few - many - few projects, then the
classification is usually considered as acceptable. This means that (few) projects
responding to reaDy critical emergencies are highlighted, giving to them the highest
chance ofbeing fmanced. The (few) projects not responding at all to regional policies are
also highlighted, giving to them the lowest chance of being fmanced. Finally there is an
intermediate class ofprojects (which includes many projects) which are almost equivalent
in terms of urgency. for which a fmer evaluation is needed (see below). Usually, the
funds are considered as weil distributed if they cover apart, but not all, of the central
class.

4.1.3. Calculating the utility score of a project

This is the traditional part of the mixed threshold-score system. A utility score is attached
to each project on the basis of a weighted performance on each criteria. This utility score
represents the intrinsic features of the project. Examples of the indicators used for this
purpose are:
• effectiveness (project ability of solving specific problems);
• efficiency (in economic terms);
• other indicators regarding law and administrative parameters.

The utility score of each project is calculated in the traditional way; l.e., by first
normalising the indicators and then attaching a weight to each of them. Indicators here
stand for the general data function of the project (or the context). It is clear that two
different indicators cannot be summed up immediately (which is necessary for achieving

101
a final score) because they are expressed in different measurement units and often vary
within unlimited intervals. The normalisation serves to transform each indicator into a
range always between zero (minimum importance) and one (maximum importance).
The weights. representing the relative importance of indicators. are then multiplied by
the normalised scores and summed up to determine the overall utility score of a project.
Weights have always been a critical point in this type of scheme. since they are subjective
and difficult to assess. At the same time they perform an important role in respect of the
fmal result. However. one of the main advantages of the mixed threshold-score system is
that it reduces the relevance of the weights. The overall utility scores are decisive for the
fmancial grant only within a class, and in particular only within the class for which there
are insufficient funds for fmancing all projects in that class (usually the intermediate
class). For the higher and lower class, fmancing of all or no projects respectively occurs
due to the thresholds, independent of the utility scores, and, therefore, of the weights.

4.1.4. Calculation ofthe rank order

The overall rank order of the projects depends, fIrst. on their classification and. second,
on their utility score within a class. The rank order is obtained by progressively
numbering the projects of the high class and then those of the intermediate and low
classes, according to a decreasing utility score inside each class. It is worth pointing out
that usually there are projects in a class which have a lower utility score than other
projects in a lower class.

4.1.5. - Distribution offunds

Loans are assigned by fIrst satisfying the applications classified in the highest priority
class and then going down to the following classes, taking into account the utility scores
of each project. If the allocated sum is not sufficient to cover all requests in a class.
projects are fmanced on the basis of their utility ranking within a class. Otherwise. all
projects in a class are fmanced, implying that utility scores and weights become
irrelevant.

4.2. Other specialjeatures ojthe assessment system

The computation of a single rank order valid for the whole Region represents the
simplest formulation and has been usually followed by the Lombardy region up to now.
If loan assignments have to be made at the provincial level. then the same procedure can
be applied. thereby changing the reference context from the regional to the provincial
level. In all other respects, the procedure remains the same.
At the end of the evaluation. some tests are usually necessary to verify the results.
This especially servesto make Sure that, even if resources are abundant, only effective
projects are financed. Therefore. some fmal tests on the projects have been identified.
For example. it is possible to exclude from fmancing (or to move to the immediately
lower priority class) any projects that do not reach a pre-established minimum score. or a

102
minimum return (such as an efficiency level, calculated as the ratio between some of the
project features and the amount of the loan).

IWtJ-admissible Exduded from

financing

Select priority dasses for each project

and each indicator
P = priority dass:
P = 1 ~ maximum priority (alarm)
Sort projects into three
P = 2 ~ medium priority (attention)
priority dasses
P = 3 ~ minimum priority
f = total funds available
Within each dass: sort the projects by
calculating utility scores E P= overall expenditure for projects of
priority P

Finance projects of priority P,

according to their score, until
depletion of funds available

;>--~--..... c END )

Figure J. Block diagram of the fund allocation proce ss.

5. Assessment of the Road-Network sector: a case study

5.1. Description 0/ the new proposal

In this section, an application example is considered for projects for the road-network
sector. The evaluation is based on four objectives typical of the road-network system: (I)
road safety; (2) pollution; (3) accessibility;(4) traffic congestion.
The indicators which correspond to these objectives substitute the indicators shown in
Table 4 above. This allows for the consideration of the technical aspects of road

103
planning, as weil as a subdivision into priority c1asses based on the "emergencies" of the
context. The decision process follows the five steps shown in Section 4. Indicators and
thresholds are described below.

5.2. Priority classijication (critical state ofthe context)

5.2.1. Safety
A high level of danger of an existing road is one of the main reasons that can lead either
to the replacement of a road or to the modification of it. As it is difficult to defme critical
levels of accidents in absolute terms, the evaluation is based on average territorial levels,
such as the regional level average. In particular, to keep the number of accidents,
fatalities and injuries into account, a parameter called equivalent accidents rate has been
used. This parameter is defmed as a linear combination of the individual factors
considered and measured in accidentslyear . km:

Equivalent accidents rate = accidents + injured people x 20 + fatalities x 150

The coefficients used in the linear combination are those proposed in Anas (1984).
The average data on the number of accidents is available on a regional or provincial basis
from the data bases of ISTAT (Italian Bureau of Statistics). The data for 1993 for
Lombardy is shown in Table 5 (Aci, 1994).

Table 5. Reference values for the number of accidents in Lombardy.

Fatalities Injured Equivalent
Accidents/km peopleJkm peopleJkm accidents
State roads 0.600 0.060 0.900 27.6
Provincial roads 0.150 0.015 0.250 7.4
Municipal roads outside the town 0.030 0.004 0.045 1.5
Average value 0.175 0.020 0.270 8.6

A comparison between the average data and the project data leads to the subdivision
into priority classes. This approach can be used to establish thresholds for safety on
national, provincial and municipal roads. The result is:

Indicator: accidentslkm·year
Reference: regional average of accidents/km·year on roads of the same
type of the project (national, provincial, municipal)
Threshold: - alarm threshold ::. 2 x reference
- attention threshold = 1.2 x reference

The use of average reference values, different for each type of road, enables the traffic
on the road itself to be taken indirectly into account, traffic being the variable that affects
more the overall number of accidents.

104
5.2.2. Pollution
In assessing the critical state of the pre-existing route. an evaluation of the noise and air
pollution occurring in the project area is necessary. While it is relatively easy to quantify
noise pollution. for air pollution it is more difficult to know what proportion is due to the
road section considered by the project.
Noise estimates are based on equivalent noise level (NLeq), calculated on the basis of
traffic data and of the geometrical and construction features of the road. This leads to:

Indicator: equivalent noise level

Reference: limits provided by law: 60 dB(A} (valid for generic residential
areas) or other limits for specific areas
Threshold: - alarm threshold = 15 dB(A} beyond the limit
- attention threshold =5 dB(A} beyond the limit

Among the four objectives taken into account, pollution, and air pollution in particular,
is the factor for which an appropriate assessment is more difficult to make. Therefore,
pollution is accounted only for projects for which the applicants are able to provide the
necessary data. In any other case pollution is ignored, which means that projects are
always put into the lowest priority dass for this objective.
Air pollution takes the concentrations ofCO (petrol engines) and S02 (diesel engines)
into account. Legal thresholds are available and set the attention and alarm thresholds for
both of the parameters. Therefore, the dass formulation is:

• Indicator I: Air pollution: CO

Reference: limits provided by law
Threshold: - alarm threshold = 30 mg/m 3
- attention threshold = 15 mg/m 3
• Indicator 2: Air pollution: S02
Reference: limits provided by law
Threshold: - alarm threshold =250 J.lg/m 3
- attention threshold = 125 J.lglm 3

A project is put into the fIrst or second priority dass if either CO or S02 exceed their
own alarm/attention thresholds (an OR condition).

5.2.3. Accessibility
The present state of public transports, together with local situations such as connections
to isolated mountain villages, are taken into account. This is consistent with the most
recent regional policies, and it is also appropriate in the specific context of Lombardy,
which is characterised by a very dense and branched road network. Therefore, it appears

105
appropriate to consider accessibility in terms of local networks and of public transports.
The threshold structure is:

• Indicator I Isolated villages

Threshold: - alann threshold: absence of road connection
- attention threshold: inadequate connection, especially for
basic facilities such as schools, hospitals or public offices
• Indieator 2 Public transportation (by road)
Threshold: - alann threshold: impossible transit of buses
- attention threshold : strong difficulties in transit of buses
• Indieator 3 Regional railway service (RRS)
Threshold: - alann threshold : lack of aeeess to an RRS station
- attention threshold: strong diffieulties in aeeessing a RRS
station

Even in this ease, the global class for aeeessibility is obtained with an OR condition.

5.2.4. TrafJic congestion

Of the four objectives, traffie eongestion is the one that is more closely related with the
teehnieal problem of road planning; as a result, eongestion has already been widely
analysed and robust indicators are available for use. The most irnportant aspeet in traffic
eongestion is the relation between the road capacity and the traffie on the road. Thus, the
knowledge of the theoretical capacity (HCM, 1985) of a road is neeessary, together
with the availability of traffie eensus infonnation (Provineial Administration, 1995; Anas,
1990). Road capacity, ealculated in vehicles per hour, is to be compared with the traffie
levelleading to a Volume of Traffic/Capacity ratio. This leads to:

Indicator: Volume of Traffic/Capaeity

Threshold: =
- alann threshold l.5
- attention threshold = I

5.3. Calculating the utility score 0/ a project

On ce the eritieal state of the problem has been established, the rank order within eaeh
class is obtained on the basis of a nonnal weighted sumo The weighted sum is earried out
with respeet to four indieators, S, P, A and C, which eorrespond to the four objeetives:
Safety, Pollution, Accessibility and Congestion. The indicators are expressed in a
normalised scale between 0 and I, and are in turn obtained by eombining sub-indicators
SI' S2' ... , PI' P 2,··· which quantify the eritieal state ofthe present situation (drawn from
parameters already used for thresholds) or the effeetiveness of the proposed solutions.

106
These indicators, as weH as their utility functions and weights, are the result of interviews
with the regional officers of Frisl. The data on each indicator needs to be supplied by the
fund applicants.

5.3.1. Safety
The indicator SI is based on the equivalent accidents per year, which is the same
indicator used for the classification above. As regards the project effectiveness (S2)' this
is measured through a simplified approach, wh ich distinguishes only qualitative levels:

Effectiveness judged as: sure good weak null

2/3 1/3 o

S2 is an indirect measure of how many accidents recorded in the past are expected to
be avoided by tbe proposed projects. Tbe fmal indicator is tberefore: S = SI X S2. The
nonnalisation is based on the alarm threshold, which is equal to twice the regional
average value (see Figure 2).

Equivalent accident rate

5
1 . __________~--

: accidents/year·km
2 x regional average

Figure 2. Utility function for safety.

5.3.2. Pollution
For noise pollution, the most appropriate calculation methodology would lead to a
subdivision of tbe population affected by noise into noise buffers, or areas of equal noise
levels around tbe noise source before and after the project is irnplemented. To sirnplify
the approach, only one noise class is considered (65 dB). Tbis serves to measure tbe
number of people subjected to a noise level bigher tban 65 dB. Tbe project effectiveness
is then defined as the percent improvement in noise (PNoise)' obtained as:

where Nbefore and Nafter are the number of inhabitants exposed to noise levels higher tban
65 dB before and after tbe intervention. If PNoise is negative, no score is assigned.
The indicator on air pollution is obtained in the same way, by counting tbe number of
people exposed to air pollution levels higher than 15 mg/m 3 for CO and 125 flg/m 3 for
S02 (before and after intervention). This leads to the PI and P2 indicators. The overall

107
indicator Pair is the maximum between the two. Finally, the partial utility assigned to
pollution levels is computed as:

P =P noise . W noise + Pair' Wair

where the W i are the weights for noise and air pollution (see below).

5.3.3. Accessibility
Three indicators relating to villages, public road transport and the Regional Railway
System (RRS) are considered and combined in a weighted linear fashion. The
accessibility score for villages (AI) is obtained considering the number of people (N)
living in a village and comparing this number with a reference number of inhabitants Nref
= 200. This level represents the minimum of inhabitants necessary for considering the
new road as effective:

For public road transport, the number of trips (per day and direction) related to the
road section under examination is considered. The A 2 indicator is defined as:

A2 =Min {B x trips I tripsref; I}

where:
- B is a qualitative indicator of the expected irnprovement from public transport with
B=1 for strong direct irnprovements (example: enlargement of very narrow streets with
constant traffic difficulties) and B=O.5 for low or indireet improvement (example:
congestion reduction for opening of a bypass, while the bus still travels through the
town).
- tripsref is a reference value of trips, set at 200 tripslday·direction (this corresponds to
two lines running during the day, from 6 a.m. to 10 p.m., with a passage every ten
minutes, or six lines with a passage every half an hour).

For the RRS score (A 3), the level of service of the station connected by the project is
assessed in a similar way:

A 3 =Min {trips I tripsref; 1 }

Since the frequency on a railway line is usually less than that of a bus, tripsref is set at
a level equal to 32; ie. a full score for a train every 30 minutes. Therefore the total
accessibility indicator is:

A = W village' AI + W public transport' A 2 + W RRS . A 3

108
5.3.4. Congestion
Congestion (Figure 3) uses an effectiveness indicator related to the percent capacity
increase obtained with the project. This indicator reaches the maximum value of I for a
capacity which doubles the existing one.

Capacity ·increase
c
1

o 1 2 C After / C Before

Figure 3. Utility functions for congestion.

5.4. Final classification

The fmal priority dass is obtained considering both thresholds and overall utility scores.
A project is put in the first dass if it exceeds at least one of the alarm thresholds. A
project is put in the second dass if it does not exceed any alarm threshold, but exceeds at
least one of the attention thresholds. The remaining projects go in the last dass.
Within a class, the weighted sum discriminates hetween projects. The weights shown
in Table 6 are used for this purpose. This set of weights is the result of aseries of
interviews with the regional officers involved with Frisl. The weights are meant to
represent the present policy of the Regional Administration in terms of road planning.
The column "sub-weights" in Table 6 aUows the computation of S, P, A and C
starting from SI' S2' ...• PI' P 2•... as descrihed in the earlier sections. The ''weights'' are
used to compute the overall utility score as:

Sc =S x 40% + P x 15% + A x 25% + C x 20%

During the fmal phases of the evaluation. it is also possible to fmd anomalous cases of
projects which appear in a high priority class but. as regards the utility score, seem to he
rather ineffective. These projects can he evaluated individually and. if necessary, moved
to a lower class. In addition. the efficiency test can he userl to reduce the probability that
very expensive projects. in relation to their henefIts, are fmanced. Furthermore. it is also
possible to consider the occurrence of project spillovers. such as projects which generate
a new critical state; e.g., the case of pollution. In all these cases, however, the
introduction of an automatic mechanism of redassifIcation is not advisable. Individual
judgements based on a case-by-case basis are preferred here (Dept. Environment, 1966;

109
HCM, 1985). This task could be managed by the CTAR team (a regional technical-
administrative board) whose task is the quality control of fmancial aspects of the regional
administration.

Table 6. Weights usedfor computing utiUry scores.

Objectives Weight Indicators Sub-weights

Sarety 40%
Accident rate Expected reduction of accidents/year (100%)
Pollution 15%
Noise pollution Reduction of exposed inhabitants WNoise=70%
Air pollution Reduction of exposed inhabitants WAir= 30%
AccessibiIity 25%
Villages Inhabitants connected WVillages = 35%
Public transport Bus trips travelling through WPub. Transp. = 40%
RRS Running trains WRRS=25%
Congestion 20%
Capacity Capacity increase (100%)

5.4.1. Data requirements

The new Frisl system requires several pieces of data. These are supplied by:
The applicant:
• geometrical data of the street and its capacity (HCM, 1985);
• traffic and congestion data (average daily traffic and/or standard rush hour traffic);
• accidents, injuries and fatalities;
• noise data and CO and S02 concentrations;
• villages to be connected and number of related inhabitants, public transport trips on
road and railway.
Regional administration:
• regional average values (e.g. accidents, injuries, fatalities on average per km);
• references for the thresholds (law standards for pollution; optimal service level for
public transport, etc.).

6. A case study: 1992 Frisl Road Network Projects

The new system for the road network has been tested on the 40 projects fmanced in 1992
for which all data were available. Due to the radical change in the criteria used, the rank
order looks somewhat different from the original one. In particular, the desired division
into a few-many-few project distribution has been achieved. This division is capable of
highlighting true priorities and of giving projects in the top dass the highest chances of
being fmanced. The results are summarised in Table 7. The fmal rank order is shown in
Figure 4, where the decreasing score within each priority dass can be easily seen. It is
also interesting to note the correspondence between the number of projects in the high
dass and the general situation of the road network in Lombardy, characterised by an

110
average good level of accessibility, problems of congestion and a high accident rate. The
first six projects occupy this position due to "safety" concerns, while most of projects in
the intermediate dass justify their position due to congestion. This appears to be
coherent with the present regional road policy, which mainly aims at reducing the
accidents rate and, subordinately, at improving routes capacity.

Toble 7. Number ofprojectsfor each ofthe 3 priority classes.

Priority Safety Noise and Accessibility Congestion Total
pollution
high 8 0 2 2 12
medium 4 9 9 21 22
low 28 31 29 17 6

':: [ •••••••• New.fu;~d«Ore ~~ • Ori~~ ~T ~ .~.~ A

70.0 :.;-: -;- :~ -;-: -;-:--;. ;-: -;-:-;. ~ .;-: -;- :~ ;-: ~:--;. ;-: -;-:-;. ~ i-: -;- :
i"' 60.0 :.;- : -;- 4--; -;-: ~ :--;. ;-: -;-: -;. ~ .;-: -;- :--.; -;-: 4- :--.;. ;-: -;-: ~ ~ .;-: -;- :
50.0 • . ~-: --ls: :--.; .:-: ~:--.;. ~: -.:-:~ ~ ~: ~~ .:-: ~:--.;. ~~ ~:~ ~ ~: ~ :
~ 400 " . , , 6 6 . " " . " ' " 1:>, , , , , , , , , , , , , , , , , , , ';'
ii: . .,...-, --r- ,~ ~-r- 'ß r--a tr ,-,- r--""'1 -r-8--r- I~ -r-, -r- ,---r" I:J;'-l ~ ,--r r"'"""""'1 .,...-,---,-y
'6.' I!I, , , '6.' , 6 . , , , , , , , , , , , , , , , , , ,A, , , , , , ,
30.0 ~6~ I!I-• •-, ~ ,~ .--f!,. -'-f!-• .!!~ ~'--ts: '-6. .,..-n~ ,-/}. ~ ~,~ t:s- ~,~ ,
200
.
10.0
, , , , , , ' . ' 6' , , , , ,
~,~
""""'

I
--
--,-- -- --, --
, I , , ,
~,-

, ,
•.
,
..
,~ ~ -,~ ~ ~,~,~

I ,
"""""""'"
,-+ - - - - , - -
I , , • ,
" ! •••
iI~' ,A, , , , '6.' , '61:>,

~
, , I I I I
~.-.-

--,-- - - --, --,-+ - - '.-.-.

t , I ,
••
I I
'.,
~ ~,~ ~ ~,~,

I I ,
,
I
,./\, , , ,
I
-w.-.-.,.
, I , I I I

0.0 +-t---..-.-t-+-t---..-.-t-+-t---r-1-t-+-t---..-.-t-+-t---r-1-t-+-t---r-1-t-+-t--+-t-t-+-t--+-t-i

Figure 4. Total score for the J 992 projects.

It is again important to stress that the first priority dass indudes some projects with a
rather low overall utility score. Nevertheless, these projects are considered to be able to
solve a single but critical problem, and are thus suitable for financial support. The project
labe lied Cremia, for instance, refers to a little mountain village not yet connected to the
road network. The project Cantu' is oriented towards a major re-organisation of the
public transport system in the Cantu' area. The situation in this area, characterised by
very low accessibility, is rather exceptional compared to the rest of Lombardy. However,
this aspect is sufficient to guarantee a high probability of financing the project, even if it
has a rather low utility score.

111
7. Conclusrons

This paper originated from a specific request included in the regional law which
established the Frisl system (Reg. Lombardia, 1991): arevision and analysis of Frisl after
three years of application. The Frisl is one of the most relevant financial mechanisms for
Lombardy. The large number of applications - between 500 to 900 per year -
demonstrates the success of Frisl. Yet, the fact that available funds only covered 56% of
the admitted applications in 1994 makes it clear that selection is a very critical issue. The
new assessment system is characterised by a mixed formulation thresholds-utility scores.
This exploits the simplicity and immediateness of the weighted sum, while reducing,
through a threshold classification, its extreme compensation effects. Thresholds allow an
easy subdivision of projects into priority classes, leaving the finer distinctions within
classes to utility scores. This methodology originates from traditional multi-criteria
analysis (cf. Keeney and Raiffa, 1976), but simplifies the approach a great deal to meet
the requirement of simplicity and ease of use that represented the fundamental
specification expressed by the regional Authorities. At the time that this article was being
written (spring 1997), other Italian Regions were considering this type of evaluation,
also for other sec tors such as waste disposal management.

Acknowledgements
This research was made possible as a result of the active co-operation of all members of
the Workgroup set up by the Istituto Regionale di Ricerca della Lombardia (IReR) and
co-ordinated by Dr. Tarulli and Dr. Moriggi. The contributions provided by Dr. Gelmini
of the Centro Studi Traffico in Milan for the Road Network sec tor, and by Dr. Bianchi of
the Istituto di ricerche Ambiente Italia, for the Waste Disposal sec tor, are greatly
appreciated. A number of regional officers provided substantial contributions to this
work. Thanks go to Dr. Cesca and Dr. Salemo (Road Network), Dr. Bertani and Dr.
Gherardi (Waste Disposal) and Dr. Bonifacio and Dr. Cattaneo (Frisl Office).

References
Aci (1994), Localizzazione degli incidenti stradali (Localisation of road accidents), XLVII-L
conference oftraftic held in Stresa, Istat, Roma, 1991-1994.
Anas (1984) Piano decennale del/a viabi/ita di grande comunicazione (Ten year plan for
national road network), (Iaw 531/82), Anas, Roma.
Anas (1990), Censimenti dei traffico sul/a rete statale (Traffic census on national roads), Anas,
Roma, 1985-90.
Dept. Environment (1966), Layouts of Roads in Rural Areas and Roads in Urban Areas,
Department of the Environment, Scottish Development Department, The Welsh Office.
HCM (1985), Highway Capacity Manual, National Research Council.
Keeney, R., H. Raiffa (1976), Decisions with Multiple Objectives: Preferences and Value
Tradeoffs, lohn Wiley, New York.

112
Provincial Administration (1995), Censimenti deI traffico sulla rete provinciale (Traffic census
on provincial roads), Provincial Administrations, Bergamo, Brescia, Corno, Cremona,
Mantova, Milano, Pavia, Sondrio, Varese, 1982-95.
Reg. Lombardia (1991) Regional Law n. 33 of December 14th "Modifiche ed integrazioni della
legge regionale 31 marzo 1978, n. 34 'Norme sulle procedure della programmazione, sul
bilancio e sulla contabilita della Regione e successive modificazioni. Istituzione del Fondo
Ricostituzione Infrastrutture Sociali Lombardia (FRISL) (Modifications and supplements of
regionallaw n. 34 of March 31 st, 1978, 'Law upon procedures of planning, of budget and of
accounting of the Region' and following updating. Institution of the Reconstitution Fund for
Social Facilities in Lombardy (FRISL)) (BUR, supp!. ordinario n. 51 - December 19th, 1991).
Reg. Lombardia (1994) Istruzioni generali per l'accesso al Frisl (General Instructions for the
Frisl TooI), in the following editions:
- 1992 (BUR, 4° supp!. straordinario al n. 6 - February 8th, 1992)
- 1993 (BUR, 3° supp!. straordinario al n. 2 - January 16th, 1993)
- 1994 (BUR, 5° supp!. straordinario al n. 36 - September 10th, 1994)
(containing the legislative measures regarding Frisl, a template for the application and all
related instructions).
Reg. Lombardia (1995) Assegnazione dei contributi Frisl (Assignment of Frisl Loans), in the
following editions:
- 1992 (BUR, 7° supp!. straordinario al n. 34 - August 22m!, 1992 and 1° supp!.
straordinario al n. 9 - March 2nd, 1993)
-1993 (BUR, 3° supp!. straordinario al n. 9 - March 4th, 1994)
-1994 (BUR, 2° supp!. straordinario al n. 15 - April 13rd, 1995)
(containing a description of the assessment system used in each sector , a list of the filed
applications, as weH as ofthe admitted and financed projects).

113
A farm multicriteria analysis model for the economic
and environmental evaluation of agricultural land use 1

Carlo Giupponi 1 and Paolo Rosato 2

I Dipartimento Agronomia Ambientale e Produzioni Vegetali
2Dipartimento Territorio e Sistemi Agro- Forestali
Universita di Padova, ltaly

Abstract
The paper presents a model that produces and evaluates alternative farming systems. The
economic and environmental viewpoints are considered within different hypotheses of agricultural
policy and farmer's decision-making process. The model is composed of three parts: a) a
multiobjective programming model far the simulation at farm level of policies and farmer's
decision-making; b) a model for the environmental impact assessment of the results of the
previous simulations; c) a multi-attribute evaluation of the simulations from both the economic
and environmental point of view. The model has been applied in the area of the Venice Lagoon
Basin (VLB), located in northern Italy. The economic-environmental evaluation demonstrates that
there is clear synergy between the tendency of the farmer to minimise the risk of the cropping
systems and the necessity to reduce the environmental impact of farming techniques. This is
particularly useful for promoting low-impact agricultural practices in the situations dominated by
traditional production methods whose environmental implications appear very variable.

Keywords: agriculture; environment; impact indices; multicriteria analysis.

1. Introduction

The process of development and spread of eco-compatible agricultural practices IS

sustained by appropriate technical and scientific inputs and by efficient tools of technical,
economic and environmental evaluation (Marangon, 1992; Ridgley, 1995). These tools
have been the subject of intense development for some time and a great deal of progress
has been made within the different disciplines involved. Yet the characteristics of the
problems faced c1early show the need for a tight integration of the evaluation phase. This
can be satisfied with models which integrate physico-environmental phenomena with
technical and economic aspects. The advantages offered by such a combination are
obvious and can be summed up by the possibility of simulating and fully evaluating
alternative scenarios; under this viewpoint integrated models become useful development

1 P. Rosato was responsible for the economic-farm evaluations (part I). C. Giupponi set up the
environmental evaluations (part 2). Both were involved in the joint economic-environmental analysis
(part 3). The authors would like to thank the anonymous referees for their suggestions which contributed
greatly towards improving the paper.

115
tools of agro-environmental policies (Giupponi and Rosato, 1993). The aim of this paper
is to demonstrate the potentials of such an integration of economic and environmental
models, referring to a case study carried out on the watershed of the Lagoon of Venice.
A multicriteria (MCA) model is presented in this paper which produces and evaluates
alternative cropping systems from the economic and environmental point of view. The
model is composed of three sub-models: a) a multiobjective (MOA) programming model
for the identification of cropping systems and the estimation of economic-farming
indicators; b) a non-point source pollution model for the simulation of environmental
impacts; c) a multiattribute analysis (MAA) model for the joint economic and
environmental evaluation. The model incorporates the farming system with the physical
environment (especially soil and water) and can supply estimates of both the
environmental and economic indicators. The evaluation procedure also includes a
"social" evaluation phase where, starting from trade-offs between private fmancial
objectives and quality of the water resources, compromise solutions are explored.

2. The farm programming model

2.1. Background

The farm programming model was designed around the part-time farm of the VLB, as
this type of farm is very widespread and covers a good part of cultivated agricultural
land. It is also very important from the environmental point of view because it is
widespread in areas of economic development and diffuse housing; i.e., where the
environmental implications of agricultural activities are particularly evident. In these
farms, cropping operations are carried out almost entirely by contractors, and the owner
plays a purely organisational and supervisory role. The cropping system is extremely
simplified and essentially includes arable crops such as maize, wheat, sugarbeet and,
more rarely, alfalfa.
The 1992 agricultural policy reform of the European Union (CAP) places new
decisional problems on this type of farm, linked to the possibility of choosing between
different systems of income support and subsidised programmes for reducing
environmental impact of production techniques. Concerning this, the CAP reform poses
problems of choice that require the adaptation andlor development of the modeL~ used up
to now. The decision-making process is complicated by new components which are only
in. part manageable with the c1assical linear programming traditionally used in farm
management. Currently, the choice of farm cropping system involves two aspects:
• the choice among four alternative farming systems: general or simplified scheme of
income support (EC Reg. 1765/92; EC, 1992a) and traditional or low environmental
impact techniques (EC Reg. 2078/92; EC, 1992b);
• the choice of land use (e.g. how much maize, soybean, sugarbeet, etc.).
Resolving a problem of this type requires discrete choices (e.g. type ot' income
support system), continuous choices (e.g. crop allocation), as weil as the incIusion of

116
many facets of management processes. This can be solved with multiobjective integer
linear programming (see next section).
The technical-economic data used in the model were taken from a farm survey. As
regards the eco-compatible techniques. data came from experiments carried out at the
Veneto Agricultural Oevelopment Authority's Oossetto Farm in Vallevecchia di Caorle
(Venice province).

2.2. The farm model

The model used for the farm simulation is a mixed integer multiobjective linear
programming model that includes choices of the continuous and discrete types (Azzolin
and Rosato, 1995). Tbe formulation of a model is as folIows:

I) Optimise Z = f(GM, R)
wiili .
2) GM =Lij xijPii ;=1 •..• 1 j=I, .. ,J
3) R=TAO
4) Lij ahij xij ::; bh h = 1, .. , H
5) Li xij= Sj
6) Sj = Oll (integer variable)
7) Lj Sj = 1
8) Xij ~ 0

Where:
GM = farm expected gross margin;
R = expected risk;
xii = amount of farm land (%) allocated to land use ; and farming system j;
pij = gross margin achievable with one hectare of land use i and farming system j;
TAO = total absolute deviation of standardised yields (see below);
ahij = resource h necessary for one hectare of land use i and farming system j;
b h = total resources available per hectare of farm land;
Sj = farm land allocated in farming system j.

Equation I summarises two functions: a) the multiobjective, that guides the choice of
crop allocation; b) the multiattribute. in respect to which the farming system is identified.
Moreover, equations I. 2 and 3 identify a multiobjective programming model that
simulates the behaviour of a farmer who aims at maximising his gross margin and
minimising the risk of the cropping systems. The objective function of the model has
been formulated according to the results of a survey among the farmers of the VLB
(Bombana and Rosato, 1996).
Risk was estimated by means of a MOT AO formulation (Minimisation Of Total
Absolute Deviation; Hazell and Norton. 1986) where instead of variability of gross
margins the yields variability was taken into consideration. To eliminate the scale effect.

117
the yields of the different crops were standardised statisticalll MOT AD is the best
known linear method for the estimation of the variability of wholc farm productivity from
crops performance. It avoids the c1assical quadratic formulation, based on variance-
covariance matrices (Markowitz, 1959), utilising the absolute deviations from the
average observed in a determined temporal horizon T.

9) TAD = L, (Z,+ + Z,') with Lij (Yij,- Yij)Xij - z,+ + Z,· = 0 t = I, .. , T

where:
TAD = Total Absolute Deviation;
Z,+, Z; =absolute deviation in year t of the me an of period T;
Yij, =standardised yield of land use i with farming systemsj in year t;
Yij =average standardised yield of land use i with farming systems j in period T.
MOTAD is much appreciated because it makes direct reference to the farmer's
observations and is utilisable with few data (t) and with asymmetrie distributions 3•
Equation 4 represents the use of farm resources. Equations 5, 6 and 7 summarise the
device used to formulate the aspects relating to the discrete choices. The combined
action of equations 6 and 7 activates only one of the j farming systems and, therefore, by
means of equation 5 only land belonging to that farming system is made available for the
production processes. In this way the solution of the model gives the best farming system
and the associated cropping system expressed in percentage.
As regards the functional form of the objective function it was hypothesised that the
farmer can decide on the basis of three possible alternatives: a) maximisation of gross
margin; b) minimisation of risk; c) combined optimisation of income and risk. The first
two hypotheses present no computational problems. Formulating the combined
optimisation of income and risk is more complicated as it supposes the choice of a
method of synthesis between the two objectives. Within this context it seemed reasonable
to use goal programming (GP, Charnes and Ferguson, 1955). With GP it is possible to
represent the attitude of a decision-maker who seeks a compromise between more than
one objective by minimising the standardised and weighted differences between the
model's feasible solutions and a pre-defIned reference state. The so-called ideal point
(Zeleny, 1973) was assumed as reference state (goal or aim), identifIed by optimising the
gross margin and risk separately. This point is not feasible but is a reasonable reference
point. The assumed weights are unitary, and two separate minimisation procedures were

2 To ea1culate the index of risk. referenee was made to the yields and not to the ineome. The rationale is
that the economic data used to calibrate the model refer to aperiod of transition between the old system
of ineome support and fixed priees and the new one of varying priees with subsidies. During this period
there were big anomalies in the prices that would have produeed a substantial over-estimation of the
variability of the profitability of the erops.
3 In this ease the distribution of the yields eould be eonsidered normal and symmetrie. therefore
minimisation of the total variability coincides with minimisation of the negative deviations from the
average.

118
hypothesised: additive (WGP) and MinMax (MINMAXGP) (Romero and Rehman,
1989).

2.3. Results

The solutions obtained with the simulation of the reform of the EU agricultural
commodities market (EC Reg. 1765/92) are reported in Table I. When the farmer acts to
minimise the variability of farm income (MinRisk model). the model selects the general
support scheme with the farmland used for growing wheat (50%). soybean (58%. catch
crops4 included) imd sugarbeet (25%). Set-aside. bound to the general support scheme.
occupies nearly 7.8% of the farm land. Optimising the cropping system against risk
privileges the combination of production processes with low yield variability. At the
same time, it highlights support schemes that. proportionally. ensure higher eamings
from direct compensations. since these are not exposed to the risks connected with
cultivation. Within this approach the gross margin becomes fixed at nearly Lit. 1.6
million per hectare with an index of risk equal to approximately 0.65 • Nitrogen
applications are just over 100 Kglha.
If it is assumed that the farmer pursues concurrently the maximisation of gross margin
and the minimisation of risk (ADDGP). the optimal cropping system changes
significantly: set-aside and soybean as first crop disappear. maize is grown on 50% of the
Utilised Agricultural Surface (UAS) and there is an expansion of soybean as second crop
(+9.2% of the UAS). With this cropping system the gross margin rises to Lit. 1.8 million
per hectare (+17%) and the index of risk to 0.938 (+60%). It should be noted that.
comparing the solution minimising risk. the increase of one percentage point in the index
of risk gives an increase in income of only 0.3%. The MMGP solutions are only slightly
different. A slight contraction of the land under soybean as catch crop is observed in
favour of maize as catch crop (+3.8% of the UAS). However. there is an appreciable
increase in the index of risk (+77.6%). while the gross margin remains more or less
constant (+0.5%). It should be noted that the expansion of cereal crops determines a
huge increase in nitrogen use (+69% on average) that reaches 183 units of nitrogen per
hectare in the last solution iIIustrated. Lastly. maximising gross margins (MGM)
produces a sugarbeet-cereals cropping system. with maize substituting for soybean as
catch crop. This system provides almost Lit. 2 million per hectare of gross margin
(+6.6% compared to the MMGP solution). but is extremely risky (+95.7% more so than
MMGP) and nitrogen use is very high (+50%).
The model. in the four solutions identified. proposes a very intensive cropping system
with ample recourse to catch crops. It is interesting that wheat and sugarbeet remain
steady in all the farming systems identified. The presence of winter cereals is due to two
reasons; the first linked to the possibility of a second crop. the other tied to the lower

4Catch crops are crops cultivated after a main crop in the same agrarian year.
5The index provides for a minimum equal to 0 realisable with yields constant over time. Obviously, low
values are preferable.

119
variability of wheat yields compared to those of maize which are frequently affected by
summer drought. The permanence of sugarbeet is due to its high profitability.

Toble I. Solutions ofthe simulation ofthe CAP reform (EC ReK. /765/92).
Solutions
Min risk AdditiveGP MinMaxGP Max gross
margin
MR ADDGP MMGP MGM
Farming system (*) G S S S
Soybean (%) 17.2 - - -
Sovbean 2° eroD (%) 40.8 50.0 42.2 -
Maize (%) - 25.0 25.0 25.0
Maize 2° erOD (%) - - 3.80 50.0
Wheat(%) 50.0 50.0 50.0 50.0
Sugarbeet (%) 25.0 25.0 25.0 25.0
Alfalfa (%) - - - -
Set-aside (%) 7.8 - - -
Gross marl!;in (.000 liras/he) 1571 1839 1849 1971
ofwhieh from subsidies (%) 35.8 33.4 33.2 31.1
Index of risk 0.586 0.938 1.009 1.975
Nitrol!en distribution (kl!/he) 106 175 183 275
* G = general; S = simplified. The shaded va1ues refer to the "ideal solution".

Variations in the objectives attainment are relatively small as regards income (±11 %
compared to the average), while they are significant in the index of risk (around 60%).
This means that considerable variations in the risk of cropping systems and farming
system are accompanied by modest income variations, suggesting the possibility of
finding acceptable compromise solutions.
The solutions, including the possibility of adopting eco-compatible agricultural
practices subsidised by the EU (Reg. 2078/92), are a bit more uniform than the previous
ones because of the more rigid rotational constraints imposed by the regulations. The
basic trend which emerges is that of the general support scheme being preferred where
there is a high aversion to risk, and the simplified scheme being preferred when
maximising gross margin has priority (see Table 2). However, the advantage of adopting
eco-compatible farming practices remains stable. They are chosen in all the solutions
identified, since the subsidies for eco-compatible farming practices are higher than the
corresponding reductions in crop income. The gross margin which can be obtained with
these techniques is on average 11 % higher than those obtained with traditional cropping
techniques. Furthermore, with the increase of the portion of earnings due to direct
subsidies (+14.5%), the index of risk of the cropping systems declines by 20% on
average.
With low propensity for risk the growing of legumes prevails (40.6%), mainly
soybean (33%), followed by wheat (31 %) and sugarbeet (25%). With this cropping
system the index of risk is slightly 10wer than that of the corresponding solution
"1765/92" (-6%), while the gross margin is decidedly higher (+ 16.1 %). It is also

120
interesting that, in this case, eco-compatible farming practices allow nitrogen applications
to be halved.

Table 2. Solutions ofthe simulation 0/ eco-compatible aRriculture (Ee ReR. 2078/92).

Solutions
Min risk AdditiveGP MinMaxGP Max gross
ADDGP MMGP margin
MR MGM
Farming system (*) EG ES ES ES
Soybean (%) 33.0 24.4 17.8 9.0
Soybean 2° crop (%) - - - -
Maize (%) - 22.0 33.0 33.0
Maize 2° crop (%) - - - -
Wheat (%) 31.0 28.6 24.2 33.0
Sugarbeet (%) 25.0 25.0 25.0 25.0
Alfalfa (%) 3.4 - - -
Set-aside (%) 7.6 - - -
Gross margin (.000 Iiras/he) 1824 1960 2022 2092
ofwhich from subsidies (%) 53.1 46.7 45.3 43.8
Index of risk 0.551 0.816 0.944 1.128
Nitrogen distribution (kglhe) 50 91 107 119
* EG = eco-compatible under general scheme; ES = eco-compatible under simplified scheme. The
shaded values refer to the "ideal solution".

The solution to the MMGP model includes a partial substitution of legumes and wheat
by maize (22%). Alfalfa disappears, soybean is reduced by 26% in comparison to the
previous solution, and wheat by 7.7%. Simultaneously, there is an increase in the gross
margin (+7.5%). Compared to the corresponding solution" 1765/92", the performance is
6% higher for gross margin and 23.7% 10wer for risk. These trends are confmned with
the ADDGP solution, where soybean and wheat decline further in favour of maize. The
index of risk increases by 17.7% and the gross margin by 3.2%. In comparison to the
corresponding solution "1765/92", the index of risk is similar, while income is 10%
higher. With the maximisation of gross margin, the expansion of cereal crops reaches the
limit imposed by rotational constraints with the re-expansion of wheat (+36.4%) at the
cost of soybean, which is halved. The gross margin rises slightly (+3.5%), whereas the
increase of risk is quite pronounced (+ 19.5%). The direct ratio is therefore conflrmed
between income and cereal crops and the inverse one between risk and oleaproteaginous
crops. All in all, the eco-compatible solutions dominate traditional ones, as is also
demonstrated by the fact that the ideal solution to model "2078/92" dominates that of
model" 1765/92" in a Paretian sense. The solution "2078/92" is shown, lastly, as being
particularly efficient for curbing reductions in income when the priority is to reduce the
risk of the cropping plans. The difference in income between solutions "2078/92 and
"1765/92" is 16% when risk is minimised, and drops to 6.1 % when gross margin is
maximised.

121
3. Environmental evaluation of the farming models

3.1. Background

An important problem when comparing alternative farming systems is their enormous

complexity and the many aspects that should be evaluated. This determines firstly the
need for a complex multivariate approach (e.g. analysis of risks for human health,
erosion, water pollution, etc.), followed by that of defining a comprehensive
methodology to allow concise judgements. In this field, the simulation models of
agricuItural non-point pollution sources (Giupponi, 1995), if adequately calibrated and
validated on data collected by agronomic and environmental research, become
irreplaceable tools for describing the functioning of agricultural systems, both real and
hypothetical. Interest in an environmental evaluation of hypothetical cropping patterns
can have different motivations, for example the usefulness of having indications of future
agricultural systems in a given area, following regulatory and market changes.
The objective of this second part of the work was to evaluate and compare the farm
scenarios identified by the multiobjective methods in terms of environmental impact.
Figure I shows a scheme of the approach. Farming systems were defined not only in
terms of crop percentages and cultivation systems, but also by crop rotations to obtain
more realistic resuIts from long term (30 years) simulations carried out with the non-
point source pollution model. Long term simulation allowed the calculation of impact
indices over average annual data of pollution phenomena.

3.2. Definition 0/ the environmental impact indices 0/ alternative

/arming systems

The set of impact indices adopted aims at supporting decisions among alternative
farming systems on the basis of a detailed evaluation of the possible risks deriving from
nutrient and pesticide losses from cultivated fields. Regarding the environmental risk,
there is concern about the degradation of water resources (e.g. eutrophication from
nutrients of agricultural origin), of the soil (e.g. erosion) and of the air (e.g. volatilisation
of pesticides or ammonia). For human risks, the concerns regard direct or indirect health
consequences (e.g. pesticide accumulation in the food chain). A more comprehensive
proposal of a set of indices of environmental impacts has been defined in Giupponi
(1998), whilst here only water pollution is dealt with. This is in fact the major source of
concern in the Veneto area, especially because of pollution problems in the Venice
Lagoon and Adriatic Sea.

122
 I t - - - - - - - I Catalog of alternative
rotations
Definition of realistic farming
systems for long term simulations

I--_ _ _~ Crop percentages, rotations,

cultivation systems

Analysis

Figure 1. Flow chart of environmental evaluations.

Five types of risks were analysed to assess the impact of agricultural activities: (1)
risks to drinking water resources, (2) toxicity for mammals; (3) toxicity for aquatic life;
(4) eutrophication; and (5) risks to fanners health.
The indices were calculated on the basis of the inputs of pesticides and fertilisers in
the cultivation systems and their possible dispersion in the environment. Estimates were
based on a simulation model of agricultural non-point source pollution. The GLEAMS
model (Leonard et al., 1987) was adopted on the basis of extensive previous experience
(cf. Giardini et al., 1992; Morari and Giupponi, 1998). The indices have only a relative
meaning and are suitable only for comparison within a given set of systems of alternative
cultivation. For this reason, they are re-scaled between one and zero, where the value of
zero corresponds to the least impact. The impact indices used are the following:

REGULATORYDRINKABILITY INDEX: RD = Lei

y
where Lei is the total number of events for which the drinkability threshold is
exceeded (EC, 1980) in observed or simulated daily concentrations, and y is the
number of years of the time series. Four sub-indexes can be calculated: RD NR (for

123
 nitrates in runoft), RD NL (for nitrates in leaching), RD pR (for pesticides in runoff) and
RD pL (for pesticides in leaching).

MAMMAL TOXICITY INDEX:

where l; is the annual average of the quantity of each active ingredient lost into water,
and AD1;, is the corresponding chronic toxicity of that active ingredient (Grilli, 1994).
Two sub-indices can be calculated: MTR (for pesticides in surface water) and MTL
(for pesticides released into groundwater).

NON-MAMMAL TOXICITY INDEX: NTR (for pesticide runoff) NTR =L e 2

y
where rez is the total number of events in which LCso values are exceeded in surface
water by at least one of the active ingredients used.

NTD (for drift effects) NTv =L (d; . p. LC~;)

where d; is the annual dose of pesticide, LC50; the corresponding toxicity, and p the
estimated drift fraction, which is 0.01 for open field applications with downward
spraying on crops less than 25 cm in height, 0.02 for taUer crops, and 0.1 for lateral or
upward-directed spraying on fruit crops (EPPO, 1993).

EUTROPHICA TION RISK INDEX: ER = Lye3

where re3 is the sum of events in which the 10 mg r l total phosphorus limit or 30 mg
r l total nitrogen limit is exceeded in surface water (Marchetti, 1993).

FARMER'S RISK INDEX: FRA (acute toxicity) FRA =L (d; . LD;~;)

FRc (chronic toxicity) FRA =L (d; . AD1;-' )

where d; is the annual dose and W50; (rat) and ADI/ are the corresponding
parameters for acute and chronic toxicity, for each active ingredienl

Applying this set of indices aUows the inputs and outputs of the agro-environmental
simulation model to be elaborated and to obtains a set of comparative indicators. In turn,
they can be used with multicriteria evaluations or coupled with economic indicators of
the agricultural systems being compared.

124
3.3. Application 0/ the evaluation indices to the studied case

The two sets of alternative farm scenarios for the small farms on the Veneto Plain were
examined to simulate their environmental implications. The crop-allocations percentages
for the eight efficient solutions used were examined. In light of the knowledge about the
cultivation systems and environments of the Veneto Plain, the following assumptions
were made (Table 3):
• intensive-cropping techniques were hypothesised for the first four scenarios deriving
from the market reform (EC Reg. 1765/92), whereas reduced-input techniques were
represented in the second four, deriving from EC Reg. 2078/92;
• the use of irrigation was simulated for the scenarios with catch crops (two crops per
year; see footnote 3), based on normal practice on the Veneto Plain;
• the agronomic use of bovine liquid manure was represented in farming systems
MMGP and MGM (EC Reg. 1765/92), whose crop allocation is typical of farms with
livestock;
• as previously mentioned, the crop allocations identified using multiobjective
techniques were elaborated on in light of the cropping systems (rotations) in which
they could realistically be organised. A set of possible rotations was thus defmed that
can be adopted on the Veneto Plain (four-year, three-year and two-year), on the basis
of which the percentages hypothesised for the different crops were re-read (see
second part of Table 3. Even if it is not certain that farmers would really adopt rigid
crop-rotation schemes, this was done to obtain suitable scenarios for long-term
simulations, and to avoid agronomie nonsense, such as the simulation of sugar beet
monoculture.

Thirteen sets of input parameter files were bullt to run the GLEAMS model on a 30-
year daily weather data set and pedologie eonditions of a representative site of the
Veneto Plain6 • Tbe simulation scenarios described entire rotations based on a
combination of five crop types, set-aside and catch crops, with two intensity levels and
either with or without the use of liquid manure.
The results in terms of pesticide and nutrient losses were used to calculate the impact
indices, which were then attributed to the farm systems according to the percentages of
each crop and rotation in every scenario. The results of this phase showed relative1y
relevant leaching events, with moderate phenomena of surface runoff and negligible soll
losses, which were consistent with previous experimental and modelling data. With these
patterns of soil water balance, phosphorus losses were generally low and unaffected by
different crops. Maize showed the widest field of variation and maximum values for
nitrogen releases, thus confirming that this crop is a potential origin of high releases

6 The representative site from which environmental data were taken is the Experimental Farm of the
Faculty of Agriculture of the University of Padova. This location was chosen for its central position in
Venice Lagoon Basin. the area where the territorial extrapolations of the final stage of the project are
planned. For details of the results see Giupponi and Morari. 1995.

125
(especially where livestock manure has been spread); however. it is also a crop that can
respond effectively to the introduction of low-input techniques. Sugarbeet confirmed its
high level of nutrient uptake. with no link between a higher nutrient availability and
coincident increasing releases. Soybean released not negligible quantities of nitrogen (in
particular organic N after harvesting) on which it is more difficult to intervene. since
such releases are not linked to human inputs; for this crop it appears that ways to limit
releases must be sought in tillage reduction.

Table 3. Alternative production scenarios (according to the crop allocations defined in section /.1) for
the smallfarms on the Veneto Plain; aRronomic inteR ration in terms ofinputs and possible rotations.
FARM SCENARIOS
EC Reg. 1765/92 EC Reg. 2078/92
MR ADDGP MMGP MGM MR ADDGP MMGP MGM
CROP ALLOCA TION % %
Soybean (S I) 14.0 0.0 0.0 0.0 32.5 25.0 18.0 9.0
Soybean 2° crop (S2) 39.0 50.0 42.2 0.0 0.0 0.0 0.0 0.0
Maize (MI) 0.0 25.0 25.0 25.0 0.0 22.0 32.0 33.0
Maize 2° crop (M2) 0.0 0.0 3.8 50.0 0.0 0.0 0.0 0.0
Wheat(W) 49.0 50.0 0.0 0.0 32.5 28.0 25.0 33.0
Barley (Y) 0.0 0.0 50.0 50.0 0.0 0.0 0.0 0.0
Sugarbeet (B) 23.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0
Set-aside (A) 10.0 0.0 0.0 0.0 10.0 0.0 0.0 0.0
INPUT LEVEL high high high high low low low low
IRRIGATION yes yes yes yes yes no no no
USE OF MANURE no no yes yes no no no no
TYPEOF ROTATION* % %
B-W-S-A 40.0 0.0 0.0 0.0 40.0 0.0 0.0 0.0
B-W/S2-W/S2-W/S2 52.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
B-W/S2-M I-W/S2 0.0 100.0 0.0 0.0 0.0 0.0 0.0 0.0
B-Y/S2-MI-Y/S2 0.0 0.0 84.8 0.0 0.0 0.0 0.0 0.0
B-Y/S2-MI-YIM2 0.0 0.0 15.2 0.0 0.0 0.0 0.0 0.0
B-YIM2-M 1-YIM2 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0
B-S-MI-W 0.0 0.0 0.0 0.0 0.0 88.0 72.0 36.0
B-W-S-W 0.0 0.0 0.0 0.0 0.0 12.0 0.0 0.0
B-W-MI-MI 0.0 0.0 0.0 0.0 0.0 0.0 28.0 0.0
B-W-S 0.0 0.0 0.0 0.0 45.0 0.0 0.0 0.0
B-W-MI 0.0 0.0 0.0 0.0 0.0 0.0 0.0 48.0
W-S 8.0 0.0 0.0 0.0 15.0 0.0 0.0 0.0
W-MI 0.0 0.0 0.0 0.0 0.0 0.0 0.0 16.0
*) Meanings of symbols are stated in the first part of the table (crop allocation). Dashes separate crop of
different years. slashes separate crops of the same year (main crop 1 catch crop); e.g. B-W/S2 means:
first year sugarbeet. second year winter wheat as main crop and soybean as catch crop.

Results of the model applications to the set rotations were subsequently weighted to
give information about farming scenarios, by means of the percentage values reported in
Table 3. Absolute values obtained for each index were normalised between zero
(minimum value caIculated for the eight farm scenarios) and one (maximum value).

126
 Marked differences were obtained between the two groups of hypothetical systems
(Table 4). Those regulated by market reform showed, as expected, higher impact indices.
The RDpL index presents a value in scenario MGM (EC Reg. 1765/92) that is at least 3
times that of any other scenario, conftrming the environmental concern aroused by
farming systems that use livestock manure and intensive cropping. The RD NR index is
equal to zero in all scenarios, since events of surface runoff with nitrates exceeding the
concentration limits for drinkability were never simulated.
Something similar happens for the MT R index in the low-input scenarios (EC Reg.
2078/92), and therefore does not give rise to concern about chronic toxicity for surface
waters. In no case is the highest index found in the low-input scenarios; this can always
be attributed to scenario MGM and in some cases to MMGP (EC Reg. 1765/92).

Table 4. Impaet In
. d'lees.
FARM SCENARIOS
EC Reg. 1765/92 EC Reg. 2078/92
IMPACT INDICES MR ADDGP MMGP MGM MR ADDGP MMGP MGM
RDPL 0.20 0.31 0.22 1.00 0.04 0.09 0.09 0.04
RDNL 0.18 0.37 1.00 0.93 0.24 0.63 0.52 0.31
RDpR 0.53 0.77 0.77 1.00 0.52 0.63 0.65 0.52
RDNR 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
MTR 0.32 0.45 0.40 1.00 0.00 0.00 0.00 0.00
MTL 0.34 0.65 0.67 1.00 0.22 0.37 0.39 0.22
NT 0.55 0.75 0.81 1.00 0.15 0.18 0.17 0.13
ERN 0.97 0.98 1.00 0.85 0.53 0.58 0.57 0.54
ERp 0.46 0.81 0.84 1.00 0.12 0.31 0.32 0.22
FRA 0.58 0.79 0.84 1.00 0.09 0.11 0.11 0.09
FRC 0.61 0.79 0.80 1.00 0.21 0.24 0.23 0.20

4. Economic-environmental evaluation

4.1. Background

The economic and environmental evaluation of the cropping systems described was done
using multiattribute analysis (MAA). As reported above, this analysis allows pre-defmed
alternatives to be selected with reference to a given set of parameters (attributes)
expressing the criteria involved in the choice. MAA consists of a fairly large number of
algorithms and procedures which start from a set of not-dominated alternatives. These
procedures, which identify the alternative that maximises a utility or value function, differ
widely over the amount and quality of necessary data and sophistication of the
calculations (Keeney and Raiffa, 1976; Saaty, 1980; Hwang and Yoon, 1981, Nijkamp er
al., 1990, Vincke, (992). In this paper, on the basis of the amount and quality of the
available information and characteristics of the decision problem, a linear and additive
ranking procedure was adopted (Crawford, 1973; Einhorn and McCoach, 1977; Otway
and Edwards, (977). The approach might appear simplistic but it has the merit of being

127
clear and, especially in the absence of appropriate and reliable utility functions, verifies
the motivations of the rankings obtained in a direct and intuitive way.
The procedure was applied to the analysis matrix containing the performances of the
four cropping systems obtained with the multiobjective model with respect to the
evaluation parameters indicated previously. Mathematically:

MAX TU s = WfusFU s + WeqsEQs s = decision scenarios

with FUs =WgrnsGM + WrisRI + WtrasFRA
EQs = WrdsRDs+ WrntsMTs + WntsNTs + WersERs
RDs = WplsRD pL + WnlsRDNL + WprsRD pR
MTs = W1sMTL + WrsMT R
ERs = W nsERN + WphsERPh
and Wfus + Weqs = I
Wgms + Wris + Wfras = I
Wrds + W rnts + W nts + Wers = I
Wpls + Wnrs + Wprs = I
W1s + Wrs = I
Wns + Wphs = I

where: TU = total utility;

FU = farmers' utility;
EQ = environmental impactlquality;
GM = gross margin;
RI = economic risk index;
MT =mamma) toxicity;
RD = risk for the runoff waters in terms of drinkability;
NT =toxicity for aquatic life;
ER = eutrophication risk;
FRA = risk to the farmers for pesticides used;
W = weights.

4.2. The evaluation matrices

The first step in the procedure is the construction of standardised evaluation matrices
containing the performances of the different alternatives. These matrices were
standardised O-MAX to keep the initial variability of the evaluation parameters unaltered
and transformed in such a way as to associate a greater utility to the highest values of the
attribute (see Tables 5 and 6).
The transformation was made applying a linear function and included only those
parameters which had a decreasing utility with the increase of the parameter (RI and
environmental impact indices). It was assumed that the contribution of the single
decisional parameters to the total utility would be linear. This choice seemed reasonable
within the interval of variation of the parameters.

128
 Toble 5. Evaluation matrix ofthe standardised and transformed so/utions to the model" J765/92 ".
SOLUTIONS
ATTRIBUTES MR ADDGP MMGP MGM
Gross margin (GM) 0.797 0.933 0.938 1.000
Index of risk (RI) 1.000 0.822 0.786 0.297
FRA 1.000 0.795 0.745 0.582
RDpL 1.000 0.884 0.973 0.197
RDNl 1.000 0.812 0.185 0.257
RDpR 1.000 0.760 0.760 0.530
MTL 1.000 0.687 0.670 0.336
MTR 1.000 0.865 0.924 0.319
NT 1.000 0.796 0.740 0.548
ERN 0.877 0.867 0.849 1.000
ERJ>h 1.000 0.654 0.617 0.461

Toble 6. Evaluation matrix ofthe standardised and transformed so/utions to the model "2078192 ".
SOLUTIONS
ATTRIBUTES MR ADDGP MMGP MGM
Gross marl!;in (GM) 0.872 0.941 0.967 1.000
Index of risk (RI) 1.000 0.765 0.652 0.488
FRA 0.999 0.979 0.980 1.000
RDpL 1.000 0.947 0.950 0.998
RDNl 1.000 0.601 0.717 0.927
_l@>ß 0.995 0.893 0.873 1.000
MTL 0.996 0.843 0.823 1.000
MTR 1.000 0.999 0.998 1.000
NT 0.974 0.951 0.962 1.000
ERN 1.000 0.947 0.957 0.994
ERPh 1.000 0.819 0.802 0.909

4.3. Assessment 0/ the weights

The assessment of weights to the different criteria was done with respect to some
alternative scenarios connected to the farmer and the environmental characteristics of the
site of the farm. Regarding the farmer, two limiting situations were considered (see Table
7). The first provides for someone totally dedicated to maximising gross margin and
therefore indifferent to income variability and to any health risks. In contrast, the second
assurnes that the decisions are also taken on the basis of prudence and therefore gives
some importance to restricting income variability and tries, even if to a limited extent, to
reduce the use of toxic production factors. The weights that represent these management
scenarios have been drawn from the results of a study aimed at analysing the
characteristics ofthe farmer's decision-making process (Bombana and Rosato, 1996).
The weights of the environmental scenarios were defIned by evaluating the relevance
of the different possible impacts in the different environments of the Venice Lagoon
Basin (see Figure 2).

129
 For this purpose three large areas in the north, centre and south-east of the basin were
identified. The northem plain has a relatively wetter climate, soil lying above layers of
gravel and therefore highly penneable. This means that the water flow has a consistent
component of vertical movement that feeds an aquifer of enonnous importance for
drinking water. Between the northem and central plain there is a belt of springs
originating from this groundwater, which are the source of the many rivers that fonn die
hydrography of the basin. The central plain is crossed by these rivers (Dese, Marzenego,
Tergola, Zero) and by several canals that become suspended above the level of the
surrounding land on entering the eastem plain, which lies mainly below sea level, and are
therefore managed by a network of pumping stations for mechanical drainage.

. h(S.
e . anagement scenarIOs a nd welRI
Tab17M
TYPEOFFARMER
A TIll.1 BUn::s INSENS ITIVE TO RISK AVERSE TO RISK
GROSS MARGIN (GM) 1.0 0.7
EcONOMIC RlSK (RI) 0.0 0.2
R1SK TO HEALTH (FRA) 0.0 0.1
AGGREGATED TOTAL 1.0 1.0

• Euganean Hills
• Aqu~er recharge area
Spring bell
o Middle plain
o Areas belowsea level
o Venice Lagoon

North
(j)
. UTM32
MeIers

25.~

Figure 2. The Venice Lagoon Basin.

Regarding the environmental impact indices, the frrst step was to define relative
weights, varying according to the specifications of the three areas. Successively, the four
aggregated environmental indices were standardised within each area. Table 8 shows the
weights of the environmental impact indices in the three areas.

130
 Tabte 8 EnVlronmen(a
. . h(s.
scenarIOS and welgl
Environmental scenarios
Attribute Northern plain Central plain Eastern plain
Aggregate Elementary Element. Aggreg. Element. Aggreg. Element. Aggreg.
RDPL 0.30 0.20 0.15
RD RDNL 0.30 0.20 0.15
RDPR 0.40 0.60 0.70
TOTAL 0.20 0.15 0.05
MTL 0.60 0.40 0.30
MT MTR 0.40 0.60 0.70
TOTAL 0.45 0.40 0.25
NT TOTAL 0.25 0.15 0.20
ERN 0.70 0.70 0.70
ER ERPh 0.30 0.30 0.30
TOTAL 0.\0 0.30 0.50
TOTAL 1.00 1.00 1.00

The RD index has decreasing weights going from the northern to the eastern plain.
This follows the decrease in the use of water resources in the three areas for drinking
purposes. A similar trend was considered for the weights of the toxicity index MT.
However, the decrease is less marked being related to an impact factor directly
correlated with human health.
The index that identifies risk for aquatic life (fIsh in particular) (NT) was weighted
according to the use of water bodies for fIshing. In the north, the water courses have
fairly good quality that make them suitable for aquatic life and, therefore, for fIshing and
the raising of rainbow trout. The waters of the central belt, because of pollution from
various sources, especially urban and industrial, have limited potential for fIshing and
aquaculture. Lastly in the eastern plain, the water quality is gene rally poor, but this is in
proximity to the lagoon, where fIshing and aquaculture are important. On the basis of
these considerations the weights descend from the northern to eastern plain.
The eutrophication risk was considered to increase in proximity of the lagoon where
this phenomenon is of particular concern. During nutrient transport (nitrogen and
phosphorus) in the water network various phenomena can contribute to self-purification
of the runoff waters. Therefore, releases of the same magnitude contribute more to the
lagoon pollution as they are closer to it.
When the importance of the different types of impact on the northern plain were
evaluated, MT had the highest weight, RD and NT an intermediate weight and ER the
lowest. In contrast, for the eastern plain, the highest weight is that of the ER index,
whereas MT and NT have intermediate values and RD the lowest.

4.4. Discussion

The results of the multiattribute evaluation are shown in Table 9, for the model relating
to EC Reg. 1765/92 and in Table 10, for the model EC Reg. 2078/92. These tables
report the performances of the different solutions identifIed in Section 1.3. with respect

\3\
to farmer's utility (FU) and aggregate environmental quality (EQ). As regards the policy
designated by EC Reg. 1765/92 the situation changes in line with the scenarios chosen.

Table 9. Economic and environmental evaluation ofthe solutions to model" J765/92 ".
STA"ffi OF TIfE AGGREGATE INTERVAL VARIA nON WEIGHTS
INDICATORS (Wr:ofWFlJ)
SCENARIOS SOLUTIONS BQ FU INITIAL FINAL
MR 0.991 0.797 110 0.40/0.60
Farmer insensitive to risk ADDGP 0.781 0.933 0.39/0.61 0.16/0.84
of the northem plain MMGP 0.742 0.938 *
MGM 0.428 1.000 0.15/0.85 Oll
MR 0.974 0.797 110 0.43/0.57
Farmer insensitive to risk ADDGP 0.797 0.933 0.42/0.58 0.23/0.77
of the central plain MMGP 0.780 0.938 *
MGM 0.532 1.000 0.22/0.78 Oll
MR 0.957 0.797 110 0.47/0.53
Farmer insensitive to risk ADDGP 0.803 0.933 0.46/0.54 0.29/0.71
of the eastem plain MMGP 0.785 0.938 *
MGM 0.632 1.000 0.28/0.72 Oll
MR 0.991 0.858 110 0.17/0.83
Farmer averse to risk ADDGP 0.781 0.898 0.16/0.84 Oll
of the northem plain MMGP 0.742 0.892 dominated
MGM 0.428 0.819 dominated
MR 0.974 0.858 110 0.19/0.81
Farmer averse to risk ADDGP 0.797 0.898 0.18/0.82 Oll
of the central plain MMGP 0.780 0.892 dominated
MGM 0.532 0.819 dominated
MR 0.957 0.858 110 0.2110.79
Farmer averse to risk ADDGP 0.803 0.898 0.20/0.80 0/1
of the eastem plain MMGP 0.785 0.892 dominated
MGM 0.632 0.819 dominated
* Solutions not dominated but situated in a non-convex region of the efficient frontier

Assuming a farmer insensitive to risk on the northern plain, the solution which is
favoured is obtained by minimising fmancial risk (MR). As will be seen in the following
scenarios, a strong relationship exists between fmancial risk and environmental impact
and, therefore, although at a different level, the solutions which minimise risk are also
those with lower environmental impact. The additive compromise solution (ADDGP)
wins when the weight of the farmer's utility is twice as high as that of the environmental
impact. The solution which maximises gross margin (MGM) is preferred when the
environmental impact has a marginal role. It should be noted that the solution of MinMax
compromise (MMGP), while not being dominated, is never selected. This is due to the
fact that the introduction of the variables expressing environmental impact determines a
"local non-convexity" in the efficient frontier in correspondence to the solution MMGP.
Therefore, by utilising a linear utility function, it is always rejected. This phenomenon is
accentuated for a farmer sensitive to risk, leading to a true Paretian dominance.

132
 Table 10. Economic and environmental evaluation ofthe solutions to model "2078192 ".
STATEOF THE AGGREGATE INTERVAL VARIA nON WEIGHTS
INDICATORS (WEofWFU)
SCENARIOS SOLUTIONS EQ FU INITIAL FINAL
MR 0.993 0.872 1/0 0.994/0.006
Farmer insensitive to risk ADDGP 0.900 0.967 dominated
of the northem plain MMGP 0.898 0.941 dominated
MGM 0.992 1.000 0.993/0.007 0/1
MR 0.994 0.872 1/0 0.964/0.036
Farmer insensitive to risk ADDGP 0.922 0.967 dominated
of the central plain MMGP 0.922 0.941 dominated
MGM 0.989 1.000 0.963/0.037 0/1
MR 0.995 0.872 1/0 0.924/0.026
Farmer insensitive to risk ADDGP 0.927 0.967 dominated
of the eastem plain MMGP 0.925 0.941 dominated
MGM 0.983 1.000 0.923/0.027 Oll
MR 0.993 0.9\0 1/0 Oll
Farmer averse to risk ADDGP 0.900 0.905 dominated
of the northem plain MMGP 0.898 0.909 dominated
MGM 0.992 0.898 dominated
MR 0.994 0.9\0 1/0 Oll
Farmer averse to risk ADDGP 0.922 0.905 dominated
of the central plain MMGP 0.922 0.909 dominated
MGM 0.989 0.898 dominated
MR 0.995 0.9\0 1/0 Oll
Farmer averse to risk ADDGP 0.927 0.905 dominated
of the eastem plain MMGP 0.925 0.909 dominated
MGM 0.983 0.898 dominated

In the central plain, on average more environmentally resilient, the weight intervals
within which the MR and ADDGP solutions win narrow down, whereas they widen
when solution MGM is preferred. The trendis accentuated in the eastern plain. This
shows that the alternatives considered are better suited to tackle groundwater pollution,
which is more vulnerable in the northern plain where the model favours less intensive
cropping systems. This, however, does not exclude impact on surface waters. Taking a
farmer averse to risk, two alternatives (MMGP, MGM) are dominated and therefore the
choice is reduced to the solutions MR and ADDGP. This dominance is due to the
correlation between fmancial risk and environmental impact mentioned above. In this
scenario, this is accentuated by introducing health risk of the operator in the farmer' s
utility function. As regards the two non-dominated solutions, a situation qualitatively
similar to the previous one is found, with the MR solution widely favoured in all
environments, especially in the northern plain. Lastly, it should be noted that the effect of
the management scenarios in ordering the alternatives is much greater than of the
environmental scenarios.
Taking the solutions to the model 2078/92 the situation is notably simplified. If the
farmer is insensitive to risk, both compromise solutions are dominated by the MGM
solution that, amongst other things, is widely favoured in all the environmental scenarios.
This because the rules of EC Reg. 2078/92 imply a substantial uniformity of the

133
environmental impact of the alternative solutions identitied, while reasonable ditferences
exist as regards gross margin. When, instead, the farmer becomes sensitive to risk, then
the MR solution dominates all others and therefore the problem of choice disappears.
The differences between the solutions are in any case fairly modest, as the rules of
2078/92 heavily intluence crop allocation.

5. Conclusions

The analyses demonstrate that the CAP reform has c1early increased the need for support
tools for farm management. In the past they were essentially used for the optimisation of
land use resources. The income support schemes introduces by CAP require the tools
used up to now to be adapted and developed further. The simulations have shown that
the general support system is appropriate where there is strong aversion to risk and the
simplified scheme when maximising gross margins becomes more important.
Furthermore, minimising risk favours leguminous crops while the spring cereals are
preferred when maximising gross margins. The steadiness of wheat and sugarbeet in all
the farming systems identified is worth noting, as is the obvious suitability of adopting
eco-compatible agricuItural practices.
Nitrogen treatments increase with earnings requirements and, adopting the measures
provided for by EC Reg. 2078/92 would, on average, halve them. Under the
environmental profile the analysis has demonstrated differing trends between indices
relating to nutrients and those relating to pesticides. In the high-input rotations the
thresholds are exceeded more frequently because of pesticides, except when livestock
manure is used. In the low-input rotations, exceeding the thresholds, when it occurs, is in
general due to releases of nitrates in leaching water. Nitrates never exceed the threshold
levels in runoff waters. The risks of toxicity for mammals are generally almost nil in the
low-input rotations, but high in the rotations with catch crops where there is consistent
use of chemicals. The herbicides, although being used on a plain, move mainly with
surface runoff when they are released. Consequently, the events exceeding the thresholds
and the higher total quantities affect the surface waters. The thresholds of toxicity for
fish are never exceeded in the runoff waters and the risk of eutrophication of the surface
waters varies less between high- and low-input rotations, especially as regards
phosphorus releases, whereas those of nitrogen are heavily intluenced by the use of
animal wastes. The trends of the index of risk of acute and chronic toxicity for the farmer
follow those of the indices of toxicity for mammals fairly weil, even though they are
calculated on doses used instead of on releases. Regarding the energy index, the main
intluence is generally due to nitrogen releases, the intluence of pesticides being normally
from 10 to 100 times lower; for this reason, similarly to the eutrophication indices, there
is little difference between high- and low-input systems.
Lastly, the joint economic-environmental evaluation showed that there is a c1ear
synergy between the tendency of the farmer to minimise the risk of cropping systems and
the need to reduce the environmental impact of agro-technologies. This consideration is

134
particularly useful in the promotion of low impact farming practices in a situation
dominated by traditional production techniques with highly variable environmental
implications. In the scenarios dominated by EC Reg. 2078/92 the situation is very
homogeneous under the environmental profile and therefore the conflicts are more or
less non-existent. From the methodological point of view the work appears to have
demonstrated a good potential of multicriteria analysis in integrating economic and
environmental evaluations. This integration is essentially based on the connection
between multiobjective models, which identify the cropping systems, and multiattribute
models, which serve for their evaluation. Clearly many questions still remain open. First
of all, the effect of the double evaluation, first multiobjective and then multiattribute,
which gives rise to dominance phenomena that, in the present study, are particularly
clear. This confrrms the need to incorporate algorithms of environmental evaluation
directly into the farm models. Unfortunately, at present these algorithms are often too
complex to incorporate into matrices of linear programming and therefore the distinction
between the economic evaluation phase and the environmental one is, at present,
justified.

References
Azzolin, F., P. Rosato (1995), La gestione dei nuovi problemi di sce/ta dell'azienda agraria:
programmazione lineare multi-obiettivo a numeri interi, Studi di Economia e Diriuo, 4: 593-
624.
Bombana, M., P. Rosato (1996), Farming Objectives and Environmentallssues in the Venice
Lagoon Water Basin. Working paper, WP96-4, University of Minnesota.
Charnes, A., R. Ferguson (1955), Optimal Estimation of Executive Compensation by Linear
Programming. Management Science, 1(2): 138-151.
Crawford, A. B. (1973), Impact Analysis Using Differentially Weighted Evaluative Criteria, in 1.
L. Cochrane, M. Zeleny (Eds.), Multiple criteria decision making, University of South
Carolina, Columbia Press.
EC (European CommunitieS) (1980), Directive relating to the quality of water intended for
human consumption. (801778IEEC), OJ.E.C. L 229: 11-29.
EC (European Communities) (I 992a). Council regulation (EC) No 1765/92 of 30 June 1992
establishing a support System for producers of certain arable crops, OJ.E.C., L 181: 12-20.
EC (European Communities) (1992b), Council regulation (EC) No 2078/92 of 30 June 1992 on
agricultural production method.l' compatible with the requirements of the protection of the
environmentand the maintenance ofthe country.l'ide, OJ.E.C. L 215: 85-89.
Einhorn, HJ. W. McCoach (1977), A Simple Multiattribute Procedure for Evaluation.
Behavioural Science, 22: 270-282.
EPPO (1993), Decision-Making Scheme for the Environmental Ri.l'k Assessment of Plant
Protection Products, EPPO Bulletin, 23( I ):81.
Giardini, L., C. Giupponi, A. Marani, M. Tamaro (1992), Modelli di simulazione per le scelte
agronomiche in relazione aHa qualita delle acque, Rivista di Agronomia, 24(4): 714-720.
Giupponi, C. (1995), Modelling Agriculture and the Environment: Crop production and Non-
Point Source Pollution, European Journal of Agronomy 4(4): 403-412.

135
Giupponi, C. (1998), Environmental evaluation of alternative cropping systems with impact
indices of pollution, European Journal of Agronomy, 8:71-82.
Giupponi, C., F. Morari (1995), Agro-Environmental Evaluation of Alternative Farm
Management Systems FoUowing the European Community Reform of Agrieultural Polie)"
Working paper, WP95-3, University of Minnesota.
Giupponi, C., P. Rosato (1993), Environmental and Economic Simulations for Predicting
Agricultural Non-Point Source Pollution, in M. McAleer, A. Jakeman, (Eds.), International
Congress on Modelling and Simulation Proeeedings, M.S.S.A.-I.M.A.C.S. 705-710.
Grilli, S. (1994), Impostazione della problematica scientifiea e metodologia di lavoro. In:
Valutazione delta tossicitii per /'uomo degli antiparassitari ad uso agrieolo: il eontributo
della Regione Emilia-Romagna, Ass. Agricoltura Reg. E.R.
HazeIl, P.B.R., R.D. Norton (1986), Mathematical Programming for Economie Analysis in
Agriculture, New York, Macmillan.
Hwang, C. L., K. Yoon (1981), Multiple Attribute Deci.~ion Making - Methods and
Applications. A State-of-the-Art Survey, New York, Springer-Verlag.
Keeney, M., H. Raiffa (1976), Decisions with Multiple Objective, New York, Wiley & Sons.
Leonard, R.A., W.G. Knisel, D.A. Still (1987), GLEAMS. Groundwater Loading Effects of
Agricultural Management Systems, Trans. ASAE, 30(5): 1403-1418.
Marangon, F. (1992), Agricoltura a minor impatto ambientale ed economia dell'azienda agraria.
un approccio mediante analisi multiobiettivo, Rivista di Economia Agraria, 4: 545-585.
Marchetti, R. (Ed.) (1993), Ecologia applicata, Milano, Citta Studio
Markowitz, H.M. (1959), Portfolio Selection, New York, Wiley & Sons.
Morari, F., C. Giupponi (1998), Effects of Four Cultivation Systems for Maize on Nitrogen
Leaching: 11; Model Simulations, European Journal of Agronomy, 6: 113-123.
Nijkamp, P., P. Rietveld, H. Voogd (1990), Multicriteria Evaluation in Physieal Planning,
Amsterdam, North-Holland.
Otway, H., W. Edwards (1977), Application of a Simple Multi-Attribute rating Teehniques to
Evaluation of Nuclear Waste Disposal Site: A Demonstration. RM-77-31, Schloss
Laxemburg, IIASA.
Ridgley, M.A. (1995), Determining Rural Land Use Goals: Methodological Primer and an
Application to Agroforestry in Italy, The Environmental Professional, 17: 209-225.
Romero, C., T. Rehman (1989), Multiple Criteria Analysis for Agricultural Decisions,
Amsterdam, Elsevier.
Saaty, T. (1980), The Analytic Hierarchy Process. New York: McGraw-Hill.
Vincke, P. (1992), Multicriteria Decision Aid. New York: John Wiley & Sons.
Zeleny, M. (1973), Compromise programming, in 1. L. Cochrane, M. Zeleny (Eds.), Multiple
criteria decision making, University of South Carolina, Columbia Press.

136
 Part 111
Multiple actors and multiple perspectives
in land-use decisions
Spatial conflicts in transport policies: an exploration of
the perspectives of regional and local actors

Euro Beinatl •2 and Michiel van Drunen l

1 Institute for Ellvironmelltal Studies and 2Faculty of Ecollomics

Vrije Ulliversitei~ Amsterdam, l11e Netherlallds

Abstract
Transport policies in the European Union are directed towards the integration of national and
local policies while maintaining national and regional identities. Since few policies are able to
satisfy all needs and demands of all areas involved, spatial con11icts often occur. This paper
discusses the nature and characteristics of these conflicts and introduces a framework for their
study. The discussion is based on a large infrastructure project in the Netherlands. The paper
addresses spatial conflicts on the basis of the results of the public participation process, and on
the basis of a value function model assessed for individual spatial units. The objections raised to
the projcct by regional authorities, local administrations and citizcns are explored to analyse the
spatial distribution of objections and the reason for the intensity and distribution observed. The
value function models are used to analyse the altractiveness of the proposed plan from the
perspcctive of each unit. DiiIcrences between units and thc reasons for contlicts are then
identified. Con11ict resolution measures are brietly discussed within the framework of the project.

Keywords: spatial cont1icts, decision analysis, transport policies.

1. Introduction
The transport seetor in the European Union (EU) is a souree of employment and a
erucial faetor for the eeonomie growth of Europe. The foeus of EU transport polley is
manifoldt, but infrastrueture development still plays a fundamental role (cf. Commission
of the European Communitics, 1993). Infrastrueture is meant to inerease the aceessibility
01' European regions, to reduee the eonstraints for eeonomie growth, and to aehieve a
higher level of efficieney of the transport seetor.
Infrastrueture, however, is a blunt-edged instrument: it addresses the eompetitiveness-
elTicieney nexus only indireetly and has several side effeets, sueh as environmental and
social impaets (ef. Beinat, 1997b). Therefore, the planning and eonstruetion 01' new
infrastrueture eomponents always raises the problem of balaneing short- and long-term

1 A policy is defined as an action scheme which inc1udes activities to be carried out in a time horizon to

achieve spccific goals (cL Goodwin and Wright, 1991). A policy provides the orientation for the
activities to be carried out under the policy framework (e.g. measures to curb private transport), which
are further spccified into plans and programmes (e.g. a road-tax scheme or a rai!way improvement
scheme), and then implemented through aseries of individual projccts (e.g. a rai! link between two
cities).The term policy in this paper will inc1ude all these stages, unless differently specified.

139
costs and benefits. In addition, large plans like the Trans-European Networks (TEN),
affect large areas and give rise to a spatial distribution of costs and benefits. Since
regions share different identities, environmental conditions, sodal and economic
characteristics, spatial conflicts are likely to occur.
The paper has two main objectives. The frrst is to analyse the reason, nature and type
of the differences between spatial actors (actors which are assodated with a geographical
entüy, such as a region) indirectly, on the basis of their reactions to a proposed transport
plan. The second is to explore the value system of the actors directly, by assessing a
value function model for each policy unit. The Betuweroute railway in the Netherlands is
used as an example to show the spatial distribution of the actors, their different
perspectives and how this drives the evaluation.

2. A characterisation of spatial conflicts

2.1. Location and scale

Spatial decision problems are characterised by a location and ascale dimension. The
location dimension refers to the spatial differentiation of land, the spatial distribution of
impacts and spatial differentiation of evaluation perspectives. The first indicates the
obvious fact that po lides allect areas with dillerent spatial features. Spatial impacts
indicate that the costs and benefits of a policy are not uniformly spread in space. Spatial
evaluation perspectives indicate that dillerent individuals perceive and evaluate the same
impacts differently at different places (for instance, the different tolerance levels for noise
in large and small urban centres).
Differentiations in the spatial features of the land and in the spatial distribution of
impacts imply the appearance of spatially dependent evaluation norms. This dependence
is captured, for instance, by environmental protection measures which are spatially
differentiated or based on land-use characteristics. However, the perceptions, concerns
and priorities of people at different places are not a simple consequence of spatial
dilIerentiatons of socia~ econornic and environmental conditions; they have to be explored ex
1I0VO for every new evaluation (cf. the practice of Environmental Impact Assessment and

Strategie Impact Assessment; Therivel and Partidario, 1996; Wathern, 1988).

The scale dimension indicates that different policy impacts are appropriate for
evaluations at different spatial scales. For instance, the national consequences of large
infrastructure are addressed in terms of GNP increase, changes in greenhouse gas
emissions or modification to tbe landscape structure. At tbe local level, the same
infrastructure will cause, for instance, noise pollution and housing relocation. Even when
the same type of impact is important at various scales (for example, landscape ), scale
dependent features emerge. For instance, the changes in landscape structure are better
appraised at a large scale, while visual intrusion is meaningful at a local scale.

140
2.2. Conflicts

The notion of conflict usually distinguishes between intra-personal and interpersonal

conflicts (cf. Bogetoft and Pruzan, 1991). Intra-personal conflicts refer to the need, in
most cases, to accept poorer results for some criteria in order to achieve beUer results
for others. An intra-personal conflict occurs when none 01' the possible choices available
is best on all counts. Interpersonal conflicts occur when individuals disagree on a course
of action: what is best (or good, or acceptable) for someone is not so for someone else.
Interpersonal conflicts may arise due to one or more of a variety 01' factors including
(adapted from Bogetoft and Pruzan, 1991):

Value system factors:

• individuals have different values and objectives;
• they employ different criteria when representing their objectives;
• they have different preference relations (such as different weights);
Impact distribution factors:
• they are likely to be affected by different costs and benefits 01' the policy;
• the distribution 01' costs and benefits is perceived as unequal and unjust;
Uncertainty factors:
• actors disagree as to the likcly outcomes of an action;
• they may hesitate about the importance 01' some policy outcomes;
• there may be insufficient evidence or understanding of the policy outcomes;
• there may be uncertainty about the combined effects with olher policy measures;
Process factors:
• actors have difficulty in communicating with each other as to their values, objectives,
criteria, preferences and expectations;
• the role played by the actors is not satisfactory for all actors.

The notion 01' conflicts, as underlined by all these factors, is extremely genera1. It is
also clear that a number of definitions of the term conflict can be given. In the rest of the
paper, the term conflict will indicate a disagreement Oll tlte outcome 01 the decisioll
withill a givell decisioll cOlltext. This definition is fully specified when:
• the outcome is defined (such as the selection of a plan within a set of alternatives);
• the decision context is defined and agreed upon by all actors;
• the term disagreement is defined (e.g., disagreement on the best alternative; on the
ranking of alternatives; or on an attractive solution, etc.).

This definition is an end-of-the-pipe notion of conflicts, which are identified when the
desired outcomes für al1 actors are known: disagreements on individual factors do not
necessarily lead to conflicts. Also, this concept 01' conflict is outcome-oriented, rather
than process-oriented. However, in negotiation and dispute resolution literature (cf.
Susskind and Cruikshank, 1987; Nagel and Mills, 1991) major efforts are dedicated to

141
shaping the decision process and to dealing with all factors that are a source of
disagreement, so that the probability or intensity of disagreement on outcomes is
minimised.
The outcome-oriented concept of conflict is useful for ex-post conflict analysis, to
explore the causes of conflicts. It is also a useful support [or the re-design of pollcy
options. When the preferred outcome for an actor is identified through an analytical
model, [or example a multicriteria model, then it is possible to establish a link between
results and o~iectives, impacts, criteria, weights, etc.. Conflict analysis in this ease aims
at revealing the reasons for confliets by assessing the influenee of individual faetors. This
evidence can be used to explore alternative solutions (for instance, the design of
mitigation or eompensation sehemes) which eould diminish the confliets.

2.3. Spatial conflicts

Far spatial decision problems, it is useful to introduee the eoneept of policy units. A
polley unit is an area (an administrative area such as a region, or another area
homogeneous [or polley purposes, such as a river basin) to whieh a eoherent evaluation
perspeetive ean be associated. Although in such an area several decision actors may be
present, it is often possible to associate a unit to a polley position whieh is eharaeteristic
o[ the unit, and which differentiates it [rom the others. Units are often the levels at which
decisions are made (by national, regionalor local authorities) and where deeision powers
for projeets of publie interest reside.
Spatial eon11icts arise between poliey units when their preferred outcomes does not
coincide. If ac tors within a unit da not share the same point o[ view, contlicts may also
arise within units. However, these con11icts do not have a spatial dimension and will not
be eonsidered further in this paper.
The location and scale dimensions suggest a distinction between vertical and
horizontal con11icts. Vertical conflicts arise between polley units nested into eaeh other,
mainly due to the seale eiTeets. Examples are the eon11iets between the national,
provincial and local authorities on the aeeeptability of the Betuweroute railway (see
below). Horizontal eonfliets oecur between pollcy units at the same level, for instance
between loeal authorities eompeting for the alloeation of funds. Horizontal eonfliets
oeeur mainly due to loeation effeets.

3. A metbodological framework for spatial conflict analysis

A theory whieh systematically addresses spatial eonfliets is not available at present.

Rather, there is a set of interdependent theories and methodologies which deal with
single aspects of spatial con11icts (cf. among others, Densham, 1991; Armstrong, 1994;
Douven, 1996). Broadly speaking, one reeognises three main approaches: spatial analysis
and planning, deeision analysis and organisational theory. Spatial analysis and planning
foeus on the spatial eomponent of adecision problem (cf. Maguire et al, 1991).

142
Geographical Information Systems (GIS) play an essential role in this context due to the
amount and complexity of the information needed for spatial problems. Decision analysis
aims at an analytical and systematic exploration of the value systems 01' individuals
involved in the decision. It provides methodologies for problem structuring, value
judgement assessment, decision recommendation and sensitivity analysis (cf. among
others Rosenhead, 1989; Keeney, 1992; Vincke, 1992). Organisation theory focuses on
the relationship and interaction between actors in adecision process, including the role of
procedures, information flows and behavioural effects (Hersey and Blanchard, 1988;
Simon, 1991; Teisman, 1992; Huigen et ai., 1993).
The common ground of these approaches can be referred to as Collaborative Spatial
Decision Support (CSDS; NCGIA, 1996). The components of CSDS are organised as in
Figure 1 (adapted !"rom Burgoyne, 1994). The figure distinguishes between several actors
which participate during the structuringldesign, evaluation and negotiation phases. At
each phase, an actor is characterised by the roIe played, the value system driving the
actor's behaviour and Ihe set of facls which allows an actor to take a position. As an
example, at the evaluation stage some actors have a specifie power to influence the
evaluation, while others may only indirectly alTeet the outcome of this phase. This
corresponds 10 the "roles" dimension. An aetor will need evidence on the outcomes and
impacts of policy options to be able to assess the attractiveness of a poliey from the
actor's perspeetive. This eorresponds to the "facts" dimension. Finally, the position of
actors depends on their priorities, concerns and world views. This eorresponds to the
"values" dimension.

"Role s"
(fnfluence,
Stiltus, po'Nf::rl

'Values"
(world-views. objectives
pocrillts, conet:rnsl

"Facts "
(Information, knowledge.
e:vldel"jce)

Strucrurlng Evaluation Negotlation

and d@slgn

Stages of t he de cisi on proces s

Figllre 1. Re[erellce framework [or stakeholder analysis (adapled [rOlli Burgoyne, 1994).

143
 This framework is based on the following hypo thesis:· (1) the possibility of finding one
or more time windows within which; (2) the actors involved in the process can be
identified; (3) their value system, that is their concerns, expectations, points of view,
objectives etc., can be made explicit; and (4) sufficient evidence on the consequences of
the policy action(s) is available, so that each actor can determine if a policy measure is
attractive or unattractive.
Within this framework, the approach to conllict analysis is synthesised in Figure 2.
The methodology aims at the construction of a value model of the unit which measures
the attractiveness of a policy option from the unit's perspective.
The methodology distinguishes between five steps for each unit. The first is the
exploration of the concerns, expectations and objectives of a unit, usually described in
the form of qualitative statements. The second is the derivation of appropriate evaluation
criteria, which are the measurement rods for assessing the impacts of the policy. The
third is the impact analysis, which derives an impact profile for each policy option for
each unit. The fourth is the assessment of the attractiveness, or unattractiveness, of each
impact with respect to single concerns. The final step is the combination of these
evaluations into an overall attractiveness score for a policy option. Conflicts can be
detected by comparing the results of different units. This evidence can be used to inform
the actors on their relative position, to specify for which aspects the positions are more
incompatible, and whether this is related to subjective differences or to impact
differences.

---4 .Identifypolicy oplions ~ Idcntify agolS and ""liey unitil ..•. 1

Vnlln
Viilf •••

-
Unlt2 UitlH· •. ····. ..

List ooncems, expectatioßS,

objedivcs, ete.
Specify operational evaluation
criteria
,.
I· ., Estimale/predi"~ impacts of
poliey action
Rate attractiveness of
individual impacts
Evaluate overall allractiveness
ofoptions

~
I Identify conilicls I
Fig"re 2. Conj1ict analysis [ramework.

The approach differentiates between subjective information (concerns, criteria,

weights, etc.) and objective evidence (environment al impacts, economic impacts, etc.)
and derives from these two components the unit's position on a policy option. When
positions are compared, the separation between "facts" and "values" (cf. Keeney, 1992)
darifies the reason for the difference (or conflict) and supports the conflict resolution

144
stage. Since the model produces an analytical representation of the units positions, a
feedback process can be used to isolate the subset of factors (impacts, weights, etc.)
which are the most responsible far conllicts. This can be exploited to improve the policy
or to generate alternative measures, far instance including impact mitigation or
compensation for some actors.

4. A ca se study: the Betuweroute railway

The Betuweroute is a railway project which will connect the port of Rotterdam in the
Netherlands to Emmerich in Germany. The Betuweroute is meant to improve the
hinterland connections of the Rotterdam harbour, to increase the quality and availability
of goods transport to the rest of Europe, and to meet the future demands of freight
transport. The proposed railway will follow a route of about 130 km from Rotterdam to
Germany (Figure 3). Although some of the area around Rotterdam is urbanised, most of
the area affected is agricultural, pasture land or natural area.
The two-track high-speed railway is dedicated to freight transport and includes only
one intermediate station (labelIed CUP in Figure 3). The plan is designed to support the
activities of the port of Rotterdam, where the benefits are likely to concentrate. Other
areas affected by the routing decision can benefit only indirectly, depending on the
degree to which they can exploit the positive spillover of the harbour activities. Distant
areas, which are less dependent on Rotterdam tor business and employment, will see
fewer (if any) benefits from this project, although they are affccted by the environmental
and social costs. The Betuweroute project is characteriscd by a quite clear spatial
distribution of costs and bencfits, which is one of the main reasons tor the conllicts which
emerged during the decision process.

Figure .1. Map olthe Betuweroute ,.ai/way.

Scveral administrations and dccision actors are involved in this plan: the European
Union, the Dutch Government, the Ministry of Environment, the Ministry of Transport,
the Dutch Railway company, the Provinces of Gelderland and South Holland, 27

145
municipalities, several Non-Governmental Organisations (NGOs) and special interest
groups.
The main stages of the decision process are illustrated in Figure 4. Public participation
started within the Environmental Impact Assessment (EIA) procedure. Citizens, local
communities and organisations had the possibility to review and express their remarks on
the proposed routes, on their impacts and on the EIA study. EIA is mandatory for the
corridor se1ection, while previous decisions were not subject to EIA (such as the choice
of railway as opposed to road or water transport). This means that public participation
started after the strategic decisions were made.
The EIA was the basis of the Key Planning Decision (Phase 4), which presented the
selected corridor. Three subsequent stages were necessary to specify the final route and
the construction details. All stages ineluded public participation and appeals. In this step-
wise process, the scope of the decision is progressively reduced and public participation
at a certain stage does not affect past decisions. Ideally, each step is based on the public
recommendations from the previous one. In practice, the pace imposed by the national
authorities on the decision process made this impossible, resulting in overlaps and
confusion between procedures. This undermined the trust of citizens and reduced tbe
expectations of tbe effectiveness of participation (Boom and Metze, 1997).

1 91 EIA (environmental assessment and evaluation of alternative routes)

Key Planning Decision 1 (alternative "AlS" is selected)

Key Planning Decision 4 (changed on the basis of public participation)

7 94 Preliminary Draft Route Decision

1/95 Report Committee Hermans

----+ Public participation or appeal
Draft Route Decision 12/96 "Tracewet" procedu re
1/95 Other major events
Route Decision

Figure 4. The decision process for the Betuweroute.

Tbe project was initiated and firmly supported at the national level (especially by the
Ministry of Transport and the railway company). Tbe province of South Holland, wbere
Rotterdam is located, maintained a low profile and a ratber neutral position. Tbe
province of Gelderland, elose to the German border, bas been very critical of the project
due to tbe bigh regional environmental costs and the lack of benefits. Tbe municipalities
have reactcd in different ways, ranging from strong opposition to favourable judgements.
This depended on tbe magnitude of local impacts, but also on tbe specific attitude of
local communities. Interest groups and NGOs bave been active at all levels, and tbeir
views bave oiten been ineluded in tbe local, regional or national perspectives. All

146
eonsidered, a distinet position eould be assoeiated with national, regional and loeal
aetors, making it possible to eonsider them as spatial poliey units.

5. Analysis of the reactions to the proposed route

Most of the reaetions, objeetions and remarks resulted from the Draft Route Deeision,
when the projeet and the route details were presented. Provinees, munieipalities,
individual eitizens and interest groups had the possibility 01' expressing their eomplaints
to the projeet initiator - the national authorities - in eonformity with the law tor routing
decisions (TB, 1986). Table 1 shows the eount of eomplaints c1assified into eight c1asses:
1. Proeedures: role and influenee of Ioeal authorities and citizens, strueture and sequenee
o[ stages in the proeedure, alternatives eonsidered;
2. Living environment: air, soil, water and groundwater pollution;
3. Noise: noise and vibration disturbanee [or residential, working and reereation areas;
4. Nature and landseape: effeets on Hora and fauna; landseape and visual effeets; damage
to areas of high historical and arehaeologieal value;
5. Health and risks: risks of aecidents in the proxirnity of erossings of urban areas,
espeeially for transport o[ ehernieals or inflammable materials;
6. Aeeessibility: ehanges in the road patterns, severanee of loeal eommunities and
modification of the urban and loeal fabrie;
7. Damage: damage to individual properties, sueh as demolished buildings, and to
eeonomie aetivities, sueh as agriculture;
8. Other: the temporary disturbanee during eonstruetion, the teehnical ehoiees for
bridges and viaduets, ete.

Table 1 provides an aggregated overview of the remarks (see TB, 1996, for the full
list). The table distinguishes between eomplaints [rom the provinees of South Holland
and Gelderland, those of the 27 munieipalities affeeted (12 in South Holland and 15 in
Gc1derland) and the eomplaints of citizens and/or loeal groups in eaeh municipality.
Citizens' remarks are further c1assified into general and area speeifie remarks, whieh
relate to a speeific loeation.
Figure 5 shows the pereentage 01' eaeh type of eomplaint for provinees, munieipalities
and eitizens. Provinees show a diiIerent pattern eompared with munieipalities and
eitizens. Noise and proeedural re marks, however, aeeount for about 50% of all remarks
in all eases. The main diilerenee between the distribution 01' municipality and eitizen
re marks refer to aeeessibility and property re marks (e.g. demolitions). The laUer are
emphasised by citizens, dircetly affeeted, while aeeessibility and loeal planning
eonstraints are better addressed by Ioeal authorifies.

147
 Table 1. Summary o[ the complaints/remarks o[ 27 municipalities alld 2 provinces to the Betuweroute
plan (adapted [rolli Beillat et af., 1998).
PSH MSII CSII-lI CSII-s PG MG CG-!! CG-. TOTAL
Procedures (pal1icipation, timing of decisions, ... ) No 0 25 338 5 3 37 842 86 1336
% 0 25 42 2 23 26 30 7
Uving Environment (air poßution, water pollution, . .. ) No 0 5 20 5 0 2 110 23 165
% 0 5 2 2 0 1 4 2
Noise (houses and other buildings) No 5 30 148 55 3 34 686 471 1432
% 56 30 18 24 23 24 24 41
Nature and landscape (ecology, landscape, history, .. ) No 4 8 25 42 5 12 47 125 268
% 44 8 3 18 38 9 2 11
Health and risks (accidents, ... ) No 0 6 37 11 1 7 125 78 265
% 0 6 5 5 8 5 4 7
Aecessibility (Iocal severanee, route, .. ) No 0 18 18 59 1 41 41 270 448
% 0 18 2 26 8 29 1 23
Damage (individual properties, activities,.. ) No 0 4 125 40 0 5 588 90 852
% 0 4 15 17 0 3 21 8
Other (distumanee during construetion, . .. ) No 0 5 100 14 0 4 383 16 522
% 0 5 12 6 0 3 14 1
Tolal No 9 101 811 231 13 143 2821 1159 5288
% 100 100 100 100 100 100 100 100
Legend:
PSH =Province of South Holland;
MSH = Municipalities in South Holland;
CSH-g = Citizens in South-Holland Municipalilies: general remarks;
CSH-s =Citizens in South-Holland Municipalities: area specific remarks;
PG =Province of Gelderland;
MG = Municipalities in Gelderland;
CG-g =Citizens in Gelderland Municipalities: general remarks;
CG-s = Citizens in Gelderland Municipalities: area specific remarks.

~"o
HellltUIf1ClIhkt~...
- ..'"
Prooad.. ..

...........
.,. ---
.~

F igllre 5. Percelltage o[ remarks per dass ( Ielt: Provinces; middle: Mun icipafities; right: citizellS).

The number and spatial distribution of complaints correlale with some basic features
01' the project. As shown iri Table 2, the remarks of the municipalities are positively
associated with:
1. the length of the track in the municipality;
2, the noise impacts (number of houses exposed to higher levels than 57dB(a»;
3. the number of buildings demolished.

148
 More precisely, the environment-related remarks of the municipalities correlate with
the main project impacts, while the procedural and socio-economic remarks do not show
any significant association wilh these impacts. This indicates that socio-economic and
procedural concerns were raised independent of the costs suffered in the area. A similar
pattern, but less marked, can be observed for the citizens' remarks (see Beinat, 1988, for
details).
The predominant procedural complaints from the municipalities (insufficient
involvement of local authorities at the early stages of the planning; insufficient analysis of
radical alternatives) were raised independent of the impacts associated with an area. This
indicates a wish to participate at all stages of the design, planning and evaluation of the
project, independent of the extent of the consequences expected.
Concerning the citizens' objections, there is a positive correlation between the number
01' remarks and the position 01' the municipality in the east-west corridor (which
corresponds to the municipality number) and with the noise levels.
An attempt to relate the distribution of objections with the share of political votes in
the municipality has also been attempted. Although several significant associations can be
found, no dear trend emerges 1'rom this analysis. Similar negative results were obtained
for age distribution, urban-rural distinctions, income distribution, etc..

Table 2. Pearson correlation coefficients between area features, impacts and remarks [rom citizens and
lIIunicipalities (Bold: signijicant 95%; Italic: signi[icant 90%). The lIIunicipalities are nUlllbered east-
west froml (Rotterdalll) to 27 (Zevenaar).

Municipalilies Cilizens (g) Citizens (s) Citizens (total)

Area features
Municipality number (1-27) 0.19 0.43 0.41 0.43
Inhabitants 0.19 -0.08 -0.04 -0.07
Ml\jor impacts
Lenglh of track 0.34 0.23 0.19 0.22
Noise (houses > 57 dB(a» 0.46 0.43 0.43 0.44
Demolitions 0.62 0.25 0.23 0.25

From the data in Table 1 it is dear that most remarks co me from the Province,
municipalities and citizens in Gelderland. The distribution 01' remarks is shown in Figure
6. The province of Gelderland, its municipalities and citizens are generally more opposed
to the project compared to the province oi South Holland.

149
 100%

80%

60%
_ Gelde,tsnd
mSouttl Hollond
40%

20%

0%
P,ovinces Municip. Citi:zens~ g Citizens-s

Figllre 6. Percentage oi remarks in Gelderland and Soutlt llollan(l.

This trend is illustrated in detail in Figure 7 (re marks of the municipality

administrators) and Figure 8 (remarks of citizens in the municipalities). From both graphs
it appears that municipalitics in Gelderland (from 13 to 27) oppose the project with
higher intensity. This trend is acccntuated for the citizen rc marks. The municipality
rc marks focus especially in 10ur areas. This spatial distribution can be partly justified by
thc impact distribution in these areas. Figure 9, tor instance, shows a trend of a subset of
impacts which can clearly be associated with the trend in Figure 7.

40,-- - - - - - - - - - - - - - - - - - - -- - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,

OSocio-econom ic (Tot) Municipotitill!ls

35
IllEnvironmenl (T 01)
.Procedures (ToQ
30

25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

Figllre 7. Rcmarks oimunicipalities grouped into socio-ecollomic, procedural and envirollmental

remarks. rite Ilumbers in the x-axis correspond to the municipalities along the proposed line, numbered
irom 1 (Rouerdam) to 27 (Zevenaar). rhe maps shows the spatial distrilmtion oi the remarks (tlte darker
tlte area, the lliglter the number oi remarks).

150
 1600 .-----------------------------------------------------------~

O Socio-economic (Tol) Citizens

1400
IllErMlOnme nl (Tol)
. P,ocedUle. (T 01)
1200

1000

800

600

1 ~ 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 ~1 22 23 24 25 26 27

r (27)

Figllre 8. Remarks o[ citizens grouped illto socio-economic, procedural and ellvirollmental remark5.
71ze numbers in the x-axis correspond 10 the municipalities along fhe proposed fine, numbered [rom 1
(Rotten/am) to 27 (Zevel/aar). nIe maps shows the spatial {/istributio/t o[ the remark5 (the darker the
area, the higher the number o[ remarks).

1~r---------------------------------------------------------_.
D Demolitlons (10 d~mol lllons ,li nhabllllmts)
900 O Lenghth of lOltway "0'" "m/inhabl'anl)
«Nolse (10' bU1ld1r'lgsJintlabl1anis l
800

700

600

500

400

1 2 3 -4 5 6 7 8 9 ,0 " 12 13 14 1$ 16 17 18 19 20 2' 22 23 24 25 26 27

Figllre 9. Maill impacts [or the 27municipalities (norma!ised per ntlJ/lber o[ inhabitallls ill each
municipality). rhe numbers ill the x-axis correspond to the mUllicipalities along Ihe proposed !ine,
numbered [rom 1 (Rouen/am) 10 27 (Zevenam).

Municipality remarks, and citizen remarks in particular, tend to increase for locations
dislant from Rotterdam. The likely reason is that the economic impulse of the project will
peak in Rouerdam, while distant areas are less likely to benefit, making their costs-

151
 benefits balance less favourable. This simplistic explanation is not questioned by the
various economic studies which have been made.
First of all, although the economic eonsequences of the project are expected to be
positive, their magnitude remains uncertain (ePB, 1993, 1995; Hermans, 1995). The
clTectiveness of the railway for promoting a modal shin from road to railway is also
doubtful, since it will probably depend on external factors such as fuel price (Hermans,
1995). The diflleulties in predicting economic effeets are accentuated for the regional and
local seales. The number of jobs created or the ability to generate business for some
areas are hard to anticipate, making it ditlicult to perform any type of informed regional
and loeal assessment of benefits. In addition, recent studies on the regional e1'fects of
transport corridors (Bruinsma et al., 1997) show that the impacts of large transport
infrastructure, like highways, on regional economic growth are not signifieant (apart
1'rom the positive e1Iects to the transport sector). The loeal environmental and social
costs are, on the other hand, clearly recognisable. In a situation characterised by clear
costs and uncertain benefits, the acceptability of the projeet depends on the degree to
wh ich the costs can be minimised.
The differences which emerge 1'rom this analysis highlight a scale and location
differentiation 01' the position of policy units. However, an extensive statistical analysis of
these objections resulted in very few significant associations, like those presented in
Table 2. There are several possible explanations for this, mostly related to procedural
biases.
In some areas, public participation was guided, supported, and stimulated by the local
authorities, resulting in a high number of complaints. Although it can be argued that this
is an indication of high concern, the number oE complaints t'or some areas probably
includes an amplification factor which makes the comparison between areas ditIicult.
This distortion meehanism plays a role especially for eitizen remarks.
Objeetions also include a strategie behaviour component, which is relevant in light of
the evaluation process applied to the Betuweroute decision. At the level oE the Draft
Route Dccision (when the objeetions were analysed) only a few aspects of the projeet
eould still be modified. Many units eonsidered the plan at that stage as unattraetive, and
the ehanges still possible were not su1'ficient to modify their position. Sinee the national
aUlhoritics were determined to proeeed, the only sensible strategy for the Ioeal
communities was to request mitigations and compensations. Depending on several
factors, like the likelihood of obtaining the requested mitigation, this determined the type
and number 01' remarks prescnted, distorting the relationship between the objeetions and
the unattractiveness of the plan. This distortion ean be positive, and amplify the number
of re marks, but also negative, and reduce the complaints. This may have happened for
units expericnced with the proccdures, having being involvcd in similar evaluations in the
past. Some considered the elTectiveness of these parlicipalory tools as poor and did not
invest time and resourees in this approach. Interviews with the loeal authorities
highlighted that othcr informal channels werc often prefcrrcd to aehieve thc desircd
outeomes at the loeallevel.

152
6. Analysis of the value system of the units

6.1. Structuring the evaluation: concerns, criteria and impacts

The intrinsic (un)attractiveness 01' the project can be analysed with a multicriteria value
function model. This analysis was carried out ex-post and was limited to the Province 01'
Gelderland and to a samplc of eight 01' its municipalities 2 . The goal is to highlight
concerns, criteria and weights applied by each unit, as much as possible independent 01'
procedural distortions. Value 1'unction models were based on direct interviews3 . The
model requires, for each unit, the specification of:
1. the concerns 01' the unit;
2. the evaluation criteria which make concerns operational and measurable;
3. the estimation 01' impacts ror each or the criteria;
4. the assessment 01' the attractiveness 01' each impact separately;
5. the assessment 01' the relative importance 01' different impacts;
6. the aggregation rulc which leads to the overall attractiveness 01' the plan.

Figure 10 illustrates the main concerns 01' the national, regional and local policy units.
Some concerns are shared by all units (tür instance, landscape concerns), while some are
unit-specific and depend both on the scale and location 01' the unit. To assess the
attractiveness 01' the Betuweroute against these concerns, evaluation criteria need first to
be defined, so that the impacts of the railway can be measured in terms of the concerns
01' the policy units.
Table 3 shows concerns, criteria and main impacts lür the nine units considered in this
paper. Some concerns and criteria are common to all units (e.g. noise, landscape). Some
are relevant only for the Iocal or regional levels. Air pollution due to road traffic and
traffie eongestion, for instanee, are more meaningful at the regional level, while at the
locallevcl their consequences are hard to assess.
The constructed scales shown in Table 3 are simplified versions 01' those used in the
assessments. The scalc 1'or vihrations, for instance, distinguishes between:
1. Levels imperceptible tür people, with no damage to structures and buildings and no
c1Tects on industrial activities sensitive to vibrations.
2. Disturhance to industrial activities, or long-term potential damage to buildings and
constructions, or disturbance to pcoplc.
3. Levels which disturb peoplc, a1'fect buildings and structures and prevent the
operations 01' industrial activities sensitive to vibrations.

2 Municipalitics arc indicatcd with MI, ... , M8 and the Pruvince with P. The value model constructed for
cach unit is here reproduced only rar scientific purposes and to demonstrate the potentiality of the
method. No conclusions on the official position of the unit may be drawn [rom these results.
3 Interviews were either eonducted face-to-face, or by telephone, or with the help of questionnaires

spccifically designed for this purpose.

153
 NATIONAL CONCERNS CONCERNS OF OEl.J)ERlAND CONCaNS OF A MUNICIPAUTY
- .strengthen the position ol Rotterdam - . minimize noisc BInd vibrations
dccrca.sc thc tUtets of private transport . pruerve lnindSC8pc.
(pollution et< .. ) - avoid demolitioM
A minimizc costs of the

. . . . • Beluweroute:
- stfengthcn thc
• keep Ihe qu.lity of the
- minimiz.e environmental imDI'c:tslsW'
- avoid C\lmulating of impacts
· achieve high levels of compensation i nfJll$htldurc
• prevent separation of natural ",rea$

Figltre 10. A simplijied represelllation of the position of the main policy units.

Table 3. Overview of concems, criteria antI impact profiles for the provi/lce ami eight selected
mU/licipalities. Empty cells i/ldicate that the correspo/lding criteria are /lot included in the unit 's model.

......................... 9.~~.~~~~..(~..'!.~..~!EI2~~.~ ..~~.~~~.Ü?~~E~2............. "" p MI M2 M3 M4 M5 M6 M7 M8

Concerns
Effects on Demolition (n. buildings demolished) 175 15 3 4 17 5 72
peoplc Noise (houses exposed 10 > 57dB(a» 423 28 26 32 34 11 3 24 52
Vibrations (constructed scale) 3
Accidents (km line in unit) 81 8 5.4 7.2 5.4 2.6 1.8 0.2 7.3
Accidents (n. dangerous crossings) 0 1 2 1 1 2
Environm. Landscape (km in charact. landscape) 47 8 0.3 3 0 0 0.4 0.2 5.8
Landscape structure (constructed scale) 3 1 2 2 2 2 2 2
Nature (important areas affected) 12 0 1 2 0 1 2 1
Air pollution (constructed scale) 1
Traffic congestion (constructed scale) 1
Cultural heritage (n. sites affected) 11
Socio· Agriculture area lost (ha) 790 69 50 70 80 20 20 2
economy Economic impacts (constructed scale) 0 0 0 0 0 0 0
Temporary hindrance (years) 6
Planning constraints (constructed scale) 2 2 2
Costs for the unit (Million Guilders) 2
Scale for VibratIOns: 3=hlgh effcct; 2=moderate effect; l=no effect.
Scale for air pollution: 1=increase; O=no effect; -1= decrease.
Scale for traffic congestion: l=increase; O=no effect; -1= decrease.
Scale for landscape structure: 1=line bundled with highway; 2= partly bundled; 3=line in open area.
Scale for economic impact: 2=structural, positive; l=positive; O=none; -l=negative.
Scale for planning constraints: 2=high, significant constraint, 1=moderate constraint; O=none.

154
 Most impacts are calculated on the basis 01' a Geographical Information System.
Starting from the concerns of a unit and the criteria selected, a GIS was used to model
the effects 01' the railway and to cxtract thc impact profile tor the unit selected. An
examplc is shown in Figure 11 far two units with the same evaluation criteria.

Demolition 17 15
Noise 34 28
Accidents (line) 5.4 8
Acidents (crossings) 0 2
Landscape 0 8
Landscape structure 2 3
Nature 0 0
Agriculture 80 69
Economic impact 0 0

Figltre 11. l!xample o[ re[erence map [ar compUling impacts Jor two policy units. Impacts in bald are
calculated with GIS models.

6.2. Attractiveness of the railway for a policy unit

6.2.1. The value function model
Thc attractivcness of the railway for a policy unit was hased on an additive value
I'unction model. If Vi is the unit and X=(x/, ... ,x,/) is the impact profile of thc railway in
Vi, the attractiveness uf X is defined as:
" (1)
V(Xi)=V(X;, ... ,x:J= ~w~v~(x~)
f.f
where, w/ is the weight of criterion .i far unit Vi, and v/(xij) is the attractiveness 01' the
individual impact Xij . The assessment of the individual value I'unction v/(xij) was based on
direct rating on a seven-point value scale, while weights w/
were assessed through direct
rating 01' swing wcights (cL Beinat, 1997a).
Table 4 shows the attractiveness scores for the individual impacts, that is the v/(xij)
terms. Scores are normalised between +6 (impact level meets a strategic objective of the
unit), O=(neutral impact) and -6 (the impact level is unacceptahlc for the unit). For each
impact, only the positive (0/6) or negative (-6/0) range is used.
The 0 level (the neutral impact) was selected by the assessors. In most cases it was
associated to the status-quo, that is to thc projcction of the current situation into the
future. For nature and landscape structure same units have chosen a different reference
level, corresponding to the cxpected impact of a previous design 01' the railway. For M4,

155
for instance, the fact that no natural areas were atIected becomes a very positive
outcome (instead of a neutral outcome) compared to a previous design which had
significant ecological impacts.
By comparing Table 4 with Table 3 it is dear that similar impacts for ditIerent units
may receive different attractiveness judgements. MI and M4, for instance, score 15 and
17 demolitions, respectively. The attractiveness scores, on the other hand, are -6 and -2.
There are two possible explanation for this difference. The first is simply that demolitions
count less in M4. The second is the fact that demolition is only a proxy indication of a
more fundamental concern. The specific buildings involved, or the people who inhabit
these houses, or the location of the buildings in the urban fabric, are aspects which
contribute to the (un)attractiveness of demolitions. The number of buildings, therefore, is
only one of the indicators which are considered by a unit when assessing the
(un)attractiveness of demolitions. This indicator is used for all units because is easily
measurable and comparable, and because it is considered as a very important indicator by
all units.
The overall attractiveness of the railway was computed with a weighted sumo The
results are shown in Figure 12. The zero-line corresponds to the neutral profile for each
unit, while the -1 line is associated to the worst impact profile for each unit (worst
impact on all criteria). The overall attractiveness of the railway in a unit is normalised
between these two levels. All these units consider the Betuweroute as an unattractive
project. The level of unattractiveness is highest for M8, while M2 is the dosest to the
neutral levd. Figure 12 also shows the number of objections presented by the units,
normalised against the largest number within the group (that of M8). The attractiveness
score and the number of objections registered do not show any significant relationship.

Table 4. Attractiveness scores (-6=l1Iuxil1lul1I negative; +6=muxil1lul1I positive).

Concems Criteria (and impact descriptors) P M1 M2 M3 M4 M5 M6 M7 M8

Effectson Demolition (n. buildings demolished) -1 -6 0 -2 -2 -2 -5
people Noise (houses exposed to > 57dB(a) -2 -5 -3 -5 -6 -5 -4 -4 -6
Vibrations (constructed scale) -5
Accidents (km line in unit) -1 0 0 -1 0 -3 0 -2 -5
Accidents (n. dangerous crossings) 0 -1 -6 -3 -2 2
Environm. Landscape (km in charact.landscape) -1 -3 -3 -5 0 0 -5 -5 -6
Landscape structure (constructed scale) 0 1 -1 6 -6 2 3 -3
Nature (important areas affected) -3 3 -4 -3 6 0 -4 -5 -3
Air pollution (constructed scale) -3
Traffic congestion (constructed scale) -5
Cultural heritage (n. sites affected) -2
Socio- Agriculture area losl (ha) -1 -4 -2 -6 -3 -3 -2 -3
economy Economic impacts (constructed scale) 0 0 0 -6 -6 0 -2
Temporary hindrance (years) -5
Planning conslraints (constructed scale) -4 -4 -5
Costs for the unit (Million Guilders) -5

4 The comparison of levels of attractiveness is possible if the 0 and the -1 scores have the same
substantive meaning for all units. While this is a reasonable assumption for the neutral profile (the 0),
this is a rather strong assumption for the worst possible profile (-1).

156
 1 _•......•_.•.....•......... _......•..•. _........_-_•.......... _ .. _ -..••. _.•............ _...•.
1lI0bjections
0.81-::::-:--- : - - - - -- - - - -- - - - -
. Attractiveness
0.6 t -- - ---------Irnm-- - -- - - -

0.4 ~
~
0.2 'm ~
w ~
0 f-+- f-+- :-..- . -+-- f-+- ~ f-+- f-+- f-+-
·0 .2 I- I- - - f-- f-- I- I-

·0 .4 f-- I- I-

·0.6 '--

·0.8t - -- - - - -- - - -- - - - - - - <
· 1 Province M1 M2
.................................. M3 M4 - ...........
_................................ M5 _ ........................
M6 M7 __ _........
MB _

Figure 12. Overall attracliveness score for the eight 1IIunicipalilies and for Ihe province. The number of
objeclions regislered ill these units is shown with positive bars.

6.2.2. Weights
The overall attractiveness depends on the weights associated with each criterion by each
policy unit. Within an additive value model, these weights are trade-offs between impacts
and do not necessarily re fleet a concept of relative importance (Beinat, 1997a; Bana e
Costa and Vansnick, 1997; Keeney, 1992).
Figure 13 shows a recalibration of the original weights so that they can be compared
between units. For each unit, the bars show the improvement in the overall attractiveness
of the railway obtained by:
• (noise) reducing the number of houses exposed to higher levels than 57dB(a) from 25
to 0, or;
• (landscape) shortening the route in valuable landscapes from 1 km to 0 km, or;
• (accidents) shortening the route in the municipality from 1 km to 0 km.
In all situations there is an impact reduction and thus an improvement of the
attractiveness of the impacts. The relative size of this improvement is linked to the
weight 01' noise, landscape and accidents, respectivell.
To compare the trade-offs tor different units, the value model of so me of them had to
be extrapolated to include the profile noise=25; landscape=1; accidents=1. This occurred
for M2 (landscape), M5 (noise), M6 (noise and landscape), M7 (landscape and
accidents). The value functions 01' these units are linearly extrapolated outside their
original range. This is a major simplification, which probably explains the extreme

5 It is dear that this interpretation of weights is meaningless without specifying the quantities which are
compared and the initial profile from which impacts are reduced or increased. In the case of linear value
functions, like in this case, trade-offs depend only on the quantities and not on the starting point.

157
weights of M7. This unit is affected by landscape and accident impacts much lower than
those considered here.
Apart from M7, Figure 13 shows that the weight of noise (25 houses) is the highest
for all units and that the weight of landscape (1 km) is higher than (or equal to) that of
accidents (1 km). The differences between weights are the result of a different subjective
assessment of the same impact, but possibly also of a slightly different interpretation of
the impact itself (cf. Section 6.2.1. above).

__.__ ._._-_._-------------_.----.--.--.--.. _....--. __ ....-----.--. ----------.------.............._._._.........-----.--------,

,.....
IIINoise (25 Buildings) ...
•.•+---- - - - -- - - -- - - - - ----1 1-- - - ;
-Landscape (lkm)

0Accidents (1km)

2 •• + - - - - - - - - - - -- - -- - - - - ,;:rl 1--- - ;

,-. +---- - - -- - - -- - - -- ---1. 1----;

•.• ~J-=--fil----

p Ml M2 M3 M4 M5 M6 M7 MB

Figure 13. Irade-offs between three criteria: noise, landscape and accidents. For each unit, each bar
indicates the improvement in the overall attractivelless obtained by reducillg the noise impact 0/25
houses, or the landscape impact 0/1 km, or the accidents impact 0/1 kill. Noise (25 houses affected) is
taken as a reference ami is assigned the score 0/1.

7. Contlicts
The analysis shows that all units considered in this study rate the projcct as unattractive.
Since the project at the national level was considered as highly attractive (although there
is no comparable value model from where this conclusion can be deduced), there is a
contlict between the national level and the regional and local levels, which would fare
better with the status quo. In the Beluweroute project, other conllicts occurrcd also
between national and local levels of the other region involved (South Holland) and
between municipalities and provinces.
The degree of contlict is the result of the distribution and intensity of impacts, of the
different unattractiveness of the impacts as perceived by the units, and of the weights
assessed by each unit. To reduce contlicts in such a situation, two basic strategies can be
considered:

158
1. To mitigate or compensate impacts so that the impact profile of the project moves
towards the neutraliine.
2. To introduce new elements into the plan, such as additional projects or infrastructure,
so that positive outcomes for locallevels can be enhanced.

In the first case, mitigation and compensations are meant to re du ce the

unattractiveness of the project for the regional and local units. Table 5 shows an
overview of the critical impacts for each unit. The scores shown are the w/
v/(xi) values,
that is the weighted attractiveness scores. A high negative score, like those in bold, is the
result 01' either a bad impact, or of a low attractiveness of the impact, or of a high weight,
or 01' a combination of these reasons. In any case, reducing the impact corresponding to a
high critical score produces a (comparatively) higher reduction 01' conl1icts.

Table 5. Critical scores normalised between -10 and 0 for each unit. Bald numbers indicate weighted
atlractiveness scores lower than the unit's average (** corresponds to attractive impacts).

Concerns Criteria (and impact descriptors) P MI M2 M3 M4 M5 M6 M7 MB

Effects onDemolition (n. buildings demolished) -4 -10 0 -3 -4 -5 -8
people Noise (houses exposed to > 57dB(a» -10 -8 -7 -8 -10 -9 -7 -10 -10
Vibrations (constructed scale) -5
Accidents (km line in unit) -8 0 0 -2 0 -4 0 -4 -9
Accidents (n. dangerous crossings) 0 -1 -8 -6 -5 0
Environm. Landscape (km in charact. landscape) -6 -6 -2 -7 0 0 -8 -6 -10
Landscape structure (constructed scale) 0 ** -2 ** -10 ** ** 0
Nature (important areas affected) -5 ** -8 -6 ** 0 -6 -8 -2
Air pollution (construclcd scale) -7
Traffic congestion (constructed scale) -9
Cultural hcritage (n. sites affected) -3
Socio- Agriculturc area lost (ha) -2 -7 -3 -9 -2 -7 -4 -1
economy Ecooomic impacts (constructed scale) 0 0 0 -6 -8 0 0
Temporary hindrance (years) -6
Planning constraints (constructed -5 -5 -3
scale)
Costs for the unit (Million Guilders) -4

A limit situation would be reached when all negative impacts are negligible for all
units considered here. If the project remains attractive for the national leve~ this would
probably be co me an acceptable solution for all units. There was a suggestion, for
instance, that a wholc section of the railway could be routed through a tunnel, making
the project neutral at regional and locallevels. However, cost and feasibility arguments
made this pro pos al unattractive for the national level and the plan was abandoned.
An alternative to complete mitigation is to introduce positive impacts for the local and
regional levels. An attempl in Ihis sense has been made by the regional level, which
considered the extension of the plan 10 include aseries of ancillary projects, such as a
multimodal terminal 1'or combining rai! and water transport. The revised plan increases
the overall environmental impacts, but it is expected to generate significant economic
bencfits (as many as 8000 jobs). In this new situation, the combined projects

159
(Betuweroute + regional infrastructures) hecome attractive from the Prövince's
perspective.
It is interesting to notice that, even on the hasis of very simple value models, the
effectiveness of rnitigations, compensations and revised plans to reduce conflicts can be
easily assessed. This analysis can be also used to suggest these measures and explore a
wide range of revisions which could support the negotiation phase and the design of
improved plans (see Beinat, 1998).

8. Discussion and conclusions

This paper illustrates an approach for the analysis of the perspectives of spatial actors for
the evaluation of transport policies. The approach combines spatial analysis and decision
analysis and offers a framework for the study of the perceptions of multiple actors. The
example of the Betuweroute shows that objectives, concerns and their importance are
spatially differentiated. The analysis of the impacts of the project also demonstrated that
project acceptance does not simply depend on the magnitude of impact hut also, perhaps
especially, on the evaluation perspective of the spatial actors involved. This can be used
to understand the genesis, nature and severity of spatial contlicts, hut also to isolate
those issues which generate more conflicts.
The positive reactions received from those who experimented with this approach
suggest three main added values. The first is the possibility 01' revealing con11icts earlier,
before they crystallise and jeopardise the interaction hetween actors. The second is the
possibility of narrowing the focus and increasing tbe eITectiveness of tbe searcb for
contlict resolution measures. The third is an increase in the transparency and
communicability of the perspectives and positions of different actors.
There are also a number of critical issues which deserve special attention. An
important one is the delicate balance between the stage at whicb a value funetion model
can be reasonably assessed and the stage at whieh its results would be most useful.
The case study demonstrates that local units want to participate in the planning
proeess from the early stages. Disc10sing contliets at this stage would be most useful, so
tbat poliey measures ean be designed to prevent eontliet as mueh as possible, and to take
into aeeount the needs and expectations of multiple actors. However, this ease also
shows that when there is insufficient evidenee on the positive outcomes of a plan, loeal
units stress tbe negative ones. Tbe laek of preeise evidence is eharacteristie of strategie
evaluations (ef. ThCrivel and Partidario, 1996) at early planning stages. Tbe risk,
therefore, is of transforming early participation into early opposition.
Following a struetured value analysis may inerease the aeeeplability of the evaluation
and participation. However, a value function model tor loeal unils can be assessed when
loeal effeets are sufficiently c1ear. This usually oecurs c10se to the final stages 01' the
design, when modifications to tbe plan are difiicult to implement or require costly re-
designs.

160
 The right balance between the moment at which participation starts, the content of the
decision at thil.t stage, and the impacts which can be calculated at that stage are critical
factors which determine the effectiveness and usefulness of analytical conflict analysis.

Acknowledgements
This research is part of the DTCS project (Spatial decision support for negotiation and
con11ict resolution of environment al and economic effects of transport policies), financed
by the European Commission DG12-D5 (contract ENV4-CT96-0199). The project is
developed by IVM (Amsterdam, NL), CERTE (Canterbury, UK), CESUR (Lisbon, P)
and MIP (Milano, I). The authors would like to thank all partners in DTCS for their
contribution to the development of the contlict analysis methodology. The spatial data
used in this paper are provided by the Province of Gelderland (NL), which is gratefully
acknowledged. Special thanks go to Margaret Jones (RENU) for the impact assessment
and GIS models. The paper expresses only the authors' views.

References
Armstrong, M.P. (1994) Requirements for the development of GIS-based group decision support
systems, Journal of the American Society Jor Information Science, 45(9): 669-677.
Bana e Costa, C.A, Vansnick J.C. (1997) A theoretical framework for measuring attractiveness
by a categorical based evaluation technique (MACBETH), in J. Climaco (Ed.) Multicriteria
Analysis, Springer, Berlin, 15-24.
Beinat, E, M. van Drunen, M. Jones (1998) The Betuweroute railway project: an analysis of
spatial conflicts, in Beinat, E. (Ed.) A metlwdology for policy analysis and spatial conj7icts
Jor transport policies, Institute For Environmental Studies, Vrije Universiteit, Amsterdam.
Beinat, E. (1997a) Value Junctions for environmental management, Kluwer, Dordrecht.
Beinat, E. (1997b, Ed.) Transport po/icies in the European Union: economy, environment and
space, Report 97/15, Institute For Environmental Studies, Vrije Universiteit, Amsterdam.
Beinat, E. (1998, Ed.) A methodology Jor policy analysis and spatial conj1icts Jor transport
po/icies, Institute For Environmental Studies, Vrije Universiteit, Amsterdam.
Bogetoft, P., Pruzan, P. (1991) Planning with Multiple Criteria: Investigation, Communication,
Choice, North-Holland.
Boom, H., M. Metze (1997) S/ag om de Betuwe Route, Uitgeverij Balans.
Bruinsma, F.R., S.A Rienstra, P. Rietveld (1997) Econornic impacts of the construction of a
transport corridor: a multi-level and multi-approach study for the construction of the Al
highway in the Netherlands, Regional Studies, 31(4):391-402.
Burgoyne, J.G. (1994) Stakeholder analysis, in C. Cassell, G. Symon (Eds.) Qualitative methods
ill orgallisatiolla/ research, Sage, London.
Commission of the European Communities (1993) Trans-Europeall lIetworks, Office for Official
Publications of the European Communities, Brussels.
CPB (1993) De macro-economische effecten van de Betuweroute. Werkdocumellt 52, Centraal
Planbureau, Den Haag.
CPB (1995) Economische etIecten van de Betuweroute op basis van recente informatie,
Werkdocument 79, Centraal Planbureau, Den Haag.

161
Densham, P. J. (1991) Spatial decision support systems, in D. Maguire, M.F. Goodchild, D.
Rhind (Eds.) Geographical Information Systems: principles and applications, Wiley, New
York.
Douven, W. (1996) Improving the accessibility of spatial information for environmental
management, PhD thesis, Free University of Amsterdam, Amsterdam, The Netherlands.
Goodwin, P., G. Wright (1991) Decision Analysis for Management Judgement, John Wiley &
Sons, Chichester.
Hermans, L.M.L.H.A. (1995) Commissie Betuweroute, Projectdirectie Betuweroute, Den Haag.
Hersey, P., K.H. Blanchard (1988) Management of organisational behaviour, Prentice-Hall,
NJ.
Huigen, J., P.H.A. Frissen, P.W. Tops (1993) Het project Betuwelijn: spoorlijn of bestuurlijke
co-productie? KUB Report, Tilburg.
Keeney, R.L. (1992) Value-Focused Thinking, Harvard Universily Press, Cambridge.
Maguire, D., M.F. Goodchild, D. Rhind (1991, Eds.) Geograpltical Information Systems:
Principles andApplications, Wiley, New York.
Nagel, S.S., M.K. Mills (1991, Eds.) Systematic analysis in dispute resolution, Quorum, New
York.
NCGIA: National Center for Geographical Information and Analysis (1996)
hUp://www.ncgia.edu/research!iI7/I-17_home.html.
Rosenhead,1. (1989, Ed.) Rational analysis for a problematic world, Wiley, Chichester.
Simon, H.A. (1991) Bounded rationality and organisationallearning, Organisation Science, 2:
125-139.
Susskind, L., J. Cruikshank (1987) Breaking the impasse, Basic Books, New York.
TB (1996) Tracebesluit Betuweroute, SDU uitgeverij, Den Haag.
Teisman, G.R. (1992) Complexe besluit vorming, PhD thesis, Erasmus University, Rotterdam.
Therivel, R, R.M. Partidario (1996) The practice of strategic envirollmental assessment, Earthscan,
London.
Vincke. Ph. (1992), Multicriteria Decision-Aid, Wiley, Chichester.
Wathern, P. (1988) Environmental Impact Assessment, Routledge, London.

162
An algorithm for computing multiple attribute additive
value measurement ranges under a hierarchy of the
criteria: application to farm or rangeland management
decisions

Diana s. Yakowitz l .2 and Mark Weltz l

IUSDA-Agricultural Research Service. Southwest Watershed Research Center
2Department 0/Agricultural and Biosystems Engineering. University 0/Arizona
Tucson. Arizona. USA

Abstract
Adecision tool is described and applied to the problem of evaluating fann or rangeland
management systems with respect to both economic and environmental criteria. The method
quickly computes the possible range of value from the most optimistic to the most pessimistic
(best to worst) for any given hierarchy of the multiple attributes. The method is to be applied
after commensurate attribute values have been determined for each alternative without requiring
one to specify or determine explicit weights on the attributes. The decision tool is particularly
useful for examining alternatives from numerous decision-making viewpoints or by multiple
decision makers. The importance order of the criteria or attributes at any tier in the hierarchy
can be changed and the value range computed again using a simple algorithmic method that
does not require a linear programming sol ver. This solution method makes it easy to determine
the result of modifying priorities in portions of the hierarchical architecture without
recalculating the contributions of unaffected parts. The method is applied to the problem of
determining a possible replacement fann management system for several fields in western
Iowa. Environmental and economical improvement over the current farming system is indicated
by several of the alternative systems. Current projects are underway to develop indices of
rangeland and soil health.

Keywords: decision making, multiple attribute, multiple objective, agriculture,

environmental management.

1. Introduction

This work is prompted by the need for tools that can be easily and quickly understood
and sensibly applied to multiple attribute decision making situations that occur in land
management. The tool presented in this paper, however, is not limited to applications
for land management. It is readily applicable to any decision-making situation that can
be similarly described. Examples among the problems that could be considered include:
determining which among a suite of feasible farm management practices should be

163
preferred on a given field; or determining which field from among several feasible
locations is best for a given management system; ranking rangeland watersheds or other
ecosystems for the purpose of determining where limited funds and time should be
concentrated. Project or portfolio selection is analogous to the latter, while product
comparison or suitability to the former. Alternative facility layouts and transportation
planning are other decision making areas that the method described herein could
benefit.
The development of our decision support tool was encouraged by statements in Dyer
et al. (1992) that simple, understandable, and usable approaches for solving multiple
criteria decision making (MCDM) or multiple attribute utility theory (MAUT) problems
are still needed. The added complications of multiple decision makers as weIl as
concerned andlor involved parties in environmental decision-making, mandated that
much of the subjectivity (single decision maker perspective) involved in existing
MCDMlMAUT methods be eliminated.
We consider a problem that has been formulated as a hierarchical multi-attribute or
multiple criteria decision problem. The development here is related to earlier work of
Salo and Hämäläinen (1992) and Yakowitz et al. (1992, 1993a) that considers the
impact on an additive value function caused by allowing the attribute weights to vary
subject only to an importance ordering of the criteria. These works fall under the
category of partial information or interval weights in MAUT (Fishbum, 1965;
Kirkwood and Sarin, 1985; Claessens et al. 1991; Bana e Costa and Vincke, 1995;
Hazen, 1986 contains a discussion and numerous references).
Most techniques proposed in the literature for assessing weights, either solicit
weights directly, or seek to discern them indirectly, from the decision maker
(Goicoechea et al., 1982; Keeney and Raiffa, 1976; Saaty, 1980 are examples). The
resultant ranking of the alternatives are often extremely sensitive to the importance
order of the attributes, and therefore the weights. The calculations for examining this
sensitivity are straightforward in the case of a simple ordinal priority ranking of all of
the attributes (Yakowitz et al., 1993a). However, the complexity imposed by a
hierarchical architecture of the criteria at first appears to be difficult to overcome in a
simple way. Yet, it is just this type of structure that is needed to define many complex
decision making problems.
Under a hierarchical structure, the decision priorities alone do not necessarily imply
an ordinal ranking of each individual attribute. Therefore, examining the solutions of
the linear programs given in Yakowitz et al. (1993a) and elsewhere, which are
described in the next section, without considering the possible added freedom (or
relaxation of the weights) due to the hierarchical structure, does not provide the full
range of value of the additive value function. The method presented here does. It allows
one to quickly compute the range of values from the best to the worst for a
hierarchically arranged multiple attribute problem under the assumption of an additive
value function. The method allows the decision maker (DM) to quickly assess the most
optimistic and most pessimistic DM viewpoint given multiple importance orders of the

164
attributes at any tier in the hierarchy. We assume that the attribute values for each
alternative are already in common units and common range as determined by some
multiple attribute or scoring method. For example, the Analytic Hierarchy Process
(AHP) of Saaty (1980), or scoring functions such as those of Wymore (1988) or others
could be used to assess the attribute values or convert data to unit common values and
ranges.
Computing the range of value of an additive value function makes it possible for
various stakeholders or interested parties to determine the sensitivity of the ranking of
alternatives to the hierarchical order. Often this order is a compromise between various
objectives. Examining the range of the value function under other possible hierarchical
scenarios, may reveal an alternative choice that is favored by each party and eliminates
the need for the often contentious task in group decision making of discerning the
weights explicitly.
We begin by defining what we mean by a hierarchy. This is followed by abrief
description ofthe algorithm for computing the range ofthe value function. Algorithmic
details are then developed, followed by a detailed description of the calculations needed
for a specific hierarchical structure. The generalization to more complicated
hierarchical structures is readily apparent. This method is then used for evaluating four
alternative farming systems with respect to specific economic and environmental
decision criteria. The concluding section includes a discussion of current projects under
consideration using the evaluation tool developed in this paper.

2. Defining the hierarchy

A generic hierarchical architecture for a multi-attribute decision problem is illustrated

in Figure 1. Note that we have assumed a quite generic structure where branches of the
hierarchy may terminate at different tiers or levels. Dummy elements may, of course, be
added at intermediate levels if equal depth branches are desired. The Major Goal is at
the highest level (Tier 1). This level could be choosing a Sustainable Agricultural
System from a finite number of alternatives for a given farm/ranch or region; or
choosing the best TrajJic Plan from among several alternative plans. The subsequent
levels (Tier 2 through N) contain sub-elements of the parent or previous levels. Thus,
for example in the sustainable agriculture problem, Tier 2 could include environmental,
economic, and social sub-goals. Subsequent levels of the environmental branch could
then include surface water, sub-surface water, species diversity, and soil, followed by
criteria including fertilizer and pesticide impacts, and erosion under the proper
category. Of interest here is the effect of changing the importance order of the elements
ofthe hierarchy on the range ofvalues ofthe additive value function. For example, one
may want to know how alternative farm or range management alternatives would be
ranked if environmental criteria were given a higher priority than economic criteria.
However, the farmer or rancher may want to give a higher priority to the economic
criteria considered. Importance order is often the key issue of contention when there are

165
multiple stakeholders. In all figures, it is assumed that the importance order is from left
to right. That is, for elements in the same tier, that emanate from a common branch, an
element to the left has a higher priority and therefore more "weight" in the decision
making process at that level. No assumption is made regarding the priority relationship
between elements emanating from different branches even ifthey are in the same tier.
As is the case in most multi-attribute solution methods, the goal of the methodology
is to determine the value of an additive value function that can be used to rank the set of
alternatives.

Tier 1

Tier 2

Tier 3

TierN

Figure 1. Generic decision hierarchy.

An additive value function takes the following form:

V(w, v) = I:; WjV j ,

where i ranges over the terminal elements of each branch in the hierarchy; V indicates
the value assigned to the alternative with respect to each terminal element; and the
weights, w, are consistent with the hierarchy and normalized so that they sum to a finite
number (assumed to be I in all calculations given below). We emphasize that V is a
function of both the individual criterion values determined for each alternative and the
weights (unknown) for each attribute or criteria. Since we are primarily concerned with
the effects on the above function caused by changing hierarchical element priorities, we
will assume that for each alternative, v j is fixed for all i. We refer to the above as V,or
with subscript, Jj , when wishing to distinguish between alternatives (in this case the
values, v, will be double subscripted, vij)'
The algorithm for assessing the fuH range from the best to the worst under our

166
assumptions begins at the lowest tier of each branch of the hierarchy. Best and worst
additive values are computed for each element using analytic solutions to two simple
linear programs that maximize and minimize V at the parent element over all weights
consistent with the importance order of the decision elements. These same programs are
used at intermediate elements, substituting the maximum (or minimum) values
previously computed as the values for those elements that have descendent elements,
until the main or first tier is reached. Altering the priority at any level requires redoing
only those calculations that occur after that point to the main or first tier. This fact
makes it easy to examine the effects of changing priorities or decision maker
preferences, which may be especially useful if there are more than one decision maker
or affected party involved.

3. Aigorithmic details

3.1. Computing best and worst subvalues /or the lowest tier 0/ each
branch

Given the priority order of the criteria at the lowest tier of each branch, best and worst
additive values can be found without requiring the decision maker to set specific
weights for each ofthe criteria (see Claessens et al., 1991; Salo and Hämäläinen, 1992;
or Yakowitz et al., 1992, 1993a). Referring to the notated branch at Tier N of Figure 1,
it is assumed that if i < j then criterion i has a higher priority than criterion j (i.e.
criterion 1 has higher priority than criterion 2 and so forth). Since there are m criteria,
the importance order suggests that we should require that the weights, w;, i = 1, m, have
the following relation:

Therefore, given the importance order and the criteria values for alternative j, the
best (worst) composite score that alternative j can achieve is determined by solving the
following linear programs (LPs):

Best (Worst) Additive Value:

max(min)w V j = L;:/W;V;j

s.t. L;:/W; = 1
w/~ Wl ~... ~ Wm ~ o.

The best additive value is found by maximizing the objective function while the
worst additive value is found by minimizing the objective function. The first constraint
is a normalizing constraint. The second, fixes the importance order and restriets the

167
weights to be positive. The above linear programs can be easily solved analytically. For
k=l, ... ,m, let

Skj = llk 1:;; J•••• k Vij' (1)

Then, the best or maximum additive value, Max V, for alternativej is given by :

Max ~= maxk {SJ. (2)

The worst or minimum additive value, Min V. for alternative j is given by :

Min ~= mink {SJ. (3)

For proofs see Yakowitz et al. (1993a) or Fishburn (1965). The above is equivalent
to evaluating the objective function at the extreme points of the constraint set, which
occur at the points {I ,O,O ... ,O}, {I 12, 1/2,0, ... ,0}, {1/3, 1/3, 1/3,0, ... ,0} , ... ,
{l1N,11N, ... ,11N}, and selecting the vectors that produce the min and max. All other
feasible weight vectors can, of course, be written as linear combinations of these N
vectors. This result is also known as Paelinck's Theorem and is proved and described in
Claessens et al. (1991).
In the case of equal importance of some criteria, there are strict equalities in the
importance order constraint set, i.e. wj=wj + 1 for j in a subset, i, of the integers 1
through m. If we define K = { 1, 2, '" , m} \ j, then the above formulas far Max and Min
Vapply if k is restricted such that kE K.
Ca1culations for cases in which it is desired to specify the level of preference
between criteria (for example, criterion 1 is to have a weight at least twice that of
criterion 2) require a modified definition of (1). Suppose that one supplies constants Ci ~
1, i=2, ... ,m, that imply the following relationships:

Then ifwe let c J = 1 the required modification to (1) is as folIows':

___l""-__I ~=I Vij n ~=i Cr (4)

I 7=1 n ~=
.
I Cp Ci
CI

For each alternative, the solutions (2) and (3), using (1) or (4), determine the
maximum and minimum additive value possible for any combination of weights that are

I This fonnula corrects a typographical error on page 175 of Yakowitz, Lane and Szidarovszky (1993a).

168
consistent with the importance order of the criterialattributes. Having these two
objective values available immediately alerts the DM to the sensitivity of each
alternative to the weights possible with the current importance order of the criteria.
These values can be displayed graphically (illustrated later) in the form of side by side
bar graphs with the best value for each alternative at the top of each bar and the worst
value at the bottom.
An alternative that exhibits little difference between the best and the worst values
indicates that this alternative is relatively insensitive to the weights consistent with the
given importance order. Additionally, if the worst value of one alternative is greater
than the best value of another alternative, then clearly, that alternative dominates the
other alternative. These alternatives are a subset of those that dominate in additive value
with respect to a given importance order and are said to strongly (Yakowitz et al.,
1993a) or absolutely (Salo and Hämäläinen, 1992) dominate. With respect to the given
importance order and in the absence of strong or absolute dominance, one can
determine dominance in additive value of Alternative j by Alternative k if and only if
Si/c ;?Sij for all i=1,2, ... ,m, and Sjk > Sij for at least one i (see Yakowitz et al., 1993a,
for a discussion and theorems).

3.2. Computing the best and the worst values for a multi-level hierarchy

One way to account for a hierarchy of the criteria and still provide the range from the
best to the worst composite scores is to include additional constraints in the LPs given
above. For example, suppose we have a three tier hierarchy, and each element i in Tier
2 is composed of Ij sub-criteria in Tier 3, the terminating level. Let Vj,k, j and wj,k'

k= 1, ... ,/j indicate the values (scores) for alternative j, and sub-weights (unspecified),
respectively, associated with sub-criteria k of criteria i. Then, the following two
constraints for each i are added to bestlworst LPs to account for this hierarchy:

Wi = Wi,l + Wi,2 + ... + Wi.I,

Wi,l ~ Wi,2 ~ ... ~ Wi,/, ~ O.

The objective functions of the bestlworst LPs for alternative j are then replaced by:

Again, one can obtain the range from maximum to minimum without the need to
specify weights or sub-weights. Linear modifications due to more general hierarchical
considerations are easily made in this manner. Note that the above formulation makes
no assumptions regarding the ordinal ranking of attributes on different branches, a
desirable feature in cases when it is not possible or undesirable to prioritize attributes
across branches. An explicit linear program for computing the maximum and minimum

169
V for any hierarchy can be formulated and solved. The notation needed to indicate each
level ofthe hierarchy, however, becomes very cumbersome. Solving the resulting linear
programs explicitly for each alternative is not necessary, however, since a simple
algorithm that considers each portion of the hierarchy in an optimal manner is
developed here. The algorithm, which could also be described in dynamic programming
(Sniedovich, 1992) terms, is more suitable for examining the effects of changing
priorities than a linear programming model. Calculating min and max V is an intuitively
simple procedure when performed from the lowest tier up. The contribution of each
level is solved optimally in an iterative manner yielding the optimal objective values
(min or max) once the top tier is reached. To illustrate this fact, the solution procedure
will be described for the four tier decision hierarchy of Figure 2 and then applied to the
problem of considering an alternative farming system on a small watershed (field size)
in western Iowa.

3.3. Algorithm for computing the range of values under a given hierarchy

The following procedure is described to obtain the range from the best to the worst of
the additive value function under the hierarchy illustrated in Figure 2. The procedure
for any other hierarchical variation is handled in a similar manner and will become
transparent.

Tier 1

Tier 2
v(2.2)

Tier 3

max (min) v(3.1.l) v(3.1.2) v(3.1.3) v(3.1.4) max (min) v(3.3.l) v(3.3.2)

Tier 4
v(4.1.1.1) v(4.l.l.2) v(4.1.1.m) v( 4.3.1.\) v( 4.3.1.2) v(4.3.l.n)

Figure 2. Decision hierarchy Jor algorithmic explanation.

170
 For each alternative under consideration, we assume that the associated value for the
terminal elements has been determined by some means. Thus, for Figure 2, the values
indicated by v(2.2), v(3.l.2), v(3.1.3), v(3.1.4), v(3.3.2). v(4.l. 1. 1) ,.... v(4. 1. 1.m),
v(4.3. 1. 1) ,... , v(4.3.1.n), are known for each alternative and are in a common range. All
indices are with respect to the hierarchy of Figure 2, which indicates the inputs and
ca1culations required. Ca1culations start at the lowest Tier in the hierarchy. In this case,
Tier 4.

Tier 4.
Compute for each alternative j,
8(4. 1. 1)kj = 11k.E;= I.... k vi4.1.l.i), k=l, ... ,m,
and 8(4.3.1)kj = 11k.E; = I.... k vi4.3.1.i}, k=l, ... ,no

Then according to Eq. (1) and (2),

max (min) vi3.1.1) = max (minJt {S(4.J.J)t}, and
max (min) vi3.3.1) = max (min)k {S(4.3.J)t}.

Tier 3.
Compute the following for each alternative j:
8mox(3.1)kj= 11k.E;=I .... kvi3.l.i), fork=l, ... ,4,
with vi3.1.1)= max (vi3. 1. 1)), and
Smin(3.1)kj=1lkI;=I .... k vi3.l.i), fork=1, ... ,4,
with vi3.1.1)= min (vi3.l.l)).
Smax(3.3)kj = 11k.E; = I.... k v13.3.i), k=1,2,
with vi3.3.1)= max (vi3.3.1)), and
Smil3.3)kj = 11k I; = I.... k v13.3.i), k=1,2,
with v/3.3.1)= min {v/3.3.1}}.

Then:
max (min) vi2.1) = max (min)k {Smox(m;nl3.J)t}, and
max (min) vi2.3) = max (minJt {S(3.3)t}.

Tier 2.
Compute the following for each alternative j:
Smax(2)kj = 11k.E; = I.... k vi2.i), for k=1, ... ,3,
with vi2.1)= max (vi2.1)), and
vi2.3)= max (vi2.3)), and
Smil2)kj = 11k I; = I.... k vi2.i), for k=l, ... ,3,
vi2.1)= min (vi2.1)), and
vi2.3)= min (vi2.3)).

Then:
Best (Worst) ~ = max (min)k {Smax(minl2)t}.

171
 A bar graph indicating the range from the Best V, the top of each bar, to the Worst V,
the bottom of each bar, for each alternative would aid the decision maker by indicating
possible domination and the sensitivity of each alternative to the priorities in the
hierarchy. As discussed and described in Yakowitz et al. (1992, 1993a), the alternatives
can be ranked based on the average of the Best and the Worst ~. Clearly from the end
of Section 3.1 above, this ranking preserves all dominance re1ationships and provides a
means of ranking non-dominated alternatives that takes into account all possible DM
perspectives with respect to the current importance ordering. It is weil known that for
non-dominated alternatives it is possible to have rank reversal even for weight sets that
are consistent with the hierarchy importance orders. Tberefore, ranking based on the
midpoint can be viewed as less risky than ranking based on a single set of weights since
the midpoint ranking includes information from the most pessimistic as weil as most
optimistic viewpoints considering the entire decision hierarchy structure.
Changing an importance ordering in any tier in the hierarchy, requires only
recalculating appropriate max and min values in the tiers that appear above the point at
which the change is made. For example, suppose one wishes to consider the scenario in
which the elements previously ordered in Tier 2 of Figure 2 are reversed. Tben, only
those calculations indicated under Tier 2 given above, need to be computed again.
Other scenarios that reflect the differing priorities of interested parties or multiple
decision makers could be quickly examined. If one or two alternatives stand out as
doing weH under multiple decision scenarios, one would have a defensible basis for
supporting these alternatives and avoiding unnecessary argument. In fact, if two
alternatives are non-dominating and changing priorities still produces midpoint values
that prefer one alternative over the other, then ranking based on the midpoint values is
less risky in the sense that it preserves the ranking of the majority of decision makers
that agree with one or the other of the decision hierarchy structures examined.

4. Evaluating alternative farming systems for a field in

western Iowa, USA

Under the management of National Soil Tilth Laboratory of the US Department of

Agriculture - Agricultural Research Service(USDA-ARS) is the Deep Loess Research
Station located near Treynor, Iowa. Watershed #1 is a 32 hectare field that has been
planted annually with corn (Zea mays L.) on the contour since research on the
watershed was initiated in 1964. Tbe tillage practice in current use is known as deep
disking. Tbe current system was used to determine baseline values of each of the
attributes. To improve farm net return as weH as concern for nitrate concentrations in
the ground water, a corn and soybean (Glycine max L.) crop rotation is being
considered along with four different tillage practices: deep disking (DD), chise1 plow
(CP), ridge tiH (RT), and no till (NT). As an indication of the benefit obtained by
changing the cropping system alone, an alternative that uses the no till practice but

172
continues the planting of only corn (NT-c) is also included. The alternatives are
compared based on predicted amounts of sediment, nutrients, and pesticides leaving the
field, which have an impact on surface and ground water quality, as weil as the farmer's
predicted net returns for each of the five alternative management systems. Figure 3
illustrates the structure ofthe decision hierarchy and the attributes used to evaluate each
management system.
Table 1 contains the values for each attribute for each alternative determined using a
USDA-ARS decision support system (see Yakowitz et al., 1992, 1993b). Each
management system was simulated for 24 years and the average annual values of the
predicted amounts of the listed attributes were scored compared to the predicted
amounts for the existing management system. All scores range from 0 to 1 for all
attributes. A score of better than 0.5 indicates that that management system improved
with respect to that attribute over the current system, while a score of less that 0.5
indicates a less desirable result.

Table 1. Attributes and values or scores for four simulated farm management systems on a field in
western lowa.
Alt. #1 DD Alt. #2 RT Alt. #3 CP Alt. #4 NT Alt. #5 NT-c
net income 0.67 0.82 0.71 0.82 0.66
nitrates (s) 0.51 0.98 0.91 0.99 1.00
nitrates (ss) 0.74 0.66 0.73 0.52 0.30
phosphorus 0.38 0.93 0.81 0.95 0.96
sediment 0.34 0.97 0.91 0.99 1.00
pesticide A 0.92 0.81 0.97 0.85 0.44
pesticide B 0.94 0.79 0.98 0.86 0.44

Applying the algorithmic steps detailed above, the results for the importance orders
(left to right on common branches) indicated in Figure 3 are illustrated in the Figure 4a
bar graph. When economic improvement has a higher priority than environmental
improvement, a11 alternatives except #5 do better than the baseline or current
management system (indicated at the 0.5 point), even at their worst scores. Alternatives
#2 and #4 attain the highest best scores. Ranking based on the midpoint of the bars
(average of best and worst) puts Alternative #2 on top, Alternatives #3 and #4 tied in
second place, fo11owed by Alternative #1 and then #5. This order of the alternatives
never ranks an alternative above one that it is dominated by (Yakowitz et al. 1993a).
Figure 4b indicates the results when the importance order at the first branch (tier 2) is
reversed. In this case, environmental improvement is given a higher priority than
economic improvement.

173
 Figure 3. Decision hierarchical structure for farm management system selection.

a. Economic over b. Environmental over

Environmental Improvement. Economic Improvement.

--11
best

·-11
11
Total score Total score
I 1
0.9 0.9
0.8
worst 0.8
0.7 0.7

...............................• ......~~~~~~~~ ..... ~;~.

0.6
...............................
0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
o L-_~_.J-_.J...._-'-_...I O ....._..L..._......_--"--'--.........I
#IDD #2RT #3CP #4NT #5NT-c #IDD #2RT #3CP #4NT #5NT-c

Alternatives Alternatives

Figure 4. Value range from best to worst for farm management system problem. Results when economic
improvement is preferred over environmental improvement (a.) and results when environmental
improvement is preferred over economic improvement (b.).

174
 While all ofthe first four alternatives do at least as weIl as the baseline, Alternative
#3 could be preferred over the others since it has both a high best and worst score. The
range of values possible for Alternative #3 indicates that this alternative is the least
sensitive to the weights. In other words, this alternative is affected the least by any
particular decision maker's weight preferences given the importance order structure of
the hierarchy of Figure 3. Alternatives #2 and #3 perform weIl with respect to both
goals. In fact, the farmer and the community would benefit by adopting any one of the
alternatives considered in this example study and would benefit the most with respect to
the given criteria by adopting either Alternative #2, RT, or #3, CP. Other decision
hierarchy scenarios that take into account both on and off farm impacts, long and short
term risks, additional criteria, and other alternatives could be examined to provide
further confidence in a choice.

5. Other applications, extensions and concluding remarks

The method described herein explicitly calculates the full range of values possible for
an additive value function subject to the priorities of a hierarchical decision structure.
The procedure involves the solutions to two simple linear programs and a solution
method not requiring an LP solver was presented. This procedure also minimizes the
number of calculations needed to examine the effects of changes to the hierarchical
structure and can lend insight into the evaluation process. The methodology has the
advantage of including all possible decision maker points of view, from the most
pessimistic to the most optimistic, within the structure ofthe hierarchy.
Some remarks with respect to the bar charts and evidence of domination or non-
domination need to be addressed. First, ranking based on the midpoint is a basis for a
decision but other factors should be considered. For example, it is clear that if the
results produce a lengthy sized bar (additive value range) for an alternative, that
alternative is sensitive to weights consistent with that hierarchy. Therefore an
alternative that has a high and narrow range from best to worst does weIl and is less
risky with respect to the feasible weights for the given hierarchy. If the bar for one
alternative contains the range of the bar for another alternative, then clearly those
alternatives are non-dominating and rank reversal with respect to specific weight
vectors can occur. It is also possible that ifthe bars oftwo alternatives overlap, one may
or may not be in a situation of domination, and rank reversal is possible in this case too.
However, these are exactly the alternatives that one requires a method for ranking. In
this latter case, overlapping bars, one can determine additive value function domination
under a hierarchy only by looking at all of the extreme point solutions, which is not a
trivial problem in a deep hierarchy. The method presented here for determining the full
range an then ranking captures all possible decision maker opinions on the weights and
uses all of this information by ranking based on the midpoint. When there is not a clear
preference for one alternative over another (overlapping situation) one can observe
what happens to the results when one considers a change in the hierarchy at an upper

175
 level (in the example presented in the last section, this meant switching the order of
importance of the economic and environmental sub-categories). If this switch still
supports one alternative over another based on the midpoint, then it is further support
for this ranking since now, not only does this ranking capture the majority of possible
decision makers that agree with the first hierarchy but also those that agree with the
second. The ranking is in this sense less risky.
The method proposed above is well suited for determining a quality index for
various purposes. For example, currently under development is the determination of a
decision hierarchy that will properly define an Index for Rangeland Health. This index
could then be used as adecision aid in determining, for example, where remedial action
is needed. or to which areas limited funding might be allotted or which land areas are
best suited for a given purpose or for purchase or exchange. A hierarchical structure
that includes Biodiversity Conservation, Soil Stability and Watershed Function, and
Animal Production at the second tier level followed by those elements that define or
contribute to these topics is being proposed. This work is being conducted by
cooperative research between scientists with the Agricultural Research Service (ARS)
and the Natural Resources Conservation Service (NRCS) of the United States
Department of Agriculture (USDA). In addition, the methodology described in this
paper has been proposed for use in the development of a SoU Quality Index by the
above agencies for all regions of the Uni ted States.
Index ranges can be compared and contrasted in many different ways. For example,
given a specific hierarchy structure, the index range for a number of rangeland sites can
be viewed side by side. Alternatively, a single site's index ranges for multiple decision
hierarchy structures or scenarios could be viewed in a single graph. Temporal
comparison of the index range for a single site given a single hierarchical structure may
also be of interest for long term planning.
As illustrated in the farming system problem above, the method can be a valuable aid
to decision makers especially in the case of multiple decision makers or stakeholders.
The ability of the method to take into account other viewpoints and to examine the
impact of many scenarios on the ranking of the alternatives, could make it a strong
negotiation tool between conflicting parties.
We suggest that the procedure introduced here could also be a valuable tool in the
analysis of hierarchies with more than two levels which according to final remarks in
Lootsma (1996) has several open research issues. The effect of changes to the structure
such as splitting criteria on the additive value measurement range is easily examined for
any permutation, addition or deletion to the existing structure. Furthermore, this
information is available at each branching point in the hierarchy for each alternative and
can be used to determine the impact of any changes. A generic spreadsheet macro
program has been developed to apply the method.

176
References
ßana e Costa, C.A.,PH. Vincke, (1995), Measuring credibility of compensatory preference
statements when trade-offs are interval determined. Theory and Decision, 39(2): 127-155.
Claessens, M.N.AJ., F.A. Lootsma, FJ. Vogt, (1991), An elementary proof of Paelinck's
theorem on the convex huH of ranked criterion weights. European Journal of Operational
Research 52: 255-258.
Dyer, 1.S., P.c. Fishburn, R.E. Steuer, J.W. WaHenius, S. Zionts, (1992), Multiple criteria
decision making, multi-attribute utility theory: the next ten years. Management Science
38(5):645-654.
Fishburn, P.C. (1965), Analysis of decisions with incomplete knowledge of probabilities.
Operations Research 13 :217-237.
Goicoechea. A., D. Hansen, L. Duckstein, (1982), Multiobjective Decision Analysis with
Engineering and Business Applications. John Wiley and Sons, New York.
Hazen, G.ß. (1986), Partial information, dominance, and potential optimality in multiattribute
utility theory. Operations Research 34(2):296-310.
Keeney, R.L., H. Raiffa, (1976), Decisions with Multiple Objectives: Preferences and Value
Trade-offs. John Wiley and Sons, New York.
Kirkwood. C. W., R.K. Sarin, (1985), Ranking with partial information: a method and an
application. Operations Research 33(1):38-48.
Lootsma, F.A. (1996), A model for the relative importance ofthe criteria in the Multiplicative
AHP and SMART. European Journal ofOperational Research 94: 467-476.
Saaty, T.L. (1980), The Analytic Hierarchy Process. McGraw Hili, New York.
Salo, A.A., R.P. Hämäläinen, (1992), Preference Assessment by Imprecise Ratio Statements.
Operations Research 40(6): lOS 3- 1061.
Sniedovich, M. (1992), Dynamic Programming. Marcel Dekker, New York, NY.
Wymore, A. W. (1988), Structuring System Design Decisions, Proceedings ofthe International
Conference on Systems Science and Engineering, ßeijing, China /C. Weimen, ed. July
25-28, International Academic Publishers, 704-709.
Yakowitz, D. S, L. J. Lane, J. J. Stone, P. Heilman, R. K. Reddy, B. Imam, (1992), Evaluation
of Land Management Effects on Water Quality Using Multiobjective Analysis within a
Decision Support System, Proceedings of the A WRA First International Conference on
Ground Water Ecology. Miami. FLI April. 265-374.
Yakowitz, D.S., LJ. Lane, F. Szidarovszky, (1993a), Multi-attribute decision making:
dominance with respect to an importance order of the attributes. Applied Mathematics and
Computation 54: 167-181.
Yakowitz, D.S., lJ. Stone, LJ. Lane, P. Heilman, J. Masterson, J. Abolt, B. Imam, (1993b), A
decision support system for evaluating the effects of alternative farm management practices
on water quality and economics. Water Science and Technology, 28(3-5): 47-54.

177
Using decision panels to evaluate hydropower
development projects

Fred Wenstop and Arne J. Carlsen

Norwegian School 0/ Management
Sandvika. Norway

Abstract
The construction of hydropower plants in Norway is notoriously controversial. They impact
rivers, waterfalls and lakes and affect activities such as hunting, angling, skiing and tourism etc.
The local community is usually divided on the issues. People appreciate the income from sales
of electricity and related taxes, but resent the negative environmental impacts. Usually different
stakeholders take quite different stances on the prospects, making the decision process thorny.
In addition, central authorities and national organizations also have their say, adding to the
complexities of the decision process. The paper describes a controversial case where the object
was to recommend whether or not to carry out a planned extension of an existing hydroelectric
plant in the scenic Sauda valley on the west coast. A traditional cost-benefit analysis had
already been carried out. To complete the picture, we employed decision panels in an extended
cost-benefit analysis (ECBA) to identify a range of willingness-to-pay (WTP) for non-market
environmental goods. The WTPs were supposed to be representative of important stakeholders
in the controversy. They were inferred by constructing multicriteria utility functions for the
panels. We also hoped that the rationality of value-focused multicriteria thinking would help
mitigate controversies among the stakeholders. Three parallel and independent decision panels
were created. The participants ineluded persons from the Ministry of Environment, the
Norwegian Water and Electricity Board, the local government, the regional environmental
authority and the development agency. The sessions with the decision panels were computer
interactive. Six criteria were selected from a list to represent the most crucial environmentally
related concerns: submerged areas, river stretches with reduced discharges, new roads,
transmission lines, affected cultural sites, reduced fishing and hunting activities. In addition, the
costs of development was used as adecision criterion. The panels made trade-off judgments of
hypothetical situations described by criteria pairs. Their responses were given on an interval
scale and analyzed with regression analysis. The panels performed the trade-off analysis
diligently and with high self-confidence. The implicit WTPs were computed and used as a
process feedback to the participants. The process revealed systematic preference differences
both within and across the panels, the largest one being ofthe order of 10. The differences were
consistent with known stakeholder interests. The WTPs of the local panel were e10se to the
WTPs for the local population which was estimated by contingent valuation in aseparate study.
This indicates a degree of consistency between the decision analysis results and the sentiments
of the population. The ECBA showed that, according to the preferences of the local panel, the
extension project should be built. The preferences of the two other panels, however, pointed the
other way. The analysis, therefore, served to uncover and explain the roots of the existing
controversy. Another conelusion was that the methodology of multicriteria decision making
(MCDM) lends itself as a natural tool to systematize and valuate the extern al (non-market)

179
effects of hydropower development projects. In this way, the MCDM process can be a useful
instrument in extended cost-benefit analyses (ECBA).

Keywords: cost-benefit analysis, decision panels, willingness to pay, multicriteria

analysis.

1. Introduction and background

Norway is abundant with mountains and waterfalls. ConsequentIy, nearly 100 % ofthe
electricity is produced by hydropower. The first hydropower station was built in 1886
only two years after the first station in the world was built in Wisconsin, USo The
country therefore has a long tradition of hydropower development. The total average
production capacity of the network is at present 110 TWh/year and another 30
TWh/year ofhydropower may be developed in the future ifthere is political willingness
to do so (I TWh = 109 KWh). But exploitation of the remaining rivers is becoming
increasingly controversial because of environmental impacts on wildlife, fish stocks,
outdoor recreation, drinking water supply, and concerns about conservation of pristine
nature.
In the early period of hydropower development, extensive hydropower projects were
built without much protest. In the last few decades, however, there has been mounting
opposition to further development. A climax occurred in connection with the building
of the Alta power station in northern Finnmark, when demonstrators tried to stop the
workers building the station and 600 police officers were subsequently called in to
remove them. This is unusual behavior for normally law-abiding Norwegians, and the
incident prompted the Norwegian Parliament to ask for the development of a master
plan for the remainiQg hydropower sources. The mandate was to rank the hydropower
stations "so that the economically sounder and less controversial projects can be
developed before the more expensive and more controversial ones". The ranking took
into account 11 environmental dimensions. To weight the dimensions was at that time
considered politically too controversial, and the official policy was ''that they were all
equally important". A study by Carlsen et al. (1993) ofthe Parliament's ranking ofthe
532 hydropower projects included in the Master Plan, however, shows that the implicit
weights were in fact very different. Moreover, the external total environmental cost for
the group of the highest ranking projects was approximately 30% of the production
costs.

1.1. Hydropower production and land-use

Hydropower development has multifaceted impacts on land use. Water reservoirs are
most obvious. They often cover extensive land areas with water bodies with changing
levels through the seasons, leaving ugly rims when the water is tapped low and the ice

180
unsafe in winter, preventing normal crossing. But hydropower also impacts many other
forms of land use. New roads are usually built in the construction phase with negative
consequences for recreational interests and landscape conservation, yet nevertheless
they do facilitate transportation. Frozen lakes are sometimes important as roads in the
winter, so that vehicles can reach places that are inaccessible in the summer. Unsafe
reservoirs hinder transportation. They mayaiso interfere with wildlife migration routes
and hunting. In northern NOJ;way, especially, there have been intense, heated conflicts
with Lapps when hydropower development was threatening the traditional migration
routes oftheir reindeer.
Conflicts with agricultural interests are common when arable soil or forests are
submerged. On the other hand, darns may protect against floods, so that more land can
brought under the plow. Rivers are commonly diverted to collect as much water as
possible for the power plant, making old riverbeds and waterfalls dry and lifeless. In
addition to the obvious esthetic consequences, traditional river fishing of trout and
salmon is hurt. The river water mayaiso become more polluted because discharges
become less diluted. This again can make the supply of drinking water more precarious.
Hydropower development is notoriously loathed by tourists who do not enjoy seeing
the landscape and historical sites distorted by power plants, rock fills, power lines,
reservoirs and the like. On this background, the Ministry of Environment has identified
11 dimensions of possible conflict with other land-use interests: transportation, hunting,
fishing, agriculture, reindeer herding, water supply, water pollution, flooding, nature
protection, recreation, and historical sites (Wenst0p and Carlsen, 1988).

1.2. Decision-making processes in hydropower development

Because of the increasing awareness of land-use conflicts, new projects are subject to
careful scrutiny by the authorities before a development pennission can be granted. In
the early days of hydropower development, an application was written on just a few
pages. Today, an application usually requires a meter or more worth of documents.
Even small and relatively uncontroversial projects may take years to process and
evaluate in the decision-making system. The most important legal framework is a Water
Regulation Law of 1917 and a Water Resources Law of 1940 which is now under
revision. In addition, there is the recent Plan and Building Law. This law is especially
important in connection with the consequence analysis of the project, and stipulates that
the information provided with the project application be decision orientated.
An application for pennission to build a hydropower station is addressed to the
Norwegian Water and Energy Administration (NVE). However, it is the Government
which grants the pennission in most cases, whereas large controversial projects are
decided upon by the Parliament. An application is supposed to consist of a technical
part and apart which describes the positive and negative external effects of the project.
Based on this, NVE is supposed to perform a holistic evaluation before its
recommendation subsequently goes to the Government. There are, however, no written

181
guidelines as to how the holistic evaluation should be carried out. The result is that
more or less ad hoc procedures tend to be used. More systematic and consistent
evaluation processes are therefore called for, and the methodology of multi criteria
decision making (MCDM) lends itself as a natural tool to systematize and valuate the
external (non-market) effects ofthe project. In this way, the MCDM process is used as
an instrument in an extended cost-benefit analysis (ECBA) to provide a holistic
evaluation ofhydropower development projects.

2. Methodological approach

2.1. Comprehensive weighting through constructive processes

Hydropower development has environmental and other effects that concern many
stakeholders. It is therefore important for the holistic evaluation process to make it
possible for most stakeholder groups to have a say. In his book on environmental
philosophy, Attfield (1994) calls in such situations for Comprehensive Weighing, a
process whereby all relevant impact domains are valued so that the weights (or prices)
are open to public discussions, thereby facilitating communication and debate about the
proper weights to be used in a given decision situation. One way to do this, is to
conduct Contingent Valuation (CV) surveys where a large number of people, typically
about 1000, are interviewed about their willingness to pay (WTP) to avoid various
environmental impacts. The aim is to estimate the total WTP of the population so that it
can be compared to the economic advantages of the project. The method has been
described in reviews by Cumming et al. (1986) and Mitchell and Carson (1989). The
approach has the merit ofbringing forth "objective" values that are presumed to exist in
the population. If we believe in democratic principles that stipulate that decisions
should be made in accordance with the values held by the people, this is exactly the
kind of infonnation needed. Cowen (1993) refers to this as the preference sovereignty
norm. He wams, however, that the nonn is incoherent and that it "does not even have a
positive marginal product in judging welfare or making social decisions". His major
concern is that the population does not really know what is best for them or for future
generations. Other authors have discovered more pragmatic, but still troubling problems
with the CV method. Brown and Slovic (1988) find it susceptible to influence from
cognitive and contextual biases. Kahneman and Knetsch (1992) have argued that CV
responses denote moral sentiments rather than economic values. In a penetrating paper,
Gregory et al. (1993) state that the major problem with CV is that preferences for
environmental goods do not exist apriori in the minds of common citizens. Attempts to
measure something that does not exist will lead to arbitrary biases and random effects.
To increase validity, they advocate that preferences should be constructed through
suitable processes. They examine techniques derived from multi attribute decision
analysis and recommend that such methods be used. Several applications have been

182
made in environmental and energy poliey. Ralph Keeney is an originator of the method
(Keeney and Wood, 1977; Keeney et al., 1986, 1987). The first major applieation in
Norway was earried out to identify the optimal level preparedness against oil spill
pollution in eoastal waters (Wenstf21p 1983).
We do not think that the arguments above are strong enough to reeommend the
termination of CV praetiee. On the eontrary, we argue that an estimate of the values of
the population is certainly an important piece of information for the decision makers.
Therefore, the CV methods should rather be improved to lessen the impact of biases. In
a political decision making proeess, however, it is obviously useful to have information
about the values of the most important stakeholders as weIl, especially when such data
can be brought forth with more accuracy (and less costs) than WTP estimates of the
population. An added benefit is derived if the partieipation of important stakeholders in
a rational and holistic ECBA process facilitates the real decision-making process.
We therefore reeommend that both CV and stakeholder MCDM is carried out as part
of the ECBA of hydroelectric development projects. Both approaehes produce WTPs
that are relevant to the decision making process, although they need not be estimates of
the same parameters. CV estimates the total WTP of the population, while MCDM
produces WTPs that are representative of various stakeholder groups.

2.2. Decision panels

The task of adecision panel (DP) is to implement comprehensive weighting through a

constructive process. The main purpose is to construct a preference model for the
participants which allows for their WTPs to be inferred. But there is also another
purpose whieh should not be forgotten; the weighting proeess as an educational
experience for participants in the real decision process. By inviting important
stakeholders to participate in the panels, the aetual decision making process may be
facilitated. The multi-criteria strueturing of the problem and the discussions that usually
take place in a DP encourages the participants to adopt a holistic attitude where
different values are represented in a rational way as different criteria weights.
In a eonstruetive process, weights may be identified by many means. See Beinat
(1997) for an overview. We have found pairwise trade-offs of criteria particularly
useful for weight elicitation. It was therefore used here, as described in Chapter 3. The
initial weight estimate is in turn used to tentatively rank different policies, giving the
panel important feedback as to their soundness. The panels then reassess the weights,
until they feel there is consistency between the relative weights and the consequential
ranking of the alternatives. But there is still one more important feedback instrument.
While the weights belong to the deeision problem and express how important the
criteria are in the decision context, WTPs have a context independent meaning and can
be communicated to the outside world. If costs are one of the decision criteria, the
WTPs implied in the weights can be computed from the relative weights. The panels
thereby also get information in monetary terms about the consequences of their

183
weighting, which provides a new opportunity to review them. They finally arrive at a
consistent set of rankings, and thus a "corrected" set of weights attributed to the
different criteria.
In a CV survey, the respondents are asked how much they themselves are willing to
pay for environmental goods. To answer, they have to take their own budgetary
limitation into account. The WTP of the society is then found by statistical inference
and aggregation over all individuals. With decision panels, this has to be done in a
different way. The personal financial situation of the participants is an artifact that
should not influence the results. Therefore, the panelists are told not to let their private
pecuniary situation influence their valuations. Instead, they are required to judge
direct1y what the society should pay for environmental goods, for instance through
increased taxes.

2.3. Roles 0/ the panelists and composition 0/panels

Most people play several roles in their professional life, and sometimes the roles are
conflicting. People therefore get used to changing hats and acting according to the hat
they wear at the moment. Which hat to wear is an important issue with regard to
decision panels as weil, where the mIes of the game should be made clear from the
beginning. It is important for the panel to have a good understanding of the role it is
supposed to play in the valuation process. A DP member may in principle play three
alternative roles, the role of a politician, a stakeholder, or that of a citizen.
In the role of a politician, the panelist is requested to try and interpret and represent
the wishes of the population as a whole: "What does the population at large think the
society should pay for the environment?" This is not necessarily easy, and real
politicians would probably be best at it. But, on the other hand, politicians are usually
not willing to sit in a DP, since it may limit their freedom of choice Iater. We do not
know of any application of this sort, but it might certainly be interesting if suitable
panelists could be found.
In the stakeholder role, the panelists represent their organization and interprets and
express its particular viewpoints on the value issues, e.g. an environmentalist
organization: "What does my organization think the society should pay for the
environment?" This approach would probably create a range of weights that would
throw usefullight on the decision problem and compensation issues. The main danger is
that it allows for strategie behavior which would distort the results. We know of no
applications to environmental and energy problems.
In the citizen role, the panel members represent themselves, and express their own
personal preferences and values, amended by the insights acquired through their
professional work and organizational life: "What do I think the society should pay for
the environment?" This encourages introspection which in turn increases the probability
that the results are truer than in the more artificial setting of assuming the role of a
stakeholder representative. This approach appears to be most common, and is the one

184
adopted here. It is of course also vulnerable to strategie behavior, but based on our
experience we judge the problem to be small. People are both eager to and capable of
expressing sincerely their own values as they are developed and formed through
discussions with the other panelists. Open discussions based on own values make it
more diffieult to pursue hidden agendas.
The DPs may be composed according to two different strategies; the panels may be
made homogenous or heterogeneous. In the first ease, people with similar background
are put in the same panel. This makes discussions easier and results more readily
forthcoming. The risk is that not all relevant issues will be brought up and important
arguments not heard. With this strategy, it is important that several and different panels
are used, so that a relevant range of results is produced. It has been employed to assess
the values exposed to the risk of oil spills (Wenst0p, 1983) and a valuation of the end
impacts ofroad traffic (Wenst0p et al., 1997). With the second strategy, the panels are
composed of representatives with a known "high" valuation of a particular good (and
who would be likely to overstate their valuations), and others with a known "low"
valuation (and who would be likely to understate). Such compositions may be
advantageous in order to handle and control possible strategie misrepresentations of
preferenees among panel members. It also creates stimulating diseussion within the
panels, some of which can be lengthy at times. If the members of a panel are unable to
reach agreement on values, the set of weights of the different fractions of the panels is
entered into the protoeol as separate results. This is the strategy that was adopted in the
present application.

2.4. Weight elicitation

It is convenient, but not necessary, to construct the preference strueture with computer
assistance. In the case described below, we used a simple spreadsheet program,
Pro&ConW, which uses trade-offs in the sense of Beinat (1997) as the major vehic\e for
establishing relative weights between the decision criteria. Pro&Con W uses adecision
table as the starting point for the construetion process. The weight elicitation proeess in
Pro&ConW is a stepwise procedure:
1. The panel is presented with adecision table [x\] describing adecision problem. The
table lists the decision criteria and provides estimates of their scores for each
decision alternative. i k is the score on criterion i of alternative k.
2. The panel is asked to specify a utility funetion for each criterion.
3. The panel is asked to select two criteria for trade-off judgment. (The two first criteria
are proposed by the program. )
4. The panel is presented with a graph with two random points and asked to which
degree they prefer one over the other. They may respond with "strong", "moderate"
or "slight" preference or "indifference". The responses are coded numerically as 3, 2,
1, or O. Alternatively, they may choose to move the points until they are indifferent.
See Figure 1 where the two points are shown as a dot and a rectangle.

185
5. Step 4 is repeated several times with different pairs of criteria. Pro&ConW proposes
the sequence ofpairs (1;2), (2;3), .. ,(n-l;n), (n;I). The sequence may, however, be
overruled by the panel.
6. The coordinates and responses are written down in a protocol. Linear regression
analysis is used to compute the implicit relative weights, and a graph shows 95%
confidence intervals. The panel is informed about their consistency (R 2 adjusted) and
may choose 10 change some oftheir responses, ifthey deern this appropriate.
7. The utilities ofthe alternatives are computed, using an additive utility function. (Like
in most WTP studies, we simplify by assuming independence between the criteria.)
8. While the weights express the importance of the criteria in the decision problem, the
implicit WTP for one unit of a criterion, has a problem-independent meaning. If one
ofthe criteria is monetary (such as construction costs), Pro&ConW computes WTPs
for all criteria as an important feedback to the panel. If the .monetary criterion is
called X m, and the worst and best value in the decision table of a criterion is denoted
XO and x* , then the WTP for one unit of criterion· i is calculated as:
• -x",
W i X'" 0
WT~ = - - .- - 0 '
. the we1'ght 0 f cntenon
H ere, W 1S . . I..
i
WIll X, -X,
9. The panel performs sensitivity analysis, using all available information, and arrives
at a final set of weights.

Proc:.sing ProtDcoI Une nO.:3

3.00

I
2,50 -------~------~-------~--------
I

E
::!!. 2,00
I
-------~------~-------~-------~
I

!
I I
• I I

•
1
I
I

1 50 -------~------~-------~--------
• I I

•
I I

11.00 -------I----_--~-------~------_.
I I I

.5
_______ L ______ J I _______ J I _______ _
0.50
I I
I I
I I

0,00 +--__-+___-+-___+--__ I
~

0.00 5.00 10,00 15.00 20.00

River length (km)

Figure 1. Trade-o.f/ analysis: The dot and the square can be moved until the panel is indifferent
between them.

186
3. The case: Sauda hydropower development

The impacts of hydropower development are highly site specific. To make our
application more generally useful, we selected as case a so called storage project which
is rather typical of future development projects ofhydropower in Norway.
A storage project consists of a large dam that effectively blocks and limits the amount
of water that can flow downstream. Behind the dam, water accumulates to form a
reservoir. A large volume of water allows the storage project to draw on this water
whenever power generation is required. Developers are presently preoccupied with
improving storage projects and increasing their efficiency. One way to do this, is to link
up existing projects to new river diversions.
The Sauda project provides a typicaJ example. In the mountainous area surrounding
the small municipality of Sauda with 5300 inhabitants and situated by the Sauda Fjord
in south-westem Norway, an upgrading and extension project is being planned. The
existing plant is at present producing approximately 1 TWh on an average basis per
year. The deveJopment consists of a basic project and six possible diversions. Table I
shows the production and cost for the different alternatives. The new projects will
increase the production volume considerably.
Several combinations of the six different diversion projects can be chosen, but to
simplify the study we only considered the basic project. This means that the decision
problem in the present analysis is whether or not to build the basic project. Table I is
included to show how optional diversions typically may increase the effect of a new
development.

Table 1. Expected energy production and costsfor the basic project and six possible diversions. Yearly
costs are based on 40 years and 7% interest rate.
Alternative Production Investment costs Yearly costs
GWh/year USO USO
Basic project 674 431 32
Optional diversions (showing extra production and costs)
Mardals-Sagelva 32 8,7 0,65
Ab,,/Sandvatn 99 21,5 1,61
Nedre Ab" 61 20,8 1,56
Oalselv 206 46,2 3,46
Hamrab" 86 32,8 2,46
Lingsvang 129 30,0 2,25

3.1. Environmental impacts ofthe basic project

In the basic project, two existing reservoirs will be extended but no new reservoirs will
be built. The most important impacts are related to the reduction ofwater flow in rivers.

187
Effort has been made to limit the building of new roads by drilling many of the new
tunnels without new roads. Other roads will be buHt in elose cooperation with local
interests. In some of the places, the impacts of reduced water flow will be abated by the
use of small river dams where fish can survive.
The impacts of hydropower development are manifold and it is achallenge to try
describe and measure it in a simple and decision-oriented way. The most important
objectives were identified as nature conservation, cultural sites conservation, enhancing
recreation and reduction of costs.

Nature conservation
Four criteria were chosen to describe nature conservation.
1. Impounded area (km2): Extra ofland submerged by the extended reservoir. Since the
panel's WTP will depend on the type ofland that is impound~d, pictures were used
to show what the area looked like.
2. Length o/rivers with reduced discharge (km): The reduced dischargewill influence
the vegetation in the relatively flat river basin. The area has botanical value since the
plants here are relatively unique.
3. Length 0/ new roads (km): Since the development area is located up in the
mountains, some new roads have to be buHt. These may improve transportation, but
when nature conservation is the objective, they are a disadvantage and therefore
treated as a negative impact.
4. Length 0/ new power transmission fines (km): To transport the electricity, new
power transmission lines have to be buHt with consequential negative visual impacts
on the landscape.

The advantage of the four criteria is that they are measured in an objective way.
There can hardly be any discussion as to how big the impounded area will be. The
disadvantage is that they are to some degree only indicators of more essential
objectives, thus making weighting more difficult.

Conservation 0/ cultural sites

Cultural sites are primarily old summer farm remains and therefore rather homogenous
and of equal value. They may be directly or indirectly afflicted by hydropower
development. The impact is direct when the cultural object is submerged or cannot
function in the original way. This is the case when an old water driven saw is in a
tributary were the discharge is drastically reduced. Indirect influence occurs when the
landscape around the cultural object is disturbed.
5. Number 0/ cultural sites ajJ1icted.

Recreation
The mountains where the hydropower project is being planned is a popular recreation
area, with fishing and hunting as main activities. Surveys have been carried out to

188
measure the extent of present use, and estimates have been worked out for the expected
decline should the basic alternative be realized. Two criteria were identified:
6. Fishing days (reduction in days per year).
7. Hunting days (reduction in days per year).

Costs
8. Construction costs (mUSD/year) were computed at 7% interest rate based on an
expected life time of 40 years (7% is the official discount rate).

3.2. The decision table

All estimated consequences of the basic project were measured relative to the present
situation. Theyare shown in Table 2, which constitutes the decision table that served as
the input to the weighting process.

Tllble 2. The decision table with estimated scores.

Criteria Basic Present
project situation
Impounded area (km2» 2,6 0
River length (km) 18,9 0
Newroads (km) 17 0
Transmission line (km) 26,5 0
Cultural sites (number) 38 0
Fishing (days/year) 1600 0
Hunting (days/year) 200 0
Costs (mill. USD/year) 32 0

Onlyenvironmental criteria and costs were included in the present study. The Sauda
development project will have impacts on agriculture, forestry and tlood protection as
weil. These criteria would have to be included in the cost-benefit part of a
comprehensive evaluation of the project.

4. Tbe decision panels

Our panel members were supposed to act as concerned citizens helped by whatever
expert knowledge they might have acquired through their professional work. Although
we chose participants who are stakeholders in the project, they were specifieally told
not to try and retleet any official viewpoints of their organization, but rely solelyon
their personal judgments as to what the society should pay for the environmental goods
at stake. To further limit the possibilities for strategie behavior, we eomposed the panels
to induee open diseussion. The panels were asked to try and reach consensus in their

189
valuations, but if this did not work out, we would register the different opinions as
separate sets of weights.

4.1. Panel composition

We created three panels with three persons in each. According to our experiences, a
panel of three persons is preferable. More persons will complicate the discussions by
silencing some, and letting others dominate, and make it more difficult to reach
consensus. Panelists were selected who had good knowledge of the Sauda project
and/or were involved in the permission process. Each panel consisted of people who
were expected to hold opposing viewpoints, and since environmental impacts are site
specific, there was at least one person with good local knowledge in each panel as weIl.

Panel 1: Norwegian Water and Energy Administration (2 persons)

Ministry of Environment ( Energy and pollution, 1 person)
Panel 2: Local politician ( 1 person, chairman ofthe community board)
Local environment administration ( 1 person, county)
Local hydropower developer (1 person, Sauda development corp.)
Panel 3: Ministry ofEnvironment ( 1 person, nature conservation)
Ministry of Oil and Energy ( 1 person, hydro power)
Hydropower developer ( 1 person, State Hydropower Development)

4.2. Process

A DP interview sessions lasted approximately 5 hours, with the following agenda:

1. We started with a short introduction to the MCDM method and the idea of
computing WTPs based on the weights, of which every person had already received
adescription. Some of the participants were in principle skeptical to the idea of using
WTP as a way to valuate the impacts, finding it too limited.
2. We then went on to discuss the expected consequences ofthe Sauda project, and how
to represent them in terms of decision criteria (Table 2). A video film had been
produced which showed the present scenery of the area, and what it was likely to
look like after the development. The latter pictures were made by artists who
manipulated the original ones, and they turned out to be effective. The panels were
very interested in discussing these matters. They basically accepted the proposed
criteria, but remarked that a measure of bio-diversity should have been included.
They were uncertain of whether the estimate of loss of fishing days was correct,
since many parameters affect the behavior of fish and fishermen, and the local
panelists in particular were skeptical about the expected reduction of recreation
activity as a consequence ofthe hydropower development.
3. Pro&ConW was then used to elicit preferences. We were impressed by the acuteness
displayed in the selection of responses to the questions, with resulting high degrees

190
 of consistency. Within each panel, the participants were encouraged to try and reach
consensus through discussions and compromise, but we were prepared to end up
with different sets of weights if the values held by the participants turned out to be
irreconcilable.
4. Finally, we reviewed the results, focusing on the WTPs. Printouts were made, and
the participants were encouraged to tell us later if they on afterthought found them
unacceptable.

5. ResuIts

5.1. Estimated WTPs tor environmental impacts

Panel 2 could not reach a consensus and produced a majority and a minority set of
weights. In panel 3, one person adjusted bis weights retrospectively. Table 3 shows
weighted averages for those panels.
According to Table 3, the annual WTP of panel 1 to avoid submerging one km2 of
land, is $25 000. Further, the annual WTP to avoid reduced discharge in one km river is
$100000, and to avoid one km new road and one km new transmission lines are worth
$25 000 and $13 000, respectively. To avoid destruction of one cultural site (e.g. old
summer fann tuft) is worth $18 000 per year. Finally, the loss attributed to one day of
tishing or hunting for one person is valued at $770 and $2670, respectively.

Ttlble J. WTPs per unit 0/ the criteria elicited/rom the three panels.
Criteria I WTP (1000 USD/year) Panel 1 Panel 2 Panel 3
Impounded area (km2) 25 4.8 65
River length (km) 100 Il 34
New roads (km) 25 8 28.5
Transmission line (km) 13 7.67 24.2
Cultural sites (number) 18 4.67 12.8
Fishing (days/year) 0.77 0.17 0.33
Hunting (days/year) 2.67 0.30 1.5

Tbe table reveals a considerable range of WTPs across the panels. While there is
reasonable agreement between Panel land Panel 3, the valuations of Panel 2 are
considerably lower. Tbe minutes from the discussions reveal that this is not due to
uncertainty, but to real differences of opinion. Panel 2 is the local panel. Given that the
hydro-power project will provide the local community with considerable revenue in
terms oftaxes from the sale of electricity, one interpretation is that the lower values are
due to strategic behavior. Judging from the discussions in the panel, a more credible
interpretation is that they simply know better what environmental values are at stake,
and therefore have made a more appropriate assessment.

191
 By combining Table 2 and 3, it is possible to compute a total environmental cost of
the basic alternative. The result is shown in Table 4. The annual construction cost of the
basic alternative is estimated at 32 million USO. Our estimate of the environmental
costs is thus approximately 10 % of this. This is low compared to the Master Plan
where the Parliament through its ranking implicitly attributed a 30% environmental cost
to the group of least controversial projects, but it agrees weH considering that the Sauda
project is one of the environmentaHy sounder projects of this group (Carlsen et al.,
1993). This comparison should not be carried too far, however, since the project
descriptions available to the Parliament were much cruder than what is now available in
Sauda. We have also compared our results to a CV study carried out by Navrud (1993)
where the sum ofWTPs agree weH with panel 2 in our study.

Tllble 4. Estimated annual environmental costs 0/ the basic alternative, based on the WTP 0/ the panels.
Criteria Impact Environmental cost (1000 USD/a)
Panel 1 Panel 2 Panel 3
Impounded area (km2» 2.6 65 12 169
River length (km) 18.9 1890 208 643
Newroads (km) 17 425 136 485
Transmission line (km) 26.5 345 203 641
Cultural sites (number) 37 666 173 474
Fishing (days/year) 1591 1225 270 525
Hunting (days/year) 203 542 61 305
Sum 5158 1064 3241

5.1. Extended cost-beneflt analysis

Our estimate of the external (environmental) costs allow for an extended cost-benefit
analysis. The major benefits accrue from the production of electricity which can be sold
to the consumer, at prices which depend on precipitation and temperatures. The
Norwegian electricity market is now deregulated. The price of electricity is therefore a
function of supply and demand. There is no official estimate of future prices, but
analysts we have contacted suggest $0.05 per KWh as a realistic estimate.
Other benefits accrue from increased employment in the area. Strand and Wenstep
(1993) estimated the social benefit ofthis to be $1.2 million per year. Another benefit
arises from increased discharge of freshwater into the fjord. which wiH reduce the
periods when ice is an obstaele to catamarans. Navrud (1992) has estimated this benefit
to $10,000 per year.
An economic loss to farming and forestry is also expected. and costs will be incurred
to seeure a new water supply. These are estimated by Navrud (1992) to be respectively
$13,500, $33 and $1,125 per year. The full picture is shown in Table 5. We see that the
whole project is elose to break-even. For panel 2, the benefits outweigh the costs, while
the scale tips the other way for panels 1 and 3.

192
Table 5. Extended cost-benefit analysis 01 the basic project in the Sauda hydropower development plan
(JOOO USD/year).
Benefits Internat costs
Etectricity production 33700 Construction costs 32300
Increased ernptoyrnent 1230 Fanning 14
Fjord trafik 10 Forestry
Water supply
Sum 34940 Surn 32315
Net internat benefit 2625

Externat costs Panel 1 Panel 2 Panel 3

5158 1064 3241

6. Conclusions

The Sauda hydropower development plan is still in process, and a formal application
has not yet been made. Our conclusions are therefore tentative and reflect primarily
experiences from our part of the process.
First of a11, there is a noticeable reluctance on part of the panelists to embrace the
concept of value focused thinking. They prefer to discuss consequences and engage in
such a discourse with great enthusiasm. In contrast, when asked to engage in a value
focused discussion, they tend to hesitate in front of the challenge. However, when they
got used to the idea, they displayed an acuteness of perception that revealed an intimate
fami1iarity with the trade-off issues involved. Thus, when differences of opinion arose,
these were to a lesser degree due to uncertainty, than to real differences of preference.
We have in other previous studies used homogeneous panels. The heterogeneous
panels which w! are used in this study created more discussions within the panels. In a
larger perspective, this is useful since it gives the panelists, who also are participants in
the real process, a chance to exchange viewpoints on values and not only on
consequences which tends to characterize the public debate. This may provide a basis
for better mutual understanding. This is of course hard to demonstrate and therefore a
bit speculative.
Decision panels often end up with different sets of weights. This is a manifestation
of the range of preferences among the stakeholders, and is as such useful information. It
poses problems, however, ifthe different sets ofweights lead to different recommended
courses of action, as in this case.
With regard to a strategie misrepresentation of preferences, we honesty believe this
not to be a problem in the present study. The faet that the panelists ended up with
differing preferences that were compatible with their stakeholder backgrounds, even

193
though they were required to aet as eItIzens, is not surprising and do not indieate
strategie behavior. Peoples' preferenees are obviously formed by the environment they
work in and the sort of issues they deal with daily. This will manifest itself in a
valuation proeess, but makes it obligatory to include people with expeeted opposing
views in the proeess.
1nitially we provided a list of possible deeision eriteria, but the panels had
eonsiderable influenee on the eriteria that were eventually used. Some of the
partieipants had strong and clear opinions regarding essentialness of the eriteria. There
were some objeetions to the concept of a limited and quantitative set of criteria to
represent the multifaceted issues involved, but the aeeeptanee gradually improved when
it was understood that the criteria could be viewed as indieators of more fundamental
underlying amenities.
The interactive weight elieitation process with pairwise trade-offs worked weil, but
the interpretation of weights tend to be a bit elusive. The eomputation of WTPs was
therefore particularly effeetive and prompted a feedback and reeonciliation process.
Although we were able to compare our WTPs with those from other studies, this was
not brought into the elicitation process, and there was no cross information between the
panels. They had to rely solelyon themselves.
1t is reassuring that the WTPs of the loeal panel were in general agreement with
those of the local population. We advise linking the output of deeision panel elicitation
processes to those from other studies through some of the criteria so that the overall
results can be calibrated.
Since the Sauda development plan is still in process, we know Httle about the
specific impact of our part of it, or of the possibilities for getting aeeeptance for using
deeision panels as anormal proeedure in the required holistie project applieation
evaluation proeess. Although comments are favorable, they are not deeisive. 1t does
seem to be a growing interest, however, in eonneetion with reeonditioning of hydro-
electric plants whieh was inspired by our work (Tangen 1996).

References
Attfield, R. (1994), Environmental Philosophy: Principles and Prospeets, Avebury, Aldershot.
Beinat, E. (1997), Value Junetion Jor environmental management, Kluwer, Dordrecht.
Brown, T.C., P. Slovic (1988), EtTects of Context on Economic Measures of Value, in
Peterson, G. (Ed.) Integrating Eeonomie and Psyehologieal Knowledge in Valuations oJ
Publie Amenity Resourees, State College, PA, Venture, 23-30.
Carlsen, A.J., 1. Strand, F. Wenstop (1993), Implicit Environmental Costs in Hydropower
Development: An Analysis ofthe Norwegian Master Plan for Water Resources, Journal oJ
Environmental Eeonomics and Management, 25: 201-211.
Cowen, T. (1993), Tbe scope and limits of preference sovereignity, Eeonomies and Philosphy,
9: 253-259.
Cummings, R. G., D.S. Brookshire, D. Schulze (\ 986), Valuing Environmental Goods:
Assessment oJthe Contingent Valuation Method, Rowman ans Allanheld, Totowa.

194
Gregory, R., S. Lichtenstein, P. Siovic (1993), Valuing environmental resources: A
constructive approach, Journal of Risk and Uncertainty, 7: 177-197.
Kahneman, D. J. Knetsch (1992), Valuing Public Goods: The Purchase of Moral Satisfaction,
Journal ofEnvironmental Economics and Management. 22: 57-70.
Keeney, R. L., E.F. Wood (1977), An Illustrative Example ofthe Use of Multiattribute Utility
Theory for Water Resource Planning, Water resources Research, 13: 705-716.
Keeney, R.L., J.F. Lathrop, A. Sicherman (1986), An Analysis of Baltimore Gas and Electric
Company's Technical Choice, Operations Research, 34: 18-39.
Keeney, R.L., 0. Renn, D. von Winterfeldt (1987), Structuring Germany's Energy Objectives.
Energy Policy, 15: 52-362.
Navrud, S. (1992), Kvantifisering av miljekostnader av vannkraftutbygging i Sauda, ENCO,
1300 Sandvika, Norway.
Navrud, S. (1993), Miljekostnader av Vannkraftutbyggingen i Sauda: en betalings-
villighetsundersekelse. ENCO, 1300 Sandvika, Norway.
MitcheII, R.C., R.T. Carson (1989), Using Surveys to value Public Goods: The Contingent
Valuation Method, Resources for the Future, Washington, DC.
Strand, 1., F. Wenst0p (1991), Miljekostnader og samjunnselconomi. delprosjekt 7. Department
of Social Economics, University of Oslo.
Tangen, G. (1996), Decision Making Support applied to Hydropower Plant Upgrading, ITK
report 1996:51-W, NTNU, Trondheim, Norway.
Wenst0p, F. (1983), Evaluation of Oil spill Combat Plans by Means of Multi Criteria Decision
Analysis, in Stigum, B., Wenst0p, F. (Eds.) Foundations of Utility and Risk Theory with
Applications. Reidel, Dordrecht.
Wenst0p, F., A.J. Carlsen (1988), Ranking hydro-electric power projects with multi-criteria
decision analysis, Interfaces. 18:36-48.
Wenst0p, F., A.J. Carlsen, 0. Berlgand, P. Magnus (1997), Valuation of Environmental goods
with Expert Panels, in 1. Climaco (Ed.) Multicriteria Analysis, Springer Verlag, Berlin.

195
Improving decision-making for land-use management
key ideas for an integrated approach based on MCA
negotiation forums

Andrea Nardini l
Center for Research on Environmental Sciences EULA-Chile
Universidad de Concepcion, Concepcion, Chile

Abstract
The aim ofthis paper is to contribute to making the Multicriteria Analysis (MCA) approach
a better aid to environmental decision making than it has been to date. In order to do so, the
current problems and weaknesses are first discussed in relation to Environmental Impact
Assessment, Cost-Benefit Analysis and MCA as evaluation tools in support of land-use
management. The unsatisfactory handling of different views and interest conflicts arising
from the many players involved is identified as a negative common structural feature. lt is
also noted that a suitable framework through which these basic approaches could be
integrated is still not available. Accordingly, an integrated Decision Analysis approach that
could overcome some of these weaknesses is outlined and the role MCA would play in it is
discussed. The key components of this approach are: i) the significant role given to public
participation by means of a specific administrative procedure which provides a formal forum
for negotiation based on an MCA framework; ii) the use of specially designed evaluation
indices to represent impacts within the c1assic two-entry MCA evaluation matrix. A case
study on an important Chilean irrigation project is taken as an illustration to show some of
the ideas proposed.

Keywords: Multicriteria Analysis; Environmental Impact Assessment; Cost-Benefit

Analysis; public participation; negotiation; decision making; integrated evaluation;
evaluation indices.

1. Introduction

Decision making related to land use management certainly poses very difficult and
challenging problems. There are 'technical' problems in assessing and predicting
complex and uncertain environmental and socio-economic consequences.
Furthermore, conflicting and sometimes powerful vested interests must be dealt with.
This is particularly important in developing countries, where the financial aspects
related to multinationals very often outweigh any other consideration.

I Current address: Ambiente Italia, via C.Poerio, 39 - 20129 Milano, Italy.

197
 It is difficult to find definitive answers about how land-use management is
performed. Satisfactory established methods are seldom available (Warner 1996),
especially in developing countries, particularly concerning the creative phase in
which possible alternative actions (projects, plans, etc.) are defined. The evaluation
phase can, nevertheless, rely to a much greater extent on formalized procedures and
weIl established techniques.
Theoretical developments on evaluation techniques and methodologies to support
decision making have been made mainly at the project level. In response to a wide
consensus gained over recent years on the need to formally evaluate plans, policies
and programs (Therivel, et al., 1992; Glasson et al., 1994; Ortolano and Shepherd,
1995), new methods have been devised, particularly in the field of land-use
management (see for instance Warner, 1996; Johnson et al., 1994; Yin and Pierce,
1993; Goodenough, 1992; Baird and Ive, 1989).
These and other quantitative approaches are, in all cases, based to some extent on
techniques belonging to: i) Environmental Impact Assessment (EIA) (Huang, 1989;
Conacher, 1994; Warner, 1996), ii) Economic Valuations or social Cost-Benefit
Analysis (CBA) - mainly performed in developed countries (see, for instance, Asian
Development Bank 1988), and, increasingly, iii) Multicriteria Analysis (MCA) (see,
for instance, Edwards 1977; Janssen 1992; UNESCO 1994).
The evaluation phase certainly plays a fundamental role in the decision process
related to land-use management. However, in spite of the considerable existing
theoretical corpus and practical experience gathered regarding the three basic
evaluation approaches (EIA, CBA and MCA), they are still fraught with structural
weaknesses (which are reviewed briefly later on); and they are stilliargely developed
and applied in an independent fashion, with no inter-approach integration.
There is the need and room for an integrated approach which takes advantage of
the combined potential of EIA, CBA and MCA to support environmental decision
making, and land-use management in particular. The driving principle behind such
integration is to support a participatory and transparent decision making (DM)
process in which the divergent views of the many players involved are considered,
and the conflicts of interest addressed. This paper presents an ideal approach along
this line, and is centered on the evaluation phase.
It is important to point out that the issues discussed here are viewed essentially
from a practical rather than a theoretical perspective. In practice, actual decisions
depend much less on the power of analytical tools than on an ability to understand
the reasons for divergence among the players, and on the ability to create new
alternatives which could be "socially acceptable". According to this, the focus of the
following discussion is on the Multicriteria approach to the problem at hand, rather
than on the Multicriteria techniques which can support the DM process.
First, the main weaknesses in current evaluation practices are highlighted. Then,
the key features of an integrated approach are identified followed by a discussion on

198
how to implement some of them; the Multicriteria approach in particular. A Chilean
irrigation project is taken as a case study to briefly iIIustrate some of the concepts
introduced.

2. Weaknesses in current evaluation practices

In developing countries, environmental degradation and social conflicts are

commonly the results of decision making processes about contentious projects (see,
for instance, Long, 1994). Together with industrialized countries, where decisions
related to land use (e.g. the selection of waste disposal sites) are increasingly
influenced by spontaneous public concern, experience indicates structural
weaknesses in current evaluation practices which are all based on EIA, and/or CBA
and/or MCA. In order to better understand this evidence, let us briefly consider these
weaknesses.

2.1. The case 0/ Environmentallmpact Assessment

EIA is a widely used procedure for identifying and evaluating the future
environmental consequences of a current or proposed action, this being a project or
plan. A number of authors have detailed the various weaknesses of project-Ievel EIA
(Therivel et al., 1992; Glasson et al., 1994; Ortolano and Shepherd, 1995; Wamer,
1996). Two considerations seem particularly relevant. On the one hand, EIA is
undertaken when it is too late to actually incorporate environmental considerations at
the creative stage. In fact, at the point at which an EIA is commissioned, the type,
scale, and location of a project have often already been decided on the basis of
economic, technicalor political criteria. In this context, the public has no chance to
gain early influence in the way decisions are made. Consequently, both the planning
authority and the proponent tend to see public reaction as unproductive opposition,
while the public feels excluded from participating in decisions where its interests are
at stake. The causes for this can be found in the very conception of EIA. As its name
implies, it is evaluation, and as such it comes after the object of the evaluation has
been defined. Indeed, the norms and procedures concerning EIA also conform, in
many countries, to this perspective. The EIA's current main purpose is to ensure that
negative impacts do not pass an acceptable threshold (sometimes objective - e.g.
when legal standards are defined - but often subjective - e.g., when landscape quality
is reduced). EIA was not conceived, and is not regulated and practiced, as a tool to
provide a complete balance of strengths and weaknesses of alternative actions. It can
be argued that all this corresponds to "bad EIA practice". However, there is broad
practical evidence that this is the way things very often go, especially in developing
countries (see for instance Nardini et al., 1997, for a Chilean analysis).

199
2.2. The case ofCost-Benefit analysis

The traditional economic approach to project appraisal - Cost Benefit Analysis

(CBA) - has been (and is being) constantly extended and improved for use in
environmental decision making. In particular, techniques have been developed to
value environmental goods and services for which a market does not exist, thus
allowing one, in particular, to 'intemalize extemalities' (see for instance Dixon and
Hufschmidt, 1986; Johansson, 1987 and Munasinghe, 1993). CBA, together with the
ancillary practice of the economic valuation of impacts, have been criticized on
ethical, technical and practical-political grounds. In particular:

i) The economic valuation of environmental, socio-cultural and health effects, is

quite often difficult, sometimes impossible, and almost always questionable
(see, for instance, Ehrendeid, 1988, conceming biodiversity). Contingent
Valuation (CV) exercises - probably the most popular valuation technique for
such items - suffer from a number of methodological and ethical problems.
Strategic, information, 'starting point' and 'hypothetical' biases are some of the
first type of problems (Tietenberg, 1992; Jacobs, 1992). The latter type are
related, amongst others, to the fact that in CV practices the public is generally
addressed as an aggregate of individual consumers which, as Sagoff (1988)
argues, is far from being the attitude of the public when dealing with
environmental problems.
ii) The discount rate, necessary to aggregate the stream of future benefits and
costs, is quite an intricate point and is basically unclear. As a consequence,
there is always room for arbitrariness in the final result: a project which would
be discarded at a given discount rate, might look acceptable at a lower one
because future benefits are given more weight.
iii) Efficiency - the criterion underlying CBA- is theoretically not incompatible
with sustainability (Tietenberg, 1992). The existence, however, of exhaustible
natural and environmental resources for which it is impossible to find
substitutes (for instance landscape quality, or wildemess), and the difficulties
associated with the choice of the discount rate strongly challenge this
theoretical possibility. Even without taking a strict sustainability position, it is
often suggested, as a consequence, that the field of action of CBA should be
limited by a set of empirical constraints on the use of natural resources (see for
instance Munasinghe, 1993).
iv) Efficiency is theoretically not incompatible with equity, provided that the
maximum net social benefit is fairly redistributed among individuals. Real
world evidence, however, demonstrates how hard it is to apply this concept in
practice.

200
v) Most important, interest conflicts are, in practice, ignored because the CBA
framework does not strictly allow for public participation and negotiation. It
can be advocated that some valuation techniques, like Contingent Valuation,
do interact with affected people; but it is so only to extract the information
needed to estimate their own willingness to pay, without an attempt to drive
people into a negotiation process necessary for conflict resolution.

2.3. The case 0/ M ulticriteria Analysis

An alternative approach is given by Decision Analysis, a set of theories and

techniques designed to help individuals and organizations make complex decisions,
especially those decisions that involve uncertainty and a multiplicity of effects (von
Winterfeldt and Edwards, 1986). In this context, Multicriteria Analysis (MCA)
techniques (see UNESCO, 1994 for a review) constitute the natural tool to address
environmental decision problems and, specifically, land-use management. The main
advantages of the MCA approach in environmental decision making are: i) its non-
economic valuation character, ii) its capacity to deal with multiple and conflicting
issues, iii) the help it can provide in structuring complex decision problems, thus
increasing transparency.
However, MCA theories, techniques and applications are often too complicated
mathematically to allow the public to be involved effectively (Bisset, 1988). This is
reinforced by the fact that MCA is generally not included in an administrative
(statutory) procedure, such as that ofthe EIA system. In addition, as is made clear at
a later point, the information presented in the evaluation matrix is often not suitable
in addressing interest conflicts by involving the public direct1y. Finally, common
MCA techniques make it difficult to really take into consideration the extensive
infonnation produced by EIA studies.
On the other hand, the MCA approach maintains no links, from a theoretical point
of view, with the traditional and pervasive criterion of economic efficiency, i.e. with
Cost Benefit Analysis. There may be a link in the case where MCA is used to
support a cost-effectiveness approach. The link is, however, theoretically
unsatisfactory. In fact, efficiency strict1y requires maximizing the difference between
benefits and costs, where these are all expressed in economic terms (conceptually
they are willingness to pay). Furthermore, cost-effectiveness is only a special case of
the MCA approach.

201
3. An integrated approach to decision making for land-
use management

The study of several real world problems related to land-use management together
with the broad array of Decision Analysis applications2 leads one to the immediate
conclusion that it is worth making an effort to exploit the positive and attractive
features of EIA, CBA and MCA, while at the same time compensating for their
deficiencies. This can only be achieved through suitable integration.
A line of reasoning is developed below. First, the key features that should be
included in an ideal approach to decision making for land-use management are
identified. These features should be regarded as requirements. The way some of
them can be attained is discussed later on. Finally, examples are given to clarify the
concepts presented.

3.1. Key features of an integrated approach to decision making

8. Strengths-weaknesses balance. Similarly to the CBA approach, a summary of

strengths (benefits) and weaknesses (costs) has to be performed as far as possible.
This means that both positive and negative impacts should be considered at the same
time. All possible types of impacts should beincluded: economic, bio-physical,
aesthetic, socio-economic, socio-cultural, health, political, etc. Specific approaches
for assessing such diverse impacts are currently being developed and tested in the
EIA context. Consequently, decision making becomes what it really is, i.e. a trade-
off between conflicting criteria. In the end, this trade-off should represent, as in
CBA, a social judgment, even in the case of (large) private projects, because they
will affect people's welfare to some extent.

b. Comparison 0/ alternatives. The people affected by the consequences of the

implementation of a given project need to express a judgment about it. The approach
that makes sense to them is to perform comparisons within a set of project
alternatives (including a do-nothing alternative). The presence of alternatives is
indeed apre-requisite to the reaching of negotiated solutions.

c. Scenarios. The future outcomes of a project are not completely determined even
when the project alternative is thoroughly defined. The outcomes also depend, in

2 Applications inc\ude, among others, decisions on water resource planning and management
(Ortolano, 1976), land-use regulation and management (Edwards, 1977), coal prospect evaluation
(AlIet, 1986), acid rain policy analysis (Watson, 1986), energy policy options (Jones et al., 1990;
Renn et al., 1993), rural development (Petry, 1990), sewage sludge management problems (Renn
et al., 1993), and road developments (Tamura et al., 1994).

202
fact to a large extent, upon uncontrollable and uncertain variables. The evaluation of
the project alternatives can be carried out only by specifying what future pattern of
such variables is assumed to occur; this pattern is called a scenario. Generally,
different scenarios are to be defined and considered. The evaluation should then be
carried out in parallel for each one of them. By definition, it does not make sense to
select the more desirable scenario because the occurrence of a scenario is
independent ofthe human will. Therefore, the criteria that can be adopted must be of
the type: 'select the alternative that, under the most probable scenario (or, rather,
under the most adverse, etc.), performs better' .

d. Multicriteria approach. The evaluation problem is instantly put as a Multicriteria

decision problem in which the strengths and weaknesses of all the considered
alternatives are shown simultaneously. All the information that is relevant to the
evaluation of the alternatives should be presented concisely in an evaluation matrix:
i.e. a table in which each column corresponds to an alternative, and each row to an
impact. All of this information is suitably represented and, of course, accompanied
by all the relevant documentation.

e. Impact Assessment. "Environmental Impact Assessment and Statement" are no

longer appropriate names. Instead, Impact Assessment (lA) should be used. This
would refer to a study relying on sectoral approaches and techniques like Social IA,
Environmental IA, etc., as weil as Economic IA (i.e. economic evaluation). lAs
should consider and analyze all different types of impacts, offer a comprehensive
and concise view ofthe performance of each alternative, involve the affected people,
and be quantitative as far as possible. Finally, the IA should be carried out
simultaneously für all the considered alternatives.

f. Negotiation /or conflict resolution. Conflict resolution should be recognized as a

fundamental element and should be treated as such. In fact, any decision making
process implies, to a greater or lesser degree interest conflicts between (and within)
different social groups. Even in the ideal case of a project characterized by a unique
positive economic impact (e.g. an income increase), conflicts potentially arise as a
result ofthe uneven distribution of such an impact among the people.

g. Role 0/ CBA. Extended Cost-Benefit Analysis (ECBA) should be integrated into

the DM process for three reasons: i) CBA's appeal given by its link to welfare
economics (i.e. the belief that economic efficiency corresponds to a socially

203
desirable situation l ); ii) CBA's power, since market prices used in any economic
evaluation, tend to be - at least conceptually- the result of aglobai bargaining
process between everybody for everything4 ; and iii) CBA popularity: decision
makers are too used to CBA to simply abandon it. However, CBA should be applied
- particularly in the case of projects proposed by a public body- only after interest
contlicts have been addressed and resolved. It should be noted that even in the case
where there is only one accepted alternative, it makes perfect sense, from the
perspective of efficiency, to compare such an alternative to the do-nothing choice.

h. Publie involvement and dynamie decision proeess. The public can greatly
contribute in the screening and scoping, in identifying and assessing impacts,
creating alternatives, proposing mitigation measures, etc. Furthermore, the success
of the project often depends strongly on its acceptance by the people involved. The
evaluation of the alternatives cannot proceed independently of conflict resolution;
and this, in turn, cannot be achieved if not through organized public involvement.
Decision making is indeed a process which must be more open to public
involvement, although sufficiently structured to avoid unmanageable situations.

It is important to point out that interesting advances have been reached in some
industrialized countries - especially in the Netherlands - concerning some of these
issues. However, the situation in most countries, especially developing ones, is very
far from this ideal picture.
The following sections consider some of the key aspects outlined above but in
greater detail. In particular, the role of MCA in the evaluation phase is discussed
further. A more precise definition of impact is proposed, and the issue of measuring
and representing the impacts in a form suitable for public involvement - i.e. through
specially designed evaluation indices - is discussed.

3.2. The MCA-based evaluation phase: a negotiation forum

Let us assume that, from a conceptual point of view, the alternative to be selected is
the one which maximizes the global welfare of all people directly or indirectly
affected by the proposed action. However, real problems are characterized by the
presence of several groups of people with contlicting interests and often
irreconcilable positions. Consequently, no analytical aggregation of public

J It is very hard to state that efficiency in itself is not a desirable criterion. The problem is that the
potentially 'desirable' efficient situation often implies aredistribution of the benefits which, in
practice, hardly ever occurs. Therefore, undesirable side effects actuallY very often accompany
efficiency.
4 Many actual market prices are, however, pretty far from this figure owing to the imperfections in
the real markets.

204
preferences would be likely to enjoy a measure of success, for that would not tackle
the conflicts. As already noted in the introduction, what really matters in the context
of multi-player problems considered here, is the capacity that a Multicriteria
approach can provide in understanding the reasons for divergence amongst the
players, and to create new alternatives which could, in the end, be "socially
acceptable" .
This is, for instance, the emphasis given by Jones et al. (1990) and Renn et al.
(1993) in their decision analysis applications to energy planning. Multicriteria
techniques can of course support the decision making process. In particular, MCA
can increase the degree of transparency. To be of some help, however, such
techniques should be perfectly understandable by the people involved. This is hardly
Iikely for methods that are more complicated than the simplest of rankings (in
particular, the weighted sum). In any case, this particular issue is not particularly
relevant in the following discussion. Therefore, in order to develop a simple and
feasible procedure, reference is made explicitly to ranking methods without
excluding the possibility of adopting alternative techniques. A corresponding process
could then be carried out by following these steps (this issue is developed further in
Nardini,1997b):
I. Every interest group operates through a formal representative by way of
suitable meetings and questionnaires. They are asked to express judgments
about the relative importance assigned to the impacts al ready identified
through IA and public participation. Each group is explicitly asked to take a
social perspective, i.e. to seriously consider the impacts on, and the position of,
the other groups. This can be achieved through sm all group meetings, field
surveys, interaction with experts, or dialogue among stakeholders (see Webler
et al., 1995). Groups are made aware of the fact that their value judgments will
be made public. Legal actions can in any case be undertaken in a later phase,
should the negotiation process not lead to a satisfactory trade-off solution.
II. On the basis of such information, an MCA session (or aseries of sessions) is
developed in which a ranking of the alternatives is obtained for each group,
according to its own preferences (weightings). The main reasons for the
diverging views are identified and analyzed with the support of sensitivity
analysis.
III. Conflicts are addressed by identifying, proposing, designing and assessing
mitigation and compensation measures, thus creating new improved
alternatives. In addition, the evaluation criteria can be modified, integrated or
refined. In this process, a win-win strategy should be pursued. Indeed, if a
project is to be accepted and undertaken in a democratic setting, it means that a
net benefit would result (in a broad sense, not just economic), i.e. what is
gained outweighs what is lost. This holds for public as weil as private projects
because projects are to be judged from a social perspective. In line with the

205
 idea of economic efficiency, it would then be possible - at least conceptually -
to redistribute this net benefit in some way among all the impacted interest
groups. This would mean that no one would end up worse off than before the
project (win-win solution). Often, however, it is not possible to find win-win
alternatives in practice. It is exact1y in the~e cases that there is a powerful need
for negotiation and conflict resolution with the aim of (a) achieving at least the
feeling that a win-win solution has been found, or (b) convincing those who
are going to lose something that in some sense it is a fair or just outcome (e.g.
when the property rights on some kind of resource were badly defined, or
when one person's "sacrifice" is greater than the compensation in gains to
someone else who was previously worse off, etc.). If the project is well
conceived, it should be possible to identify alternatives which are preferred to
the do-nothing alternative in one of the senses just specified (actual win-win,
'win-win feeling', 'fair sacrifice'). Such improved alternatives can be defined
as 'socially acceptable'. On the contrary, if no such socia//y acceptable
alternative(s) can be identified, then the project should be discarded.
IV. Finally, if some socially acceptable alternative is found, the DM can process
the information about relative preferences for the interest groups according to
different criteria (e.g. weighted average based on numerical entity of the
groups, or based on the reciprocal of their income, etc.), and thus derive a final
ranking. Finally, the economically most efficient alternative could be selected
from the ones preferred according to the MCA analysis.

In this process it is essential to reach an understandable and agreed-upon

representation of the problem at hand by presenting the information concisely in an
evaluation matrix. In this context, rather than looking for more refined MCA
techniques, it is preferable to refine the elements around which the evaluation is
performed, i.e. to better define and represent the impacts. MCA techniques can in
any case greatly increase the degree oftransparency. In fact, given a set ofjudgments
about the relative importance of the impacts, the ranking of the alternatives should
not give rise to objections, ifthe conceptual model ofthe problem at hand is correct.
When objections do in fact arise, it often means that new objectives, hitherto
disregarded or simply ignored, must be introduced. If no new objectives or other
clear reason for disagreement can be identified, then the evaluation outcome should
not be changed. This mechanism fosters transparency.

3.3. Evaluation indices

Any modification to the environment and to the socio-cultural-economic system is

identified, in EIA practice, as an 'impact'. Though this is certainly a reasonable ·first

206
step, it is not completely correct in order to address the central issue of multiple
players' views and associated interest conflicts.
In line with the 'value focused thinking' concept ofKeeney (1992), the impacts to
be considered in the evaluation phase are only those to which some of the
stakeholders clearly attach some kind of va1ue (in a generalized sense), i.e. those
characterized by a change to someone's satis/action, for better or for worse (positive
and negative impacts, respectively). This definition might sound too anthropocentric,
but it will not be so as long as non-use values are included (these issues are discussed
further in Nardini 1997a; for a classic discussion of values see, for instance,
Munasinghe 1993, or Pearce 1993).
Considering all, and only, those impacts with which a satisfaction is clearly
associated is the first basic step in producing a fruitful negotiation forum. However,
this is not enough. Too often discussions and opposition are fed by the unclear
measurement and representation of some impacts. The second step that should
therefore follow is the suitable measurement of the impacts and their representation
in the evaluation matrix. This calls for building and using evaluation indices capable
of representing the degree of satisfaction that each given interest group experiences
in relation to a given impact. The degree of satisfaction varies in the range defined
by the two alternative actions which perform worst and best from that point of view.
Figure 1 shows a conceptual scheme for constructing evaluation indices. The
discussion is based on Nardini (1998 forthcoming) and it is a prototype proposal for
an approach that is explicitly designed for project evaluation. It aims mainly to make
a point about the relevant aspects that are to be considered when constructing an
evaluation index, and about the way such aspects can be organized into a logical
process of aggregation.
Notice that one of the reasons why evaluation indices have often been considered
misleading, and thus disregarded, does indeed lie in the lack of a recognizable
structure.
Let the term attribute denote adescriptor of a quality relevant in characterizing a
state component of the socio-economic-environmental system. The set of attributes is
such as to characterize the considered component from the point of view of the
values (in a broad sense) associated with it.
Each attribute is measured by a suitable indicator which is a numerical variable
that transforms the verbal, visual or whatever description into a number over a
suitable scale (see French, 1988, for a rigorous definition of scales and Volta and
Servida, 1992, for an interesting discussion about scale properties). Generally
speaking, each indicator is a function of space and time. For instance, the water
quality of a recreational lake (e.g. temperature) depends on the point considered
inside the water body, and possibly both on the date and on the hour ofthe day.

207
 Global group Instantaneous Instantaneous Instantaneous
satisfaetion group individual loeal individual
satisfaetion satisfaetion satisfaetion
objective group index index sub-index qualitylabundance indicators
Temporal Group aggregation Spatial quality & Aggregation Miero-seale
aggregation (over individuals) aggregation abundance of (spaee and time)
aggregation indieators aggregation

d ~ (a)
Co (I,a)
d~ (a)

d~ (a)

Figure I. General structure proposedjor any evaluation indexJ

5 To maintain a general approach, but to comply with space limits for the paper, only a minimum

number of elements is represented. It should not be difficult, however, to imagine how to complete
the figure. For instance, in the fourth column from the right (index), the items si2). so(3) .... so(N-l)
are missing. They correspond to the other individuals, apart from the first (I) and last (N), belonging
to the interest group considered.

208
 However, the indicator 'summer, daytime, surface-averaged temperature', would
be a better choice than 'instantaneous, local temperature'. Indicators, indeed, are
often the result of a spatial-temporal aggregation at a micro-scale. The general i-th
indicator, at time t (of the macro-scale), and site p (of the macro-scale), is denoted in
the following by the symbol d~ (P) (first column on the right ofFigure I).
The values assumed by the set of n indicators are displayed in the first column on
the right of Figure I. Just two sets of values for the indicators are included in the
figure, both for time t=O; the upper set represents site p=a, while the second set
represents site p=b (etc.).
A suitable aggregation of the indicators results in the quality magnitude c,(l,p)
which is a measure of the intensive component of the instantaneous satisfaction
experienced by individual I at instant t in relation to site p. This aggregation
incorporates the subjective value judgments of the individual I. Associated with
quality is an attribute to measure the extension measured by the indicator a,(l,p).
Extension may vary with time t, even ifthe quality keeps constant.
An aggregation of quality and extension determines the instantaneous local
individual satisfaction measured by the sub-index s,(l,p), shown in the third column
from the right. For instance, the relative importance given by an environmental
group to the impact on an untouched ecosystem will depend not only on its
ecological status (quality) but also on its area (extension). Satisfaction typically
increases with extension, at least until some threshold is reached. If the spatial
pattern of quality is also relevant, then this should be incorporated in the definition
of quality by considering additional suitable attributes and indicators (this is,
however, one of the issues deserving of major research effort).
An aggregation of the values assumed by the sub-index over all the sites p to
which individual I is sensitive is denoted index s,(l) (third column from the left). lt
measures the instantaneous individual satisfaction.
By aggregating the index over the N individuals of the group, the group index s, is
obtained (second column from the left) which is a measure of the instantaneous
group satisfaction at time t, and takes into account the number of people and the
differences in their judgment within the group under consideration.
Finally (first column on the left), a suitable macro-scale temporal aggregation of
the group index values determines the objective, denoted by J, which is a measure of
the global group satisfaction. It is this magnitude that the corresponding group is, at
the very end, willing to maximize and that allows one to perform a preferential
ordering of different alternatives from the particular point of view of the impact
considered. In the following, the term 'evaluation index' is used synonymously with
objective.
It is worth noting that at first sight the above may seem to be an application of
Multi-Attribute Utility Theory (MAUT from Keeney and Raiffa, 1976). What is
proposed here, however, is quite different. In fact, it is intended to be applied to the

209
construction of each individual evaluation index and not to solve the overall decision
problem. The common point with MAUT lies in the fact that, any time an
aggregation is needed, a Value (or Utility) Function should be constructed. Indeed,
the objective, index, sub-index and quality as defined above would be called Value
(or Utility) Function in Keeney and Raiffa's terminology.
The above structure allows the construction of evaluation indices which are
suitable in support of an evaluation which fulfills the sustainability criterion. In fact,
sustainability implies that entitlements are in any case distributed in some sense
adequately towards the future. The scheme in Figure 1 considers an aggregation of
the measures of instantaneous satisfaction (index values) into the future to produce
an overall judgment of satisfaction (objective) about the impact considered. It is
exactly in this aggregation that a criterion about how to deal with future generations
is needed and must be explicitly stated. If strict sustainability is adopted, then the
future counts as the present, and therefore it makes sense to simply consider the
situation which can be forecast into a (reasonably) distant future.
Intra-generational equity, on the other hand, can be explicitly addressed in the
relative comparison of the objectives to be performed in the posterior evaluation of
the alternatives through negotiation supported by Multicriteria Analysis.
It can be noted, furthermore, that the evaluation indices available in the literature
can be derived from this same structure by adopting particularly simplified
aggregation rules at different levels. The most common case is that in which the
objective J is assumed to be coincident with quality c(t,P), of the 'expert' individual
t, for the selected site p', given the values of the (proxy) indicators at a particular
instant of time, generally not clearly specified.

3.4. Requirements tor evaluation indices

In addition to this conceptual structure, evaluation indices should satisfy a number of

requirements in order to fulfill their role. Only two of them are considered here (a
more detailed analysis is presented in Nardini, 1998 forthcoming):

1 - Communicability. There must be an intuitively meaningful representation of the

numerical value of an evaluation index; otherwise, it would be impossible for the
public to understand and manage the information presented in the Evaluation Matrix.
This representation can rely on visual, verbal, or any other type of media
(multimedia and virtual reality offer enormous potential in this sense). For instance,
to communicate the satisfaction for the use value of a water supply service, a
numerical description could be given for the price of water, and a verbal one for the
frequency and intensity with which water shortages occur. In addition, it is perhaps
not too fantastic to imagine that a virtual communication of the taste of the water
could be supplied. In any case, the representation adopted must be such so as to

210
allow the affected group to perform a consistent preferential ranking of the
alternatives from the point of view of that particular impact. If this is not the case,
the interest group will not be able to associate a satisfaction with the index values
and associated situations, and hence will not be in any condition to negotiate trade-
offs that can be defended by rational arguments, hence reducing its capability to
change decisions as desired.
Communicability calls for using only natural (or constructed) attributes
(according to the terminology of Keeney and Raiffa, 1976) because value judgments
can indeed only be attached to such attributes. Consider, for instance, the case of the
proxy attribute 'concentration of pollutants' which is often used to describe health
impacts indirectly. Although it is c1ear that lower concentrations are better, a non-
technical expert person is not able to evaluate how important a change in these
attributes is, when compared to a change in the attributes characterizing other
impacts, such as, for instance, an income increase. It is thus impossible to trade-off
impacts from a conscious, transparent and defensible base. Furthermore, it is very
common for aperson, when given ambiguous information, to exaggerate the
negative consequences. This in turn implies greater difficulties in negotiating. The
final decision maker also faces the same difficulty. If he/she is not an expert, hislher
judgment is very likely to be biased because a non-explicit 'cause-effect model'
would necessarily, but unconsciously, be adopted. Finally, even a sectoral expert,
when considering a situation described by a set of values for the relevant proxy
indicators, would actually think more or less intuitively of amental cause-effect
re\ationship binding those indicators to the final natural attributes. Only when the
conversion from proxy to natural is immediate and unambiguous would the problem
vanish.
The problem also exists when proxy indicators refer directly to standards (as is
typical with pollution issues). When concentrations are lower than the declared
standard, it is easy to conclude that there is no problem. However, when
concentrations exceed the standards, the intuitive interpretation is basica\1y one of
fear, which is counterproductive for negotiating purposes.
There is no way to avoid the explicit or implicit conversion of proxy attributes
into natural attributes. Often, however, empirical evidence or unequivocal
interpretation is lacking and hence the (hard) problem of converting proxy attributes
into natural ones has to be faced. This is demanded of the sectoral experts who
should thus substitute the cause-effect model. To increase transparency, consistency,
and to reduce diverging views among the experts, Value Function models of the
experts can be developed, as c1early explained in Beinat (1997). To increase
acceptance by the public of experts' verdicts, interest groups should be free to se1ect
their most trusted experts.
The lack of an objective assessment of the effects (attributes) and the consequent
reliance upon experts is, of course, a potential source of divergence. It is, however,

211
reasonable to expect that: (i) the experts will be very careful in producing their
verdicts as they belong to a 'scientific community' which gives them credibility; (ii)
they should not be reluctant to discuss their opinions with other experts hence
possibly smoothing out differences, or at least accepting an averaging of the
assessments. On the contrary, the divergence which can arise from intuitive non-
expert interpretation of the same factual, but proxy, information is more unlikely to
be remediable.

2 - Internal coherence. A particular degree of satisfaction is associated with each

numerical value of an evaluation index. This corresponds to a given 'situation'
described by the natural attributes. It is very likely that different situations will have
the same numerical value associated with them. This must imply that the affected
group considers them the same in terms of satisfaction. In addition, if situation A is
preferred to situation B, then the numerical value of the index must be greater for
situation A than for B. An index for which the above holds true for all possible
situations can be qualified as 'internally coherent'. Therefore, the mathematical
structure ofany index should always be that ofa Value (or Utility) Function (Keeney
and Raiffa, 1976; see also Beinat, 1997 for a review, and Beinat et al., 1994 for an
example of application), because it is this structure that, by definition, guarantees
internal coherence. It can be observed that many of the classical environmental
indices the literature (e.g. Brown et al., 1970; Smith, 1989) do not fulfill this
requirement, or at least it is not explicitly stated that they do so. Beinat (1997)
presents a complete review, including original contributions, of the assessment
techniques which can be used to build Value Functions based on questionnaires
presented to experts or stakeholders.

4. Examples (rom the Laja-Diguillin canal ca se study

Some examples are presented in the following sections. They refer to the Chilean
irrigation 'Laja Diguillin Canal' project (for a critical analysis see Nardini et al.,
1997; fortechnical issues and original references, see Nardini and Montoya, 1993;
Nardini and Montoya, 1995; and Castro, 1996). The project consists of the
construction of a new irrigation canal of 65 cmls of capacity that would derive water
from the river Laja (the main tributary of the Biobio river, VIII Region, Chile) and
draw it into the adjacent Itata basin to irrigate 65,000 ha.

4.1. An index for the 'Salto dei Laja'

The project would result in a diminution of the flow of the river Laja during
summertime, among other relevant impacts. In particular, the attractiveness of the
beautiful 'Salto deI Laja' waterfalls would thus be affected. Therefore, there would

212
be a loss in the recreational satisfaction (Use value) for the potential visitors. The
main natural attribute needed to characterize a situation to which it is possible to
assign recreational satisfaction is certainly the visual aspect of the falls. This depends
on the river flow, that is the variable which naturally relates the management of the
Laja water resource system to this recreational impact.
Flow reduction and recreational activities occur almost only during summertime.
Therefore, it is reasonable to consider as causal factor the summer-averaged flow Qt
depending on year t. This indicator is thus the result of a micro-scale temporal
aggregation. Hydrological magnitudes are intrinsically uncertain. In this example, a
simple approach has been adopted. It is assumed that the flow Qt is characterized by
a cyclo-stationary probability distribution (with aperiod of one year) due to the
natural annual meteorological cycle. It is thus possible to define the expected value
Q of Qt. With each given Q, a corresponding picture is associated (it is assumed, for
simplicity, but not far from reality, that only one view is relevant). The image
identifier I(Q) is the indicator actually used.
Let us now consider the conceptual structure of the index (with reference to
Figure 1). Quality ell,p)=e(l) measures, in this case, the psychological perception
that individual I has in relation to the visual aspect of the fall. The site dependence
(P) is dropped because there is just one Salto deI Laja. The time dependence is
dropped because it is assumed that this perception does not change from summer to
summer (the dependence on the image identifier, and thus on the picture, is active,
but omitted for simplicity, as in Figure 1). Extension is not relevant because the falls
are a local object the spatial extension of which does not change within the project
alternatives. Thus, the sub-index coincides with quality, and the index coincides with
the sub-index (no spatial aggregation is needed). As a simplification, it is assumed
that the number of visitors would not change significantly; only quality can change.
Therefore, in order to pass from the index s(l) to the group index s only an averaging
over the different types of persons belonging to the group "visitors" has to be carried
out. Finally, temporal aggregation to obtain the objeetive J is very simple if a strict
inter-generational criterion is adopted, according to which future benefits count as
much as present ones. Under this assumption it is J=s. Of course, it has to be
specified that J represents the recreational satisfaction which vi si tors would
experience for every year in the future.
The mathematical structure of this objeetive is very simple. For a generic
alternative characterized by an expected summer-averaged flow Q and image
identifier I(Q), quality e(l) is a scalar value function (VF) vA.Q). The group
aggregation necessary to obtain the group index s=v(Q) was carried out in the
simplest (approximated) way: instead of aggregating the vA.Q) VFs, the answers
obtained through a survey were averaged, and the VF was built directly on the basis
of these latter figures. Figure 2 shows the set of situations considered. They are
represented in terms of the natural attribute "visual aspect" which is unequivocally

213
and immediately bound to the expected summer river tlow Q, or image identifier
I(Q).

Image I'

Figure 2. The nine "situations" considered in building the VFs v(/{Q)).

The images, however, are not real pictures. They are the result of a computerized
manipulation of only one original picture.
Subjective preferences for a set of people were then assessed through a pilot
survey organized according to the "mid-value point" method for the VF assessment
(see, for instance, Beinat, (997). An approximated version was adopted because this
method is suitable for continuous indicators, while the images were available as a

214
finite number only. The VF v(Q) built on the basis ofthe averaged answers is shown
in Figure 3 as a function ofthe image identifier I(Q).
This very simple example shows that the general conceptual structure for the
evaluation indices (Figure I) provides a useful framework to systematically build an
index which fulfills the requirements. It does not imply, however, that the index is
complicated. On the contrary, the final product should and can be very intuitive.

11(1)

0.75

0.5

0.25

0*-+-+-+-+-+-+-~4
1 2 3 4 5 6 789
Image

Figure 3. VF v(l) buHt on the basis ofthe averaged answers.

4.2. Summary ofthe Impact Assessment study: the evaluation matrix

The construction of evaluation indices should be made part of any Impact

Assessment study and presented in a form suitable for negotiation purposes. This
form is the evaluation matrix. An example is given in Figure 4, again in relation to
the Laja-Diguillin Canal project. For each impact, the satisfaction component Y is
indicated in the second column, as weil as the affected group G. In the third column,
a summarized verbal explanation ofthe type ofrepresentation (natural or constructed
attributes) adopted is supplied. The notation AO, AI, etc. in the header for columns
four and five denotes the corresponding project alternatives (as many as desired),
where AO denotes the do-nothing alternative. In the cells of these columns the
representation of satisfaction is given both in intuitive terms and numerical terms
(value ofthe objectives).
This latter representation is a useful guide for quickly capturing the preferential
ordering of the alternatives with respect to each individual impact. This is the
information which can be utilized directly by Multicriteria Analysis techniques.
In this example only negative environmental impacts (associated with a reduction
in river flow) are considered for illustrative purposes. Of course, positive impacts
associated with the project also exist and should be considered in a real evaluation
where strengths and weaknesses (as previously required in Point a) are offset against
one another. Otherwise there would not be any problem in selecting the best
alternative (AO dominates AI).

215
 Image I' Image 16

,
,
,, Compom:nt :Y Explanation AO 1\1
Group :G
V: lle&l ~h js..lisrllC1ion in Long term Morbid ilY: Morbidil)':
eJl.peri-M!::irtg~ giYCTl StIlte'; hcalth statc a$ a low 0.9 mediu'l' 0.6
Comp l il:~"'ry ~ l. l ue) consequence of
,
J, fi.sl1 p.-ottin Growth: Gm ..",:
G : Community nearb~\'Cf uj,.
(2S0 propie) intalo:.e good mM;um-~ood

V: Monetary capital Ne. certain

(i nd i ~ atisf..:tion in using , , incomc for SO% 0.9 5004 0.6
services! Uw: \lalue: and rOT 03, 1'1 future
~peri'CI\Cini I. aiven slatc:
J, YEar
ComplemenLaI')' va.lut" from 5«\Irily)
(%of\!)e
G: flCL (90.000 has) polentialt ' ....

,,

0 0
Y: A ailabililY ora Visual aspecl
recreational s.c:rvice orS,llodel 1.0 0.4
I SatisrllCtion in u~inK 1C1'\o'~: U.. Laja falls as a
J, .... Iuel funelion oflhe
expected
G: Visilors of Saho dei Laja
SUml11<:f rivef
(40.000 da;ly prescncelyear)
flow

0
Y: En ironmentalgood
(s,liSrlKlion in oeJripericncin8 a givm aturalit)'
stile: ExisierKe atld Option "al~
of lhc OiXuillin tiver ccos}§(cm)
0.2 0.1

/0
J. C :En,,'ironMtl'ltalist$
Quantil)' ofl;re
(80.000 indi"idual5l <s'fUitativ ilY.
.. rj~hness of
species. ,
densily) "

, ,
,
I

r':;:.. ••:-, ....:.f

" .
~

*",,' ~"~:~)'- - .;:/J

'. ~

Figure 4. An example of an Evaluation Matrix (modified from Castro, 1996).

216
 The first objective J j measures the health satisfaction resulting from a reduction of
fish proteins in the diet of a community located near the Diguillin river. This impact
would be a consequence of the diminished productivity of the river ecosystem due to
a change in water quality and regime produced by the building of a water reservoir
(this reservoir would contribute to feeding the Laja-Diguillin Canal). Only two
natural attributes are considered as an example: morbidity and growth.
The second objective J 2 measures the Use value satisfaction received by the
members of the agricultural group 'Asociaci6n Canalistas dei Laja' (an association
representing the farmers using irrigation canals fed by the Laja river). It is indirectly
associated with the net income they would earn each year into the future (see Nardini
1998, and for a theoretical background, French, 1988), and expressed as a percentage
of the potential net income corresponding to the absence of a water supply deficit.
The project is in fact very likely to cause a water supply deficit. Notice that the
numerical values of the intuitive representation differ from those of the objective
representation (on the right). This is because the first values are those assumed by
the natural indicator d : 'yearly deterministic net income', while the latter ones refer
to the objective measuring the satisfaction, i.e. they are the values assumed by a
Value Function with argument d and characterized by decreasing marginal utilities.
Furthermore, the income percentages vary between the worst and best situations
among the alternatives, where the worst does not necessarily correspond to zero
income. On the contrary, the VF assumes, by design, a zero value corresponding to
the worst situation, and a maximum value (unity) corresponding to the best
alternative amongst those considered.
The third objective J3 corresponds to the impact on the Salto dei Laja already
analyzed.
The fourth objective J 4 measures the satisfaction associated with the Existence and
Option values that environmentalists assign to the impacted stretch of the Diguillin
river. It is a simplified version of the index developed by Senior (1996). The idea is
that natural attributes for the satisfaction component are the ecosystem 'naturality'
and 'quantity of life' expressed in terms of 'richness of species - equitativity - life
density'. The first one is represented through a picture which takes into account only
the physical-morphological aspect (in this example, a reduction in the summer river
flow). 'Quantity of life' is represented by a schematic drawing from which richness
of species, and equitativity and life density can be captured visually (the last two
attributes are defined inside each taxon).
This example is only illustrative, and much can be done to improve it. Moreover,
as al ready noted, multimedia and virtual reality can improve the intuitiveness of the
representations by orders of magnitude. In any case, it should not be forgotten that
the evaluation matrix must always be accompanied by a detailed report. This must
present the original factual information (proxy attributes), present the conceptual and
mathematical structure of the indices, prove the fulfillment of the requirements,

217
explain how the prediction of the values for the indicators has been performed,
carefully guide the interpretation. of the information presented by explaining, in
particular, all the assumptions made in aggregation steps.
It is perhaps reasonable to argue that the more rigorously one pursues the process
of trying 'to get reliable numbers', the more the inquiry process throws up
complications and sources of interpretative ambiguity. This risk should always be
kept in mind so as not to lose the main goal of the Multicriteria approach which is, I
suggest, the support it can provide in structuring the decision problem at hand. It
should be recognized, however, that although gaining insight into the construction
process raises new difficulties, only by facing such difficulties together with the
affected people can they get the sufficient understanding to actually trust the results
obtained.

5. Conclusions

Environmental decision-making, and land-use management in particular, are

characterized by interest conflicts amongst the affected groups. In order to make
successful decisions in the public interest, these conflicts should be addressed
explicitly through a negotiation approach. This should rely on a weil specified
administrative procedure in which the MCA-based evaluation would playa central
role. A basic step is trying to reach an agreed upon representation of the problem at
hand. In order to obtain this, the interest groups involved should be clearly
identified, and only the impacts to which well-specified values can be attached, i.e.
with which a satisfaction can be clearly associated, should be considered in the
evaluation matrix. Furthermore, this should contain a representation of the relevant
information which is easily and intuitively understandable by the players involved,
so as to be able to measure the satisfaction each group associates with the
corresponding impact. For this reason, specially conceived evaluation indices can be
constructed and their numerical values and intuitive meaning represented in the
evaluation matrix. Each evaluation index should fulfill, in particular, the
requirements of communicability and internal coherence. Furthermore, it should be
developed according to a conceptual structure of the type presented here so that
intensive and extensive dimensions, and spatial, individual and temporal variations
are taken into account. MCA techniques can then be used to support negotiation by
capturing the subjective preferences from the involved groups and exploring the
corresponding rankings of the alternatives; to analyze the relevance of uncertainty in
the assessments of the impacts and in the value judgments via sensitivity analysis;
and to analyze the effects of proposed mitigation and compensation measures. MCA
techniques, as al ready noted, can also greatly increase the degree oftransparency. All
this process should lead to the identification of socially acceptable alternatives.

218
 It is important to remember that the particular MCA approach outlined above is
intended to be specifically suitable for decision problems involving multiple players.
For problems characterized by a single decision maker, more refined MCA
techniques can prove to be more effective.
It might appear simpler to use the indicators directly in the evaluation matrix, or
even the attributes, rather than the evaluation indices. However, the use of evaluation
indices allows:

i) one to consider two different problems separately: a) the problem of measuring

the satisfaction of each interest group, in relation to each corresponding
impact; and b) the problem of evaluating the alternatives on the basis of the
social preferences of the groups involved, by finding an acceptable trade-off
through negotiation. The first task does involve subjective value judgments.
However these are, for a given impact, specific to the affected group itself and
are required only to allow a consistent preferential ordering of the alternatives
from that particular point of view. From this perspective, this task can be
considered essentially technical (nevertheless, it mayaiso involve sociologists
and psychologists). The second task requires ethical-moral judgments about the
relative importance ofthe impacts and inherently involves interest conflicts;
ii) one to summarize the relevant information;
iii) one to ensure that each interest group is able to rank the alternatives
consistently at least from the point of view of the impacts affecting it. This
cannot be taken for granted when a given component is described through
many attributes which may assurne different combinations of values depending
on the alternative considered, what is typical in environment related problems.
On the other hand, it is a must in order to have a chance to defend one's own
interests;
iv) the interest groups to acquire a deeper insight into their inner value structure,
thus stimulating a fruitful leaming process. The pure exercise of building an
evaluation index through a elose interaction among each interest group,
experts and analysts certainly constitutes an educational and insight-achieving
process that can strongly contribute to focusing on the relevant issues and, at
the very end, to making better decisions;
v) one to produce the quantitative information that is needed to apply the simplest
MCA technique (weighted sum). This is perhaps the only one that people can
understand intuitively and, therefore, accept. This position is equivalent to
assuming that there an overall Value Function exists with an additive structure.
This fact, as Keeney (1992) argues, is not far from reality, once having
identified the fundamental attributes, i.e. those describing the impacts to which
values are elearly attached;

219
vi) one to simplify the representation of the information by taking advantage of
the internal coherence of the indices: a given situation can be replaced by a
different one, possibly fictitious, which requires less information in order to be
displayed while beingjudged the same as the original one.

Alternatively, the following negative consequence can be envisaged if evaluation

indices are not used. Apriori, there is no rule which unequivocally establishes the
number of attributes to be adopted in order to characterize a given impact. An
increase in the number of attributes could of course change (improve) the degree of
refinement in the description of the impact (e.g. by increasing the number of
variables used to describe water quality). If the attributes have not been previously
aggregated into an evaluation index, however, different selections of the number of
attributes adopted to describe a given (environmental) impact, colild lead to different
rankings of the alternatives, but this would occur not because the relative importance
assigned to that impact is now higher, but just because each attribute in any case
receives a part ofthe final score (apart from the case when special decision rules are
adopted). This is likely to lead to arbitrary elements in ranking and, therefore, in
decision making.
Of course, this is mainly an idealized proposal which could serve to orient future
improvements in the legal framework and in the practice of land-use management
decision making. Interesting practical cases that rely on similar approaches, although
mainly focused on the participatory issue, have been described for instance in Renn
et al. (1993), Webler et al. (1995). Last but not least, it is interesting to note that the
decision making process which took place in the case of the Chilean Laja-Diguillin
Canal project in practice included the same features as outlined in the approach
proposed here. This occurred, however, without prior organization or support from
the govemment, thus wasting time, energy, money, and achieving a much lower
level of consensus and success than would otherwise have been the case.

Acknowledgments
This paper is the fruit of research project 93.91.15-1.1 of the Direcci6n de
Investigaci6n of the Universidad de Concepci6n (Chile) to whom I would like to
offer thanks. The research would have not been possible without the support of
Carmen Teresa Senior, Claudia Briones, Juan Carlos Castro and Heman Blanco,
whom I want to thank warmly.

220
References
Allet, EJ. (1986). Environmental Impact Assessment and Decision Analysis, Journal of
Operational Research Society, 37(9): 901-910.
Asian Deve10pment Bank (1988). Guidelines for Integrated Regional Economic-
Environmental Planning: A Review of Regional Environmental Development Studies in
Asia, Asian Development Bank Environment Paper 3 (I).
Baird, I.A., J.R. Ive (1989). Using the LUPLAN Land-use Planning Package to Implement
the Recreation Opportunity Spectrum Approach to Park Management Planning, Journal
of environmental management, 29:3.
Beinat, E. (1997). Value functions for environmental management, Kluwer, Dordrecht.
Beinat, E., P. Nijkamp, P. Rietveld (1994). Value functions for environmental poHutants: a
technique for enhancing the assessment of expert judgments, Environmental Monitoring
and Assessment, 30: 9-23.
Bisset, R. (1988). Developments in EIA methods, in Wathem, P. (Ed.) Environmental
Impact Assessment, Routledge.
Brown, R. M., N. I. McClelland, R. A. Deininger, R. G. Tozer (1970). A water quality
index. Do we dare?, Water Sewage Works, October: 339-343.
Castro, J. (1996). Multicriteria Analysis in the Environmental Field (CiviI Engineering
degree Thesis, Dep. of Civil Engineering, Universidad de Concepci6n, Concepci6n,
Chile; in Spanish).
Conacher, AJ. (1994). The integration of land-use planning and management with
environmental impact assessment: some Australian and Canadian perspectives. Impact
Assessment, 4: 347-373.
Dixon, 1., M. Hufschmidt (1986). Economic valuation techniques for the environment, The
Johns Hopkins University Press, Baltimore, Maryland.
Edwards, W. (1977). The use of multi attribute utility measurement for social decision-
making, in D.E. Bell, R.L. Keeney, H. RaitTa (Eds.) Conj1icting objectives in Decisions,
Wiley, Chi chester: 247-276.
French, S. (1988). Decision Theory. An Introduction to the Mathematics of Rationality, Ellis
Horwood, Chichester.
Glasson, J., Therivel, R., A. Chadwick (\994). Introduction to Environmental Impact
Assessment: principles and procedures, process, practice and prospects, UCL Press,
London.
Goodenough, R. (1992). The Use of Environmental Impact Assessment in the Management
of Open Space and Recreational Land in Califomia, The International journal of
environmental studies, 40:2-3.
Huang, S.L. (1989). Aland suitability approach for integrating impact assessment with
development planning, Impact Assessment Bulletin, 8:233-249.
Jacobs, M. (1992). The limits of Neoclassicism: Towards an Institutional Environmental
Economics, Centre for the Study of Environmental Change-Mimeo, University of
Lancaster, UK.
Janssen, R. (1992). Multiobjective Decision Support for Environmental Management,
K1uwer, Dordrecht.

221
Johansson, P.O. (1987). The Economic Theory and Measurement of Environmental Benefits.
Cambridge University Press, Cambridge.
Johnson, A.K.L., R.A. Cramb, and J.R. McAlpine (1994). Integrated Land Evaluation as an
Aid to Land-use Planning in Northem Australia, Journal of environmental management,
40.
Jones, M., Hope, C. and R. Hughes (1990). A Multi-attribute Value Model for the Study of
UK Energy Policy. J. Opl. Res. Soc., 41(10):919-929.
Keeney, R., and H. RaitTa (1976). Decisions with Multiple objectives: preferences and value
tradeofJs, Wiley, New York.
Keeney, R. (1992). Value Focused Thinking, Harvard University Press, Cambridge.
Long, W. (1994). Fury Greets Chile's Plan for a River, Los Angeles Times (Tuesday, Feb.15,
1994).
Munasinghe, M. (1993). Environmental Economics and Sustainable Development. World
Bank Environmental Paper nO.3.
Nardini, A. (1997a). Integrating Environmental Impact Assessment, Cost-Benefit Analysis,
Multicriteria Analysis: a step forward through the valuejocused identification of
impacts, Proc. ofthe IV InterAmerican Conference on the Environment, December 8-11,
1997, Universidad S. Bolivar, Caracas.
Nardini, A. (1997b). A proposal for integrating Environmental Impact Assessment, Cost
Benefit Analysis and Multicriteria Analysis in Decision Making, Project Appraisal,
September 1997.
Nardini, A. (1998). A systematic approach to build evaluation indices for environmental
decision-making with active public involvement, Journal of Environmental Management
(forthcoming).
Nardini, A, and D. Montoya (1995). Remarks on a min-max optimization technique for the
management of a single multiannual reservoir aimed at hydroelectric generation and
water supply, Water Resources Research, 31(4): 1129-1135.
Nardini, A., and D. Montoya (1993). A decisional model for the integrated management of
the lake Laja-river Laja system, with reference 10 the project Laja-Diguillin Canal, Serie
Monografias Cientificas, Centro EULA, Concepcion, Chile; in Spanish.
Nardini, A., H. Blanco, C. Senior (1997). Why Environmental Impact Assessment did not
work in the case of the Chilean project Laja-Diguillin Canal?, Environmental Impact
Assessment Review, 17(1).
Ortolano, L., A. Shepherd (1995). Environmental Impact Assessment, in Vanclay, F. D.A.
Bronstein (Eds.) Environmental and Social Impact Assessment, Wiley, London.
Pearce, D. (1993). Economic Values and the Natural Wor/d, Earthscan, London.
Petry, F. (1990). Who is afraid of choices? A proposal for Multicriteria Analysis as a tool
for Decision-making support in Development Planning, J. of International Development,
2(2).
Renn, 0., Webler, T., Rakel, H., DieneI, P., B. Johnson (1993). Public participation In
decision-making: A three-step procedure, Policy Analysis, 26: 189-214.
SagotT, M. (1988). The Economy ofthe Earth, Cambridge University Press.

222
Senior, C. (1996) A proposal of indices for the evaluation of the impact of large hydraulic
works on the ecosystems, PhD thesis, Centro Eula, Universidad de Conception,
Conception, Chile (in Spanish).
Smith, D.G. (1989). A new fonn of water quality index for the great lakes, Water scienee &
Teehnology, 21(2): 123-127.
Tamura, H., S. Fujita, H. Koi (1994). Decision Analysis for Impact Assessment and
consensus fonnation among conflicting multiple agents, The Scienee 0/ the Total
Environment, 153.
Therivel, R. E. Wilson, S., Thompson, D., Heaney, D. Pritchard (1992). Strategie
Environmental Assessment, Earthscan, London.
Tietenberg, T. (1992). Environmental and Natural Resourees Economics, HarperCollins
Publishers.
UNESCO (1994). Multicriteria Decision Analysis in Water Resources Management,
Bogardi Land H. Nachtnebel (Eds.), UNESCO, Paris.
Volta, G., A. Servida (1992). Environmental indicators and measurement scales, in A.
Colombo (Ed.), Environmental Impact Assessment, ECSC, EEC, EAEC, Brussels and
Luxembourg.
Von Winterfeldt, D. W. Edwards (1986). Decision analysis and behavioral research,
Cambridge University Press, Cambridge.
Warner, M. (1996). Integration of Regional Land-use Planning with Environmental Impact
Assessment: A Practical Land Suitability Assessment Approach. Impact Assessment,
14(2): 155-189.
Webler, T., H. Kastenholz, 0. Renn (1995). Public participation in impact assessment: a
Social Learning Perspective, Environmental Impact Assessment Review, 15(5): 443-463.
Yin, Y., J.T. Pierce (1993). Integrated Resource Assessment and Sustainable Land Use.
Environmental management, 17(3).

223
 Part IV
Spatial information in
land-use management
Multi-criteria and multi-objective decision making for
land allocation using GISI

J. Rooald Eastmao, Bong Jiang1 and James Toledano

The Clark Labs/or Cartographic Technology and Geographie Analysis
Clark University. Worcester. Massachusetts. USA

Abstract
Geographie Information Systems (GIS) are designed for the acquisition, management, analysis
and display of georefereneed data. As sueh they have elear implieations for informing the
spatial decision making process. Subsequently, recent developments in GIS software and in the
conceptual basis for decision making have led to dramatic improvements in the eapabilities of
GIS for resource allocation. These developments are reviewed through an examination of
procedures for Multi-Criteria and Multi-Objective land allocation in GIS. Special emphasis is
given to the problems of ineorporating subjective expertise in the context of partieipatory
decision making; the expression of uneertainty in establishing the relationship between
evidenee and the decision to be made; proeedures for the aggregation of evidenee in the
presence of varying degrees of tradeoff between criteria; and procedures for eonflict resolution
and conflict avoidanee in eases ofmultiple objective decision problems.

Keywords: Geographie Information Systems (GIS), land allocation, multi-eriteria analysis,

multi-objective analysis.

1. Introduction

Geographie Information Systems (GIS) are software systems designed for the
acquisition, management, analysis and display of geographically referenced data. Based
on an organization of map layers3 (e.g., a roads layer, a streams layer, an elevation
layer, and so on), GIS allows one to undertake a wide variety of logical and
mathematical analyses and then display the result in either a map or tabular form.
Figure 1, for example, shows the result of modeling vulnerability to drought as a
mathematical function of aseries of map layers, such as rairifall, market accessibility,
per cent of area cash cropped, and so on. The figure also shows one of the input layers

1 Sections of this chapter have been adapted from Eastman (1997).

2 Current address: Department ofGeography, Tbe University ofIowa, Iowa City, IA 52242, USA.
3 Tbere are two basic forms of GIS that encode layers in very different ways. In the first, known as
vector, the boundaries or courses of features are delineated by aseries points joined by Iines. Attributes
for these features are then stored in aseparate database. In the second, known as raster, the map region
is subdivided into a fine matrix of pixels (typically too fine to see as separate entities), for which
attributes are recorded as a matrix of numeric values. In this chapter, all illustrations are based on a
raster logic, and thus the term image will often be used interchangeably with that of layer.

227
(market accessibility) and the specific analytical tool that was used to create this
derivative mapping of vulnerability. Such tools have revolutionized the monitoring and
management of natural resources. Thus it is not surprising that in recent years
considerable interest has been focused on the use of GIS as adecision support system
(see Eastman et al. 1993a, 1994; Eastman 1997; Carver 1991; Janssen and Rietve\d
1990; Honea et al. 1991). In this chapter we focus on recent deve\opments in GIS für
the purposes of multi-criteria and multi-objective land allocation, and their

_ .. _---.... --
implementation in the IORiSI for Windows raster GIS software system.

I ~.r_~~~~

Figure 1. An example oj environmental modeling using a Geographie Injormation System. In this

example. the vulnerability map at the top was eomputed as a junetion oj the jaetors eontained in a
seriesoj elementary map layers. One ojthese. the market aeeessibility jaetor. is indieated at ,he
lower-right - a map eomputed as a eost distanee jaetor aeeountingjor the jrietion effeets on
transportation. The analytieal tool that was used to eompute the result is shown on the boffom left.

2. Definitions

Although there is a fairly extensive literature on decision making in the Management

Science, Operations Research and Regional Science fields (sometimes linked together
under the single name Decision Science), there is unfortunate\y a broadly divergent use
of terminology (e.g., see RosenthaI 1985; Belton 1990). Accordingly, we have adopted

228
the following set of operational definitions which we feel are in keeping with the thrust
of the Decision Science literature and which are expressive of the GIS decision making
context.
Decision: decision is a choice between alternatives. The alternatives may represent
different courses of action, different hypotheses about the character of a feature,
different classifications, and so on. We call this set of alternatives the decision frame.
Thus, for example, the decision frame for a zoning problem might be [commercial
residential industrial]. The decision frame, however, should be distinguished from the
individuals to which the decision is being applied. We call this the candidate set. Thus
extending the zoning example above, the set of all locations in the image that will be
zoned is the candidate set. Finally, adecision set is that set of all individuals that are
assigned a specific alternative from the decision frame. Thus, for example, all pixels
assigned to the residential zone constitute one decision set. Similarly, those belonging
to the commercial zone constitute another. Thus another defmition of adecision would
be to consider it the act of assigning an individual to adecision set. Alternatively, it can
be thought of as a choice of alternative characterizations for an individual.
Objective: an objective includes purposes and perspectives of adecision making". It
serves as the defining role as to how the decision is structured. Purposes define number
of alternatives to be considered and the nature of the decision set; perspectives
determines decision rule (to be discussed later): what criteria will be chosen, how they
are evaluated, and how the final decision is made. For example, we want to determine
areas suitable for timber harvesting, which is the purpose that indicates the decision set
to be selected; our perspective in the selection of suitable areas is to minimize the
impact of harvesting on recreational uses in the area. The choice of criteria to be used
and evaluated would be quite different from that of a group whose primary concern was
profit maximization. Therefore, objectives are concerned not only with a final selection
of the decision set, hut also with issues of motive and social perspective. Decision
makers' perspectives serve as a guiding force in the development of a specific decision
procedure and decision rule.
Criterion: a criterion is some basis for adecision that can be measured and
evaluated. It is the evidence upon which an individual can be assigned to adecision set.
Criteria can be oftwo kinds:factors and constraints, and can pertain either to attributes
of the individual or to an entire decision set. A factor is a criterion that enhances or
detracts from the suitability of a specific alternative for the activity under consideration.
It is therefore most commonly measured on a continuous scale. For example, a forestry
company may determine that the steeper the slope, the more costly it is to transport
wood. As a result, better areas for logging would be those on shallow slopes - the
shallower the better. Factors are also known as decision variables in the Linear

41t is important to note here that we are using a somewhat broader definition ofthe term objective than
is typically encountered in the mathematical programming literature. In mathematical programming,
the term objeclive is often used to refer to a specific objective function. However, any specific objective
function is shaped by a perspective and thus is consistent with the use ofthe term here.

229
 Programming (LP) literature (see Feiring 1986) and structural variables or control
variables in the Linear Goal Programming (LGP) literature (see Ignizio 1985). A
constraint serves to limit the alternatives under consideration. As such, it serves to
remove those individuals from the candidate set. A good example of a constraint would
be the exc1usion from development of areas designated as wildlife reserves. Another
might be the stipulation that no development may proceed on slopes exceeding a 30%
gradient. Thus constraints are expressed iri the form of a boolean (logical) map: areas
exc1uded from consideration being coded with O's and those open for consideration
being coded with l's.
Goals: in addition to criteria, we commonly express some characteristics that the
decision set (i.e., the set of chosen alternatives) must possess. For example, we might
require that the total area of lands selected for development be no less than 5000
hectares, or that the decision set consist of a single contiguous area. These requirements
are often called goals (Ignizio 1985) or targets (RosenthaI 1985) and form an important
part of adecision rule. However, they are also commonly known as constraints, which
can be confusing. Technically, they are constraints that apply to the solution rather than
the alternatives under consideration. Thus we have avoided this use of the term in favor
of the less ambiguous term goal.
Decision rule: the procedure by which criteria are selected and combined to arrive at
a particular evaluation, and by which evaluations are compared and acted upon, is
known as adecision rule. Adecision rule might be as simple as a threshold applied to a
single criterion (such as, all regions with slopes less than 35% will be zoned as suitable
for development) or it may be as complex as one involving the comparison of several
multi-criteria evaluations. Decision rules typically contain procedures for combining
criteria into a single composite index (known as an aggregation procedure) and a
statement of how alternatives are to be compared using this index. For example, we
might define a composite suitability map for agriculture based on an average (or
perhaps a weighted average) of the suitabilities of soils, slope, and distance from
market. The rule might further state the goal that the best 5000 hectares are to be
selected. This could be achieved by choosing that set of raster cells, totaling 5000
hectares, in which the sum of suitabilities is maximized. It could equally be achieved by
rank ordering the cells and taking enough of the highest ranked cells to produce a total
of 5000 hectares. The former might be called a choice function (known as an objective
fonction or performance index in the mathematical programming literature - see
Diamond and Wright 1989) while the latter might be called a choice heuristic.
Cboice Function: choice functions provide a mathematical means of comparing
alternatives. Since they involve some form of optimization (such as maximizing or
minimizing some measurable characteristic), they theoretically require that each
alternative be evaluated in turn. In cases where goals specify qualities that the aggregate
set of chosen entities must possess, the number of these alternative sets may be very

230
large5• However, in some instances, techniques do exist to limit the evaluation only to
likely alternatives. For example, the Simplex method in Linear Programming (see
Feiring 1986) begins with a feasible solution and then converges on an optimal solution
while specifically avoiding unnecessary evaluations.
Choice Heuristic: choice heuristics specify a procedure to be followed rather than a
function to be evaluated. In some cases, they will produce ~ identical result to a choice
function (such as the ranking example above), while in other cases they may simply
provide a elose approximation. Choice heuristics are commonly used because they are
often simpler to understand and easier to implement.
Evaluation: the actual process of applying the decision rule is called evaluation.
Multi-Criteria Evaluations: to meet a specific objective, it is frequently the case
that several criteria will need to be evaluated. Such procedures are called Multi-Criteria
Evaluations (Voogd 1983; Carver 1991, Zeleny 1982). Another term that is sometimes
encountered for this is modeling. However, this term is avoided here since the manner
in which the criteria are combined is very much influenced by the objective of the
decision.
Multi-Objective Evaluations: while many decisions we make are prompted by a
single objective, it also happens that we need to make decisions that satisfy several
objectives. These objectives may be complementary or conflicting in nature (Carver
1991: 322). With complementary objectives, or non-conflicting objectives, land areas
may satisfy more than one objective - i.e., an individual can belong to more than one
decision set without some inherent conflict. Desirable areas will thus be those which
serve these objectives together in some specified manner. For example, we might wish
to allocate a certain amount of land for combined recreation and wildlife preservation
uses. Optimal areas would thus be those that satisfy both of these objectives to the
maximum degree possible. With conflicting objectives, objectives compete for the
available land since it can be used for one or the other, but not both. For example, we
may need to resolve the problem of allocating land for timber harvesting and wildlife
preservation. Clearly the two cannot coexist. Exactly how they compete, and on what
basis one will win out over the other, will depend upon the nature of the decision rule
that is developed.
In cases of complementary objectives, multi-objective decisions can often be solved
through a hierarchical extension of the multi-criteria evaluation process. For example,
we might assign a weight to each of the objectives and use these along with the
suitability maps developed for each to combine them into a single suitability map
indicating the degree to which areas meet all of the objectives considered (see Voogd
1983). Such is the special concern of Multi-Attribute Utility Theory (Edwards and
Newrnan 1982) - a variant of Multi-Objective Evaluation with strong procedural

5 For example, there are over a trillion (1.73 x 10 13 to be exact) combinations of 100 items taken 10 at a
time. In the context ofraster GIS where pixels are the basic items, and where the number ofpixels may
easily be in the millions. the number of combinations quickly becomes impossibly large to calculate.

231
 similarities to Multi-Criteria Evaluation. However, with conflicting objectives the
procedure is more involved.
With conflicting objectives, it is sometimes possible to rank order the objectives and
reach a prioritized solution (Rosenthai 1985). In these cases, the needs of higher ranked
objectives are satisfied before those of lower ranked objectives are dealt with. However,
this is often not possible, and the most common solution to conflicting objectives is the
development of a compromise solution.
Undoubtedly the most commonly employed techniques for multi-objective
evaluation are those involving optimization of a choice function such as Mathematical
Programming (Fiering 1986) or Goal Programming (Ignizio 1985). In both the concern
is to develop an allocation of the land that maximizes or minimizes an objective
function subject to aseries of constraints (strictly, goals). Mathematical (e,g., Linear
Programming) and Goal Programming procedures were developed in the field of
Operations Research for the purpose of determining the levels of control variables to
apply in optimizing specific outputs. Thus they are strictly mathematical fitting
procedures, similar to regression techniques. However, by treating alternatives as
boolean variables (i.e. variables that can assume values of only 0 or 1 - the latter
signifying that the alternative will be chosen), these fitting procedures can be used quite
effectively for decision problems.
While a variety of attempts have been made to apply Mathematical and Goal
Programming procedures in GIS (e.g., Campbell et al. 1992; Carver 1991; Janssen and
Rietveld 1990; Wright et al. 1983), we do not consider them in this chapter for several
reasons. First, they are severely limited in the size of problem they can address - a size
that is prohibitively small in the context of raster GIS. Second, they require
considerable skill and experience on the part of the analyst in order to extend these
essentially single-objective evaluation procedures to the problems of multi-objective
resource allocation. Third, these procedures require specialized software that is
generally expensive and difficult to find. Finally, the procedure by which the solution is
determined is generally opaque to the end user. The mechanics of the Simplex
procedure, for example, are difficult (if not impossible) to follow for someone without
a background in matrix algebra. As a consequence, attention has been focused in this
chapter on the productive use of heuristic procedures, which can offer comparable
power with fewer limitations.

3. Multi-criteria evaluation in GIS

In GIS, Multi-Criteria Evaluation (MCE) is most commonly achieved by one of two

procedures. The first involves boolean overlay, whereby all criteria are reduced to
logical statements of suitability and then combined by means of one or more logical
operators such as intersection (AND) and union (OR). The second is known as weighted
linear combination (WLC) wherein continuous criteria (jactors) are standardized to a
common numeric range, and then combined by means of a weighted average. The result

232
is a continuous mapping of suitability that may then be masked by one or more Boolean
constraints to accommodate qualitative criteria, and finally thresholded to yield a final
decision.
While these two procedures are well established in GIS, they frequently lead to
different results. The reason is that they make very different statements about how
criteria should be evaluated. In the case of boolean evaluation, a very extreme form of
decision making is used. If the criteria are combined with a logical AND (the
intersection operator), a location must meet every criterion for it to be inc1uded in the
decision set. If only a single criterion fails to be met, the location will be excluded.
Such a procedure is essentially risk-averse, and selects locations based on the most
pessimistic strategy possible - a location succeeds in being chosen to the extent that its
worst quality passes the test. On the other hand, if a logicalOR (union) is used, the
opposite applies - a location will be inc1uded in the decision set if only a single criterion
passes the test. This is thus a very optimistic strategy, with (presumably) substantial risk
involved. Now compare these strategies with that represented by WLC (Voogd 1983:
120). With WLC, factors are combined by applying a weight to each followed by a
summation ofthe results to yield a suitability map:

where S= suitability
wi = weight offactor i
Xi = criterion score of factor i

This procedure is not unfamiliar in GIS and has a form very similar to the nature of a
regression equation. In cases where boolean constraints also apply, the procedure can be
modified by multiplying the suitability calculated from the factors by the product of the
constraints.

where c'=
]
criterion score (0/1) ofconstraintj
n= product

WLC a110ws factors to trade off their qualities. A very poor quality can be
compensated for by having a number of very strong qualities. This operator represents
neither an AND nor an OR - it lies somewhere in between (Bonissone and Decker
1986). It is neither risk averse nor risk taking. Rather, it lies in between the extremes of
complete pessimism and complete optimism.
For reasons that have largely to do with the ease with which these approaches can be
implemented, the boolean strategy dominates vector approaches to MCE, while
weighted linear combination dominates solutions in raster systems. However, both types
of GIS have the ability to implement each. Furthermore, neither is c1early better - they
simply represent two very different decision strategies. In a very recent development,
Eastman and Jiang (1996) have introduced to GIS an extension of Yager's (1988)
Ordered Weighted Average (OWA) for Multi-Criteria Evaluation which has the ability

233
to offer a complete spectrum of decision strategies along primary dimensions of the
degree of tradeoff involved, and the degree of risk aversion in the solution. This
operator has been introduced into the IDRISI for Windows system (Version 2), and is
discussed further below. But before criteria can be combined, they need to be
standardized so as to develop comparable criterion scores.

3.1. Criterion scores

Because of the different scales upon which criteria are measured, it is necessary that
factors be standardized6 before combination using the formulas above, and that they be
transformed, if necessary, such that all factor maps are positively correlated with
suitability7. Voogd (1983: 77-84) reviews a variety of procedures for standardization,
typically using the minimum and maximum values as scaling points. The simplest, and
most common, is a linear scaling such as:

Xi = (R,-R".in) / (R".ax-R".;n) * standardizedJange, where R = raw score

If it can be assumed that the resulting measure is an expression of membership in the

decision set, then the process of standardization is essentially identical to that of
fuzzification in fuzzy sets (Eastman and Jiang 1996). This procedure has the character
of transforming any scale to a comparable one measured according to a standardized
range (e.g., 0-1). In essence, the result expresses the degree to which an area or feature
belongs to the decision set (on the basis of a single factor). As such, then, standardized
criteria are simply fuzzy set membership statements, which can be based on a range of
membership functions 8, and not just the linear model (see Zadeh 1965; Buckley 1984;
Hall et al. 1992). In the lORISI for Windows system, the FUZZY modl!lle is provided
for the standardization of factors using several fuzzy set membership functions
(Sigmoidal, J-Shaped, Linear, and User-Defined). Importantly, the higher value of the
standardized scale must represent the case ofbeing more likely to belong to the decision
set.
A critical issue in the standardization (fuzzification) of factors is the choice of the
end points at which set membership reaches either 0.0 or 1.0 . Our research has
suggested that blindly using a linear scaling (or indeed any other scaling) between the
minimum and maximum values of the image is ill advised. In setting these critical
points for the set membership function, it is important to consider their inherent
meaning. For example, if we feel that industrial development should be placed as far

6 In using the term standardization, we have adopted the terminology of Voogd (1983), even though
this process should more properly be called normalization.
7 Thus, for example, if locations near to a road were more advantageous for industrial siting than those
far away, a distance map would need to be transformed into one expressing proximity.
8 In the Multi-Attribute Utility Theory (MAUT), such function are known as value functions (Edwards
and Newrnan 1982) and serve the same purpose.

234
away from a nature reserve as possible, it would be dangerous to implement this
without careful consideration. Taken literally, if the map were to cover a range of
perhaps 100 km from the reserve, then the farthest point away from the reserve would
be given a value of 1.0 . Using a linear function, then, a Iocation 5 km from the reserve
would have a standardized value of only 0.05 . And yet it may be that the primary issue
was noise and minor disturbance from Iocal citizens, for which a distance of only 5
kilometers would have been equally as good as being 100 km away. Thus the
standardized score should really have been 1.0. If an MCE were undertaken using the
blind linear scaling, locations in the range of a few tens of kilometers would have been
severely devalued when it fact they might have been quite good. In this case, the
recommended critical points for the scaling should have been 0 and 5 km. In developing
standardized factors using fuzzy set memberships careful consideration should be given
to the inherent meaning of the end points chosen.

3.2. Criterion weights

Although a variety of techniques exist for the development of weights for MCE (e.g.,
von Winterfeldt and Edwards 1986), one of the most promising is that of pairwise
comparisons developed by Saaty (1977) in the context of adecision making process
known as the Analytical Hierarchy Process (AHP). The first introduction of this
technique to a GIS application was that of Rao et al. (1991), although the procedure
was developed outside the GIS software using a variety of analytical resources. In the
procedure for Multi-Criteria Evaluation using a weighted linear combination outlined
above, it is desirable that the weights sum to one. In Saaty's technique, weights of this
nature can be derived by taking the principal eigenvector of a square reciprocal matrix
of pairwise comparisons between the criteria. The comparisons concern the relative
importance of the two criteria involved in determining suitability for the stated
objective. Since the complete pairwise comparison matrix contains multiple paths by
which the relative importance of criteria can be assessed, it is also possible to determine
the degree of consistency that has been used in developing the ratings, known as a
consistency ratio (eR). The CR indicates the probability that the matrix ratings were
randomly generated. Saaty (1977) indicates that matrices with CR ratings greater than
0.10 should be re-evaluated. In addition to the overall consistency ratio, it is also
possible to analyze the matrix to determine where the inconsistencies arise. In the
mRISI for Windows system, the WEIGHT module permits all of these operations to be
done - entry ofratings, calculation ofweights, and analysis ofinconsistencies.
We have found that the AHP method is very useful when direct evaluation of criteria
weights are difficult (especially when the number of criteria increases) in the context of
group decision making. Since pairwise comparison can become a focal point for group
consensus building, the method is very efficient in weight evaluation.

235
3.3. Evaluation

Once the criteria maps (factors and constraints) have been deve\oped, an evaluation
stage is undertaken to combine the information from the various factors and constraints.
Using the Weighted Linear Combination (WLC) logic outlined above, it is a fairly
simple matter to multiply each factor map (i.e., each raster cell within each map) by its
weight and then sum the results. Since the weights sum to one, the resulting suitability
map will have a range of values that matches that of the standardized factor maps that
were used. After all of the factors have been incorporated, the resulting suitability map
is then multiplied by each of the constraints in turn to "zero out" unsuitable areas.
In the IDRISI for Windows system, the MCE module offers two logics for the
aggregation of multiple criteria. The first is Weighted Linear Combination (WLC),
while the second is the more flexible Ordered Weighted Average (OWA) previously
mentioned.

3.3.1. MCE and weighted linear combination

The WLC option requires that one specify the number of constraints, their names, and
the weights to be applied to the factors. All factors are assumed to have been
standardized to a byte (0-255 integer) range -- a surrogate for the more traditional 0-1
real number range of fuzzy sets which facilitates computation. The output will be a
suitability map as implemented using the WLC approach outlined earlier, also with a
byte range (0-255).

3.3.1. MCE and the Ordered Weighted Average

In its use and implementation, the Ordered Weighted Average approach can be usefully
thought of as a variant of WLC. In reality, WLC is a variant of OWA. With OWA, two
sets of weights are employed. The first is factor weights, which are identical to those
sued in WLC and reflect the importance of factors. The second set of weights, order
weights, applies to the ascending rank order of weighted (using factor weights) factors
and controls the manner in which these factors are aggregated (Eastman and Jiang 1996;
Vager 1988). While factor weights apply to specific factors, order weights do not apply
to any specific factor, but to those factors as determined by their rank ordering after
applying factor weights.
By varying order weights, OW A allows for a broader range of aggregation
operators. For example, in adecision making that involves three factors, order weights
(I 0 0) will assign all the weight to the factor with the lowest score, thus an equivalent
of logical AND operation will be used, producing a risk averse (i.e., pessimistic)
solution. Order weights (0, 0, I), on the contrary, will produce an equivalent of logical
OR operation and a risk-taking (i.e., optimistic) solution. Order weights (0.33, 0.33,
0.33) assign equal weights to all three factors and thus produce an intermediate solution
(i.e. risk neutral), an equivalent of WLC. In the first two operations, factors cannot
compensate for one another and only the extreme factor score (minimum in the first

236
case and maximum in the second) will be selected; the third operation, however, allows
for full tradeoff between factors, so that the relatively poor factor can be compensated
for by the higher scores of good factors. The order weights in OW A are not restricted to
these three possibilities; instead, they can take on any combination of values that sum to
1.0. Doing so expands the decision rule to a triangular decision strategy space defined
by the dimensions of risk attitude and tradeoff (Figure 2).

~ decision strategy
o
u
~

'"
t:

ANDness 0
(risk -averse) (risk -taking)

Figure 1. Decision strategy ofOWA.

In Figure 2, risk9 attitude and tradeoff are measured by ANDness and TRADEOFF,
respectively (where ORness=l-ANDness) :

ANDness=(l/(n-l)) }J(n-i) Ward,r.)

n · L(Warder -; -11 n)2
TRADEOFF = I -
n-I

in which n is the total number of factors, i is the order of factors, and Word,,_; is the
weight for the factor of the i-th order. ANDness is determined by the skewness of the
order weights, while the degree of TRADEOFF is govemed by the relative distribution
of weights between the ranked factors. In the three examples given above, the ANDness
and TRADEOFF measures are land 0, 0.5 and I, and 0 and 0, respectively.
Note that be fore applying to factor scores, factor weights are adjusted according to
TRADEOFF, so that they only retain their full effects (i.e. unchanged) when full
tradeoff (TRADEOFF equals I) is allowed. As TRADEOFF decreases, factor weights
are gradually equalized, until they are adjusted to be equal when TRADEOFF decreases
to o. Adjustment of factor weights not only conforms with the understanding of tradeoff
in multi-criteria decision making (i.e. when no tradeoff is allowed, factors are not

9 The risk concept is similar to that in the assessment of utility functions, and the relationships among
risk averse, risk neutral, and risk taking positions are also similar (see Zeleny 1982). However, here,
risk applies to the decision rules in land allocation; risk levels, though implied, are not represented by
specified probability distributions.

237
weighted, i.e., they have equal weights), but also makes OWA consistent with the cases
oflogical AND and OR operations.
What is particularly interesting about the OWA module is this continuum of
aggregation procedures that it provides. At one extreme, criteria are considered as
necessary (hut not sufficient) conditions for inclusion in the decision set -- the hard
intersection (logical AND) operator provided by the minimum. As a consequence, the
criteria are similar to constraints. However, they are not strictly constraints since they
do not affect the candidate set. At the other extreme (the union operator), each criterion
is treated as sufficient on its own to support inclusion in the decision set without
modification by other factors. Thus the level of support for inclusion in the decision set
is equal to the maximum support offered by the criteria considered. The position of the
traditional weighted linear combination operator, half way between these extremes, is
thus not surprising. This operator considers criteria as neither necessary nor sufflcient --
strong support for inclusion in the decision set by one criterion can be equally balanced
by correspondingly low support by another. It thus offers perfect tradeoff.
When applying OWA in a real-world decision question, levels of tradeoff and risk
attitude, concepts that are easier to understand, will be first solicited from decision
makers. GIS analysts can then determine the corresponding order weights to be used.
The application of OWA procedure can be very useful in the exploration of a variety of
criteria aggregation operations, particularly in decisions involving multiple decision
makers, multiple perspectives, and multiple rationales.
Once a suitability map has been prepared, it is common to undertake a fmal step of
deciding upon which cells belong to the set that meets a particular land allocation area
goal (the decision set). For example, having developed a map of suitabiIity for
industrial development, we may then wish to determine which areas constitute the best
5000 hectares that may be allocated. Oddly, this is an area where most raster systems
have difficulty achieving an exact solution. One solution would be to use a choice
function where that set of cells is chosen which maximizes the sum of suitabilities.
However, the number of combinations that would need to be evaluated is prohibitive in
a raster GIS. As a result, we have chosen in the IORISI system to use a simple choice
heuristic - to rank order the cells (using the RANK module) and choose as many ofthe
highest ranks as will be required to meet the area target. In addition, this ranking
procedure can be done using a second image to resolve the ranks of ties. The ranked
map can then be reclassified to extract the highest ranks to meet the area goal.

4. MuIti-objective decision making in GIS

Multi-objective decisions are so common in environmental management that it is

surprising that specific tools to address them have not been more extensively developed
within GIS. Until the introduction of the MOLA procedure in the IORISI system, the
few examples one could find in the literature tended to concentrate on the use of

238
niathematical programming tools outside the GIS, or were restricted to cases of
complementary objectives that could be handled through extensions to MCE.

4.1. Complementary objectives

As indicated earlier, the case of complementary objectives can be dealt with quite
simply by means of a hierarchical extension of the multi-criteria evaluation process
(e.g., Carver 1991). Here a set of suitability maps, each derived in the context of a
specific objective, serve as the factors for a new evaluation in which the objectives are
themselves weighted and combined by linear summation. Since the logic which
underlies this is multiple use, it also makes sense to multiply the result by all constraints
associated with the component objectives.

4.2. Conflicting objectives

With conflicting objectives, land can be allocated to one objective but not more than
one (although hybrid models might combine complementary and conflicting
objectives). As was indicated earlier, one possible solution lies with a prioritization of
objectives (Rosenthal 1985). After the objectives have been rank ordered, the needs of
bigher ranked objectives are satisfied (through rank ordering of cells and
reclassification to meet areal goals) before those of lower ranked ones. This is done by
successively satisfying the needs of higher objectives and then removing (as a new
constraint) areas taken from consideration by all remaining objectives. However,
instances are rare where a prioritized solution makes sense. More often a compromise
solution is required.
Compromise solutions to the multi-objective problem have most commonly been
approached through the use of mathematical programming tools outside GIS (e.g.,
Diamond and Wright 1988; Janssen and Rietveld 1990; Campbell et af. 1992).
Mathematical programming solutions (such as linear or integer programming) can work
quite weIl in instances where only a small number of alternatives are being addressed.
However, in the case of raster GIS, the massive data sets involved will typically exceed
present-day computing power. In addition, the concepts and methodology of linear and
integer programming are not particularly approachable to a broad range of decision
makers. As a result, we have sought a solution to the problem of multi-objective land
allocation under conditions of conflicting objectives such that large raster datasets may
be handled using procedures that have an immediate intuitive appeal.
The procedure we have developed (Eastman et af. 1993b) is an extension of the
decision heuristic used for the allocation of land with single objective problems. This is
best illustrated by the diagram in Figure 3a. Each of the suitability maps may be
thought of as an axis in a multi-dimensional space. Here we consider only two

239
objectives for purposes of simple explanation. Howev(!r, any number of objectives can
be used.

conflict
.,.,
.,., .,.,
N non-conflict .,., ideal point for
N
region objective 2

...
N
11)
;>
Cl
oB
unsuitable
choices 1·~
"'= e
11)
:ö'
0
0 ideal point for
c
0 objective I
0
0 objective I 255 0 objective I 255

a) b)

Figure 3. Choice heuristics for multi-objective land al/oca/ion.

Every raster cell in the image can be located within this decision space according to
its suitability level on each of the objectives. To find the best x hectares of land for
Objective 1, we simply need to move adecision line down from the top (i.e., far right)
of the Objective 1 suitability axis until enough of the best raster cells are captured to
meet our area target. We can do the same with the Objective 2 suitability axis to capture
the best y hectares of land for it. As can be seen in Figure 3a, this partitions the decision
space into four regions: areas best for Objective 1 and not suitable for Objective 2, areas
best for Objective 2 and not suitable for Objective 1, areas not suitable for either, and
areas judged best for both: areas of conflict.
To resolve these areas of conflict, a simple partitioning of the affected cells is used.
As can be seen in Figure 3b, the decision space can also be partitioned into two further
regions: those closer to the ideal point for Objective 1 and those closer to that for
Objective 2. The ideal point represents the best possible case: a cell that is maximally
suited for one objective and minimally suited for anything else. To resolve the conflict
zone, the line that divides these two regions is overlaid onto it and cells are then
allocated to their closest ideal point. Since the conflict region will be divided between
the objectives, both objectives will be short on achieving their area goals. As a result,
the process will be repeated with the decision lines being lowered for both objectives to
gain more territory. The process ofresolving conflicts and lowering the decision lines is
iteratively repeated until the exact area targets are achieved.
It should be noted that a 45 degree line between a pair of objectives assurnes that
they are given equal weight in the resolution of conflicts. However, unequal weighting
can be given. Unequal weighting has the effect of changing the angle of this dividing

240
line. In fact, the tangent of that angle is equal to the ratio of the weights assigned to
those objectives.
lt should also be noted that just as it was necessary to standardize criteria for
multi-criteria evaluation, it is also required for multi-objective evaluation. The process
involves a matching of the histograms for the two suitability maps. In cases where the
distributions are normal, conversion to standard scores would seem appropriate.
However, in many cases the distributions are not normal. In these cases, the matching
of histograms is most easily achieved by a non-parametric technique known as
histogram equalization - a procedure whereby the scaling of values is varied to achieve
equal frequencies of cases over the entire range. This is a standard option in many raster
image processing systems such as IORISI. However, it is also the case that the ranked
suitability maps produced by the RANK module in IORISI are also histogram equalized
(i.e., a histogram of a rank map is uniform). This is fortuitous since the logic outlined in
Figure 3a is best achieved by reclassification of ranked suitability maps.
As a result of the above considerations, the module named MOLA (Multi-Objective
Land Allocation) was developed in IORISI to undertake the compromise solution to the
multi-objective problem. MOLA requires the names of the objectives and their relative
weights, the names of the ranked suitability maps for each, and the areas that should be
allocated to each. It then iteratively reclassifies the ranked suitability maps to perform a
first stage allocation, checks for conflicts, and then allocates conflicts based on a
minimum-distance-to-ideal-point rule using the weighted ranks.

5. Case study: multi-criteria/multi-objective land allocation

in the Kathmandu valley, Nepal

As an illustration of the procedures outlined above. we consider a case study conceming

the expansion of the carpet industry in the Kathmandu Valley area in Nepal, and its
effects on traditional agriculture land-use. The carpet industry is one of the largest and
most rapidly growing industries in Nepal. By the early 1990s, the industry employed
over 300,000 workers in approximately 5000 registered factories. Most of these
factories are sited in the Kathmandu Valley. In recent years, considerable concem has
been expressed about the expansion of carpet industry. While it is recognized that the
production of carpets represents a major economic resource, there is concem about
encroachment on agricultural land by urbanizationlindustrialization, hence a loss of a
very traditional lifestyle in the cultural heritage of Nepal. In addition, the carpet
industry is a source ofpollutants, and is a signit~cant user ofwater resources.
This paper demonstrates the techniques of decision making in resolving this conflict
in land use. The problem is to zone 1500 hectares of current agricultural land outside
the ring road of Kathmandu for further expansion of the carpet industry. In addition,
6000 hectares will be zoned for special protection of agriculture. The problem clearly
falls into the realm of multi-objectivel multi-criteria decision making. In this case we

241
have two objectives: to protect lands that are best suited for agriculture, and at the same
time find other lands that are best suited for the carpet industry. Since land can be
allocated to only one of these uses at any one time, the objectives must be viewed as
conflicting (i.e., they may potentially compete for the same lands). Furthermore, the
evaluation of each ofthese objectives requires multiple criteria.
In the illustration that folIows, a solution to the multi-objective/multi-criteria
problem is presented, the initial procedures of which were developed with a group of
Nepalese govemment officials as part of an advanced seminar in GISJO While the
scenarios discussed here were developed purely for the purpose of demonstrating the
techniques used and the results do not represent actual policy decisions, it is one that
incorporates substantial field work and the perspectives of knowledgeable decision
makers. The procedures follow a logic in which each of the two objectives are first
dealt with as separate multi-criteria evaluation problems. The result consists of two
separate suitability maps (one for each objective) which are then compared to arrive at a
single solution that balances the needs of the two competing objectives.

5.1. Solving the single objective multi-criteria evaluations

First we consider the multi-criteria evaluation of land for the carpet industry alone.

5.1.1. Establish;ng the criteria: lacton and constra;nts

Five factors were identified by the decision making group as being relevant to the siting
ofthe carpet industry: proximity to water (which is needed in carpet washing and wool
dying), proximity to roads (for the shipment of the heavy wool material and carpet
product), proximity to power (for electricity use), proximity to the market (for market
access), and slope gradient (to reduce the cost for construction of industry sites). The
proximity to water (see Figure 4a), proximity to roads, and proximity to power line
factors are modeled by taking a streams, roads, and power line layers, respectively, and
computing the distance of each location from those objects using a module named
DISTANCE. The proximity to market, unlike the other proximity images above, was
modeled as a cost distance surface. A cost distance surface is similar to a distance
surface, except that it accounts for the effects of friction in increasing effective costs.
This is computed with the COST module, using a friction image that was based on land
cover types. Frictions were lower for good quality roads, and higher for poor quality
roads. In addition, off road frictions were also accommodated. The slope gradients were
calculated from the digital elevation image using the SURFACE module.
The following constraints were also identified in developing criteria. First, forested
areas and existing urban areas were excluded from consideration; secondly, slope
gradients greater than 100% were excluded; and thirdly, carpet industry cannot be

lo,ne seminar was hosted by UNITAR at the International Center for Integrated Mountain
Development (lCIMOD) in Nepal. September 28-October 2. 1992.

242
developed within the ring road surrounding Kathmandu. The first and the third
constraints have to do with government regulations, and the second is related to
construction cost. All three constraints are expressed by a boolean image created by
reclassifying the area, using the RECLASS module in lORISI, so that the constrained
areas have values of O's, otherwise values of l's.

5.1.2. Standardizing thefactorsfor industrial allocation

Although the constraints are directly useable in boolean form, the factors need to be
standardized before then can be aggregated in the multi-criteria process. This was
accomplished with the FUZZY module in lORISI, resulting in a set of map layers
which express fuzzy membership in the decision set. In this case we chose to express
fuzzy set membership on a byte scale from 0-255 in order to be compatible with the
MCE module used in the later stage of aggregation. All five factor maps were
standardized using a sigmoidal (s-shaped) monotonically decreasing fuzzy membership
function with two control points: a value (control point 1) at which fuzzy set
membership begins to decrease from 255, and a value (control point 2) at which fuzzy
set membership reaches O. Between the two control points, fuzzy set membership
decreases progressively from 255 to 0 in the manner of a sigmoidal (s-shaped) curve.
Table 1 contains control points used in the standardization of all five factors. To
illustrate the meaning ofthese control points, consider the case of distance to water. The
two control points, 10 and 700, indicate that an area less than 10 meters from water is
the most suitable for carpet industry allocation; if the distance is farther than 700
meters, an area is no longer suitable. Between 10 and 700 meters, as distance increases,
suitability decreases. Figure 4b illustrates the result. Figure 5 illustrates the standardized
results of all five factors and the constraints (collected together as a single image) for
the carpet industry allocation.

Tllble 1. Contro! points used in thefuzzy scaling ofthefivefactors.

Factor Control Point 1 (fuzzy Control Point 2 (fuzzy
membership 1; scaled to 255) membership 0)
distance to water (m) 10 700
distance to road (m) 30 2000
distance to power line (m) 50 5000
cost distance to market (m) 1000 50000
slope gradient (%) 2 30

243
 Figure 4. This figure iIIustrates the standardization offactors. Figure 4a (left) illustrates the distance
surface which cQntains the distance of all/ocations to the nearest stream. In this case, brighter grays
indicate farther distances. Figure 4b (right) shows the resu/t of app/ying the /uzzy set membership
/unction to create a standardized set offactor scores.

5.1.3. Establishing the Jactor weights Jor industrial al/ocation

The next stage was to establish a set of weights for each of the factors. In the nature of a
focus group, the GIS analyst worked with the decision makers as a group to fill out a
pairwise comparison matrix (see Table 2), according to the rating scale in Table 3.

Ttlble 2. Pairwise comparison matrix, Kathmandu carpet industry: rating ofthe row factor relative to
the co/umn factor.
Water Proximity to Road Market Slope
Proximity Power Line Proximity Proximity Gradient
Water Proximity I
Proximity to Power Line 1/8 I
Road Proximity 1/3 7
Market Proximity . 1/5 5 1/2 I
Slope Gradient 1/8 1/3 1/7 1/7

Tilble 3. Rating scale of AHP (1/9 = ex;tremely less imporfant, 9 = ex;tremely more importanf).
1/9 In 1/5 113 I 3 5 7 9
exttemely very strongly modennely cqually moderately strongly very extremely
strongly strongly

Each decision maker was asked in turn to estimate a rating and then to indicate why
he or she feit that way. The group would then be asked if they agreed. Further
discussion would ensue, often with suggestions for different ratings. Ultimately, if
another person made a strong case for a different rating that seemed to have broad
support, the original person who provided the rating would be asked if they were
willing tO 'change (the final decislon would in fact rest with the original rater). It has
been found through repeated experimentation with this technique that the only cases
where strong disagreement arises are cases in which a new variable is eventually
identified as needing to be incorporated.

244
 ~1 Factors

Constraints

Figure 5. Al,>ove left to right and top to bottom: proximity to water. road proximity. slope gradient.
proximity to power. and proximity to market jactors. Also above: the ring road. landcover type and
slope constraint maps overlaid as a single Boolean image. Darker areas are those that are more
suitable to the objective inquestion.

This is one of the special virtues of the pairwise comparison technique - it is very
effective in uncovering overlooked criteria. A second virtue is that it is very effective as
a tool in achieving consensus. The WEIGHT procedure contains a facility to measure

245
the degree of consistency in the ratings, and provides a direct indication of where
inconsistencies occur as well as their nature. Using this information, it is possible for
the group to develop consistent consensus weights. The following weights are
calculated from Table 2: proximity to water: 0.51; proximity to power: 0.05; proximity
to road: 0.25; proximity to market: 0.16; and slope gradient: 0.03. The consistency ratio
ofthe matrix is 0.08, within the acceptable range.

5.1.4. Undertaking the multi-criteria evaluationfor industrial allocation

Once the weights were established, the module MCE (for Multi-Criteria Evaluation) in
IDRISI was used to combine the factors and constraints. We illustrate here the two
aggregation operators provided by MCE for comparison: Weighted Linear Combination
(WLC) and Ordered Weighted Average (OWA).
WLC represents the most typical way this procedure is handled in GIS, and results in
an aggregation with full tradeoff and intermediate risk. For example, a location with
mostly high factor scores will easily compensate for a single poor factor. However, this
compensation implies risk since there exists a factor with a low score.
As previously described, OWA offers a whole spectrum of aggregation procedures,
varying in terms of the degree of tradeoff and risk aversion. For comparative purposes,
we have chosen to illustrate adecision strategy with moderate risk aversion and low
tradeoff. In this case we have used order weights of: 0.6 0.25 0.125 0.025 O. The result
is an aggregation that is not as strict as the traditional fuzzy set intersection operator
(the minimum operator), but which is dominated by the lowest scoring factor. Thus it is
quite conservative (moderate risk aversion) in that it gives strong weight to the worst
quality, but does enter some consideration of higher scoring factors.
In both cases, the MCE module sums the weighted factors (although in the case of
OWA, two sets ofweights apply). Since the weights sum to 1.0, the resulting suitability
map will be scaled identically to the factors (i.e., 0-255).
Figure 6 shows the result of applying the MCE module to the criteria for the carpet
industry. Figure 6a indicates the result of applying the WLC decision strategy whiIe
Figure 6b illustrates the effect of applying OWA with the order weights indicated
previously. Otherwise they both employ the same importance weights. Note that
brighter areas are those that are evaluated to have a higher suitabiIity. The effects ofthe
constraints can be clearly seen since they force the suitability to be zero in affected
areas. It is interesting to note the effect of the fairly conservative strategy employed
with the OWA. The OWA strategy used was one which gave quite strong influence to
the limiting factor. This can be seen quite clearly as one moves away from the rivers.
There is a sudden drop-off in suitability as the fuzzy set membership function for
proximity to water drops to zero (notice the sharp contrast in gray level compared to the
WLC result). If the strategy had been one that employed the strict minimum
intersection offuzzy sets (order weights of 1 0000), the areas outside the water factor
zone of influence would have dropped to a suitability of zero (since this would have
been the limiting factor). However, it can be seen in the fact that these areas away from

246
water are not totally black on the image that the OW A strategy used has allowed some
degree of consideration of other factors.

Figure 6. The result oJ app/ying the MCE module to aggregate the criteria Jor the carpet industry.
Figure 6a (Ieft) shows the resu/t oJ app/ying a weight linear combination (WLC), while Figure 6b
(right) illustrates the effect oJ using an ordered weighted average (OWA) with moderate risk aversion
and low tradeojf.

5.1.5. Multi-criteria evaluation for agriculturalland al/ocation

For agriculture land protection, local decision makers identified three of the same
factors: proximity to water (for irrigation), proximity to market, and slope gradient; and
a fourth factor, soil capability. Standardization of the market and slope factors was
handled identically to that for the carpet industry. However, the proximity to water
factor was handled differently. Since agricultural land receives natural precipitation,
distance to water body is not as restrictive for agriculture as for industry. As a
consequence, the evaluation of this factor is greatly relaxed. A J-shaped fuzzy set
membership function was used, with the following two control points: 10 meters (the
point at which fuzzy set membership begins to fall from 1.0) and 500 meters (the point
at which fuzzy set membership has fallen to 0.5).
For the soil capability factor, only a categorical mapping of capability was available.
As a consequence, fuzzy set membership values were assigned to each category
directly.
Weights for factors were again derived by using the WEIGHT module to evaluate
pairwise comparison ratings provided by the participants: proximity to water: 0.45;
proximity to market: 0.05; slope gradient: 0.10, and soil capability: 0.40. Only a WLC
strategy was considered for agriculture. Figure 7 i1lustrates the result, again with
brighter areas indicating higher suitabilities.

5.1. Solving the multi-objective land al/ocation problem

Once the multi-criteria suitability maps have been created for each objective, the
multi-objective decision problem can be approached. Two evaluations of
multi-objective land allocation are conducted: one with the two WLC strategies for
evaluating suitability for industry and agriculture, and the other with the OWA strategy

247
for the evaluation of suitability for industry and the WLC evaluation for agriculture
suitability.
For each of the two land allocation evaluations, the first step was to rank order the
pixels in each ofthe two suitability images (using the RANK module in lORISI). This
prepares the data for use with the multi-objective land allocation (MO LA) procedure
and has the additional effect of standardizing the suitability maps using a
non-parametric histogram equalization technique. Ranks are developed in descending
order (i.e., the best rank was 1). In both cases tied ranks were resolved by examining
the other suitability map and ranking in reverse order to the suitability on that map. This
preserves the basic logic of the uncorrelated ideal points for conflicting objectives that
is used in the resolution of conflicts.

Figure 7. The result oJ applying a weighted linear combination (WLC) strategy to the criteria Jor
awiculture usin~ the MCE module.

The second step is to submit the ranked suitability maps to the MOLA procedure.
MOLA requires the names of the objectives, the relative weight to assign to each
(which are equal in both MOLA cases), and the area to be allocated to each land use
type. The module then undertakes the iterative procedure of allocating the best ranked
pixels to each objective and then resolving conflicts on the basis of marginal rank
suitability as described earlier. This procedure is repeated iteratively until the exact area
goals are met for each. Figure 8 shows the result of the first evaluation (where both the
carpet industry and agriculture criteria were aggregated using WLC) while Figure 9
shows the result of the evaluation that used the OWA multi-criteria strategy for the
carpet industry and the WLC strategy for agriculture ll . In both cases, equal weight was
given to each objective (this can be varied) and the iteration process was stopped when
the result was within 10 hectares of both area goals. The MO LA module required only
five and six iterations, respectively, to generate these results.
lt can be observed that the two MO LA evaluations produced quite different results,
particularly with respect to the carpet industry. The second evaluation (Figure 9) is the

11 Note that the allocations for the two industries are shown here separately for purposes of c1arity.
However, in normal operation, these two results are presented as two categories on a single map, each
displayed in a different color. Since the solution is one designed for conflicting objectives. the two
allocations do not overlap.

248
more conservative with regards to the carpet industry. As can be seen, this had the
effect of removing from consideration many of the areas farther out from Kathmandu
because of the greater effect of limiting factors .

Figure 8. Land al/ocation based on the application ofthe MOLA multi-objecrive procedure using the
WLC multi-criteria strategy. Figure 80 (feft) shows the best 1500 ha al/ocation for the carpet industry
while Figure 8b (right) shows the best 6000 ha al/ocation for agricufture. Because the procedure
assumes conjlicting objectives, the al/ocations do not overlap.

Figure 9. Land al/ocation based on the application of the MOLA multi-objective procedure using the
OWA mufti-criteria strategy for the carpet industry and the WLC multi-criteria strategy far agriculture.
Figure 9a (feft) shows the best 1500 ha al/ocation for the carpet industry while Figure 9b (right) shows
the best 6000 ha al/ocationfor agricufture. Because the procedure assurnes conjlicting objectives, the
al/ocations do not overfap.

6. Conclusion

This chapter has outlined and illustrated multi-criterialmulti-objective decision making

procedures in GIS, using the special suite of decision support tools in the IDRISI
software system. These tools allow exploration of a variety of rationales and
perspectives in suitability evaluation and land allocation, thus can accommodate a
variety of decision strategies in resource management. However, further research needs
to be conducted. For example, the land allocation procedure discussed in this chapter
(MOLA) is a heuristic approach that has the advantages of being easily understood, and
capable of application to massive allocation problems (for example, the case study here

249
involved the evaluation of almost half a milIion entities). However, while the ranking
stage produces an evaluation that is global in scope, the conflict resolution stage is a
local solution. As a consequence, the comparative strengths of this approach compared
to linear programming needs further assessment. In a preliminary study, Crema (1996)
found the two approaches to produce remarkably similar results (as evaluated by the
sum of suitabilities which the linear programming comparison maximized). However, a
full comparison needs to be made which accommodates the differences between the
approaches on the issue of standardization. Further issues that need investigation
include: alternative procedures for the derivation of factor weights; further exploration
of criteria standardization approaches; the issue of contiguity in resource allocation;
trade-off in multi-objective decision problems; and finally, the incorporation of
uncertainty (database and decision rule uncertainty) and sensitivity analysis. Regardless,
it is clear that current tools offer significant advantages for the decision maker in the
allocation of land resources.

References
Belton, V. (1990) Multiple Criteria Decision Analysis - Practically the Only Way to Choose, in
L.C. Hendry, R. Eglese (Eds.) Operational Research Tutorial Paper 1990, Operational
Research Society, Binninghanm.
Bonissone, P.P., K. Decker (1986) Selecting uncertainty calculi and granularity: an experiment
in trading-off precision and complexity, in L.N.Kanal, J.F. Lemmer (Eds.) Uncertainty in
ArtificialIntelligence, North-Holland-Elsevier, Amsterdam.
Buckley, J.J. (1984) The Multiple Judge, Multiple Criteria Ranking Problem: A Fuzzy Set
Approach. Fuzzy Set and Systems, 13: 25-37.
Campbell, J.C., J. Radke, J.T. Gless, R.M. Wirtshafter (1992) An Application of Linear
Programming and Geographie Infonnation Systems: Cropland Allocation in Antigua,
Environment and Planning A, 24: 535-549.
Carver, S.J. (1991) Integrating Multi-Criteria Evaluation with Geographical Infonnation
Systems, International JournalofGeographical Information Systems, 5 (3): 321-339.
Crema, S.C. (1996) A comparison between linear programming and a choice heuristic approach
to multi-objective decision making using GIS, GISIL/S '96 Proceedings, November 19-21,
1996, Denver, Colorado, 954-963.
Diamond, J.T., J.R. Wright (1988) Design of an Integrated Spatial Infonnation System for
Multiobjective Land-Use Planning, Environment and Planning B: Planning and Design, 15:
205-214.
Diamond, J. T., J.R. Wright (1989) Efficient Land Allocation, Journal of Urban Planning and
Development, 2: 81-96.
Eastman, J.R. (1997) IDRISI for Windows: User's Guide. Version 2.0, Clark University,
Graduate School of Geography, Worcester, MA.
Eastman, J.R., H. Jiang (1996) Fuzzy Measures in Multi-Criteria Evaluation, Proceedings,
Second International Symposium on Spatial Accuracy Assessment in Natural Resources and
Environmental Studies, May 21-23, Fort Collins, Colorado, 527-534.
Eastman, J.R., W. Jin, P.A.K. Kyem, J. Toledano (1993a) GIS and Decision Making,
Explorations in Geographie Infonnation System Technology, 4, UNITAR, Geneva.

250
Eastman, J.R., P.A.K. Kyem. 1. Toledano (1993b) A Procedure for Multi-Objective Decision
Making in GIS Under Conditions of Competing Objectives, Proceedings. EGIS'93,
438-447.
Eastman, J.R., W. Jin, P.A.K. Kyem, 1. Toledano (1994) Raster Procedures for Multi-Criteria /
Multi-Objective Decisions, Photogrammetric Engineering and Remote Sensing.
Edwards, W., 1.R. Newman (1982) Multiattribute Evaluation, Quantitative Applications in the
Social Sciences, 26, Sage, Newbury.
Feiring, B.R. (1986) Linear Programming: An Introduction, Quantitative Applications in the
Social Sciences, Vol. 60, Sage, London.
Hall, G.B., F. Wang and Subaryono (1992) Comparison of boolean and fuzzy classification
methods in land suitability analysis by using geographical information systems,
Environment and Planning A, 24: 497-516.
Honea. R.B., K.A. Hake, R.C. Durfee, R.C. (1991) Incorporating GISs into Decision Support
Systems: Where Have We Come From and Where 00 We Need to Go?, in M. Heit, A.
Shortreid (Eds.) GIS Applications in Natural Resources, GIS World Inc., Fort Collins,
Colorado.
Ignizio, lP. (1985) Introduction to Linear Goal Programming, Quantitative Applications in the
Social Sciences, 56, Sage, London.
Janssen, R., P. Rietveld (1990) Multicriteria Analysis and Geographical Information Systems:
An Application to Agricultural Land Use in the Netherlands, in HJ. Scholten, J.C.H.
StillweIl (Eds.), Geographical Information Systems for Urban and Regional Planning,
KIuwer, Dordrecht, 129-139.
Rao, M., S.V.C. Sastry, P.D. Yadar, K. Kharod, S.K. Pathan, P.S. Dhinwa, K.L. Majumdar, D.
Sampat Kumar, V.N. Patkar, V.N., V.K. Phatak (1991) A Weighted Index Modelfor Urban
Suitability Assessment - A GIS Approach, Bombay Metropolitan Regional Development
Authority, Bombay.
RosenthaI, R.E. (1985) Concepts, Theory anL Techniques: Principals of Multiobjective
Optimization, Decision Sciences, 16(2): 133-152.
Saaty, T.L. (1977) A Scaling Method for Priorities in Hierarchical Structures, J. Math.
Psychology, 15,234-281.
von Winterfeldt, 0., W. Edwards (1986) Decision Analysis and Behavioral Research,
Cambridge University Press, Cambridge.
Voogd, H. (1983) Multicriteria Evaluationfor Urban and Regional Planning. Pion, London.
Wright, J., C. ReVelle, 1. Cohon (1983) A Multiobjective Integer Programming Model for the
Land Acquisition Problem, Regional Science and Urban Economics, 13: 31-53.
Vager, R. (1988) On Ordered Weighted Averaging Aggregation Operators in Multicriteria
Decision Making, IEEE Transactions on Systems, Man, and Cybernetics, 8( I): 183-190.
Zadeh, L.A. (1965) Fuzzy Sets, Information and Control, 8: 338-353.
Zeleny, M. (1982) Multiple Criteria Decision Making, McGraw-Hill, New York.

251
Map transformation and aggregation methods for
spatial decision support

Ron Janssen and Marjan van Herwijnen

Institute for Environmental Studies
Vrije Universiteit, Amsterdam, The Netherlands

Abstract
The performance of policy alternatives for environmental management is often represented in
maps. These maps are used to compare altematives and ultimately to select the one preferred.
Since most people use reference maps in everyday life, such as road maps or town maps, they
are familiar with them and consider them as an appropriate decision support instrument. In
practice effective use of maps is a difficult task for many people. Spatial evaluation methods are
designed to support this task. Spatial evaluation methods help the decision maker by structuring
and simplifying the map representations ofthe alternatives. This can be done by transformation
or aggregation ofthe information presented in a map. In transformation, maps are transformed
or combined in such a way that only the information relevant to the decision is presented. For
aggregation, indices are defined that represent certain aggregate features of the map that are
relevant to the decision. Spatial evaluation methods presented in this chapter are developed to
support the comparison of mapped policy alternatives for the Green Heart of the Netherlands.
Seven examples ofpolicy questions are used to illustrate the use of spatial evaluation methods.
It is shown that the use ofthese methods increases understanding ofthe differepces between the
alternatives. The spatial evaluation methods used do not provide the decision maker with a
single answer, such as which alternative is the best, but result in aseries of answers depending
on the specific questions asked. All methods are included in a small software package that
invites decision makers to use the methods to experiment with their problem.

Keywords: Nature management, map use, spatial evaluation.

1. Spatial evaluation

In environmental management the performance of policy alternatives is often

represented in maps. Tbe decision maker is asked to use these maps to compare
alternatives and ultimately to select the one preferred. Since most people use maps
daily, they are familiar with maps and expect them to represent reality. Decision makers
are therefore happy to use maps to support their decisions (see for example Kohsiek et
al. 1991). In practice, effective use of maps is a difticult task for many people, one
which becomes more difticult if the information density of the maps increases and the
direct link with reality decreases (Muehrcke and Muehrcke 1992). Tbematic maps used
to support environmental management usually involve a large number of variables
linked to a large number of spatial units. A major part of the information included in
these maps is usually generated by simulation models (e.g., pollution dispersion or

253
deposition models). This part of the information only represents the real world under
the specific assumptions used by the models.
It is known that people are able to process only a limited number of stimuli at one
time (Miller 1956), and that the interpretation of images is strongly dependent on map
design (Bertin 1981). Comparison of map representations of alternatives is therefore an
extremely difficult task, one which becomes exponentially more difficult if more than
two maps need to be compared at the same time.
Spatial evaluation methods are designed to support this task. They help the decision
maker by structuring and simplifying the map representations of the alternatives. This
can be done by using two main types of approaches: map transformation or map
aggregation. In map transformation, maps, or map attributes, are combined in such a
way that only the information relevant to the decision is presented. In map aggregation,
indices are defined which represent certain aggregate map features that are relevant to
the decision. Spatial evaluation methods can be considered as a subclass of the much
larger class of evaluation methods, such as multicriteria analysis, but they could also be
considered as a specific type of spatial analysis. A large variety of methods for spatial
analysis can be found in the literature (Bartlett 1975, Campbell 1991, Cressie 1993,
Haining 1994, Heamshaw and Unwin 1994, Openshaw 1991, Unwin 1981, Upton et al.
1989). The same can be said for evaluation methods (Beinat 1997, Janssen 1992,
Keeney 1992). Examples of the use of multicriteria methods to support the evaluation
of spatial problems are more rare, but can also be found in the literature (Carver 1991,
Herwijnen et al. 1993, Janssen and Rietveld 1990, Pereira and Duckstein 1993).
However, most ofthese studies do not involve a comparison of alternatives presented in
maps. Some of these studies involve the selection of an alternative, such as a location,
within one map; in other studies spatial information is aggregated before evaluation,
ignoring the spatial pattern of the performance of alternatives.
The spatial evaluation methods presented in this chapter have been developed to
support the comparison of mapped policy alternatives for the Green Heart of the
Netherlands (see also Herwijnen et al. 1997). The paper starts with abrief description
of the decision support system used to produce these maps and then illustrates the
policy objective and the policy alternatives relevant to the decision framework under
analysis. The bulk of the paper is devoted to the illustration of the use of spatial
evaluation methods. For this purpose, seven policy questions are selected and spatial
evaluation methods are used to provide an answer to them. The policy questions are:
1. What is the relative performance of the alternatives?
2. Where is one alternative better than another?
3. How does the relative performance change if comparison is linked to quality
thresholds?
4. How does the relative performance change if quality thresholds are changed?
5. How does the relative performance change if some types of nature are considered
more important than others?
6. Which of the alternatives generates the largest connected natural areas?
7. What is the best alternative?

254
 These questions are a subset of the mueh larger set of questions whieh emerged
during the ease study. However, this short list includes some fundamental issues related
to the relative and absolute evaluation of poliey alternatives. The answer to all these
questions requires some form of spatial analysis and spatial evaluation.

2. Decision support for nature management

Spatial evaluation methods were developed in this study to support integrated

environmental and nature management of specifie poliey regions. The evaluation
methods are integrated in the Decision Support System Nature Management (Kros et al.
1995). The objeetive of the DSS is to compare and evaluate the quality of the natural
environment in a region as a result of different pollution control policies. Input to this
DSS are local and regional emission patterns of S02' NOx and NH4 • The system uses
these emission patterns to calculate acid and nutrient deposition and to predict resulting
soil quality parameters. Soil quality is then used to calculate the probability that
specified plant species will occur in an area.
The model results are presented in terms of 'nature types', which are characteristic
sets of plants (such as "bog heather", "coppice and halm", etc.). All results are
presented in a grid-cell format, which, in this study, represents areas of 1 square
kilometre. The results of the model include, for each grid-cell and for each nature type,
the proportion of plants that are expected to occur under the different management
alternatives. These results represent the potential quality of the management alternatives
with a spatial resolution of 1 km2 (Latour and Reiling 1992; Kros et al. 1995). This
output is the starting point of the evaluation model. However, the performance of each
alternative depends on the policy objectives which are fixed for the natural area.
The policy objective map gives a spatial representation of the preferred allocation
and spatial distribution of all nature types in the region. This map ean be considered as
a filter map, since it usually specifies the natural objectives for a subset of the whole
region. In praetice, this implies that an additional map, called the performance map of
an alternative, can be produced by extracting from the model output a subsets of grid-
cells for which an objective is specified. The performance maps present, for each km2,
the predicted performance of the nature type specified in the policy objective. The
performance maps represent the relative quality of the alternatives and form the input to
the evaluation. At this stage, each management alternative corresponds to a
performance map. The links between the model results, the objective map, the
performance maps and the evaluation phase are shown in Figure 1. Alternatives are
evaluated according to size, quality and spatial pattern of the predicted quality of nature
and on the basis of the costs of the emission reduction instruments included in each
alternative.

255
 integral
evaluation
"t IJ I'
Cl 11 1, 1 U
(,l l l, I .' 1 11 1 1l

partial
evaluation

Il . ~

, ~ ·· 11 I I! I .~

EJ
..,

exploratory
policy objective analysis
!:.I 1_' 1 11

Figure 1. Adecision support system/ar nature management.

 Figure 1 shows that a distinction is made between integral and partial evaluation and
exploratory analysis. With integral evaluation, the information contained in a map is
aggregated through space to compute evaluation indices. These indices, which are a
synthetic representation of the map information content, are then organised in an
evaluation table. Each index can therefore be described as a non-spatial summary of the
underlying map. Partial evaluation methods evaluate each nature type separately and
results in a number of maps equal to the number of specified nature types. Finally,
exploratory analysis aims to increase understanding of the pattern of the underlying
information and focuses essentially on exploration of different patterns and spatial
distributions ofthe information contained in a map (see also Janssen and Rietveld 1990,
Herwijnen and Janssen 1996).

3. A spatial policy objective for the Green Heart of the

Netherlands

Nature management in the Netherlands often uses maps to operationalize the policy
objective for a specific region. These maps specify the composition and spatial pattern
of the required nature types in this region. The ultimate goal of policy and management
alternatives is to provide the required environmental conditions and quality necessary
for these nature types to occur (BaI et al. 1995, Ministry of Agriculture and Nature
Management 1995).
The Green Heart of the Netherlands is a relatively open space in the centre of a ring
of four large urban areas (Figure 2). The Green Heart is an open agricultural landscape
but includes areas with important natural qualities. At present nature is fragmented and
its quality is declining, leading to a decline in species diversity. To reverse this trend it
is essential to create larger connected natural areas to offer better survival chances for
rare species and to ensure sustainab1e natural qualities. The spatia1 po1icy objective for
the Green Heart, presented in Figure 3, specifies for each km2 the required nature type
which fulfils these demands (Latour et al. 1997). In Figure 3, which includes only areas
classified as nature, the soil conditions are favourable to all types of bog. Nature types
Coppice and Holm and Rough grass land have been planned for large connected areas
in the North East and Centre of the Green Heart. Most other nature types, such as bog
heather, have been planned as small isolated units (Latour et al. 1997).

4. Spatial alternatives

The basic idea ofthe DSS Nature Management is to compare spatial alternatives with a
spatial objective. The alternatives include instruments for the reduction of emissions of
S02, NOx and NH4, but differ in the intensity, type and spatial pattern of these
instruments. As mentioned before, alternatives are presented in performance maps,
which represent the relative quality of the alternatives and are the i~put to the
evaluation. Two performance maps for the Green Heart are shown in Figure 4.

257
 A
10 km

o water communities
(:::I bog heather
• rough grass-land
~ coppice and holm

Figure 2. The Green Heart ofthe Netherlands. Figure J. Spatial policy objective for the Green Heart of the Netherlands.
 The maps show the potential quality of nature resulting from two packages of policy
instruments to reduce NOx deposition. Nature quality is measured as the percentage of
plant species of the nature type specified for each km 2 that can survive under the
predicted conditions. If conditions for all plants of all nature types specified in the
policy objective are sufficient the alternative equals the policy objective and all scores
equallOO%.
In this example the decision maker is offered the two maps presented in Figure 4 to
support the selection of the preferred alternative. The information content of both maps
is very high: each map includes a performance score for 813 grid cells measured in five
c1asses. It is clear that both model calculations and policy objectives take into account
the existing physical structure of the region, such as soil type and water level, which
provides a strong link between the performance maps and the real world. However,
since these maps represent the result of a range of model calculations based on several
assumptions and simplifications, this link has to be considered an indirect one.

D othe, uses
D 0-20%
Eilll 20·40%
_ 40·60%
_ 60·80%
_ 80-10W.
(a) Low deposition alternative (b) High deposition alternative

Figure 4. Predicted performance ofthe alternatives.

All maps in this chapter are presented in black and white. Despite experiences with
the colour spectrum in science textbooks and rainbows, the mind' s eye does not readily
give a visual ordering to colours, except possibly for red to reflect a higher level than
other colours. Because there is a natural visual hierarchy, varying shades of grey show
varying quantities better than colour (Tufte 1983, pp.153-155).

259
5. Relative performance

Policy question I: What is the relative performance ofthe alternatives?

The performance maps shown in Figure 4 represent, for each alternative, the potential
nature quality. The values linked to each grid cell in these maps range from 0 (none of
the plant species in the specified nature type can survive)to 100 (all can survive). To
present these maps, values were classified in five equal classes. Classification results in
a clearer map image, even if it is a generalised image (Kraak and Ormeling 1996).
However, classification results in loss of information and so differences between
alternatives may become blurred. This may be solved by introducing more classes or
even a continuous scale as proposed by Tobler (1973). Research has shown that too
many classes reduces the readability of the map and makes it difficult, at a single
glance, to obtain an overview and understanding of the theme mapped (Kraak and
Ormeling 1996). A continuous scale also poses problems in map reproduction.
The number of classes used and the definition of the class limits are dependent on
the purpose of the map and the distribution of the data. Many methods exist to define
the classification for a single map (Kraak and Ormeling 1996 pp.143-148) and each
method serves a specific purpose. In this study, the purpose of classification is to
emphasise the differences among alternatives presented in maps, and the classification
scheme which has been used is designed accordingly.
Differences among alternatives often centre around certain values. Limiting the
number <Jf classes and, at the same ti nie, setting class limits at these values may provide
a better image of the differences among the alternatives. This is based on a difference
index used to determine the optimal class limits. The index is calculated in such a way
that the visual difference between maps is maximised. In case three or more maps are
compared, this index is calculated as folIows: if all three classes are represented, a value
of 2 is added to the index; if only two classes are represented, a value of 1 is added;
and nothing is added if all values are included in the same class. The class limits
resulting in the largest value of the difference index is assumed to represent the best
visualisation of the differences in the underlying continuous performance maps. Of
course, if only two maps are compared, the contribution to the difference index of each
grid cell can only be 1 or O.
The reclassified performance maps of the example are shown in Figure 5. As can be
seen, the overall performance of the two alternatives is very similar (34 versus 33),
while the patterns are rather different. Each map is evaluated using an evaluation index.
The index value of the map representing the policy objective is set at 100 for all
evaluation criteria. Definition of the index depends on the aspect represented in the
map. The performance index included in Figure 5 represents the average performance
ofthe alternatives. This index is independent ofthe classification used.

260
 o -=:J 100 0 -=:J 100
34 33

D otheruses
D 0·11 %
_ "·5 8 ~

• 58 · 100 ~

(b) High deposition alternative

(a) Low deposition alternative

Figure 5. Predicted performance ofthe alternatives presented in three unequal c1asses.

6. Spatial differences

Policy question 2: Where is one alternative better than another?

Because a number of underlying mechanisms are at work, even a simple question like
this is not easy to answer. The preference map in Figure 6 shows, for each km 2, which
of the two alternatives is preferred. This figure shows a pattern that is not easily
deduced from the underlying performance maps. Ifthree or more alternatives need to be
compared, preference maps can be created for each pair of alternatives. Alternatively, a
dominance map can be produced, which shows, for each grid cell, the preferred
alternative.

7. Quality thresholds

Policy question 3: What's the relative performance of the alternatives if comparison is

linked to quality thresholds?

This policy question can also be formulated as folIows: how much nature of a certain
minimum quality results from the alternatives? The answer to this question depends on
the quality thresholds used. Is a nature type considered to exist if 80% of its plant
species are found, or is 50% sufficient? Clearly this is a very complex question from an

261
ecological viewpoint, which has relevant implications for the evaluation. Since a
definite answer is not available, Figure 7 was produced by fixing a tentative threshold
of 45%. The value index shown at the top of the figure is the sum of the grid cells
which exceed the threshold. Since each grid cell in this figure represents 1 km 2, the total
area defined in the policy objective equals 813 km 2• The index thus ranges between 0
(no cells exceed the threshold) and 100 (all cells, that is a total area of 813 km 2 , are
above the threshold).

o otheluses
o Lew < High
o Low -High
• Lew > High

Figure 6. Spatial distribution ofthe ranking oftwo alternatives.

o -=:J 100 o -=:J 100

45 44

(a) Low deposition alternative

-
o

(b) High deposition alternative

Figure 7. Total area above a specified quality thresholds.

262
 Policy question 4: What's the effect of changes in quality thresholds?

As mentioned above, fixing a threshold is a complex issue, and the 45% threshold is to
some extent an arbitrary choice. Sensitivity analysis to threshold levels can be used to
specify the degree to which the maps change by changing thresholds. Positive and
negative signs, or green and red if colours are used, can be used to show which grid
cells are no longer included, or which grid cell will be included if the threshold is
increased or decreased. Figure 8 shows the result of an increase in the threshold (from
45% to 60%). The evaluation index, which decreases substantially, reflects the smaller
area that meets the higher threshold.

0.::::=J 100 0.::::=J 100

17 16

o 0111., .....
o below
Ea CMnge
_ above

(b) High deposition alternative

(a) Low deposition alternative

Figure 8. Total area before and after increasing the quality thresholds.

8. Priorities

Policy question 5: How does the relative perfonnance change if some nature types are
considered more important than others?

In the previous analyses it has been assumed that all grid cells and all nature type are
equally important. However, decision makers can have a preference for specific rare or
characteristic nature types or for protection of certain areas such as nature reserves. This
corresponds to fixing weights linked to areas or to nature types. In this application, the
relative weight of each nature type is derived from the number of bird species that use
this nature type as preferred habitat for at least part of their life cycle (Table I). The

263
weights are standardised in such a way that the average weight of the total policy area is
equal to I. Linking these weights to the spatial distribution of the nature types specified
in the policy oejective map (Figure 3) results in the priority map shown in Figure 9.
Changes in the relative weights do not influence the performance maps (Figure 4) or
total area maps (Figure 7). However, changing the weights does influence the value of
the evaluation indices linked to these maps. As is shown in Figure 7, the evaluation
index of the two alternatives considered here is almost the same if all nature types are
considered equally important (Low deposition: 45; High deposition : 44). If the weights
of Table 1 are used, these indices change considerably. While the index of the low
deposition alternative only slightly decreases (from 45 to 43), the index for the high
deposition alternative shows a substantial reduction (from 44 to 36). On the basis of
these weights, the low deposition alternative is c1early the preferred alternative. Since
the average weight for the whole policy area equals 1, this implies that the higher
weighted nature types perform badly in the high deposition alternative.

Table I. Priorities cf nature (}pes linked to number cf bird 5pecies in each nature (}pe.
weight weight
Nature type number of bird species standardised
Brackish water 15 0.79
Reed land and bush land 28 1.48
Rough grass land 25 1.32
Bog heather 18 0.95
Coppice and holm 13 0.69
Wood communities ofbog 7 0.37
Grass land 24 1.27

-
D W:O.OO
D w:076
eJ w:O.88
w· 1.46
w: 164

Figure 9. Priority map Iinked to number of bird species in each nature type.

264
 9. Large natural areas

Policy question 6: Which of the alternatives generates the largest connected natural
areas?

At present nature is very fragmented and its quality is dec\ining. This has several
consequences, inc\uding a dec\ine in species diversity. To stop and reverse this trend it
is essential to create larger connected areas to offer chances to rare species and to
ensure sustainable natural qualities. In evaluation terms, this means to evaluate and
compare the size of connected areas of each alternative.
Connected areas are calculated using King's neighbourhood function. With this
function, a cell is connected if at least one of the eight surrounding cells has the same
value. Therefore, this function highlights connected cells starting from groups of two
cells (a reference one and one of the eight which surround it) to groups of eight (a
reference one and all the eight which surround it). By scanning all the map in this way,
areas of connected cells can be identified. Figure 10 shows the size of connected natural
areas.

o-=:J 100 o -=:J 100

38 37

--
D Okm2
D 1 ·20km2
21 · 40km2
41 · 60km2
61 · 00 km2
81 · 100km2

(b) High deposition alternative

(a) Low deposition alternative

Figure 10. Size 0/ connected natural areas.

In this figure no distinction is made among the nature types, which is a valid
assumption for species that can survive in all different nature types. However if species
are dependent on nature type, connected areas have to be caIculated for each nature type

265
separately. In addition, it is also possible to generate the largest connected area for each
nature type. This is relevant if thresholds are specified for target species which need at
least a certain amount of connected space for survival. The index at the top of Figure 10
is calculated as the total natural area divided by the square root of the number of
separate areas. This reflects both size and fragmentation of the area.

10. Ranking

IPolicy question 7: What is the best alternative?

This is clearly one of the most relevant and complex questions. Each one of the maps
presented before is evaluated by means of an evaluation index, which depends on the
type of feature analysed in each map. An evaluation index can therefore be described as
a non-spatial summary of the underlying map. Since a policy alternative corresponds to
aseries of maps, its overall quality depends on the content of all these evaluation maps
taken together. Since indices are used to summarise the content of each map, a policy
alternative corresponds to aseries of indices. In turn, the overall quality of an
alternative depends on the level and variability of these indices.
Evaluation indices can be included in an evaluation table. The evaluation table can,
therefore, be described as a non-spatial summary of the decision problem and can be
evaluated using a multi-criteria approach (Janssen and van Herwijnen 1993). In the
following analysis, four c1asses of evaluation criteria are used: 1. size, 2. quality 3.
spatial pattern and 4. costs. The evaluation criteria and the evaluation table are shown in
Table 2. All scores included in the evaluation table are expressed on a 0-100 scale. The
last two columns show the value of 100, linked to the policy objective, and its meaning
(that is the situation in which the index would take on a value of 100). Fbr instance, the
policy objective for variation specifies that five nature types should be present in the
area. The maximum value of variation (100), therefore, is achieved when five nature
types are present. The score of 80 of the Low deposition alternative corresponds to the
occurrence of four of the specified nature types. The first three categories (size, quality
and spatial pattern) reflect the effectiveness of the alternatives in generating natural
qualities. The last category (costs) includes the cost of implementation of these
alternatives. The cost figures presented in this table are tentative estimates and represent
a very rough calculation of the cost of instruments to reduce NH J emissions in and
around the region (Herwijnen et al. 1997).
Since all scores representing effectiveness are measured on the same scale (0-100),
they can be combined into an overall effectiveness index. All evaluation criteria make
use of the information in the performance maps (Figure 4). This information is
combined with other information such as thresholds and weights. In addition, since
evaluation criteria use different attributes ofthe same underlying data, the scores are, to
a certain degree, interdependent especially within each category.

266
 Table Z. Evaluation table.
Low High Max. Max. represents
deposition deposition
Size
Total area 45 44 \00 813 km 2
- priority to conservation areas 46 44 100 8 \3 weighed km 2
- priority to bird species 39 33 \00 813 weighed km 2
Quality
performance 34 33 \00 all specified plant species
variation 80 60 \00 5 different nature types
naturalness 47 35 \00 5 \3 km2 sem i-natural area
Spatial pattern
connected areas/nature type 55 53 \00 15.1 km 2
connected natural areas 38 37 \00 \02 km 2
largest connected areas 46 52 \00 largest area: average
Cost
Total costs 1.85 0.59 5 5 million guilders/year
Effectiveness 46.7 42.5 100
Cost-effectiveness 25.2 72.1 20

The contribution that each of the criteria makes to the overall quality of a policy
alternative is reflected in weights. These weights, however, do not represent trade-offs,
but represent the influence of the criteria in the overall objective. The importance of the
individual criteria for the overall score depends on the function of the region. Because
the Green Heart plays an important role as a buffer between large urban areas, criteria
linked to size are considered especially important. In this application, these criteria have
been given twice the weight ofthe quality and pattern related criteria. The effectiveness
score for each alternative is calculated as the weighted sum of the scores of the
evaluation criteria (cf. Table 2). The effectiveness scores show that the Low deposition
alternative is more effective than the High deposition alternative. It is also shown,
however, that both alternatives are far removed from the policy objective (which would
correspond to an effectiveness of 100). Only if all results can be attributed to the
alternatives can cost-effectiveness be calculated as the ratio of effectiveness and costs.
In practice, adjustments will have to be made for autonomous developments. In this
case the High deposition alternative is the most cost-effective, achieving the policy
objective is the least cost effective.

11. Conclusions

The spatial evaluation methods presented in this chapter have been developed to support
the comparison of mapped policy alternatives for the Green Heart of the Netherlands.
Seven examples of policy questions are used to iIlustrate the use of spatial evaluation
methods. It is shown that the use of these methods increases understanding of the

267
differences among the alternatives. The spatial evaluation methods used do not provide
the decision maker with a single answer, such as which alternative is the best, but result
in aseries of answers depending on the specific questions asked. All methods are
included in a small software package that invites decision makers to use the methods to
experiment with their problem.
The results of model calculations have been presented in performance maps. These
maps represent the relative quality of the alternatives .and form the input to the
evaluation. From the performance maps the differences between the alternatives were
not obvious. Differences become clear by looking carefully from various angles. It has
been shown that the alternatives differ if priority is given to bird species and also that
the alternatives differ substantially in terms of their spatial pattern. Each map was then
summarised using an evaluation index and these indices were included in an evaluation
table. From this table it could be concluded that the differences in effectiveness between
the alternatives are rather smalI. However, it is important to stress that differences
between the alternatives are lost in any aggregation procedure. Since differences that
can be observed from the maps are not always reflected in the indices, maps and indices
should always be used in combination.
The results of this study offer many opportunities for further research. In the current
study, the spatial pattern of the performance maps and the policy objective is identical.
Objective and performance differ only in level of achievement. As a next step,
performance maps with spatial patterns that differ from the policy objective could also
be evaluated using distance metrics. Another open issue is the specification of the
policy objective, since this is a very complex task. Heuristics could be developed to
support the decision maker in translating non-spatial objectives into an objective map.
A clear reference point for the required natural qualities would offer possibilities for
increasing the cost-effectiveness of environmental policies. Optimisation procedures
could be developed to generate optimal combinations in intensity, spatilil pattern and
time of emission reduction and management instruments.

Acknowledgement
The research presented in this article was commissioned by the National Institute of
Public Health and Environmental Protection RIVM, Bilthoven and was included in the
MAP research project Environmental quality of natural areas. The authors would like to
thank Arthur van Beurden, Rob van de Velde and Joris Latour (RIVM) for their
contributions. The authors also thank Jos Boelens who produced the computer program
and Niels Vlaanderen for his help in including the topographical map.

References
BaI, D., H.M. Beije, Y.R. Hoogeveen, S.R.J. Jansen and P.J. van der Reest (1995) Handboek
natuurdoeltypen in Nederland, Rapport IKC Natuurbeheer nr.ll, Wageningen.
Bartlett, M.S. (1975) The Statistical Analysis ofSpatial Pattern, Chapman and Hall, London.
Beinat, E. (1997) Valuefunetionsfor environmental management, Kluwer, Dordrecht.
Bertin, J. (1981) Graphies and graphie information proeessing, Walter de Gruyter, Berlin.
Campbell, J. (1991) Map Use and Analysis, Brown Publishers, Dubuque, USA.

268
Carver, SJ. (1991) Integrating multi-criteria evaluation with geographical information systems,
International Journal oJGeographical Information Systems, 5(3): 321-339.
Cressie, A.C.N. (1993) Statistics Jor spatial data, Wiley, New York.
Davidson, D.A., S.P Theocharopoulos and RJ. Bloksma (1994) A land evaluation project in
Greece using GIS and based on Boolean and fuzzy set methodologies, International Journal
oJGeographical Information Systems 8: 369-384.
Haining, R., (1994) Designing spatial data analysis modules, in S. Fotheringham, P. Rogerson
(eds.) Spatial Analysis and GIS, Taylor & Francis, London: 13-44.
Heamshaw, H.M. and D.J.Unwin (1994) Visualisation in geographical information systems,
John Wiley, Chichester.
Herwijnen M. van, R. Janssen, A.A. Olsthoom, 1. Boelens (1997) Ontwikkeling van ruimtelijke
evaluatie methoden voor gebiedsgericht milieu- en natuurbeleid, Instituut voor
Milieuvraagstukken, Vrije Universiteit, Amsterdam.
Herwijnen, M. van, R. Janssen (1996) The use oJ risk patterns to support environmental
decision-making, Proceedings of JEC-GI 1996, Barcelona.
Herwijnen, M. van, R. Janssen, P. Nijkamp. (1993) A multi-criteria decision support model and
geographic information system for sustainable development planning of the Greek islands,
Project Appraisal, 8:9-22.
Janssen R. and P. Rietveld (1990) Multicriteria analysis and Geographical Information
Systems; an application to agricultural land use in the Netherlands, in Scholten, HJ. and
J.C.H. StillweIl (eds.) Geographical Information Systems and Urban and Regional
Planning, Kluwer Academic Publishers, Dordrecht: 129-139.
Janssen, R. (1992) Multiobjective decision support Jor environmental management, Kluwer
Academic Publishers, Dordrecht.
Janssen, R. and M. van Herwijnen (1993) DEFINITE A system to support decisions on afinite
set oJ alternatives, Kluwer Academic Publishers, Dordrecht.
Keeney, R.L. (1992) Value-focused thinking; a path to creative decision making. Harvard
University Press, Cambridge.
Kohsiek, L., F. van der Yen, G. Beugelink, N. Pellenbarg (1991) Sustainable use oJ
groundwater; problems and threats in the European Communities, Research report
600025001, National Institute ofPublic Health and Environmental Protection, Bilthoven.
Kraak, M.J., F.J.Ormeling (1996) Cartography: visualisation oJ spatial data, Longman,
Harlow.
Kros J., G.J. Reinds, W. de Vries, J.B. Latour, MJ.S. Bollen (1995) Modelling oJsoil acidity
and nitrogen availability in natural ecosystems in response to changes in acid deposition
and hydrology. Wageningen, SC-DLO.Rapportnr 95.
Latour, 1.B. and R. Reiling (1992) Ecologische normen voor verzuring, vermesting en
verdroging. Aanzet tot een risicobenadering, RIVM rapport 71190 I 003. RIVM, Bilthoven.
Latour, J.B., H. van der Vloet, RJ. van de Velde, 1. van Veldhuizen, M.Ransijn, 1. van der
Waals (1997) Methodiek voor het aangeven van de milieuhaalbaarheid van gebiedsvisies:
uitwerking voor het Groene Hart, RIVM rapport in prep., RIVM, Bilthoven.
Miller, G.A (1956) The magical number seven plus or minus two: some limits on our capacity
forprocessing information, Psychological Review, (63): 81-97.
Ministry of Agriculture and Nature Management (1995) Ecosystemen in Nederland, Ministerie
van Landbouw, Natuur en Visserij, directie Natuur, Den Haag.
Muehrcke, P.C., J.O. Muehrcke (1992) Map Use; reading, analysis and interpretation, JP
Publications, Wisconsin.

269
Openshaw, S. (1991) Developing appropriate spatial analysis methods for GIS, in DJ. Maguire,
M.F. Goodchild and D.W. Rhind (eds.) GeogrGphic Irformation Systems, Volume I.,
Longman, Harlow: 389-402.
Pereira, J.M.C., L. Duckstein (1993) A multiple criteria decision-making approach to GIS-
based land suitability evaluation, International Journal cf GeogrGphical hformation
Systems, 7(5): 407-424.
Tobler, W.R. (1973) Choropleth maps without class intervals? GeogrGphical Analysis, 5: 262-
265.
Tufte, E.R. (1985) The visual display cf quantitative hformation. Graphics Press, Cheshire,
Connecticut.
Unwin, D. (1981) Introductory Spatial Analysis. Methuen & Co. Ltd, London.
Upton, G.1.G., B. Fingleton (1989) Spatial Data Analysis by Example, Volume 2. Categorical
and Directional Data, Wiley, New York.

270
Multi-criteria analysis and Geographie Information
Systems: analysis of natural areas and ecologieal
distributions

Andrea Patrono
International Institute for Aerospace Survey and Earth Sciences (ITC)
Enschede, The Netherlands

Abstract
Evaluating natural areas involves the identification and study of the most ecologically valuable
sites so that planning can be applied that respects the areas' importance and ecological
equilibrium. A technique is proposed for combining habitat analysis and animal distribution,
focusing on its usefulness for landscape planning and environmental impact assessment (EIA)
studies. The method, which combines multi-criteria analysis (MCA) and geographic
information systerns (GIS), is applied to an EIA study. A motorway project in a complex alpine
environment near the city of Trento, northern Italy, provides an example. The plan considers a
series of feasible solutions to expand the connection between two major road networks. The
projects crosses a series of ecosystems where the presence of animals plays a key role. The GIS-
MCA approach is used in conjunction with animal distribution data for habitat suitability
assessment and connectivity analysis. The impacts on habitats subjected to fragmentation and
disturbance are evaluated and converted to a computational format, suitable for alternative
ranking for decision-making. The approach and results ofthe analysis are discussed.

Keywords: Geographie infonnation system, multi-eriteria analysis, decision support

system, impact assessment, landseape planning, habitat suitability, animal flows,
eonneetivity analysis.

1. Introduction

The presenee of humans and their aetivities generally inerease landseape heterogeneity
by modifying the natural equilibrium and ereating a built-up environment, which is
often not environmentally eompatible (Fonnan and Godron 1986; Fonnan 1995).
Eeologieally-oriented environmental policy should put forward some basie interrelated
objeetives to be achieved by landscape planning and environmental impact assessment
(EIA) studies: the evaluation of landscape "naturalness", the identification of impacts
and responses of affected ecosystems, and the protection (and, where necessary, the
enhaneement) of high-quality, sensitive, rare, ete. natural areas. The evaluation of
natural areas, or eonservation evaluation, thus assumes a eritical role in the
identification and study ofthe most ecologically valuable sites. Evaluating natural areas
involves making measurements of aseries of criteria and then deciding which areas are

271
ecologically most significant, especially considering their importance and function in
terms of spatial distribution and role within the landscape context (e.g., Margules and
Usher, 1981; Roome, 1984; Ratcliffe, 1986; Smith and Theberge, 1987; Merrill et al.,
1995).
Environmental impacts can vary in their directness, intensity and duration depending
upon the nature of the human action and of the affected biotic communities and abiotic
units. The various effects of these impacts can be quantified, allowing us to forecast and
quantify the expected changes in ecological importance, including basic, pre- and post-
intervention. The use of multiple alternative scenarios allows effective comparisons,
which result in most efficient siting and planning strategies. In recent years a number of
studies have focused on the role of spatial information systems in facilitating more
understandable and environmentally friendly landscape planning. Special attention has
been given to the potential use of geographic information systems (GIS) when
combined with multi-criteria analysis (MCA) techniques in environmental management
(e.g., Fedra and Reitsma, 1990; Schaller, 1990; Carver, 1991; Nijkamp and Scholten,
1993; Pereira and Duckstein, 1993; Eastman et al., 1995; Jankowski, 1995).
A technique to assess landscape ''naturalness'' (in terms of habitat suitability) and to
evaluate changes due to human intervention is presented here. The negative effects of a
disturbance regime created by a road, with respect to wildlife distribution, are
illustrated with an actual application: a motorway project in a complex alpine
environment in northern Italy. Because ofthe difficulties in predicting species response
to environmental conditions, the relevance of wildlife-habitat evaluation is often
neglected in landscape planning, especially in EIA studies. This study indicates the
effectiveness of the combined GIS and MCA analyses and their critical contribution to
the decision making process. The application addresses one decision criterion in
particular, which is to minimise the impact on wildlife. The final step of the study was
the rating of the proposed alternatives, using the MCA module of a multi-objective
decision support system (MODSS).

2. A tool for environmental management

GIS, MODSS and related information technologies have given us a powerful capacity
to process a wide variety of social, economic, environmental and physical data. The
integration of GIS and MCA systems, in particular, has become a central focus for the
development of general purpose land use/environmental planning tools and methods
that can assist decision makers (Wright and Buehler, 1993).
Landscape planning is probably the most obvious and widely cited field to benefit
from GIS. Spatial information and GIS are key variables in ecological analysis and
EIA, where the primary aim is to provide a synoptic view of large landscape spatial
data sets, especially for making comparative studies over space and time. Spatial
information systems cannot be considered tools as such, but rather tools for improving
the quality of planning and decision making (Haslett, 1990; Nijkamp and Scholten,

272
1993). The spatial analytic functionality ofGIS consists mainly ofthe ability to perform
deterministic overlays, spatial queries, neighbour analyses, buffer operations, etc. Such
"pure" functions are of limited use when multiple andlor conflicting mapped criteria or
objectives are involved. MCA techniques linked to or integrated within a GIS can
provide the user a valuable addition to its standard functionality (Carver, 1991; Eastman
et al., 1995). As such, this instrument becomes the basic tool for putting emphasis on
landscape value and importance, providing expert guidance in combining mapped
criteria. This also meanS that it is an aid in solving one of the key problems faced by
decision makers, that is to find the most ecologically acceptable compromise between a
set of feasible alternatives. In this last phase, MCA support is eritical beeause rules of
combination andlor comparison do not necessarily require the separate evaluation of
eaeh potential or actual combination of environmental criteria, nor the determination of
mathematical relationships among mapped criteria. Many spatial decision making
problems, such as site selection or land use allocation, require the decision maker to
consider the impacts of alternative choices along multiple dimensions in order to ehoose
the best alternative (Jankowski, 1995).

3. Ecological distribution and environmental management

3.1. Wildlife importance

Vegetation and wildlife are important aspeets of the environment and thus of landscape
planning. Living beings adapt to the surrounding environment, towards achieving an
equilibrium with its different faetors; they ean often be organised into natural groupings
(communities) with mutual dependencies among their members, and they show various
responses and sensitivities to outside influences.
Wildlife involves special problems in environmental assessment. There are often no
satisfactory methods to define homogeneous associations, highlighting ecological
quality of habitats. The problem relates mainly to the characteristics of fauna such as
flying, foraging, movement (one of the most universal problems), relationships, etc.,
which can vary within similar environments. It is therefore very difficult to study, in the
perspective of a continuum spatial distribution, wildlife-habitat relationships,
suitability, etc. This is a critical point in landscape planning and EIA. Assessment
techniques are then generally based on the evaluation of a set of criteria or
environmental characteristics, deseribing, for example, attributes of habitat suitability.
Those techniques have the following characteristies: (a) they are not necessarily linked
in a fixed deterministie way, (b) apriori, they"do not depend on animal presenee, and
(e) they have to be integrated and verified with field data (e.g., radio tracking, trapping,
hunting records, ete.) (Morrison et al., 1992). The eombined use of GIS and MCA
partially solves this problem, offering a modelling tool that can incorporate the
available environmental spatial data set and the suitability analysis in the planning

273
evaluation process. This becomes a critical advantage whenever aland use change, and
the consequent environmental modification, has to be evaluated.

3.2. Impact on wildlife

The complexity and variability of ecosystems and their resilience make precise
quantitative predictions of environmental impacts impossible. Thus, to be able to
predict the effects of proposed human interventions on ecosystems, it is necessary to
decide which factors may cause such effects and also which effects are of particular
importance. The factors that may affect biotic communities are physical interventions
(including disturbance and emissions), or changes in the abiotic environment as a result
of those interventions. In calculating effects on organisms and communities, a wide
variety of interactions may occur, which are only partia11y known. Thus, considering
the wide variety of impacts, the complex relationships among them and the relatively
little knowledge we have about them, it is necessary to select only those impacts which
are appropriate in the context of the type of disturbance regime associated with the
proposed intervention.
Any infrastructure occupies space and any land use change modi fies an equilibrium.
This inevitably results in the disappearance of some ecosystems, or parts of them. In
some cases, new, possibly more anthropogenie, ecosystems are introduced. They may
form a barrier that, together with the infrastructure, can impede the necessary
movements of a11 kinds of animal species. Thus, space occupation and barrier effect can
change the size, geographical distribution and density of a population. It must be
mentioned that, for an animal groups, an infrastructure with intense disturbance effects,
such as noise, also creates a buffer zone of negative influence on surroundings habitats.
This can be considered a sort of barrier, the permeability or friction of which can be
estimated according tQ the expected disturbance regimes.
In landscape ecology in general, and in management and planning applications in
particular, the need for habitat connectivity is evident. In landscape planning and EIA,
there is a need to consider the effects of isolation of landscape elements. Attention
should be given to the preservation of such elements of connectivity and to constructing
corridors and networks to overcome isolation effects caused by human interventions.
Nature conservationists and physical planners are becoming increasingly aware of the
importance of this process, and landscape connectivity is gaining interest as a crucial
parameter of conservation value and as a criterion for potential nature development
(e.g., Saunders and Hobbs, 1991; Knaapen et a/., 1992; Morrison et a/., 1992; Forman,
1995).
As for any other land-planning activity, a11 these analyses require spatial-geo-
referenced information as a major input, which has to be modelIed and mapped
according to behaviour, movement habits, frequency, habitat suitability, etc., of the
species analysed.

274
3.3. Wildlife mapping

The presence of a population and its growth at a particular location is affected by both
intrinsic factors (e.g., birth rate, death rate, survivorship) and extrinsic factors (e.g.,
physical environment, interactions with other species). Spatial limits are not explicit in
population ecology and mapping. Animal communities are rarely mapped because: (a)
individual organisms are often too small to be automatically detected, e.g., by automatic
recognition using remotely sensed thermal data; (b) ground-based mapping of
individual organisms over a large area is time consuming and expensive; and (c) most
populations are so dynamic that the maps of their distribution would become rapidly
outdated (Johnston, 1993).
If several difficulties are encountered in regarding to the static aspects, the
"dynamic" mapping of animal flows, which is very important for connectivity analysis,
is even more challenging. A wide number of factors can influence the flows of
organisms.
Despite all the problematic issues described above, a feasible solution has to be
found for planning and EIA purposes. One approach is to develop a simple habitat
simulation model, using a limited number of factors that are easy to find or extrapolate
from existing data sets. The model should be as simple as possible, while remaining
faithful to a realistic, if general, picture of the system. Considering literature data, field
observations, expert knowledge, etc., and with the aid of the MCA module within an
MODSS, it is possible to formulate relations (that are often impossible to extrapolate on
a purely deterministic theoretical basis) between the parameters (or combinations of
parameters) and the output (i.e., population dynamics) ofthe model (Fahrig, 1991).
The results cannot provide a real distribution map of the species being analysed, but
only a suitable potential situation. Biological data and models are rarely sufflciently
adequate or precise to predict species distribution and abundance witb little error
(Morrison et aZ., 1992). However, critical information can be derived when comparing
the result of the MCAIGIS application - providing wildlife habitat suitability (e.g., in
terms of carrying capacity) - with the animal presence as reported (e.g., in bagging or
shooting records). In terms of planning or EIA, results represent a guideline to support
environmental management, for example related to (a) high quality areas characterised
by animal presence to be protected from human disturbance, (b) high quality areas not
colonised yet but potentially highly suitable (e.g., for species reintroduction) or (c) high
quality areas actually used as flow links but potentially colonizable if left undisturbed
or if restrictions for anthropic activities are strengthened, etc.

4. The method

In the present study, the model for assessing ecological importance in terms of animal
presence and movements is based on processing of mapped criteria, corresponding to
environmental parameters. A criterion is a unique geographical attribute such as slope,

275
soil, vegetation cover, etc. A criterion map shows the geographical location and
distribution of a given criterion. This map is usually combined with other criteria maps
to generate a composite map depicting, graphically or numerically, the relative
capability, suitability, flow resistance, etc., of an area to support an animal presence or
activity. In this context, the integration of MCA technology and GIS provides an
excellent framework. MCA is applied twice: (a) for assessing functional dose-response
relationships in tenns of habitat suitability and (b) for rating the alternatives with
respect to wildlife potential/real presence.
The method applied here is shown as a flow chart in Figure 1. It consists of the
following steps:
(1 ) Selection of the criteria that contribute to the habitat evaluation and their
nonnalisation according to a method originally developed by Perco (1990) for the
province of Trento and here modified and implemented within the IL WIS GIS
(ITC, 1996). They are: precipitation, elevation, aspect, slope, temperature and land
use. Interpolation procedures were applied whenever the data available were not in
the fonn of a map.
(2) Combination of the mapped criteria to calculate and map habitat suitability - in
summer and winter- according to the evaluation of criteria roles perfonned within
the MCA module of the MODSS DEFINITE (Janssen and Herwijnen, 1992). Each
criterion map provides a specific contribution to the habitat assessment, which is
here highlighted by importance weights.
(3) Calculation of the carrying capacity (Perco, 1990) and comparison with bagging
and shooting data in order to estimate, map and compare potential and real animal
distributions. The potential habitat suitability and carrying capacity are the results
of the GIS-MCA modelling and they cover the entire study area, cell by cello
Hunting records are spatially simplified in districts; however they provide useful
infonnation abOl~t the real average wildlife distribution.
(4) Analysis of areas under impact influences caused by the proposed set of
alternatives. Technical characteristics, including expected trafiic conditions, are
integrated and combined in the GIS with the results of points 2 and 3 to evaluate
effects of noise disturbance, habitat occupation and loss of connectivity.
(5) The resulting impacts are summarised in an impact matrix. Their various effects are
then compared in tenns of damage to potential and real wildlife distribution in
winter and summer, again using MCA. The possible combinations of impacts,
suitability and season represent the criteria used to minimise the impact on wildlife
in this study. The final output provides the alternative rating.

276
 Precipitation Elevation Aspect Slope Temperature Land use

~l ~MCAJ J~
Field data ~ ~

Habitat suitability modelling

~ ~ Alternatives
Impact modelling

NruC H,,,,..1,,,,00 ~,,"Off

~
Impact matrix
MCA

~
Alternative rating

Figure 1. Flow chort ofthe method opplied.

The main goal is the maintenance of the maximum "naturalness" of the affected
areas from the wildlife point of view, preserving the environmental quality of occupied,
temporarily used or potentially suitable wildlife habitats. The method was tested for a
single species, viz. roe deer (Capreo/us capreolus L. 1758), because of information
about frequency and tendency provided by the province of Trento, apart from the
information available on behavioural/biological aspects.
It is obvious that wildlife evaluation within land planning and EIA activities cannot
be reduced to a single species analysis. However, the practical example used here based
on standard data and a flexible approach can easily be repeated for several species, e.g.
all species selected as indicators for the wildlife assessment of a given area. Perco
(1990) provided criteria importance scores for other species such as chamois and
capercaillie. The limits of full applicability are given, for example, by a total lack of
data in terms of presence and behavioural knowledge. A partial lack of ground-presence
data can affect the real distribution analysis, but it does not compromise the potential
one, thus still maintaining the effectiveness ofthe GIS-MCA approach.

s. The case study

5.1. The study area

The study area covers approximately 400 km2 in the province of Trento, northern ltaly
(Figure 2). It inc1udes part of the Val (valley) di Non and part of the Val d' Adige.
Alpine geomorphology and strong relief energy characterise the area. Elevation ranges

277
between 200m and 2000m a.s.l. Natural vegetation cover can be subdivided into units
characterised by the main wooded species. Their distribution depends on elevation
range, slope, aspect and edaphic conditions. Oak woods represent the warmer habitats.
At higher elevations, in pioneer conditions, pi ne trees are the representative elements,
whereas beech is typical of more favourable environments. Spruce and larch
characterise the highest elevations. The cuItivated wooded areas are composed alm ost
exclusively of apple orchards and vineyards. Their distribution generally covers the
alluvial deposits on the valley bottoms and on man-made terraces, i.e. the lower or
warmer parts of the slopes.

5.2. The project

The purpose of the project was to develop a road connection between the main road
network ofthe Val d'Adige and that ofthe Val di Non and Val di Sole (a little further
north), both areas characterised by strong economies, based on agriculture and tourism.
The main goal of the study was to carry out that part of the EIA analysis related to
wildlife presence, investigating their distribution and evaluating/comparing the potential
impacts caused by five proposed alternative routes.
The present narrow road passes through the town of Mezzolombardo and continues
towards the Val di Non, with a winding route that forms a strong hindrance to trafiic
both outside and within the town centre. The criteria considered by the team of planners
charged by the province of Trento to select the candidate alternatives aimed mainly at
planning a road that assures a rapid flow of trafiic, minimising the consumption of
precious agricultural land and, in general, minimising the impact on residential areas.
According to measured trafiic flows, a "two-line road solution" was adopted (average
width of 15m). The starting point, A in Figure 2, of the new road is assumed to be the
same for all alternatives, and it coincides with the turn-off at Mezzocorona from the
highway crossing the Val d'Adige. The end-of-road connection is inserted into the main
new road (now under construction) crossing the Val di Non at point B in Figure 2. As a
result of a feasibility study carried out by a team of experts, five alternatives were
proposed, which are indicated in Figure 3. The technical parameters are specified in
Table 1.

T"ble 1. Technical parameters 0/ the jive proposed alternatives.

TECHNICAL PARAMETERS Alt. I Alt. 2 Alt. 3 Alt. 4 Alt. 5
Totallength 6800m 6150m 5650m 6850m 6270m
Totallength of"open" ways 4260m 2700m 3700m 2580m 4200m
Totallength oftunnels 1700m 3 100m 1700m 3800m 1700m
Totallength ofbridges and viaducts 840m 350m 250m 470m 370m

278
 B Al ternative 1 B Alternat ive 2
...

B
A

Alternative 3 B
'" --...
Alternative 4
A

...

~ A
-~
A

B Alternative 5

A Twmels

Figure 2. The study area. Figure 3. The jive proposed alternative routes.

6. WildUfe-habitat relationship model

6.1. The approach

Technical characteristics and plans of the five alternatives, climatic data, trafiic
forecasts and wildlife data were provided by several services of the province of Trento.
The results of a preceding project were used for the elevation and ground cover data
(Patrono, 1996).
Perco's method for evaluating habitat suitability for wildlife management is a
theoretical approach based on the combination of a given set of criteria originally
developed for application in the province of Trento. An importance scale varying from
oto 10 is established for each criterion. After area, species and season are selected, the
importance of (for example) elevation is measured, attributing a score to different
elevation ranges according to their influences on animal ecology. The analysis is
repeated for all criteria considered, the scores of which are then summed to finally
provide a suitability value for the area. This can be used, for example, to estimate the
carrying capacity of the area. Transformation functions are provided for winter and
summer based on the evaluation of trophic necessities, animal biomass, etc. (Perco,
1990; see also Bobek, 1980; Jeppesen, 1990; Dragan, 1994). The original method bases
the analysis on a cumulative area estimation, which provides information for large-scale
units (e.g., basins); for EIA purposes, it is necessary to focus the analysis especially on
areas affected or influenced locally by various impacts. For this reason, the method was

279
modified and implemented within a GIS to allow a more precise spatial analysis at
single cell, or pixel, level.

6.2. The application

After an area of sufficient scope to include all Iikely impacts had been selected, it was
possible to identify expected preferences in the animals' use of different habitat
elements within their horne ranges. This was based on the following variables:
precipitation (rain and snow), elevation, slope, aspect, temperature and ground cover.
Maps of the areas potentially preferred by roe deer were prepared for the entire area
(summer and winter periods) as folIows.

6.2.1. Data preparation - mapping

A digital elevation model (DEM) was available and the land use/land cover map was
already in digital format. Slope and aspect maps were derived from the DEM, using
standard GIS functions. A map of isohyets - summarising approximately 30 years of
precipitation data - was digitised and interpolated. Climatic data of four stations,
distributed at different elevations within the study area, were used to map temperature
and snow. Both were estimated as functions of elevation (Dingman, 1994) to cover the
entire study area. Trendlines were fitted to xy plots of the stations (x elevation and y
temperature, the example shown in Figure 4), using a best-fitting procedure (R2 always
> 0.975). The DEM information was then used to extrapolate and map c1imatic data
over the entire area. Figure 5 shows the spatial distribution ofthe (summer) temperature
in the study area, used for the habitat modelling.

25

~
,-, 20
U
~
., 15

.
10
~., 5 wmter
o
.,S-
E-< -5
-10
Elevation (m)

Figure 4. Trendlinesfitted to xy plots ofthefour stations (daily mean temperature).

6.2.2. Data reclassification - standardisation

The next step was to reclassify all input data in terms of suitability, using conversion
tables as suggested by Perco (1990). An example is given in Table 2, showing ranges
and related importance values adopted to reclassify original temperature data (for more
details see Perco, 1990). This turned out to be an excellent solution for data

280
standardisation in GISfMCA-based landscape studies. This need often anses as a
consequence of the necessity to integrate data measured not only in different units but
also in different scales of measurement into the evaluation process (Pereira and
Duckstein, 1993).

Table 2. Conversion table for temperature. Different ranges are used for assessing Summer/Winter
importance va/ues.
Temp. eC) Summer Temp. eC) Winter
<10 3 < -4 1
10 / 15 8 -4 / -1 4
15 / 20 10 -111 6
20 / 25 9 1/4 9
> 25 1 >4 10

6.2.3. Data combination - habitat modelling

The major problem in combining the different criteria was the absence in the original
method of some guidelines for establishing priorities or specific aggregation rules, apart
from a simple linear additive model. In this context, MCA was applied which requires
information on the relative importance of each criterion. The pairwise comparison
method (Janssen, 1992) was used. Table 3 shows the winter and summer weight sets
resulting from the evaluation.

28\
 Tobte 3. Set 01 weights lor the set 01 criteria in summer and winter.
Criteria (summer) weights Criteria (winter) weights
land cover . 0.424 land cover . 0.4-60
elevation 0.212 elevation 0.233
temperature 0.163 precipitation 0.139
precipitation 0.108 temperature 0.098
slope 0.059 slope 0.036
aspect 0.034 aspect 0.036

The mapped criteria were then combined for both summer and winter habitat
suitability maps, using a weighted summation method (similarly to the one suggested by
Perco, 1990), easily applicable within a GIS with the map overlay functions, as folIows:

habitat( imp.) = L"

;:1
W;X; (1)

where:
n number of criteria
W; importance weight of criterion i
X; map of criterion i.

The summer map of wildlife habitat suitability is shown in Figure 6. Higher values
(elose to 10) indicate very valuable areas. Carrying-capacity maps were then obtained
by reelassifying the two weighted summations (summer/winter) according to the
transformation functions in Figure 7 (modified after Perco, 1990). The functions
estimate the ideal animal frequency per cell according to the value/score obtained by the
cell itself after the criteria combination.

Figure 6. Habitat importance map (summer). BufJer olnoise injluence is overlaid (Alt. 3).

282
 ~
35
o'E 30
-"
---~'" 25
E 20
·c 15
~
~ 10
.;;;
.,= 5
Cl 0
1 2 3 4 5 6 7 8 9 10
Importance value

Figure 7. Importance value I carrying capacity transformationfunction (modified after Perco. 1990).

6.2.4. Comparison ofpotential and real data - suitability weighting

If, for planning purposes, estimated potential suitability maps represent a basic source
ofinfonnation, comparison with real data is useful for estimating effective impacts. For
this reason, the map of hunting districts was digitised and an associated database was
built with frequency information about animal presence. For each cell it was thus
possible to establish the real density. This process ineluded some simplifications, due to
the absence of infonnation linked with the density distribution within each district.
Subtracting the real distribution map from the potential maps, it was possible to
estimate the rate of colonisation of suitable habitats. The resulting maps (winter and
summer) are an excellent source of infonnation because they can be used directly to
quantify the extent of highly valuable areas according to the computed rate of
colonisation. Using the formula

weight = (min(s)+ max(s) - s) / max(s) (2)

where
s measured density - estimated carrying capacity
min(s) minimum value of map s
max(s) maximum value of map s

it was possible to obtain a "weight map" (with values in the range 0 - I); pixels with
values elose to I indicate areas where the differences between the estimated potential
density and the field records are very small. This was the basis for creating a new
suitability map, with values between 0 and 10, which also took into account the wildlife
density measured in the field. At the end of this process, four final maps were obtained.
They highlight the habitat quality of the study area in summer and winter. Two of them
consider only the potential animal presence while the other two indicate also the
hunting and bagging data.

283
7. Impacts

7.1. Impact calculation

After suitability maps were evaluated, the impacts had to be estimated. The five
alternative tracks were digitised and rasterised and the impacts for noise, connectivity
and space occupation were calculated, as shown below.
Estimating motorway noise impact can be a complex process that may require
extensive computer processing for detailed analytical modelling. The method adopted
here was a general motorway noise model that predicts the noise level and is adequate
for most noise assessment requirements. The model can be applied equally weIl to both
high and low traffic-volume roadways (see also Mestre and Wooten 1980;
Lanzavecchia 1986; Patrono 1994). Figure 8 shows the noise level as a function of
distance from the road as calculated by the model. Trafik predictions and technical
parameters of the alternatives are the basic input for noise calculations. Investigations
have provided evidence that with a noise level of 50 dB, a proportion of the population
is still fairly adversely affected by trafiic noise. After a level of tolerance of 50 dB was
set (for all alternatives), corresponding to a buffer of 500 m (see Figure 8), the resulting
"noise influenced" areas were mapped using a distance function. Resistance factors
were used in distance calculations to automatically transform the buffer of noise
propagation from an ideal situation (e.g. no shielding factors were considered) to a
realistic one, according to the roadside environment, in particular slope and land cover.
Resistance factors were used in distance computations to simulate differences in
permeability or accessibility related to different land units or objects within the
environment considered (ITe 1996). The buffer of affected areas for the third
alternative route is shown in Figure 6.

80
70
,-., 60
CIl 50
"0
'-'
11) 40
'"
'0 30
Limit
Z 20
10
0
0 200 400 600
Distance from the road (m)

Figure 8. Model result Jor noise level analysis.

As for connectivity, a distance function was used to estimate and outline areas of
most probable organism flows affected by each of the alternatives. The approach used

284
to detect the impacts was based on a comparison of the connectivity analysis of both
pre-and post-construction situations for each alternative route. Mismatching areas were
automatically considered to be affected. The targets for the flow distance calculations
were selected first, extracting the most valuable habitats (e.g., values greater than 9 in
the 0 to 10 scale), characterised at the same time by the highest frequencies of real
presence (e.g., more than 10-15 animalsilOO ha, according to bagging records). Then
the habitat suitability map was used to set the resistance factors, derived by calculating
a direct correlation from the values of the map itself. This meant that an area with a
suitability of 10 was assigned no resistance to animal movement while an area which
scored 0 in terms of habitat suitability was considered as a barrier. The application
required two steps: (a) a distance map was first calculated using the resistance factor but
without considering the alternatives; and then, (b) the distance analysis was repeated
adding the alternatives, one by one, as barriers. The mismatching areas between the
distance maps thus highlighted the affected areas. The operation was repeated for all
five alternatives for the summer and winter impacts. Figure 9 shows the summer
distribution of connectivity importance.

Connectivity
• high
.mediwn
Dlow
D barrier

Figure 9. Map 0/ connectivity importance (summer) pre-project. Barrier elements are settlements,
quarries, large rivers etc.

The space physically occupied by an infrastructure is normally much larger than may
appear at first sight. Occupied areas were based on the technical characteristics of the
project (crossings, viaducts, etc.). A distance function buffer of impacts was used here,
calculated around each alternative according to Lanzavecchia (1986).

285
7.1. Impact assessment

The evaluation of the overall impacts caused by the emplacement of an infrastructure

on the natural environment in a given area can be seen as the task of MCA. After
predicting the potential impacts on the basis of a GIS, the decision making process
consists of ranking the alternatives and selecting the one with the overall minimum
impact (Janssen 1992). The comparison of the "naturalness" or ecological importance
of the areas affected by the different impacts was taken as the basis for calculation. In
the decision making process, the greater the "Ioss" of ecological value, the worse the
score given to the alternative (Patrono 1994; Patrono et 01. 1995). The impacts were
evaluated according to the following formula:

I=x·y (3)

where:
I impact score
x map of the butTer representing the impact effect - noise, space occupation, loss
of connectivity - (binary map: 1 impacted and 0 undisturbed areas)
y map ofhabitat suitability (potential and ''weighted'' with huoting records).

Twelve criteria were used to evaluate and rank the proposed alternatives; they were
established on the basis ofthe combination of: (a) type of impact; (b) loss of ecological
value due to impact affect; (c) season. The criteria are listed in Table 4.

Ttlble 4. Criteria to compare anti rank the jive alternatives.

Criterion Im~act Habitat value Season
1 space occupation potential summer
2 space occupation potential winter
3 space occupation real summer
4 space occupation real winter
5 connectivity cut off potential summer
6 connectivity cut off potential winter
7 connectivity cut off real summer
8 connectivity cut off real winter
9 noise disturbance potential summer
10 noise disturbance potential winter
11 noise disturbance real summer
12 noise disturbance real winter

The criteria were measured using the same method, but because of the different
extents of impacts, the final scores did not fall within the prescribed interval and thus
they had to be standardised to a common dimensionless range in order to be comparable
for the final evaluation. In EIA, normalisation of a11 the scores (e.g., between 0 and 1)

286
is a practical way to adjust the representation of different quantities. The following
formula was used for all criteria:

X norm = Yxn= (4)

where:
X no"" new normalised impact value
X score obtained by application of (3)
Xn= maximum potential impact.

The assessment of the maximum potential (or acceptable) impact is generally a

critical point in any EIA study. There are no definite rules to establish what can be
considered as the maximum ecologicallimit that is tolerable and/or acceptable within a
given area. Standards can be helpful to set limitations, but in a wider context,
considering the cumulative effect of more projects, a realistic natural limit cannot be
established apriori. Another critical point in defining the maximum impact concerns
spatial aspects. Each alternative influences a certain area, the size and characteristics of
which are not necessarily the same as in the other alternatives. An alternative may have
a wide cumulative impact, which may be weIl distributed in terms of space, and thus
environmentally sustainable. In contrast, an impact limited in space together with
cumulative affects may be critically dangerous at a locallevel. For each alternative, it is
necessary to have its maximum potential impact, whenever possible, to compare the
scores obtained with a reliable maximum related only to that particular situation, i.e.,
the area actually/effectively influenced by the alternative itself.
In the context of this study, it was possible to extract the maximum value for each
kind of impact and for all the alternatives, using the following expression:

Xn==X·Y (5)

where:
Xn= maximum potential impact for the alternative considered
X binary map of the buffer representing the impact effect
Y maximum habitat value (in this study = 10).

Applying (5), it was assumed that all areas affected were characterised by the c1ass
with the highest importance. After the maximum impacts were ca1culated for each
alternative, (5) was applied. The application of(3), (4) and (5) was then repeated for all
criteria and for all five alternatives. Table 5 shows the results of impact calculations
after normalisation (all impact values were divided by the maximum impact value). The
values in the table represent the magnitude of the impacts for each alternative. The
impacts ofthe alternatives are expressed in quantitative form and they are considered as
costs, so they are negative (i.e., the lower in absolute value, the better).

287
 Table 5. Impact matrix.
Criterion Alt. 1 Alt. 2 Alt. 3 Alt. 4 Alt. 5
1 -0.441 -0.530 -0.517 -0.509 -0.441
2 -0.532 -0.611 -0.612 -0.604 -0.518
3 -0.057 -0.074 -0.060 -0.082 -0.059
4 -0.345 -0.370 -0.359 -0.385 -0.333
5 -0.670 -0.990 -0.681 -0.950 -0.690
6 -0.959 -0.855 -0.734 -0.786 -1.000
7 -0.272 -0.259 -0.271 -0.258 -0.267
8 -0.569 -0.619 -0.590 -0.608 -0.565
9 -0.570 -0.565 -0.602 -0.508 -0.506
10 -0.648 -0.662 -0.671 -0.602 -0.596
11 -0.035 -0.037 -0.040 -0.000 -0.036
12 -0.330 -0.340 -0.358 -0.293 -0.316

As mentioned above, the DEFINITE software package (Janssen and Herwijnen

1992) was used to compute the overall impact of each of the five alternatives and to
produce a ranking. The process required the following steps:
(1) input ofthe impact matrix representing the decision problem (see Table 5);
(2) generation of a complete (/ incomplete) ranking of the alternatives;
(3) analysis ofthe sensitivity ofthe scores, expressing their uncertainty.

The procedures used to support the evaluations are described below. They were
calculated twice to evaluate two possible aspects of environmental management: (a) to
evaluate the effective impacts on the "real" presence of animals, emphasising the
importance ofthe real-weighted habitat suitability (a standard EIA approach) and (b) to
analyse the impacts as possible future disturbance in terms of land use change planning
(new parks, reintroductions, etc.), focusing the analysis on the criteria related to the
"potential" habitat suitability.

7.3. Evaluation 0/ alternatives

The problem addressed in the evaluation process is to judge the attractiveness of the
alternatives with respect to presence of roe deer in the area. The use of MCA methods
of analysis requires information on the relative importance of each criterion. Different
weighting methods were used to transform qualitative priority statements into
quantitative weights. The arrangement of criteria, from the most to the least important
(note that there are "as important as" situations characterised by the absence of the rank
number), is shown in Table 6 (see also Table 4). Ranking methods treat the order of
criteria importance established by the decision-maker as information on the unknown
quantitative weights and try to make optimal use of that information. The importance
weights determined in this way were used later in multi-criteria evaluation to rank the
alternati ves.

288
 Tuhle 6. Rank order 0/ criteria in "real" and "potential" scenarios (see also Table 4).
Prioritv rank "real" "potential"
1 3 1
4 2
2 7 5
8 6
11 9
12 10
3 1 3
2 4
4 5 7
6 8
9 11
10 12

Various combinations of weighting methods and computations of overall impacts

were applied and they resulted in almost the same ranking of the alternatives. The
results obtained can be summarised as follows: considering the extreme situations,
alternative 5 represents definitely the best solution according to the assigned relative
importance, while alternative 2 is constantly the worst. The other three alternatives are
quite similar and interchangeable, with a weaker dominance of alternative 1. Alternative
3 tends to be ranked as the worst of these three. This means that rating modifications
can be expected with small variations of the importance values. Very few differences
were detected between the "real" and "potential" ranking orders, apart the c10seness of
alternatives 1 and 4 for the second place. Considerations about the probability of
shifting positions due to small changes in the importance values were analysed in the
sensitivity analysis that follows.

7.4. Sensitivityllnalysis

Scores and priorities are often uncertain in EIA and evaluation methods involve
different assumptions. The sensitivity of the ratings obtained from changes in
evaluation methods, impact scores and weights can be analysed using DEFINITE. The
procedures performed here determine the intervals within which the rank order of a
couple of alternatives is sensitive to changes in weights. The results obtained can be
summarised as follows. Concerning the "real" analysis, alternative 5 can be
interchanged with 1 and partially with 4, by increasing the weights of space occupation
(in particular criterion (3)) and connectivity criteria, in general. Regarding the
"potential" results, the connectivity criterion (6) and the space occupation criterion (1)
seem to be the critical ones which can modify the rankings. It was also confirmed that
alternatives 1 and 4 are more interchangeable than alternative 5; in particular, while
alternative 1 has a smaller negative effect in the real situation, its affinity with

289
alternative 4 is higher in the potential scenario. As a general conc\usion, it seems quite
difficult to modify the leadership of alternative 5 without strongly twisting the roles of
the criteria.

8. Conclusions

The environmental impact assessment described here focused mainly on the impacts of
the proposed motorway project on the wildlife resources. Thus the conclusions reached
do not take into account the cost of the alternatives, the construction implications from
the engineering point of view, the negative effects on the roadside land uselland cover
types, the cost of digging tunnels and of waste disposal, etc. More considerations might
change the fmal results and this represents a limitation of this study from the EIA
standpoint. EIA always has to be considered within an integrated perspective, which
includes several disciplines to fully cover environmental and socio-economic
implications ofthe intervention.
Considering the results obtained, it is very important to notice the presence of an
alternative which systematically performed best and of a solution which constantly
performed worst. The subjectivity involved in the study leaves some uncertainty about
the second best alternative between the remaining solutions. However, with the aid of
sensitivity analysis, it was possible to identify the reasons for this uncertainty, and to
provide the decision maker with valuable insights on the effects ofuncertainty.
The method developed in this paper is able to represent the typology of the project
and its i~pacts on wildlife settings, inc\uding the spatial relationship between project
and environment. The loss of ''oaturalness'' is clearly established and calculated, using
GIS and MODSS. The method fulfilled all required evaluation phases related to the
study, avoiding complicated and more time-consuming analyses. A clear advantage is
that of providing a transparent evaluation which is reliable and easy to implement in
anY GIS. A simple interface with a MODSS, in addition, makes it possible to integrate
an MCA module with the GIS platform. Subjectivity in establishing the importance
criteria and, as a consequence, in the ranking of alternatives is part of the MCA
procedure and cannot be avoided. The only way to limit its influence is to simplify and
automate the GIS analysis as far as possible, which was one of the main goals of this
study.
In conclusion, GIS and MCA provide a powerful set of tools for processing and
analysing land information. They can be used in various manners to solve a multitude
of problems. These problems usually involve environmental management issues that
are often neglected because of the difficulties related to spatial data handling in the
decision-making processes.

Acknowledgements
Data were provided by the Provincia Autonoma di Trento (project "Informatizzazione e
trasparenza", Servizio Faunistico and Ufficio V.I.A.). Fieldwork with G. Oriolo is

290
acknowledged. The author wishes also to thank A. Stewart, A. Saldaiia and M. Dragan
for critical review of this paper and C. Brugnoli and A. Mohrer of the Provincia
Autonoma di Trento for their co-operation.

References
Bobek, B. (1980) A model for optimisation of roe deer management in central Europe, Journal
ofWildlife Management, 44: 837-848.
Carver, SJ. (1991) Integrating multi-criteria evaluation with geographical information systems,
International Journal ofGeographicalInformation Systems, 5: 321-339.
Dingman, S.L. (1994) Physical Hydrology, Englewood Cliffs, New Jersey.
Dragan, M. (1994) La capacita' Portante delI' Ambiente Carsico per iI Capriolo, Capreolus
capreolus Linnaeus, 1758, Unpublished manuscript (Tesi di Laurea) Universita' degli Studi
di Trieste, Trieste.
Eastman, J.R., Jin, W., Kyem, PAK., Toledano, 1 (1995) Raster procedures for Multi-
criteriaIMulti-objectives decisions, Photogrammetric Engineering and Remote Sensing, 61:
539-547.
Fahrig, L. (1991) Simulation methods for developing general landscape-Ievel hypotheses of
single-species dynamies, in M.G. Turner, R.H. Gardner (Eds.) Quantitative Methods in
Landscape Ecology, Springer, New York, 417-442.
Fedra, K. and Reitsma, R.F. (1990) Decision support and geographical information systems, in
HJ. Scholten, lC.H. Still weil (Eds.) Geographical Information Systems for Urban and
Regional Planning, Kluwer, Dordrecht, 177-188.
Forman, R.T.T., Godron M. (1986) Landscape Ecology, Wiley, New York.
Forman, R.T.T. (1995) Land Mosaics, The Ecology of Landscapes and Regions, Cambridge
University Press, Cambridge.
Haslett, lR. (1990) Geographie information systems: a new approach to habitat definition and
the study of distributions, Tree, 5: 214-218.
ITC (1996) IL WIS 2.0 for Windows. Step by Step - Reference Guide, International Institute for
Aerospace Survey and Earth sciences (ITC), IL W1S Department, Enschede.
Jankowski, P. (1995) Integrating geographical information systems and multiple criteria
decision-making methods, International Journal of Geographical Information Systems,
9:251-273.
Janssen, R. (1992) Multiobjective Decision Support for Environmental Management, Kluwer,
Dordrecht.
Janssen, R., van Herwijnen, M. (1992) DEFINITE. Decisions on ci FINITE set of alternatives,
Kluwer, Dordrecht.
Jeppesen, lL. (1990) Horne range and movements of free-ranging roe deer at Kaie, Danish
Review ofGame Bi%gy, 4: 1-14.
lohnston, C.A. (1993) Introduction to quantitative methods and modelling in community,
population, and landscape ecology, in M.F. Goodchild, B.O. Parks, L.T. Steyaert (Eds.)
Environmenta/ Mode/ling with GIS, Oxford University Press, Oxford, 276-283.
Knaapen, J.P, Scheffer, M., Harms, B. (1992) Estimating habitat isolation in landscape
planning, Landscape and Urban Planning, 23: 1-16.

291
Lanzavecchia, S. (1986) Guida pratica alle Valutazioni di Impatto Ambientale, Corso di
formazione sulla valutazione ambientale, Unpublished lecture notes. Milano: Settore
Energia, Italian National Research Council (CNR), 97-152.
Margules, C. and Usher, M.B. (1981) Criteria used in assessing wildlife conservation and
potential, Biological Conservation, 21: 79-109
Merrill, T., Wright, R.G., Scott, J.M. (1995) Using ecological criteria to evaluate wilderness
planning options in Idaho, Environmental Management, 19: 815-825.
Mestre, V.E., Wooten, D.C. (1980) Noise impact analysis, in G. Rau and D.C. Wooten (Eds.)
Environmentallmpact Analysis Handbook, McGraw Hili, New York.
Morrison, M.L., Marcot, B.G., Mannan, R.W. (1992) Wildlife-Habitat Relationship; Concepts
and Applications, The University ofWisconsin Press, Madison.
Nijkamp, P., SchoIten, HJ. (1993) Spatial information systems: design, modeIIing, and use in
planning, International Journal olGeographicallnlormation Systems, 7: 85-96.
Patrono, A. (1994) A Study in Environmentallmpact Assessment (EIA). Integrated methodology
to assess and predict the ecological impact 01 a motor-way project in the province 01
Trieste, MSc thesis (unpublished), Enschede: International Institute for Aerospace Survey
and Earth sciences (ITC).
Patrono, A., Fabbri, A.G., Veldkamp, J.C. (1995) GIS analysis in geomorphology for
environmental impact assessment studies, /TC Journal, 4: 347-353.
Patrono, A. (1996) Synergism ofremotely sensed data for land cover mapping in heterogeneous
alpine areas. An example combining accuracy and resolution, /TC Journal, 2: 101-109.
Perco, F. (1990) Progetto Fauna, Provincia Autonoma di Trento, Servizio Foreste, Caccia e
Pesca, Effe & Erre, Trento.
Pereira, J.M.C., Duckstein, L. (1993) A multiple criteria decision-making approach to GlS-
based fand suitabiIity analysis, Interniztionaf Journal 01 Geographicallnlormation Systems,
7: 407-424.
Ratcliffe, D.A. (1986) Selection of important areas for wildlife conservation in Great Britain:
the Nature Conservancy Council's approach, in M.B. Usher (Ed.) Wildlife Conservation
Evaluation, Chapman & HaU, London: 135-159.
Roome, N.J. (1984) Evaluation in nature conservation decision-making, Environmental
Conservation, 11: 247-252.
Saunders, D.A., Hobbs, R.J. (Eds.) (1991) Nature conservation 2: the role 01 corridors, Surrey
Beatty & Sons, Chipping Norton.
Schaller, J. (1990) Geographical Information System applications in Environmental Impact
Assessment, in HJ. Scholten, J.C.H. Still weil (Eds.) Geographicallnlormation Systems lor
Urban and Regional Planning, Kluwer, Dordrecht, 107-117.
Smith, P.G.R., Theberge, J.B.T. (1987) Evaluating Natural Areas Using Multiple Criteria:
Theory and Practice, Environmental Management, 11: 447-460.
Wright, J.R., Buehler, K.A. (1993) Probabilistic inferencing and spatial decision support
systems, in J.R. Wright, L.L. Wiggins, KJ. Ravinder, TJ. Kim (Eds.) Expert Systems in
Environmental Planning, Springer-Verlag, Berlin, 119-144.

292
Multicriterion planning of protected-area buffer zones:
an application to Mexico's Izta-Popo national park

Mark A. Ridgleyl and Gerrit W. Heif

IUniversity 0/ Hawaii, Honolulu, Hawaii, USA
lUtrecht University, Utrecht, The Netherlands

Abstract
This paper presents a methodology for the design of buffer zones around protected areas.
Although the procedure is designed to be used by the protected-area managers and hence
respond primarily to their objectives and concems, it recognizes the influence of the wider
socioeconomic and political environments in realizing those objectives. The methodology thus
helps resource managers to (I) formulate their objectives, (2) anticipate how those objectives
may conflict with ongoing or potential future changes in the wider environment within which
resource management must be conducted, and (3) obtain as efficiently as possible the
information necessary to (4) design and evaluate management options in view of such changes
and conflicts. Multiobjective optimization is used both to formulate the managers' goals
pertaining to land-use change, and to anticipate potential land-use changes resulting under
different socioeconomic scenarios. Goal programming is then employed to fashion minimum-
conflict plans, which are subsequently used as inputs into a GIS procedure that hel ps managers
design spatial land-use configurations attractive from landscape-ecological perspectives. The
prescribed land-use changes common to these plans and/or configurations may be considered a
"commitment set", and the methodology thus employed an exemplification of a robustness
philosophy.

Keywords: protected areas; multiple objectives; optimization; land use/land cover;

robustness; evaluation; measurement; GIS.

1. Buffer zones and protected areas

One could easily get the impression that the most effective way for nature and people to
coexist is simply to assign them to different places. This century has witnessed an
explosion in the number of national parks, nature reserves, protected areas, or other
tracts ofthe earth's surface similarly designated as exhibiting natural qualities worthy of
some form of protection or preservation. Although the institutional forms and
mechanisms vary, common to all is the existence of a sharp, well-defined boundary
setting off these protected areas (the term we'll use as shorthand to subsume the full
variety of such forms) from the areas around them. Inside the boundary, nature is
protected and human actions sharply circumscribed; outside it, human action has freer
rein and nature correspondingly less liberty to develop as it might.

293
 The contrast between the two sides of the boundary can be so pronounced that the
boundary itse\f, the dividing line, need not actually be visible to distinguish them. One
recalls the famous satellite image of semi-arid range land that displays two areas under
different grazing regimes abutting each other along an edge as straight and sharp as a
razor-scored etching, but with the fence that physically separates them remaining
invisible - just as in a pencil drawing we can distinguish two images through deft
shading without having to delineate their edges with the pencil point. Thus, a division
may be inferred at the juncture between dissimilar forms; marked contrast in pattern can
give the appearance that physical separation exists. But in the real world of human
beings and ecosystems, that sharp contrast is no mere visual disjuncture. It reflects a
pressure field, and pressure fields give rise to forces. Protected areas are created to
shield nature from such forces, but if the pressure is too great they may not. be able to
withstand them.
Mere\y assigning people and nature to different places is not enough to protect
nature. A boundary alone, a division of essentially zero width, is commonly insufficient
to maintain the separation, to keep what is outside from penetrating the barrier and
invading the protected. This holds almost without exception in the developing world.
Either the boundary must be fortified to prevent incursion, or those spaces must be
separated by a transition zone that buffers the incursions and vents the pressures
propelling them, or both. Strengthening the boundary necessitates resources; personne\,
vehic\es, energy, materials, technology, and finance must be marshaled and expended
for monitoring and enforcement. But the pressures may be very great, and resources
scarce, and so the design and creation of transition or buffer zones may become more
attractive than boundary fortification.
Although buffer zones can be distinguished from other areas according to a number
of other criteria (e.g., governance), we will define and treat them here in terms of land
use and land cover. Thus, aside from reducing human incursionary pressures, buffer
zones can yield valuable ecological benefits by preserving habitat continuity, hence
reducing the spatial fragmentation of ecosystems that can compromise their functioning
and increase their vulnerability. Landscape ecology has demonstrated the importance of
landscape structure - the spatial arrangement of different ecological communities and
the nature of the connections among them - in maintaining ecological integrity and
sustainability (Forman 1990; Forman and Godron 1986). For example, the length of
uninterrupted stretches of habitat is thought by many ecologists to be more important
for many aspects of ecosystem preservation than the total area of such habitat. Thus, the
maintenance of corridors and other linear elements of the landscape is a frequent goal of
regionallandscape planning. The design of transition or buffer zones adjacent to
protected areas thus affords opportunities to reap ecological payoffs that might
otherwise be lost.
This paper introduces a methodology for designing buffer zones around protected
areas, i1lustrating it in an application to Mexico's Izta-Popo National Park. The
procedure is designed for the specific use of protected-area managers, and it therefore

294
responds directly and primarily to their viewpoints and objectives. However, since
ecosystem management and decisions about natural-area preservation take place within
socioeconomic and political environments that are dynamic and pluralistic, conflicts
with those objectives are sure to arise. Therefore, the procedure goes beyond managers'
objectives and viewpoints per se to explore and respond to such potential conflicts.

2. Izta-Popo uRder siege

2.1. Motivation and context

Fifty kilometers east of Mexico City and straddling the states of Mexico, Puebla, and
Morelos, the snowcapped volcanoes of Ixtaccihuatl and Popocatepetl, part of the
southernmost stretch of the Sierra Nevada, rise to elevations of wel1 over 5200 m and
form the backbone of what is commonly referred to as "Izta-Popo" National Park. From
its constitution in 1935 until 1948 the Park extended from the summits down to an
elevation of 3000 m., encompassing an area of 59,913 ha. At that time, the boundaries
were redrawn to coincide with the 3600-m contour, reducing the park area to 25,679 ha
and leaving unprotected the 34,234 ha in the 3000-3600-m band now bordering the
park. As the century draws to a close, the region shows the land-use changes wrought
during the subsequent 50 years from population growth, urbanization, and agroforestry
development.
The Izta-Popo area (and especial1y the zone outside the park limits) is threatened by
continuous areal expansion of agriculture, urban settlement, pastoralism, forestry, and
uncontrol1ed recreation. In the last 15 years alone, the area under agriculture has
increased by 50%. Accompanying the decline in forested area is the increasing
fragmentation of that which remains. Poorly planned and controlled recreation
intensifies this process. Although the park limits are located at an elevation of 3600 m,
the area around the park that can be counted as having a direct influence on it is
considered to go down to 2500 m. In this area are 18 municipalities with a total of
694,481 inhabitants. However, towns Iying on the access roads to the park but
nominally outside this "zone of direct influence", such as Amecameca at 2420 m, also
influence the park greatly.
The main activities of the economical1y active population around the park Iie in the
primary sector (65-90%). Of the people working in this sector, those dependent on
agricultural or pastoral activities general1y have an income significantly lower than the
minimum wage in Mexico. Of the activities in the area, these agricultural, pastoral, and
forestry activities probably have the most direct effect on the ecology ofthe park and its
immediate surroundings. Traditional farmers, producing crops such as beans, pumpkins
and cucumbers. generally work on smal1 fields using traditional methods and animal
traction. Modem farming, geared towards intensive production of wheat, oats, and
potatoes, develops on larger sites making use of modem machinery, fertilizers, and
pesticides. As the area under agriculture continues to increase every year, the attendant

295
opening of large forested areas, soil modification, and ensuing erosion make it almost
impossible for the forest to recover. Extensive forestry in the region mainly concems
cutting and collecting wood, cutting Pinus branches for torches or for utensils for open-
fire cooking, collection of mushrooms, and hunting. Although these (often clandestine)
activities seem to be small-scale, their adverse effects on the forest have been
substantial. Recreation is heavily dominated by weekend visitors from Mexico City,
with tourism concentrated near the roads leading to and inside the park. Lacking
organization and facilities, most recreational activities have had significant
environmental impacts on the area (Garcia et al. 1992). So great have all these changes
been collectively, and so strong seem to be the pressures that augur continued and
accelerated change, that proposals have recently been made to effect stronger land-use
controls by lowering the borders ofthe park to the 2500-m contour, which would imply
a park area of 118,792 ha (SEDESOL et al. 1992).
Reflecting this concern, in 1991 the Department of Man and His Environment of the
Autonomous Metropolitan University at Xochimilco, in Mexico City, with World Bank
financing, began a project (SEDESOL et al. 1992) to develop a management plan for
Izta-Popo Park and its surroundings. The Department first obtained remotely sensed
imagery, classified the vegetation cover, and prepared maps of existing land use and
land cover. Then, as part of a three-week training workshop on multicriteria analysis
held at the Department in September 1995, we focused on the specific problem of
designing buffer-zone land-use configurations. The methodology that developed in
response, described and illustrated in the remainder of the paper, was designed for a
particular context, defined by the following assumptions: (1) the Park Managers (PMs)
are the "client" of the analysis, and the analysis reflects their perspective; (2) the
problem is one of designing an appropriate and realistic land-uselland-cover pattern for
a buffer zone around the park; (3) the buffer zone would comprise the 3000-3600-m
elevational band formerly part of the park; and (4) success in establishing and
maintaining the desired buffer zone would depend on land use and development
elsewhere in the municipios (a sub-state-Ievel political jurisdiction analogous to the
county in the USA) surrounding and encompassing the buffer zone and the park itself.
This last point led us to embed the buffer-zone-design problem within that of regional
development and land-use planning. For simplicity, we consider here only the
municipio of Amecameca as the regional context, and only that portion of the buffer
zone contained within it (Figure 1). The brevity of the workshop also meant that data
for many criteria, especially socioeconomic and institutional aspects, remain absent;
plausible estimates have been used for them to enable exposition of the methodology.
To design a suitable buffer zone completely encircling the park would thus require
inclusion of all surrounding municipios and the obtaining of the information that is here
onlyestimated.

296
 IAmecameca city I

Abies lorest
bar" soil/cr 0
Pinus for"S1 _

shr ub/op"n P

Grid Cf) North

+680

Figure I. Current Amecameca land use.

2.2. Overv;ew 0/ the approach

How should the park or protected-area managers (PMs) determine the best land use and
land cover for a buffer zone? The procedure it should follow can be described
informally as consisting of seven steps:
(i) First, they need to become clear about the objectives they want to achieve and to
translate these desires into operational criteria that can be used to judge weather
one proposed buffer zone is better than another.
(ii) Then, they should determine how a change from one use to another on a
particular type of land would affect those objectives. By land type, we mean land
c1asses defined by biophysical characteristics relevant to the objectives, such as
slope, soil type, aspect, vegetation cover, and so on.
Knowing the effects of these individual changes, the PMs will need to decide
what mix of changes to promote. That is, they will need to determine for each
land type, for the buffer zone as a whole, how many units (e.g., hectares) of a
given land use i should change to land use j. To arrive at such adetermination,
three steps are needed. These can be facilitated with multiobjective optimization.
(iii) Since different mixes of the same (i.e., i-to-j) land-use changes will yield different
mixes of payoffs, the manager must first decide on a desired balance of objective
attainments. This "preferred payoff mix" now becomes the managers' overall goal
for the buffer zone.
The PMs must now design a land-use mix that not only achieves the preferred
payoff mix, but does so in a contlictive and uncertain environment. Because

297
 different parties in society will be pursuing different interests with regard to land
in the buffer zone, and because the PMs will have only limited influence over
their actions in the zone, and because conditions may change in the future,
altering their preferred payoff mix, the PMs should design as robust a land-use
plan as possible.
(iv) The managers design a range of different land-use plans, each corresponding to a
different set of objectives that represents a different development scenario for the
buffer zone and the region within which it falls. Whereas the objectives
corresponding to the managers' preferred payoff mix refer only to the aspects of
the buffer zone affecting the ecological and related dimensions important for the
protected area, the objectives defined to guide the scenarios reflect other interests
deemed to be influential in the future evolution of the zone and the surrounding
region.
(v) The managers now search for a set of land-use changes that minimizes conflict
between the plan yielding its preferred payoff mix and those corresponding to the
different development scenarios. These sets indicate only the land-use changes
that should occur on land of a particular type; they do not specify the location of
such changes. Since different sets of land-use changes may yield identical or
similar payoffs, the PMs may be able to formulate a set that will accomplish their
objectives under a variety of futures. As there are a number of different ways to
define "minimum conflict", so there mayaiso be a number of different minimum-
conflict sets. In this step, the PMs find several such sets.
(vi) The PMs now must determine the best location and spatial pattern of the changes
prescribed by each minimum-conflict set. This is the task of landscape design. In
effecting it, the PMs will consider structural aspects of the landscape as they
relate to the ecological, management, and other criteria in the hierarchy. GIS is
the principal tool employed for this activity. The outcome is a set of alternative,
minimum-conflict landscape plans.
(vii) As the final step, the PMs make adecision on a course of action. Two broad
strategies suggest themselves. PMs can evaluate the alternative plans and select
one to adopt and promote. 80th discrete multicriterion evaluation (Olson 1996)
and game theory (Luce and Raiffa 1957) can offer support in this approach.
(Indeed, it is also possible to have formulated the classical rules for decision
making under uncertainty - such as minimax regret, Laplace's principle of
insufficient reason, and the maximin principle - as objective functions to be used
in place of step (v) (Romero and Rehman 1989). Alternatively, the PMs could
identify the land-use changes common to all or many of them and take actions to
implement those changes as soon as possible. Those changes constitute a
"commitment set". As the future development pattern around the protected area
becomes more clear, the PMs could identify further changes, perhaps repeating
the above procedure. Following a "robustness" philosophy (Rosenhead 1989), this
tack attempts to keep the PM's options open as long as possible. In the future, as

298
 more information becomes available, the PM will repeat these seven steps. It is
this latter tack we illustrate in the case study that folIows.

2.3. Illustration ofthe procedure

2.3.1. Identifying concerns and articulating objectives

Tbe aim of this task is to define the nature of the protected-area situation sufficiently
weil so that one knows what criteria should be considered when judging a proposed
land-use change, and that such criteria have been defined precisely and unambiguously.
These qualities are necessary for the subsequent assessment of land-use changes. This is
done by building a criterion hierarchy, a structure that identifies and defines such
criteria, articulates their interrelationships, and specifies on which ones the effects of
land-use changes will be directly assessed. Criterion hierarchy is a term used to
subsume two cIosely related structures: objectives hierarchy (Keeney 1992; Keeney and
Raiffa 1976) or value tree (von Winterfeldt and Edwards 1986), and analytic hierarchy
(Saaty 1980, 1994).
Through two iterations of role-playing, development of criterion hierarchies by
individual group members separately, and subsequent discussion, over the course of a
week the group elaborated an integrated composite hierarchy to guide subsequent
analysis (Box 1). As suggested above, these criteria pertain to the municipio-wide
development planning problem, while only the ecological criteria have direct relevance
to the buffer zone itself. Further, the group did not intend that measurable attributes
should be selected or developed for all criteria and then used directly to formulate
andlor assess options. Rather, many were meant to be areminder of the members'
earlier discussions and to clarify their concerns when it became necessary to evaluate
alternatives with respect to higher-order criteria.

2.3.2. Land c1assification and measurement 0/ effects

Two main considerations interact to guide the identification and collection of the data
needed to determine effects scores. First, data should refer to the elements represented
explicitly in the criterion hierarchy. Second, they should be scale-appropriate. An
important characteristic of current work in nature conservation and restoration is the
large scale at which ecological patterns and processes are being quantified. This is
partially a response to the recognition of regional dynamics in influencing processes
and structures that are outside the direct control of resource managers yet are crucial to
ecosystem functioning, such as nutrient cycling, hydrological patterns, and vegetation
structure. While a hierarchy of such elements can be defined (Allen and Starr 1982) -
from habitat, through regional, to geological and cIimatic factors, on the abiotic side,
and from the individual, through population and community, to landscape and biome,
on the biotic side - it is those at the landscape scale that tend to dominate ecological
considerations in buffer-zone planning. Adequate attention given to criterion relevance
and scale considerations increases the efficiency of data collection

299
CRITERIA FOR ASSESSING MUNICIP/o AND DUFFER-ZONE CHANGES

1. ECOLOGICAL CONSEQUENCES
1.1 Conservation Value
1.11 Vegetation Structure
1.12 Diversity
1.13 Pattern
1.14 Endemism
1.15 Endangered Species
1.16 Typicality
1.2 Vulnerability
1.21 Soil Quality
1.211 Water-Holding Capacity
1.212 Nutrient Availability
1.22 Risk of Structural Change
1.221 Habitat Fragmentation
1.222 Toxicity
2. ADMINISTRATIVE & INSTITUTIONAL EFFECTS
2.1 Land Tenancy & Ownership
2.2 Compatibility
2.21 National Policies
2.22 State Policies
2.23 Adjacent Municipios' Policies
2.24 Municipio Policies
2.3 Park Management & Operations
2.31 Cost ($/yr.)
2.32 Monitoring & Control
2.4 Municipio Management & Operations
2.41 Cost ($/yr.)
2.42 Compatibility with Existing Policies
3. SOCIOECONOMIC EFFECfS
3.1 Housing
3.11 Affordability (# 10w-cost units)
3.12 Supply (# units)
3.2 Employment
3.21 Level (# jobs)
3.22 Type (# ski lied vs. unskilledjobs)
3.23 Security (# temporary vs. permanent jobs)
3.24 Underemployment (# seasonal vs. year-Iongjobs)
3.3 Risk to Public Health
3.31 From domestic wastewater
3.32 From agricultural chemicals
3.4 Environmental Amenities (ha. open space)
3.5 Economic product ($)
Box I. Criterion hierarchy for use in evaluating proposed land-use changes. Possible measurable
attributes shown in parentheses.

In the Izta-Popo case, working at the landscape level means including such abiotic
and biotic factors as slope, elevation, biomass, and vegetation type. Homogeneous land
types were defined in terms of combinations of three disjunct slope (0-15%, 16-35%,
and > 35%) and elevation « 3000 m, 3000-3600 m, and > 3600 m) categories, yielding
nine c\asses in all. In tandem with different land uses and vegetative covers, the nine

300
land types allowed the determination of effects on the elements shown in the criterion
hierarchy under "Ecological Consequences". Land of the same type will typically be
found in more than one location, that at each location having a particular areal extent.
Such a spatial distribution is easily stored, retrieved, and displayed with the aid of a
geographic information system (GIS), as was used in this study.
A change of land use on a given land type will have consequences in terms of the
objectives and criteria defined earlier. These consequences will have to be measured.
Such measurements, referred to here as efJects scores, can be expressed in either
"natural" units or "value" units (von Winterfeldt and Edwards 1986). Among the
former are units of length, mass or weight, area, and money; examples are "hectares of
pine forest" and ''tons of sediment eroded". Value units are measured on artificial,
derived scales developed to measure intangibles. One mayaiso transform natural units
into value units to express the value (attractiveness, utility) of a given performance
level, measured in natural units, for a particular purpose or objective. (For
completeness, we note here that although financial and economic cost is actually
expressed in value units - since it is the result of a transformation (mapping) of a given
level or amount of natural effects (bushels of grain, say) into a value score denominated
in monetary units (e.g., dollars) - people are so familiar with these units that monetary
cost is commonly considered a "natural" measure.) A wide variety of such measurement
and scale-development methods exist; the literature on decision theory (e.g., Keeney
and Raiffa 1976; von Winterfeldt and Edwards 1986), multicriterion analysis (e.g.,
Voogd 1983; Saaty 1980), and psychophysics (e.g. Torgerson 1958) is replete with
them. What is important for the procedure presented here, however, is that such effects
scores be measured on interval or ratio scales, since other steps in the methodology
require multiplication of these scores by each other or by other scores (requiring ratio-
level measurements), or multiplication of differences between pairs of scores (allowing
interval-Ievel measurements).

2.3.3. Finding the PMs' preferred payoff mix

Guided by the ecological consequences for the buffer zone, as represented by the
corresponding elements in the criterion hierarchy, the PMs now design an optimal land-
use plan. Optimality is defined as the most preferable mix of ecological payoffs in the
buffer zone attainable from such aplan. As will be demonstrated, at this stage it is not
the plan itself - i.e., the set of land-use changes to promote - that is important but the
mix of objective attainments that could be realized through land-use change.
In the Izta-Popo case, possible land uses were categorized as cropland, shrubland,
grassland, human settlement, and two types of forest, Abies (fir) and Pinus (pine). Land
type was c1assified according to slope and elevation, as defined earlier. Some land-use
changes are prohibited according to land type: neither crops nor settlement are
permitted to be assigned to areas above 3600 m, and settlement is also restricted to
slopes below 35°. Four operational objectives, a11 pertaining only to the buffer zone (as
signified by the suffix "BUFR"), were used to fashion a plan: maximize vegetation

301
structure (z" VEGBUFR); maximize species richness (Z2' SPECBUFR); minimize habitat
fragmentation by maximizing a proxy variable, forest habitat (Z3' FORBUFR); and
minimize erosion (Z4' ERODBUFR). These correspond to elements 1.11, 1.12, 1.221, and
1.21 of the criterion hierarchy, respectively. Since the PMs also wished to protect the
integrity of the buffer zone and to minimize land uses that would increase the risk of
encroachment into the park itself, they chose to prohibit cropland (CROPBUFR) and
settlement (URBBUFR) in the zone. Thus, the cropland already there would have to
change to another use. Further, although a switch from one type of forest to another
(FOR2FOR) on the same land type would lead to a mathematically different solution to
the optimization model, such switches were deemed of insignificant payoff to the
objectives yet undesirable on other grounds. Thus, those changes were prevented. The
need to accommodate a roughly 50% increase over the present population of 36,321,
and to allocate that new population among rural settlement (RURALPOP) and high-cost
(HCURBPOP) and low-cost (LCURBPOP) urban places, drives the land conversion.
Let x be the vector of decision-variable values for Xijl' where Xijl =the amount of land
that should change from existing use i to new use j on land type t. The following
baseline optimization model was formulated to help design the land-use plan according
to these conditions.

(1) maximize {VEGBUFR, SPECBUFR, FORBUFR}

(2) minimize {ERODBUFR}
subject to (s.t.)
(3) CROPBUFR = 0
(4) URBBUFR = 0
(5) FOR2FOR = 0
(6) HCURBPOP + LCURBPOP + RURALPOP = 18150
(7) XE F

Expression (7) denotes the remaining feasibility and accounting constraints. These
include the functions of x that define the objective and state variables appearing in the
model, as weil as the land-balance constraints - the amount of land of any type t
changing from use i to use j cannot exceed the initial amount in use i on that type.
Notice that this baseline model (by definition; see Ignizio 1982) provides no
information on the relative importance of each of the four objectives, information
required for the solution of the model. Such information is contained in the operational
model presented below.
The objectives ofinterest to the PMs, (1)-(2), are functions ofvalue scores. For each
ofthe first three objectives (VEGBUFR, SPECBUFR, and FORBUFR), land uses were valued
intuitively through direct numerical assignment. Later assessments were made through
pairwise comparison with the eigenvector procedure (Saaty 1980), the relative values
obtained differing little from those initial valuations. FORBUFR is defined as a
monotonically increasing function of forest land and shrubland, with Abies and Pinus

302
valued equally and shrubland valued half as much as forest. SPECBUFR is a function of
all land uses, while VEGBUFR is a function of all but human settlement. The fourth
objective, ERODBUFR, is a proxy chosen because it affects both wafer-holding capacity
and nutrient availability, two aspects of the criterion soil quality. We used the Analytic
Hierarchy Process (AHP) (Saaty 1980) to estimate the relative erosion potential of each
land use on each land type. That AHP model had the nine land types as its single level
of criteria and the six land uses as the alternatives.
This baseline model was solved by transforming it into the following operational
model, similar to a lexicographic weighted Tchebycheff pro gram (albeit with all
weights equal to 1) (Steuer 1986: 445):

(8) lex min {D, (-Zj - Z2 - Z3 + Z4)}

S.l.
(9) [IZh· - ~lIlzh· - ~ WI] :::: D h = 1, ... ,4
(10) (3)-(7)

Let us call (8)-(10) "Model PM1". The objectives Zh, h = 1, ... ,4, refer to VEGBUFR,
SPECBUFR, FORBUFR, and ERODBUFR, respectively. With Zh· denoting the best possible
attainment of objective Zh, and ~ Wdenoting an estimate of the worst value that need be
considered (obtained from the payoff table in Table 1), the first problem in (8) together
with (9) is a traditional compromise program that minimizes the maximum deviation
from the "ideal solution" (Zeleny 1982). The second problem in (8) then maximizes the
sum of the objectives. The preferred solution shown in Table 1 is the PMs' preferred
payoffmix.

Table 1. Payofftable and preferred solutionfor Model PMl.

Objective Value of Value of Value of Value of Preferred
Optimized VEGBUFR SPECBUFR FORBUFR ERODBUFR Solution
VEGBUFR 73,446 73,446 12,241 92,703 68,755
SPECBUFR 48,964 85,687 6,120 312,173 75,791
FORBUFR 73,446 73,446 12,241 92,703 11,068
ERODBUFR 43,338 43,338 7,223 77,199 111,464

2.3.4. Formulating development scenarios

To get a feet for how their plan might conflict with Amecameca's future development,
the PMs formulate two quite different development scenarios. The first emphasizes
economic development while limiting loss of agriculture, while the second focuses on
expanded agriculture and agroforestry as the basis for future development.
Scenario 1 characterizes a regional development strategy that maximizes economic
value and employment while minimizing the amount of land to be used for high-cost

303
urban settlements (HCOSTURB), the cost for water suppl.y and sanitation (WSSCOST), and
the amount of cropland lost (CROPLOSS) from the current level of25,219 units. Omitting
constraints (3) and (4), this scenario allows agriculture and human settlements in the
buffer zone. Due to lack of readily accessible data and the tight time constraints of the
workshop, we estimated the relative values by land use for ECONVALU and EMPLOYMT,
taking account of the different agricultural productivity by land type, also estimated.
High-cost urban development is characterized by high-cost water supply and sanitation
(i.e., conventional waterbome sewerage with multiple-tap piped water to the
household); costs used are from Christmas and de Roy (1991). Thus, the baseline model
for Scenario 1 is:

(ll) mrullmize {ECONVALU, EMPLOYMT}

(l2) minimize {HCOSTURB, WSSCOST, CROPLOSS}
s.l.
(13) CROP + CROPLOSS - CROPGAlN = 25,219
(14) (5)-(7)

Compromise programming with the Tchebycheff metric (i.e., the first-priority

problem in (8)-(10» was used to solve the corresponding operational model. The
payoffs obtained are shown in Table 2.
The strategy underlying Scenario 2 is to maximize agroforestry employment
(EMPAGFOR) and revenues (AGREVENU), and to maximize total cropland (CROP). As in
the previous scenario, this one allows agricultural and urban development in the buffer
zone. Once again, the value functions for EMPAGFOR and AGREVENU were derived from
direct numerical assignment based on our intuitive assessments. Except for the objective
functions, the baseline model for Scenario 2 is identical to that of Scenario 1:

(l5) maximize {EMPAGFOR, AGREVENU, CROP}

(16) minimize {EROSION}
s.t.
(5)-(7)

The lexicographic procedure of (8)-(10) was employed once again to obtain a

solution (Table 2).

Toble 1. Comparison 0/ selected criterion values 0/ the Park Managers' pre/erred solution with those 0/
the /wo development scenarios/ormulated.
Model VEGBUFR SPECBUFR FORBUFR ERODBUFR CROP CROPBUFR
PMl 68,755 75,791 11,068 111,464 o o
Scenario I 99,300 99,300 12,241 338,314 12,927 12,927
Scenario 2 79,006 75,791 7,223 417,520 17,834 17,834

304
2.3.5. Formulating minimum-conjlict land-use plans
Table 2 shows that the PMs' plan for the park and the two development scenarios for
Amecameca would have quite different consequences. The land-use changes yielding
these consequences are also quite different, suggesting considerable conflict should the
PMs promote this plan and one of the scenarios unfold. However, the conflict implied
in Table 2 may be more apparent than real. For example, the PMs have no desire to
convert all cropland (CROP = 0), but merely wish to prevent cropland in the buffer zone.
To identify ways that it could achieve the same payoffs while mitigating this and other
sources of conflict, the PMs solve several minimum-conflict models. These models try
to find a set of land-use changes that are as c\ose as possible to those prescribed under
the alternative development scenarios while still achieving the PMs' preferred payoff
mix. Let us denote by x: the set of land-use changes corresponding to scenario s.
Solving the following model separately for each objective function indicated in (17)
identifies five different minimum-conflict plans.

(17) minimize {SUMDEV, SUMDEV I, SUMDEV 2, CHANGE, MAXDEV}

s.t.
(18) (3)-(7)
(19) ~(x)~ Zt.* h= 1,2,3
(20) ~(x).:s Z4*

(21) x+os-Ps=x: s=I,2

(22) Es (os + Ps) = SUMDEV
(23) 0 1 + PI = SUMDEV I
(24) 02+ P2 = SUMDEV 2
(25) ~i~.1 Xijl = CHANGE
", J

(26) SUMDEV I .s MAXDEV

(27) SUMDEV2 ~ MAXDEV

A sixth minimum-conflict solution can be found by using unordered lexicographic

minimization (ULM) (Schrage 1991: 298-302).
The different conflict measures corresponding to each of these models are shown in
the first four columns of Table 3. (The fifth column is not a direct measure of conflict;
it merely shows the amount of cropland.) Due to (19) and (20), each yields ecological
payoffs for the buffer zone equivalent to those attained by Model PMI; that is, they all
deliver the PMs' preferred payoff mix. The differences depend on the development
scenario emphasized. SUMDEV minimizes the sum of the deviations from both scenarios,
while SUMDEVI and SUMDEV2 only minimize the sum of deviations from scenario I or
scenario 2, respectively. CHANGE represents the continuation of the status quo, as it
minimizes the changes from present land use. This is reflected in the very sm all loss of
cropland under that strategy, as indicated by its value for CROP in Table 3. Minimizing
MAXDEV avoids unnecessarily large conflicts under either scenario.

305
2.3.6. Landscape design with GIS
For each solution ~.) corresponding to a model in Table 3, the GIS package IORISI was
used to design a spatial configuration. To do that, we employed the MCE decision
support tool available for multi-criteria evaluation. Criteria included proximity to
settlements, to roads, and to the park. Through a multi-criteria evaluation, these factors
were combined to form suitability maps from which the final configuration for each
solution was made. Criteria were of two types, factors and constraints. Factors are
continuous in nature, such as distance to settlements, and were combined by means of
weighted summation. Constraints are boolean in character, considering such factors as
whether or not a location was within the park, the buffer zone, or further down in the
valley, and served to exclude certain areas for considerations. The MCE procedure
starts by multiplying each factor by a weight and then sums the results. Then the
constraints are applied by successive multiplication to "zero out" excluded areas
(Eastman 1993). Figure 2 shows an overview of the steps followed in the GIS-based
analysis. Figure 3 shows the minimum-conflict land use when SUMDEV2 is minimized;
Figure 4 displays that when CHANGE is minimized.

Tllble J. Performance ofeach ofthe six minimum-conflict plansformulated. All measures are in land
units. For thefirstfour measures, the higher the value, the greater the conjlict.
Objective SUMDEV SUMDEVI SUMDEV2 CHANGE CROP
orModel
SUMDEV 31,675 6,103 25,572 42,024 0
SUMDEVI 32,363 2,489 29,875 42,141 0
SUMDEv2 45,599 41,336 5,262 45,889 0
MAXDEV 34,331 17,166 17,166 45,889 0
CHANGE 90,547 41,259 49,288 3,626 25,090
ULM 48,598 23,924 24,674 24,674 14,853

2.3.7. Dejining the commitment set

Each of the minimum-conflict plans has merits, yet each also has drawbacks. Because
these plans are based on only a fraction of the concerns represented in the criterion
hierarchy' (Box 1), they should be viewed only as rough guides to the most desirable
land-use pattern for both municipio development and park management. If the PMs and
municipio and other planners feel pressed to select one of these plans, then the criteria
articulated earlier can be used with a discrete multicriterion evaluation method to help
render the necessary assessments. On the other hand, if they prefer to keep their options
open and minimize their commitments, while nonetheless strengthening the buffer zone,
they might choose a robust strategy instead. Figure 5 shows such an initial commitment
set, derived from minimum-conflict models CHANGE and SUMDEV2. The land-use
changes indicated in this commitment set are common to both the CHANGE and the
SUMDEV2 strategies.

306
 I TopographicaJ map I I Landsa•
I Digilal ,,,",,in •Tn<J<k1 (DTM) I I Road•map I I Land uselco

1 1 "+~0-.
t
I Siope map I
! •
Constrain. map Le_: Factor maps rar Foetor maps for Foeto. maps fo. Conversioo maps. Le
<15. 15-35. ;> 35 mapvs. distance from and distanc:e from and distance fmm and Pi.ne. senlemenlS. Fir.
<3000. 300Q. 3600. ;>3600 proximity 10 proximity to proximity to s/uub. gm ••• cropIban:
map the pari< mads settlements soil map

I I I I
~

IOpt:~ion ~,- _ _ _ _ _ _ _ _ _--,I+-_____________...J

Figure 2. Overview 0/ the different analysis and processing steps in the GIS environment_

Abl~s forest
bare soil / e r
Pinus forest
st-lr ub/ open P
--
CJ

EZl

Snd ( J ) NOrth

1680

Figure. 3. Minimum conjlic( land-use map when SUMDEV2 is minimi=ed.

307
 Abi .. s lorut _

bar .. soil/cr D
.' Pinus forest

shr ub/op<>n P

Grid C.{) North

i680

Figure 4. Minimum conflictland-use map when CHANGE is minimized.

Abi .. s ~or ..st l1li

bar .. soil/cr 0
Pinus lor .. st l1li
shrub/open P

6rld Cl) North

.... ' .. rs
i680

Figure 5. Commitment set derivedfrom the maps in Figure 3 and Figure 4.

3. Conclusion

Planning how to eonfigure buffer zones around proteeted areas involves eonsideration
of the likelihood of eonfliet over land-use objectives and the uncertainty regarding what
those objeetives are, whose they are, and in what future environment they will be set
and pursued. Planners and resouree managers often eonfront this planning task under
severe resouree eonstraints, manifested by poor data, by searee budgets, personnei, and
the equipment to obtain what they believe is needed. and by even seareer resourees to

308
implement and sustain their plans. The methodology presented and illustrated here
represents a step forward in dealing effectively and efficiently with these challenges.

References
Allen, T.F.H., Starr, T.B. (1982). Hierarchy: Perspectives for Ecological Complexity. Chicago:
University of Chicago Press.
Christmas, J., de Roy, C. (1991). The Decade and beyond: at a glance. Water International
16(3): 127-134.
Eastman, lR. (1993). Idrisi version 4.01, Update Manual. Worcester: Clark University,
Graduate School ofGeography.
Forman R.T.T. M.Godron, M.I (1986). Landscape Ecology, New York: J.Wiley and Sons.
Forman R.T.T. (1990). Ecologically sustainable landscapes: The role of spatial configuration.
In: I.S. Zonneveld and R.T.T. Forman (Eds.), Changing Landscapes: An Ecological
Perspective. Springer-Verlag, New York. pp. 261-278.
Garcia. M., Lopez-Paniagua, J. and Rosas, M. (1992). Socioeconomicos. In: SEDESOL-Banco
Mundial-UAM-X. Programa de manejo para el Parque Nacional Ixtaccihuatl-
Popocatepetl.
Ignizio, J. (1982). Linear Programming in Single- and Multiple-Objective Systems. Englewood
Cliffs, NJ: Prentice-Hall.
Keeney, R. (1992). Value-Focused Thinking. Cambridge, MA: Harvard University Press.
Keeney, R., Raiffa. H. (1976). Decisions with Multiple Objectives: Preferences and Value
Tradeoffs. New York: John Wiley.
Luce, R.D., Raiffa, H. (\957). Games and Decisions. New York: Wiley.
Olson, D. (1996). Decision Aidsfor Selection Problems. New York: Springer.
Romero, c., Rehman, T. (1989). Multiple Crlteria Analysis for Agricultural Decisions.
Amsterdam: Elsevier.
Rosenhead, J. (1989). Robustness analysis: keeping your options open. In: J. Rosenhead (Ed.),
Rational Analysis for a Problematic World: Problem Structuring Methods for Complexity,
Uncertainty, and Conjlict. Chichester: Wiley.
Saaty, T. (1980). The Analytic Hierarchy Process. New York: McGraw-Hill.
Saaty, T. (1994). Fundamentals of Decision Making and Priority Theory with the Analytic
Hierarchy Process. Pittsburgh: R WS Publications.
Schrage, L. (1991). LINDO: An Optimization Modeling System. San Francisco: The Scientific
Press.
SEDESOL-Banco Mundial-UAM-X. (1992). Programa de manejo para el parque nacional
Iztaccihuatl-Popocatepetl. Mexico City: Secretaria de Desarrollo Social.
Steuer, R. (1986). Multiple Criteria Optimization: Theory, Computation, and Application. New
York: Wiley.
Torgerson, W.S. (1958). Theory and Methods ofScaling. New York.
von Winterfeldt, D., Edwards, W. (1986). Decision Analysis and Behavioral Research.
Cambridge: Cambridge University Press.
Voogd, H. (1983). Multicriteria Evaluationfor Urban and Regional Planning. London: Pion.
Zeleny, M. (1982). Multiple Criteria Decision Making. New York: McGraw-Hill.

309
Agro-ecological zones and Resource Management
Domains (RMDs) in relation to land use planning

Jaques Antoine, Frankjohn J. Dent, Denis Sims and Robert Brinkman

Food and Agriculture Drganization (FAD), Rome

Abstract
Land use plannin~, or at least land resources management, is one of the key issues in the
modem development of rural areas. However, only a holistic approach, which would inc1ude
both bio-geo-physical and socio-economic aspects, can deal adequately with such complex
planning problems. FAO, as UN Task Manager for Chapter 10 of UNCED Agenda 21, has
proposed a new approach to Interactive Land Use Planning for sustainable Land Resources
Management. The guiding principles are: production and conservation incentives, stakeholder
negotiation on resource management and provision of information to support negotiation and
decisions. Effective negotiation and decision making on land use cannot take place without a
knowledge base that is accessible to the stakeholders. The two essential components of the
knowledge base are information and education. The basis for interactive land use planning is a
spatial zoning of the resources across the natural land units. Zoning itself concerns a
subdivision of rural lands on the basis of the physical and biological or agro-ecological
characteristics of the various natural land units, as weil as the prevailing socio-economic
conditions. Together, they form the basis for the delineation of land Resource Management
Domains (RMDs), i.e. rural areas where both agro-ecological and socio-economic conditions
are so similar that one can expect that development- or conservation-oriented intervention
packages can be successfully transferred from one si te or farm to another. The FAO integrated
approach inc1udes Land Resources Information Systems (LRIS) as an essential tool for analysis
of land resources information. The FAO LRIS is conceived as a modular toolbox which can be
implemented at different scales. It is currently implemented using GIS databases, AEZ models
of land suitability and land productivity assessment, multicriteria tools for the analysis of land
use options and user-oriented interfaces to enable easy and transparent interaction of the user
with the system. AEZ/LRIS procedures can be used in the delineation of the RMDs which
provide a practical basis for interactive land use planning. The collaborative work of all the
actors, inc1uding decision makers, technicians and stakeholders, is essential to the success of
such integrated approaches.

Keywords: Agro-Eeologieal Zone (AEZ), Aspiration-Ied Deeision Support (ALDS),

Integrated Land Use Planning and Management, Geographie Information System (GIS),
Deeision Support System (DSS), Land Resourees Information System (LRIS), Multi-
Criteria Deeision Analysis (MCDA), Resouree Management Domains (RMDs).

This paper was prepared by Messrs. 1. Antoine, F. 1. Dent, D. Sims, and R. Brinkman, Land and Water
Development Division of the Food and Agriculture Organization of the United Nations (FAO) and is
reproduced here by perm iss ion ofthat organization.

311
1. Introduction

Natural resources world-wide and particularly in developing countries of Africa, Asia,

Latin America and the Caribbean, are under pressure from high population growth,
fragile soils in sensitive eco-regions, multiple demands on limited resources, resource-
poor farmers and limited or absent institutional support. As a result some of these
natural resources are being rapidly degraded. Urban growth has reached unprecedented
proportions. For example, in 1950 the combined population ofthe Asia-Pacific region's
5 mega-cities was 33.5 million. By 1990 the region's number of mega-cities had risen
to 21 and their combined population to over 125 million (ESCAP 1990). In many
places it is the prime agricultural land that is most susceptible to urban and industrial
encroachment. In the Philippines an estimated 24,000 ha of irrigated farmland are lost
per year due to the conversion of farms to urban and industrial development and other
non-agricultural uses (Dent 1995). Densely populated developing countries will need to
improve their ability to planning their rapidly growing urban areas, infrastructure
networks and industrial complexes. Otherwise uncontrolled urban-industrial sprawl will
eat up valuable agricultural and forest lands and destroy indispensable natural
ecosystems.
So far, mankind's response to the challenge of meeting these rapidly increasing
human demands on the natural resources and of maintaining vital ecosystems has been
weak. The root cause of the problem is deficient planning and management due to a
number of factors of both a physical and a socio-O(;onomic nature. These can be
grouped under four headings: a lack of sound information on the resource opportunities
and constraints, a lack of integration, a lack of stakeholder involvement, and
institutional weakness.
Land use planning or at least land resources management is one of the key issues in
today's development of rural areas. Land use zoning and urban planning are urgently
needed to prevent cities and their peripheral infrastructures from encroaching upon
agriculturalland (F AO 1996). Only a holistic approach, which would include both bio-
geo-physical and socio-economic approaches, can deal with such complex planning
problems.
These problems were clearly recognized by the United Nations Conference for
Environment and Development (UNCED) in calling for an integrated approach to the
planning and management of land resources. Adequate planning is essential to the
sustainable management of land resources. In Agenda 21, the document resulting from
the Earth Summit of 1992, UNCED proposed a package of policy and strategy
measures which includes integration of environmental, social and economic issues, the
active participation of local communities and the strengthening of institutions in the
process. Chapter 10 of Agenda 21 deals with an Integrated Approach to the Planning
and Management of Land Resources.
FAO, as UN Task Manager for the implementation of Chapter 10 of Agenda 21, has
developed and is promoting a new approach which is consistent with emerging

312
principles of good governance now viewed as aprerequisite to sustainable development
(FAO 1995). Sustainable deve\opment is conceived as an interactive process which can
be successful only if there is a commitment on behalf of both the government and the
people.

2. Interactive land use planning

At the core of the FAO integrated approach is the concept of interactive land-use
planning, essentially a mechanism for decision support to guide the stakeholders in
selecting the best sustainable options which are consistent with their land use objectives.
Interactive land-use planning aims at achieving optimum use of the land resources by
inducing the maximum numbers of the stakeholders to make sound land use decisions
and to apply appropriate management practices to their land. The FAO approach is
based on the three guiding principles:
1) Conservation of the production potential of the land. This implies that land is
allocated to uses for which it is suitable, and used under appropriate management
practices to ensure that natural resources are conserved and that the surrounding
environment is protected from adverse impacts.
2) Promotion of equity of land access and use. This is a participatory process which
recognizes the rights of all the stakeholders, inc\uding those who may be
traditionally disenfranchised in so me societies, such as sharecroppers, landless
farmers, rural women or ethnic groups.
3) Negotiated decision-making. This devolves decision making at the lowest possible
level that is consistent with the ability for implementation. As far as possible,
decisions on land use and management in the village should be taken by village
level institutions, within the broader framework of government land use policy.

2.1. Scope and scales ofinteractive land use planning

Interactive land use planning as a key element of sustainable land resources

management is always demand-driven, and the demand may result from a problem or a
development opportunity perceived at village or sub-national level, or from a concern
ofthe national govemment (Figure I).
Interactive land use planning has both aspace and a time dimension. Spatially, it
may be implemented at farm level, watershed, district, national or regional seal es, while
the long-time scale of sustainable production requires combinations of appropriate
techniques, site observations and monitoring. Table I summarizes some of the main
issues related to sustainable land resources management which concern FAO.

313
 National level R.spons •
• nd support

Initiatives and
EXpreS{;n of Needs

Subnational level

Il.,pon •• . lnform .don

end t.ahnlcal
•••.• t.ne.
~------------~~

Initiatives and
v
O
expreSSIOnO°fn'::
-... 1 ~O<OI ., ••
groups

Consultations

~i~ i ~ i ~i ~ Stakeholders

Figure 1. Demand-driven interactive development (DDID),

Table 1. Scales 0/ interactive land use planning,

Level of analysis Scale/spatial resolution Issue
(*)
Fieldlproduction unit <1:5,000 Productive crops and animals conservation of soil
(site specific) and water; high levels ofsoil fertility; low levels
of soil and water pollutants; low levels of crop
pests and animal diseases.
Fannlvillage 1:1,000-1:50,000 Viable production systems; economic and social
(Iocalllevel) needs satisfied; awareness by farmers.
Country (nationaV sub- 1:25,000-1 :2,500,000 Judicious development of agro- ecological
national) potential; food production and food security;
conservation of natural resources and bio-
diversity; land degradation; public awareness.
Continentlworld 1; 1,000,000-1 :5,000,000 Land degradation and desertification;
(regional! global) preservation ofbio-diversity; population
development and food security; climate change
and agricuItural potential; awareness ofregional
and global institut ions.
(*) A range of scales is indicated at any given level of analysis. In practice the actual scale of an
application is selected according to the extent ofthe area and the availability of maps.

314
2.2. Land resources information requirements

Effective negotiation and decision making on land use cannot take place without a
knowledge base that is accessible and usable for all stakeholders. The two essential
components of the knowledge base are information.
A great deal of information is available regarding the specific problems of soil
productivity and water management and the methods to conserve the resource base.
However, the information needs are becoming more diverse, growing rapidly both in
quantity and quality as the spatial and environmental and socio-economic aspects of an
increasingly complex world are considered.
The information required is geo-referenced, that is it is location specific, and is
represented in the form of data and maps with two or three dimensional coordinates.
This includes information on topography, landform, soil, climate, water availability,
land use and cover, arable land, land suitability, land productivity, population, presence
of diseases (e.g. trypanosomiasis), infrastructure, land tenure, etc. In the prevailing
conditions of rapid change in various countries, the informal, gradual accumulation of
local experience as a basis for planning and management decision becomes a slow and
expensive learning process. More formal and rapid methods of information gathering
and interpretation are then needed to support decision making with a broader and more
systematic information base (Brinkman 1994).

3. Resource Management Domains (RMDs)

Interactive land use planning can be based on a spatial zoning of the land, water and
biological resources across the natural land units. In rural planning with an agricultural
production or environmental protection focus, zoning provides an intermediate product
in the land use planning process. The zoning itself involves a subdivision of the land on
the basis of the physical and biological or agro-ecological characteristics of the various
natural land units, primarily climate, soils, landform, land cover and land use as weil as
a mapping ofthe different socio-economic conditions. Together, they form the basis for
the delineation of land Resource Management Domains (RMDs, Figure 2), i.e. rural
areas where both agro-ecological and socio-economic conditions are so similar that one
can expect that development- or conservation-oriented intervention packages can be
successfully transferred from one site or farm to another (Sombroek and Egert 1996).
RMD has no apriori bias towards high-input and high-producing agricultural land
use but considers a wide range of uses which may satisfy the objectives of the
stakeholders. These objectives may be incompatible to a greater or lesser extent, and
they may change over time. The use of multiple-goal analysis enables the ranking and
periodic reassessment of objectives of stakeholders, and subsequent optimization will
support stakeholders' decisions towards optimum uses of the defined area. The
principal aims ofRMDs are as follows (Sombroek and Egert 1996):

315
1. to identify areas where particular uses may be encouraged through development
programs, services, financial incentives, etc.;
2. to identify areas with special needs and problems, as weil as areas which require
protection or conservation; and
3. to provide a basis for infrastructure development.

SOTER: Soil-and-Terrain unit INFRASTRUCTURE

~ ~er1ts - - ho<ostlnaol'-l .... ~ i""""..:I<IJtputt
,..!-•• roilds ," •• _. be"cbofmttorurblnlnftuence
no: TniditionalLand Occupation. type

=='. . .
\ow...,lo"""
.....eid _nisb IOib ",edomi.. ting tl- MOb!< og_ure: Ind~llond o........hlp;
p~1ng; tZ. rong< "'"M9tnlOnl:c!om.....,.11ond 0_.1hIp
RDM: land Resoun:e Management Domains
rl, r2, rl ...: MUS of ditfom1l o;opIIy<iuI+.odO'4<onomlc c.ondillo""
.. tht mult oIa>mbinolioN 01 AU. SOlU, TlO unilS ond ;nln,,1NCtInI
condltloru.
Figure 2. The concept 0/ land Resource Management Domains (RMD) and their spatial relations hip
with AEZ. SOTER, TLO units and inJraslrUctural conditions: an example.

The potential benefits of RMDs are the following:

1. the avoidance of haphazard occupation of the land under consideration, which may
lead to social conflicts and damage to the quality of the natural resources system;
2. the better understanding of the objectives, priorities and requirements of the
different stakeholders, thereby facilitating an eventual consensus for actual
implementation of land use plans through reconciliation of conflicting interests;
3. the harmonization of the work of national, district and local institutions that deal
with elements of land characterization, evaluation and rural physical planning.

RMDs deal with land resources as weil as people and their social organization. These
people comprise the actual or potential land users, which may consist of individuals,
communities or govemments that have a traditional, current or future right to decide on
the use of the land. In this sense RMDs are important elements of interactive land-use
planning. The RMD concept is applicable to all geographic scales and for lands of any
intensity of use. It is used for large tracts of land, such as major river catchments and
physiographic regions that have as yet a sparse human population, as weil as delineation

316
of land management units in detailed scale land-use plans at local area level. RMDs are
applicable to multi-scale land use planning. A typical design involves province or
district level land-use plans with large land management areas, within which are nested
village land-use management plans at detailed scale and for areas of interest. An
essential element of RMDs is their dynamic character; they can and should be redefined
or adjusted in relation to changing socio-economic conditions of the area concerned and
outside influences, such as local and regional market trends.

3.1. Proposed step-by-step zoning procedure tor the delineation 0/ RMDs

The generation of RMD information is similar to the procedure proposed by Sombroek

(1994) for the delineation of Ecological Economic Zones (EEZs) as developed in
Brazil. The procedure involves four main steps:

1. Collection of maps and spatial information and entry into GIS

2. Pre-zoning activities
- Delineation of natural land units through overlaying and analysis of the various
thematic layers: climatic conditions; landform characteristics; soil conditions; land
hydrology; vegetation; bio-diversity values; current land uses; incidence of pests and
diseases; near-surface mineral reserves and mining activities; river hydrology;
population density; land ownership, formal or traditional.
- Determination of the bio-physical land qualities and limitations, for each natural land
unit distinguished. This includes attributes such as landform, physiography, slope,
soil texture, soil phase, temperature regime, moisture regime/length of growing
periods.
- Identification of agro-ecologically viable land utilization types, and determination of
their bio-physical requirements in contact with stakeholders. The land utilization
descriptions comprise sets of alternative activities available to achieve specified
objectives, usually production of crops, fodder and fuelwood. Crop specific
information is described, such as cultivation practices, input requirements, crop
calendars, utilization of main produce, crop residues and by-products.
- Characterization of the socio-economic conditions, considering infrastructure and
markets; and perspectives for physiographic sub-region or municipality, and for
areas already demarcated for specific use. Socio-economic data include farming
inputs, population, land use, crop production, food demand.
3. Zoning sensu-strictu
- Systematic comparison, through a process of matching and weighing, of the bio-
physical qualities of each identified natural land unit with the requirements of each
envisaged land utilization type.
- Modification ofthe physico-biological rating through comparison with the prevailing
socio-economic conditions.

317
- Delineation of the RMDs by regrouping land units according to the results of the
physico-biological and socio-economic analysis.
4. Post-Zoning: Interactive land use planning
- A process of land use negotiation between the various potential stakeholders on the
basis of the objective inventory and evaluation of the natural resources conditions
and their matching with land utilization alternatives, and socio-economic conditions
within the RMDs, leading to a consensus on the future use of the various units of
land.
- Implementation of the agreed future use or non-use of the land: pre-projects for
legislation, political decisions; legal, administrative and institutional execution;
demarcation on the ground, inspection and control of adherence to the decisions.

4. Agro-ecological zoning for tbe delineation of Resource

Management Domains

FAO's Agro-Ecological zoning (AEZ) methodology provides a comprehensive framework

for FAO's integrated approach to the assessment, planning and management, and
monitoring of land resources (FAO 1978). The AEZ concept essentially involves a
simplified representation of land, a complex entity, by dividing it into major sets of
individual features and allocating a separate map layer to each feature, including
topography, landform, soil, climate, vegetation, land use etc. The combination or overlay of
the maps (Figure 3) produces agro-ecological cells. The AEZ cells are the basic elements of
agro-ecological zoning and of various other kinds of land resources analysis and mapping.
They can be arranged in various ways, depending on purpose, to create zones (at small
scales) and land units (at large scales) with distinctive, defined characteristics. Each zone or
land unit has a similar combination of constraints and potentials for land use, and serves as a
focus for the implementation of practices designed to improve the existing land-use
situation, either through increasing production and limiting land degradation in cultivated
land areas, and preserving natural vegetation in forested areas.
The AEZ approach is flexible and can be applied at global, regional, national and local
scales and for different purposes. Agro-ecological zoning may be applied entirely
differently under different ecological and socio-economic conditions depending on land
cover, land use, the development priorities of a country and the policy setting of
Ministries and institutions handling natural resources conservation and management.
There are no absolute agro-ecological zonings; units are defined according to specific
objectives and needs.

318
 Thermal zones ~ _ _ _ _ _ _ _---,

LGP zones ~
Patterns of LGP zones ~
Exploratory soil map Q
Cash crop zones ~ Land resources

- .-
Forest zones
- ~
--
b overlay

L Parkland zones
- ~

I Irrigation schemes liI

11

I Tse·tse infestation areas iI

- ,-
L Provinces and districts 11
I
I Digitizing and computer conversion
to grld cell database (1 OOha cell)

I
I
I
Quantified information on soll
mapping unit composition

I
I
Agro·ecological cells : Information in computerized (GIS) form for each cell. Thermal
regime (9 zones); LGP zones (15 zones); LGP·pattern zones (22 zones); Landform (49
types); Siope gradient (6 classes); Geology/Parent material (37 types); Soil unit (128
types); Soil texture (17 classes); Soil phases (17 types); Ca,sh crop areas (19 types);
Forest areas (3 types); Parkland areas (3 types); Irrigation schemes (19 major areas);
Tse·tse infestation areas (1 type); Provinces (8); Districts (41)

Figure J. Strocture 0/ Kenya land resou,.ces database /0" definition 0/ agro-ecological cells.

AEZ pennits the development of various kinds of applications, using a common

methodological framework and the agro-ecological cells. The range of applications depends
on the type and the number of thematic data layers inc1uded in the AEZ inventory (F AO
1996). A set of tools have been developed to implement AEZ applications. The tools
essentially integrate four components:

L Geographie Information Systems (GIS) databases and analytical and visualization

tools for rapid production of infonnation products. GIS utility derives from a
capacity for dynamic functionality based on the following three main qualities:
- the physical computing capacity to manipulate data, including overlay, join,
disaggregate;
the related capacity to query the data by fonnulating hypotheses for testing
assumptions defining potential relationships and developing theoretical constructs;
- the capacity to relate two-dimensional and three-dimensional locations of earth
features along with dynamic (time) four-dimensional processes.

319
2. Models for crop growth and estimation of both potential and actual yields. Crop
modelling has proved a valuable and multipurpose tool in land resources
management which can assist in the estimation of crop yields and the prediction of
crop shortfalls due to environmental risks.
3. Remote sensing techniques to characterize and map land use patterns and monitor
land use conditions. Human intervention determined by land resource management
activities cause land use changes. These changes result from the interaction in space
and time of the biophysical factors (constraints), and human factors like population,
technology, economic conditions, etc. Evaluation of the effects of land use changes
on the conditions of the land is essential to determine the sustainability of the
practiced land uses. Remote sensing techniques offer a unique way of quickly
assessing land use situation and trends and, in particular, to detect biophysical
degradation of the land due to improper use or mismanagement in implementing
land management plans. This will assist in determining correctives measures to apply
in successive land management activities.
4. Multi-criteria decision support (MCDS) to analyze optimal land-use scenarios.
MCDS tools facilitate interactive land use negotiations. This is because feasible real-
world solutions in interactive land-use negotiations are compromise solutions
resulting from trade-offs between various conflicting objectives, in order to fmd an
efficient and acceptable balance between the requirements of the stakeholders in the
land and the resources available. Different kinds of objectives may need to be
included, expressing not only economic values of land products but also addressing
goals which cannot always be expressed in monetary terms, such as bio-diversity,
people's preferences, or minimizing risk and uncertainty. Usually scenarios have to
be run a number of times and with varying inputs in order to identify a "best" or
even an acceptable solution. The elements of a solution are not fixed values, but
variable. They are often not clear-cut values but are values expressed within certain
ranges determined by resource availability and socio-economic realities. Many
options need to be examined to generate the information and the knowledge required
for the decisions and to quantify and display the trade-offs between conflicting
objectives.

The RMD concept can be seen as an expansion of the AEZ and the EEZ approaches.
RMD delineation is an advanced AEZ application. RMD integrates all elements of the
physico-biotic environment included in AEZ with the socio-economic environment. It
then matches both of them through multiple goal analysis, thereby providing a powerful
tool for the various stakeholders (land users) to negotiate a consensus on the optimal
use or non-use of the land to be subsequently executed through legislative,
administrative and institutional action on demarcated spatial units. AEZ tools can thUS
be used to prepare the information necessary for delineating RMDs.

320
5. GIS based Land Resources Information Systems (LRIS)
as AEZ/RMD tool

Rapid development in infonnation technology in the last decade has created a unique
opportunity for the integration of the AEZ tools into computerized systems called land
resources infonnation systems (LRlS). LRlS are designed as multi-purpose systems
which can be used to generate the infonnation products quickly and economically,
inc\uding images, maps, tabular and textual reports for the elaboration of various kinds
of infonnation useful for land use decision. LRlS have literally mushroomed in various
countries in the very recent past. They are everywhere becoming an integral component
of institutionalized programs of integrated resources planning, management and
conservation (Alfaro er al. 1994 ).
ln recent years FAO has gained valuable experience in the development of such GIS
based LRlS to address issues of improved integrated land resources management at
various scales. The FAO LRlS is conceived as a modular toolbox (Figure 4). The
toolbox contains AEZ databases and models, multi-criteria analysis tools and user-
oriented interfaces to enable easy and transparent interaction of the user with the
components of the system. The components of the toolbox are being gradually
developed, to incorporate existing tools and prospective tools, in collaboration with
specialized groups in and outside FAO and in interaction with field projects.

Data collection . screen ing

Map preparation

GIS
image processing
Spatial and tabular
database
=:l
Model base

Database
management

Analysis _
~
Decision Support
System

Figure 4. Flow diagram of FAO's Land Resources Information System lOo/box.

321
6. LRIS application to AEZ and RMD for land-use
planning in FAO

The FAO integrated approach includes LRIS as an essential tool for analysis of land
resources infonnation. LRIS multi-use resource databases support a wide range of
AEZIRMD applications at different scales.

6.1. Global and regional

The foundation ofthe global FAO LRIS is agloballand resources inventory, based on
the FAOIUNESCO soil map of the world (scale 1:5,000,000), which combines
infonnation on soil, landfonn, climate resources and to a certain extent land cover to
characterize land in tenns of agro-ecological zones. This global AEZ database, expected
to be available in 1998 on CD ROM, also contains infonnation on land suitability, land
productivity, and soil fertility constraints. Infonnation in the database has been used in
delineating and estimating extents of land classes within the FAO AT2010 study
(FAO/AT2010 1995). The land classes are related to the crop production potential of
land under rainfed conditions. The AT20 10 study established seven land classes
(denoted as land classes ATl to AT7), as listed in Table 2.

T.ble 2. Land with rainfed crop production potential. developing countries (exc/uding China).
MilIion ha
Class Name Moisture Land quality Potential In use Balance
regime (LGP in
days)
ATl Dry semi-arid 75-119 VS,S,MS 154 86 68
AT2 Moist semi-arid 120-179 VS,S 350 148 202
AT3 Sub-humid 180-269 VS,S 594 222 372
AT4 Humid 270+ VS,S 598 201 915
AT5 Marginally suitable 120+ MS 518 201 915
land in the moist
semi-arid, sub-
humid, humid
c1asses
AT6 Fluvisols/g1eysols Naturally VS,S 258 64 259
flooded
AT7 Marginally suitable Naturally MS 65 64 259
fluvisols/g1eysols flooded

Total with rainfed potential 2573 721 1816

Irrigated not suitable (arid and hyperarid) land 36 36
Grand total 2573 757 1816
Source: AT 2010. LGP: length 0/growing period. S: suitable. VS: very suitable. MS: moderately
suitable.

322
 Figure 5 shows the map of AT20 10 land classes for Africa. The land classes are
characterized by agro-climate and different levels and kinds of soil and terrain
constraints on crop production.

Domin~nll~nd

~
AH : H.."id
d • ...,.
t'KUUUl.,IIOWt LAl'IIU
_An: MoiSl semiarid
_ A n : s...b-hwnid
_
AT6: glC)'90lYtlu'IIisuls
MARGI ALL Y PRODUCTIVE LAND
ATI : Dry '5fflIi ~ .id
_AT5:MDSt!iClllilllid
An: Kle)'50lsItluvisol50
UNPROOUCTIVE LAND
.'
Pani)' suilable
_Mosdy non SUill,blc-
1'1- J 01: Suilablc
_ W IIl t:f"lllacie:rs

Figure 5. Extents 0/ AT2010 land c/asses in Africa.

The AT2010 land class delineations could be used as a basis for the delineation of
resource management domains by Iinking the database with other existing global and
regional databases, including sub-national boundaries, infrastructure, population, and
developing additional layers of socio-economic factors necessary to fully define and
delineate RMDs. At the regional level RMDs are useful for the elaboration of macro-
level development programs, such as the planning of regional agricultural research
programs, bio-diversity conservation and monitoring programs.

323
6.1. National and sub-national

FAO has developed a more sophisticated and detailed LRIS-based approach for country
AEZ applications. The system consists in a package featuring enhanced models and
more detailed land resource databases (scales 1:50,000-1:1,000,000). The package
includes decision support tools developed in collaboration with the International
Institute for Applied Systems Analysis (IIASA) to deal with multiple criteria decision
analysis (MCDA) problems (Makowski 1994). The MCDA method used is the so-called
Aspiration-Reservation Led Decision Support (ARLS) (Antoine 1997). The ARLS
method is implemented in a two-stage procedure. The decision maker first analyses a
core problem which is defined in terms of the biophysical and socio-economic
opportunities and constraints in the land management units or the areas concerned to
find a compromise solution. In a second phase the decision maker interactively
introduces his/her goals and preferences to the solution to work out other solutions that
are closest to his/her aspirations and reservations.
The flow diagram in Figure 6 shows the links between the various components of the
FAO AEZlLRIS system used in arecent Kenya study. The 26 activities indicated in the
boxes represent four groups of compound activities as folIows:
• (1,2) Formulation ofLUTs and their ecological requirements;
• (3-12) Compilation of land resource database;
• (13-19) Assessment of land productivity, including cropping patterns, linkage to
livest~k and forestry production systems and soil conservation;
• (20-26) Analyzing optimalland use scenarios.
The following summarizes the use of the system in making land use choices in
Kenya districts.
The basis of this application is a set of GIS-based AEZ land resource inventories of
individual districts in Kenya. The AEZ land resource inventories combine digitized map
overlays that relate to climatic conditions, soil inventory, administrative units and
selected properties of present land use, i.e. cash crop zones, forest areas, irrigation
schemes, Tsetse infestation and game parks. The digitized data were converted to a grid
cell or raster database. Each pixel represents one square kilometer (100 ha). AEZ
computer programs are applied to the district land inventories to analyze land suitability
and land productivity including cropping patterns, linkage to livestock and forestry
production systems and soil erosion considerations. In this way aland productivity
database is generated which contains quantified information on the productivity of all
feasible land utilization types for each agro-ecological cell in the districts. The land
productivity assessment involves 64 types of food and cash crops, pastures, and 31
fuelwood species and nine livestock systems which are grouped into 26 production
commodities, including 26 crop and 10 livestock production commodities. This
database provides the input to the ALDS optimization model.

324
 LAND U..!II~"'U'"

RESOURCES
INVENTORY
(GIS)

Figure 6. An LRIS configuration as applied in arecent Kenya country study.

Tbe area of Bungoma district in the Western province of Kenya is used in this
example. The distriet is situated on the slopes and foothilIs of Mt. Elgon bordering
Uganda. The distriet enjoys good agro-ecological conditions. Presently grown cereal

325
crops include maize, wheat, barley and finger millet. The land productivity assessment
shows a total arable land potential of about 200 thousand ha representing over 60
percent ofthe entire district. About 1/3 ofthis area is of only marginal quality. There is
good potential for beans and potatoes. In addition, cash crops can be cuItivated, such as
coffee, tea, cotton, pyrethrum and sugar cane.
In the 1979 census, the recorded population in Bungoma district slightly exceeded
0.5 million people. It is projected to increase to almost 1.3 million in year 2000. In the
past the growing population density has led to increased fragmentation of holdings.
Despite generally favorable conditions the district is facing increased pressure on its
resouree base and as a result enhanced intensification of agricultural production will be
required to secure future food supplies and adequate incomes.
The model accepts user specifiable scenario parameters from a control file, reads
crop, grassland and fuelwood production potentials by agro-ecological cells from the
land productivity database, reads livestock system related data derived from herd
structure models, and simuItaneously determines land use by agro-ecological cell as
weIl as supported levels of different livestock systems, feed supplies and utilization by
livestock zone and season. The model provides a framework for specifying different
types of objective functions and kinds of constraints.
The main issue here is to analyze the potential population supporting capacity of the
district under various land use scenarios, simultaneously considering several objectives,
such as maximizing revenues from crop and livestock production, maximizing district
self-reliance in agricultural production, minimizing costs of production and
environmental damage from erosion. Population supporting capacity, as defmed here,
relates the maximum potential of soil and climatic resources to produce food energy
and protein, at a given level of technology. An intermediate level of inputltechnology is
considered in this example (FAO 1991).
The multiple objective program inc1udes the following objective functions:
1. maximize food output (average yield/production);
2. maximize net revenue;
3. minimize production costs;
4. maximize gross value of output;
5. minimize arable land use (weight of I assigned to crop and 0 to grassland);
6. minimize area harvested;
7. maximize food output (minimum yield in bad years);
8. minimize total erosion (sum of all cell erosions );
9. maximize self-sufficiency ratio (minimum of the individual commodity group self
sufficiency ratios);
10. minimize maximum erosion (largest occurring erosion per ha in a cell is smalI).

These are defined in terms of three groups of decision variables which relate to
optimal land use, livestock numbers supported, and optimal allocation of potential feed
supplies to different livestock systems, respectively:

326
(a) the land use shares, i.e., the share Xkj of agro-ecological cell j allocated to cropping
activity k;
(b) the number ofunits Lsz oflivestock system s kept in zone z, and
(c) the feed ratio f;hsz offeed item I from crop i allocated to livestock system s in period
t in zone z. These variables form the columns ofthe LP matrix, the LP activity set.

For example, the mathematical formulation of objective 7 is to maximize the level of

self-sufficiency by commodity group:

max Z = min Sh
Xkj> L.. filts h

where Sh represents the level of self-sufficiency in product group h.

The constraints relate to preferred demand baskets, crop-specific production targets,
risk aversion, economic constraints, land use by individual crop and crop group, crop
mix, input use, quality of human diet, environmental conditions, seasonal feed demand-
supply balances, feed quality, and distribution of livestock systems. The core model is
composed of equations which involve the definitions of variables and the constraints. It
can be run to produce various Pareto-optimal solutions.
Before the interactive part of multicriteria analysis starts, a user selects the criteria to
be considered. The discussion presented here is based on results obtained for the
following subset of seven criteria (namely l:Food_val, 2:NetJev, 5:Arable,
7:Food_min, 8:Eros_tot, 9:SSR_v, IO:Eros_max).
The analysis was carried out in two phases. First, a pay-off table is computed which
is composed of the Utopia and an approximation of the Nadir points. The Utopia point
is found by single-criterion optimization for each criterion separately. Then an
approximation ofNadir is found (Makowski, 1994). The ranges between the Utopia and
the current approximation' of the Nadir points are used as a guide for the evaluation of
trade-offs among criteria, providing good initial information about the ranges of criteria
values that can be expected from any rational decision. A typical starting point is to
choose criteria values at the Utopia point as aspiration levels and values at the Nadir
point as reservation levels. This solution is called a compromise solution and is
computed before the interactive analysis starts, with trade-offs between the criteria
automatically calculated based on the Utopia point and on the current approximation of
the Nadir point.
Second, an interactive multi-criteria analysis was performed to search for acceptable
solutions. The analysis is typically composed of several sessions, during which a
number of solutions are generated and analyzed. A cycle composed of analysis of
previous solutions, selection of new aspiration and reservation levels, and optimization
is conventionally called an iteration. Usually several iterations are needed to explore
one region of the Pareto optimal solution space. Although a total of seven criteria were
considered, at different stages some criteria were examined more closely than others.

327
 The results of a sampIe analysis for Bungoma distriet are given in Table 3. The first
seven rows of the table contain the criteria values obtained from solutions for which
each criterion is optimized in successive single-criterion optimization runs (step 1). The
diagonal elements of the matrix represent the Utopia values for the seven criteria (i.e.
1197.2, 1316.6, 96.2, 1010.5, 1164.9, 1337.8, 12.2). The Nadir values are found by
taking the lowest values in the columns ofthe criteria to be maximized (i.e. Food_val =
742.6, Net_rev = 783.0, Food_min = 548.4, SSR_v= 1000.0) and the highest values of
the columns of the criteria to be minimized (i.e. Arable = 165.4, Eros_tot = 3527.0,
Eros_max = 227.8).
The last five rows of Table 3 contain the criteria values resulting from a session of
interactive multi-criteria analysis involving five iterations (step 2). The user interacts
with the software tool tbrough successive screens displaying graphs of the decision
variables, using mouse clicks to make the desired changes in values of decision
variables.
The results shows an irregular pattern of variation of the decision variables within
the sequence MCD-B ... MCD-E. Generally the increase in arable land use required to
achieve higher food production and self-sufficiency ratios appears to be associated with
increased total erosion; food production, economic return and food security in terms of
guaranteed minimum production in bad years and maximum erosion vary within narrow
ranges and seem to stabilize. Table 4 contains the acreage of the various crop
commodities involved in the production. Table 5 shows the aggregated commodities,
including the livestock products, with the corresponding food supplies and the
commodity self sufficiency rates. Sugar cane is the commodity with highest acreage
increase due to expansion of arable land use.
Given that the solutions produce self-sufficiency rates above the 80% minimum limit
which was established for the scenarios, the MCD-C solution appears to be a good
choice as it represents the relatively "best" optimal combination of values of the
decision variables. The results from these and other similar scenarios could be used to
delineate and map district RMDs with respect to food production potential for food
security. The RMD maps could be created using the land resources maps stored in the
GIS database.

6.3. Local area level

FAO is working on adapting the detailed country procedure further to applications at

local area level (scales 1:5,000 to 1:25,000). So far limited experience has been gained
in field projects at that level. In recent years FAO assisted LRIS projects in Nepal,
Jamaica and Grenada to establish GIS land resources database coupled with an expert
system program called Automated Land Evaluation System (ALES) (Rossiter 1988).
The ALES program implements the FAO Framework for Land Evaluation, which also
underlies the FAO AEZ methodology, and enables an integrated physical and economic
land evaluation (FAO 1976).

328
 Toble 3.Results 0/ ALDS analysis/or Bungoma distriet.
Bungoma Food val Net rev Arable Food min Eros tot SSR v Eros max SSR
Food val 1197.2 1082.6 165.4 969.7 3206.9 1204.0 112.8 96
Net rev 931.1 1316.6 126.4 717.9 2622.1 1000.0 85.4 80
Arable 742.6 789.2 96.2 548.4 1875.3 1000.0 85.4 80
Food min 1139.3 1071.2 161.1 1010.5 3256.5 1066.7 148.4 85
Eros tot 773.0 792.5 105.8 598.6 1164.9 1000.0 29.1 80
SSR v 905.6 1044.5 157.3 654.3 3527.0 1337.8 227.8 107
Eros max 746.8 783.0 121.0 574.9 1837.6 1000.0 12.2 80

MCD-A 1027.1 1075.5 127.5 813.6 2232.1 1184.7 73.8 95

MCD-B 1074.6 1007.7 150.9 857.3 2549.0 1234.5 32.5 99
MCD-C 1090.5 1054.7 161.4 875.7 2810.2 1229.9 30.0 98
MCD-D 1066.1 1038.9 161.9 846.8 3074.1 1250.8 33.9 100
MCD-E 1082.4 1041.8 163.1 865.5 2991.7 1239.9 31.3 99

Toble 4. Total harvested areas (.000 ha) byerop eommodity.

-- -----~-~-~-~--.-_.',"~-~ ------ -~--~-------- ----

Co~~~i!L ____~~J?:-~__~S_Q-B __MCP-C. MCD-D MCD-E

Barley 20043 18724 18414 19747 18954
Maize 73497 87284 86580 87647 87079
Oats o 7 7 7 7
Rice 3623 6543 6447 6686 6793
Sorghum 2766 3602 3056 3376 3538
Wheat o 50 50 148 136
Beans 32732 37051 37536 37193 37519
Pigeon pea 1876 1849 1695 1928 1779
Cassava 3527 5339 5450 5338 5328
Sweet potato o 1656 1485 1656 1656
White potato 627 1685 1145 2474 2158
Banana 7774 9824 9889 10157 10038
Sugarcane 5831 1081\ 18251 16446 18211
Total 152296 184426 190004 192803 193195

Toble 5. Food supplies (million keal) and self-sufficieney ratios (SSR) by eommodity group.
Commodity MCDA MCD-B MCD-C MCD-D MCD-E

Supply %SSR Supply %SSR Supply %SSR Supply %SSR Supply %SSR.

Cereals 124871 95 130114 129630 98 131831 10 130674 99

Pulses 12308 95 12825 99 12778 98 12995 100 12881 99
Roots 46598 95 47577 99 47507 98 48175 100 47752 99
Sugar 16061 95 25066 149 41686 246 37300 220 41380 244
Bananas 43646 47 45481 49 45311 49 46082 50 45677 49
Fuelwood 43646 95 45480 99 45310 98 46081 100 45676 99
Meat 10126 95 10551 99 10512 98 10691 100 10597 99
Milk 109912 157 70849 101 74555 107 69951 100 69337 99
Eggs 873 95 910 99 906 98 922 100 914 99

329
7. Future activities

Upon finalizing and publicizing the new conceptual and methodological approach in
1995, FAO has been searching for new ways to provide assistance in this area
responding to requests from an increasingly greater number of countries. Initial
discussions were held with relevant officials in Asia, the Pacific, Africa and Latin
America. It is anticipated that field application of the new approach to land-use
planning will depend heavily on networking and information exchange between land-
use planners in development institutions of developing countries.
Network development was initiated in 1996. Besides this, the program as a whole
will probably involve a mix of different levels of technical assistance, depending on the
needs of individuals countries, but including workshops and training courses for single
countries and groups of countries and small funded programs. Already such programs
are being formulated or are in the execution phase in countries such as Sierra Leone,
Tanzania, and Zimbabwe. These programs are intended as the initial activities in
assisting countries to leam about the integrated approach and to develop and implement
comprehensive, long-term programs to apply it in the planning and management of their
land resources.

8. Conclusion

As the paper illustrates, there are elose links between AEZ, RMD and interactive land-
use planning. LRIS systems are cost effective tools which can produce the information
products in a timely and efficient manner, including RMDs, needed for interactive land-
use planning and land-use decision making. Until now many institutiops seem to be
more concemed with the database and hardware and software components of LRIS. The
role of LRIS as an information source to the user community to support the decision
making process in practical situations is still being neglected. In many cases there is a
wide communication gap between the technical experts producing complex maps and
other information products and the users, who need easy-to-comprehend information
products that are relevant to the tasks at hand. Lack of training of the stakeholders in the
use of the information products in their practical land use planning work is also a
constraint, which causes the products to be under-utilized or sometimes not used at all.
Appropriate training programs need to be designed to tackte this problem. Above all,
the collaborative work of all the actors, ineluding decision makers, technicians and
stakeholders is essential to the success of such integrated approaches.

330
References
Alfaro, R. et al. (1994) Sustainable Land Use Planning in Costa Rica: A Methodological Case
Study on Fann and Regional Level, in L.O. Fresco, L. Stroosnijder, 1. Bouma and H. Van
Keulen (Eds.) The Future ofthe Land: Mobilising and Integrating Knowledgefor Land Use
Options, Wiley, New York.
Antoine, 1. (1997) Multiple Criteria Land Use Analysis, Applied Mathematics and
Computation, 83:195-215.
Brinkman, R. (1994) Recent Developments in Land Use Planning, with special reference to
FAO, in L.O. Fresco, L. Stroosnijder, J. Bouma and H. Van Keulen (Eds.) The Future ofthe
Land: Mobilising and Integrating Knowledgefor Land Use Options, Wiley, New York.
Dent, F.J. (1995) Elements Under Threat: Degradation of Land and Water and Food Security,
unpublished manuscript presented at the Seminar: Towards Sustainable Food Security in the
Asia-Pacific Region, FAO, Bangkok, Thailand.
ESCAP (1990) State ofenvironment in Asia and the Pacific, Bangkok.
FAO (1976) A Framework for Land Evaluation, SoUs Bulletin 32, Rome.
FAO (1978) Report on the Agro-Ecological Zones Project (1978-1981), VoLl: Methodology
and Results for Africa, World Soils Resources Report48/1. Rome.
FAO (1991) Agro-Ecologicalland resources assessment for agricultural development planning:
A case study ofKenya, World Soils Resources Reports 7111-9. Rome.
FAO (1995) Planning for Sustainable Use of Land Resources: Towards a New Approach, Land
and Water Bulletin 2, Rome.
FAO (1996) Land and Water in Latin America and the Caribbean Region: Problems and
Solutions. An international Scheme, Towards Sustainable Agriculture, Rome.
FAOIAT2010 (1995) World Agriculture: Towards 2010, a FAO Study, FAO and 1.Wiley &
Sons, New York.
Makowski, M. (1994) Methodology and a modular tool for multiple criteria analysis of LP
models, Workingpaper WP-94-102, nASA, Laxenburg.
Rossiter, O.G. (1988) ALES: a microcomputer program to assist in land evaluation, Proceeding
International Workshop on Land Qualities in Space and Time, Wageningen, the
Netherlands, 22-26 August 1988, Pudoc, Wageningen.
Sombroek, W.G. (1994) Introduction to the philosophy, concepts and methods of ecological-
economic zoning, Manaus Workshop on Ecological-Economic Zoning in the Amazon
Region, 25-29 April 1994, Rome.
Sombroek, W.G., Egert H. (1996) What do we understand by Land Use Planning: a state-of-
the art report, Entwicklung laendlicher raum, pp 3. In Beitraege zur Internationalen
Zusammenarbeit, 30.

331
 Part V
looking ahead
Land-use trends in Europe: simulation, presentation
and scenario evaluation of future options

Henk J. Scholten l ,2, Paul Misdorp 2 and Martin van Beek l

}Faculty 0/ Economics, Vrije Universiteit, Amsterdam, The Netherlands
2Geodan, Amsterdam, The Netherlands

Abstract
This paper describes the structure ofthe Scenario Evaluation System (SES), a system developed
to generate spatial forecasts for various types of land use in Europe. The paper describes the
policy framework and the development of rural areas in Europe. However, the emphasis of the
paper is on the model itself. The model includes all relevant land-use types such as residential,
industriaI, agricultural, natural areas and water. The model can be used to forecast the pattern
and distribution of these land uses in Europe on the basis of different scenarios, and under the
influence of different policy frameworks. These predictions can be used to compare land-use for
different areas and at different times. This type of evidence is meant to support policy analysis
and the policy making process, making it possible to identify the effects on land use of
economic and environmental polices.

Keywords: Geographical Infonnation Systems, decision support, evaluation criteria,

land-use models.

1. Introduction

The development of rural areas in Europe is subject to many influences and is expected
to show many changes in the decades to come. The struggle for space - a recurring
theme in spatial planning - is a struggle among different types of land use. The question
of how the future land-use pattern will appear, for example, in the year 2025 depends
on various fields of influence.
First, the crisis in agriculture should be considered. At a European level there is a
paradoxical situation. The constantly growing productivity of soil, labour and animal
yield is the source of so many negative effects that efforts have to be made to achieve a
lower production volume. The almost autonomous trend towards an increase of
productivity raises, in particular, four fundamental issues that will lead to a less
dominant agricultural production function in the long run:
• a structural surplus of certain agricultural products;
• a decline in the income level of fanners;
• the systematic exhaustion of the soil;
• a growing public concern about agricultural production methods.

335
 Furthermore, future land use will be influenced by aseries of homogenising
processes, causing rural areas to show more and more urban features. Urban areas, in
the meantime, become greener and more mixed in land-use terms. Sub-urbanisation,
transport and industrialisation are causing a gradual disappearance of the difference
between urban and non-urban areas. Besides an evident urban push, there is also a re-
appreciation of nature and recreation space at a larger distance from the city, and of the
green structures within urban areas as image-setting elements. As an extension to this,
the claims of nature and the environment are undeniable.
Current production methods lead to great acreages with the same crops, to an
extremely high livestock intensity (scaling up), an explosive growth in the use of
nutrients (fertiliser and feed) and pesticides, a high energy consumption, and a deep
extraction of water (water consumption). These induce forms of land reallocation or
land planning extremely detrimental to the landscape. Since the negative side of these
production methods is becoming clearer from an ecological point of view, the goal is
now increasingly shifting towards sustainable, environmentally compatible agricultural
production.
These trends suggest an autonomous evolution that seems to be dictated by market
forces rather than by government policy. In reality this is not the case. Although there
are different visions regarding the relationship between market and government within
Europe, European governments have, irrespective of the country in question, a
substantial influence on land-use changes. A clear example is the agriculture sector.
Within the range of policy instruments for agriculture, subsidies and production quotas
for cultivation are widely used, indirectly controlling the type of land on which the
production takes place. There is also a more integrated, multi-facetted, spatial policy
which focuses on a mix of incentives (for example, new economic activities in
peripheral areas) and constraints (for example, compliance with minimum standards for
the preservation of environment, nature and diversity). The objective of these
instruments is to promote a mixture of land use which corresponds to a balance between
economic activities and the environment.
Within Europe, there is a growing demand for a closer collaboration between
governments and institutions as regards land-use policy, environmental management
and spatial planning. To achieve this goal, there is a strong need for :
• more knowledge/information about European rural areas and the processes that
influence them;
• the development of visions for the plausible evolution of rural areas;
• the acquisition of a flexible spatial information system, which is able to identify
possible land-use evolutions, and to stimulate consultation among governments about
land-use trends and policy influences.

The need for geographical information (covering a large part of Europe) about
present and future land uses, coupled with a flexible and user-friendly system, requires
a specific research and presentation platform. In Douven's research, a comprehensive

336
description of such a system is given (Douven, 1997). Under a contract from the
commission ofthe Dutch Public Planning Service (Rijksplanologische Dienst), a spatial
information system which satisfies these requirement was developed (Geodan, 1996).
This system can be considered as a step towards the creation of a work environment
where the user can integrate, analyse and present all data needed to assess present and
future land use. In addition, qualitative information can be added by means of texts and
images. This Scenario Evaluation System (SES) puts a heavy emphasis on GIS-
functionality in the forms of spatial data, spatial analysis and map presentation.
There is no unequivocal answer to the question of how European landuse will appear
in the coming decades. The large uncertainties about the economic and social systems,
and also about the physical and ecological systems, make it impossible to produce
precise forecasts of land-use in the medium- and long-term. However, it is possible to
sketch a number of credible and predictable future scenarios with the current knowledge
and technology, based on sector-specific knowledge and on the knowledge of the
development processes for policy, and the socio-political context. On the basis of a
number of assumptions in the political-economic sphere, a scenario-based methodology
can be used to produce aseries of coherent and plausible future land-use images (see
section 2). In the model presented in this paper, these predictions are based on a
relatively simple supply and demand model, where the land price determines the share
of land allocated to a certain land-use type (see sections 3 and 4). This type of model
can be used to generate an almost unlimited number of future images, based on
different assumptions and parameters of the supply-demand models. These future
images can be evaluated from various perspectives, such as spatial quality or
environmental quality (see section 5).

2. Conceptual model

In order to be able to simulate, evaluate and present the future land use in an acceptable
manner, it is important to gain insights into the current land use and the estimated
demand for various types of land use at a given time. Current land use, scenario
assumptions and policy choices are the main inputs for estimating future land use as
shown in Figure 1. In this structure, future land use depends on a number of
endogenous development processes that are characteristic of rural areas, and on the
external influence of socio-economic scenarios and policy measures. The endogenous
processes of rural land-use development include: industrialisation, complex formation,
scaling up, extended business development, economic innovation and marginalisation.
Scenarios (such as a certain level of economic growth and international trade) and
policy factors (such as the introduction of subsidies, levies and quotas) influence these
endogenous processes and thus the demand for a certain type of land use. 80th
scenarios and policy choices increase or reduce the demand function for certain types of
land use.

337
 Figlll'e 1. Relationship between the key elements ofthe SES spatial information system.

On the basis of this scheme, different future land-use images can be produced,
compared with each other and with the present land use. These differences can be
associated with scenario assumptions and policy choices, thus making it possible to
assess their relevance and effectiveness. The model also includes a number of
evaluation possibilities. Land-use images can be evaluated, for instance, from the
employment potentiality, or the enjoyment value. The significance of this approach for
spatial planning is evident, since certain developments can c1early lead to spatial
consequences that may or may not be desired.

2.1. Future scenarios /0' Eu,ope

The SES system is based on four scenarios: the growth scenario, the distribution
scenario, the sustainable development scenario and the crisis scenario. These scenarios
are characterised by: (1) different types of market and government influences; (2) the
role oftechnology and technological innovation; (3) the relationship between economy
and ecology; (4) the spatial structure; and (5) the spatial pattern as a combination of
different types of landuse.

2.1.1. The growth scenario

The growth scenario is characterised by a constant push towards higher levels of
agricuItural productivity. This leads to more intensive and less land-bound production
methods, production-intensifying technological innovations and higher inputs of
production materials such as feed, fertilisers and chemicals. Maximisation of the yield
is achieved, on the one hand, through scale increase (for instance, an increase of the
average farm size) and, on the other hand, through specialisation (for instance, through
knowledge-intensive and high value products). The most important consideration
behind this scenario is cost effectiveness. This implies a reduction in labour intensity
and a decline in land area.

338
 In this scenario the govemment plays a minimal role. This corresponds, for example,
to the liberalisation of agricultural policy (free trade), to the reduction of agricultural
expenditures at European level, to the reallocation of production means within the
agricultural sector, to the withdrawal of protection-stimulation measures such as quota,
set-aside, production ceilings and tariff structures.
With regard to management of the environment, environmental quality, nature and
landscape, the govemment role is also very limited. However, the se\f-regulating
mechanisms of the market do not lead to a reduction of damaging emissions or to
environmentally responsible behaviour. On the contrary, this scenario accentuates an
almost unlimited use of natural resources. With growing consumption and production,
resources are more severe\y depleted. The food chains and world trade continue to
increase transport levels and energy consumption, partly because energy becomes
cheaper in this scenario.
Nature and landscape are seen as economic goods, which implies that consumers
have to pay for their desired nature product, and that producers (for example, the farmer
of the future) are paid for this output.
Within this scenario, and for the whole of Europe, the largest amount of agricultural
land will be transformed into non-agriculturalland, primarily in favour of infrastructure
(roads), housing (urbanisation), and also nature.

2.1.2. The distribution scenario

The distribution scenario is based on the assumption that agricultural surplus is the
implicit result of an autonomous increase of productivity. While in the previous
scenario the emphasis was on maximisation, rather than optimisation, of economic
output, in the distribution scenario the core issues are efficient production and
allocation ofresources. Due to this, agricultural prices will not be lower. Consequently,
intensification, complex formation and scaling up of agricuIture will occur to a lesser
extent, compared to the growth scenario. Moreover, this scenario emphasises the
protection ofthe employment level ofthe agricuIture sector, which implies a (moderate)
protection of the European market. Since free-trade will also have to be guaranteed,
which implies a decline in product prices and income for farmers, the govemment will
have to act in compensation. However employment, rather than social security
measures, is stressed. This means, besides technology to increase productivity, also
more investment in other farming methods, such as extended agriculture (including
nature management and education).
Nature and landscape can only partially be considered as economic goods in this
scenario. The nature that is essential for physical and economic life will require
govemment intervention in the long-term. The govemment must ensure that nature be
created and managed in the most efficient way. Nature which is not so essential (a form
of "luxurious" nature) is left to market dynamics: the consumer pays for a desired
nature product and the farmer receives a payment for it.

339
 The emphasis in this scenario lies in the increase of efficiency and optimisation of
existing means and methods ofproduction (for example more intelligent use offertiliser
and less use of chemicals), which guarantees cleaner production.
With regard to the role oftechnology, it can be said that it generally falls short ofthe
possibilities it offers: the intrinsic value of labour and animal production is of
paramount importance. This results in a weaker demand for advanced technologies,
such as bio- or information technologies.
The role of the government is c1ear with regard to environment, nature, and
landscape. There is a wish for co-ordinated activities to prevent, for example,
acidification, degradation of ecosystems and the extinction of flora and fauna species.
As far as the energy price issue is concerned, this scenario assurnes amiddie position
between the energy prices of the growth- and of the sustainable-development-scenarios
(see below). Also in comparison with these two scenarios, there is the smallest
agricultural land reduction in favour of non-agricultural activities. The main reason is
that agriculture will not lose its importance in terms of employment and spatial use.

2.1.3. The sustainable development scenario

The sustainable development scenario is based on a radical change in the production
structures of the European and global scales. This will affect mobility and consumption
patterns driven by environmentally aware and ecologically responsible consumers who
want high quality and environmentally compatible products. Technologies will also
change, so that lower emission levels and cleaner input for production can be achieved.
The critical question in the sustainable development scenario is how far to limit the
exploitation of resources (energy, spaee, natural resourees), since this exploitation
corresponds to a loss, not an increase, of wealth. This scenario advocates the decrease
of the amount of energy and material per product/service unit. As a result of energy
being made expensive, businesses are compelled to increase efficiency and account for
energy scarcity. The .management of chemieals in intensive livestock breeding is an
example of this kind efficiency push: by applying a levy on phosphates, nitrogen, etc.,
which the soH and crops can no longer absorb, farmers are forced to make use of
minimal nutrient inputs. Therefore, environmental considerations are integrated into the
farm operations. A decrease in environment degradation can then be achieved, and
above all, production costs can be controlled (for example, fertiliser costs can decline
because of abundance and low prices of organic household waste and minerals from
animal manure).
In the sustainable development scenario, moderate economic growth is possible
thanks to the fact that production is carried out in an ecologically responsible manner.
Environmental limits are strictly applied especially in cases of industrialisation and
complex-formation in the agricultural sector. The relation between industry and land
management can take its own shape, but the use of cultivated biomasses and waste
products are very important in this scenario. Wastes such as straw board, waste from
wine production, stalks and other fibres can be separated from valuable parts and be

340
converted into fuel, fibre and chemicals. Carbohydrate-rich crops, such as artichokes
and sugar beet, can be converted into valuable chemicals, and even pharmaceutical
products can be made through bioconversion and fermentation. Marginalization
processes in economic and especially ecological terms will be slowed down
substantially.
The degradation of natural areas and cultural identities are major concems in this
scenario. It is realised that existing agricultural industry must be complemented with
functions such as managemen.t of nature, less intensive forms of recreation and other
forms of agricultural diversification. The govemment has, in the field of clean
production, the task of setting rules and ensuring that they will be respected. This is
especially the case in the production of agricultural goods: less will be produced
because there will be less demand for products. Oue to this, there will be a movement
from bulk production (scale increase) to quality production. Technological development
will take place within these constraints, implying that selective developments will be
pursued aimed at increasing the quality of life, environment, flora and fauna (for
example biotechnology aimed at health improvements and a decrease of environmental
pressure). Furthermore, there is a role for government in terms of nature and landscape
management. Nature and landscape are considered to be so essential for the physical
and economic life in the short- and long-term that the government must ensure
measures to stimulate conservation and proper management. The private sector can play
a role in this, but only within given limits.
As far as the extent of agricultural landuse is concemed, this scenario shows an
average land reduction in favour of non-agricultural activities; less reduction than in the
growth scenario, but more than in the distribution scenario. The land made available is
allocated to nature, in a productive as weIl as non-productive way. Allocation of
agricultural land for housing and infrastructure is less common.

2.1.4. The crisis scenario

The crisis scenario is characterised by an uncontrolled growth of high value
technological applications in core agricultural areas. This technological development,
which is solely aimed at promoting production, leads to a number of shortcomings.
First, substitution of labour by automation occurs on a large scale, causing an increase
in unemployment in the central areas. Second, the prosperity growth which takes place
in the central areas corresponds to a decline in the prosperity of peripheral areas. Low
levels of investment in the rural areas cause migration flows to the urban areas and,
consequently, to depopulation and unemployment. In peripheral areas, agriculture will
strongly attenuate or even become marginal.
Within this scenario, there is a very sharp contrast between economic and ecological
values. Automation and economic growth lead simultaneously to a strong degree of
spatial inequality. The consequence is an emphasis on scaling-up and intensification of
production processes, together with fast transports which supply the processing
industry.

341
 From the economic and also the ecological points of view there is a trend towards
marginalisation of the rural areas, especially in the more remote areas. Although
industrialisation, complex formation, and scaling up lead to marginalisation in any case,
the effects are exacerbated in this scenario. Large areas will no longer have a function
and will be impoverished due to the absence of govemment intervention. Even in
central areas a form of marginalisation will occur, due to the decrease in the quality of
production and environmental quality caused by ground-, water-, and air-pollution.
The strong industrialisation and urbanisation trends show other drawbacks. Not only
water pollution causes a threat to the public health, but also high water consumption
combined with the result of c1imate changes threatens to cause serious water shortages.
The water problem is a characteristic of the crisis scenario, especially in relation to the
agricultural sector. Scarcity of water, as weIl as waste of water, are both very
significant.
Due to high u.'1employment levels and the negative environmental effects on urban
and rural areas, economic growth, in terms ofGDP increase, does not necessarily occur.
This has consequences for land use. In the central areas, intensive agriculture will
strongly increase, while land use for other forms of agriculture will be replaced by
economic activities, housing and infrastructures. While in urban compounds agricultural
land use will be more intensive, or will be transformed to urban landuse, in peripheral
areas there is a trend towards attenuation of agriculture and increase of abandoned land.
All considered, this scenario is characterised by a strong transformation of land into
urbanised land, and an extreme tendency of some areas towards marginalisation.

3. The model for landuse in Europe

This study emphasises rural land use. Other land-use types, such as housing, industry
and services, are considered as exogenous factors which drive the development of rural
areas. In other words, the SES system does not include feedback mechanisms between
agriculture and other land-use types.
The units for spatial analysis in SES are either grid cells or NUTS-I1 areas. Larger
spatial units are used in some cases, such as regions, countries or even Europe as a
whole. These areas correspond to the spatial scales within which economic processes
and market interactions take place. The markets for agricultural production, for
instance, often function at anational, continental or even world-wide level. The model
itself cannot include supply and demand functions for this type of global market. In
such a case, the model is used in a top-down fashion: the results of other economic
models suitable for large scale analysis are considered as input for the land-use model,
which translates this information in terms of land use at a low-scale spatial level.
This section will look in broad terms into the various parts of the model, which are
shown in Figure 2.

342
 Landuse (1.,) Poliey (NUTS-II)

Data on development Minimax growth mies

progress (grid-tells) (NUTS-II)

Calculation growth
rate (Iu-type)

Legend
Calculation prite ~
~~~_~
__~I_·v~~(I_~~type~)J'"
Model Input

ICalculation step I Subsidy and taxes

(NUTS-II)

Availabil~y ofland (lu-type)

Figure 2_ Basic components ofthe SES model (lu=land use).

3.1. The models for land-use types

3.1.1. Land use for housing

A standard model of population dynamics is at the basis of the housing land-use model.
The size of the population in a spatial unit at a certain time t is equal to the size of the
population at time t-I, plus births, minus deaths, plus the net migration.
A simple spatial interaction model can be formulated for the net migration, where
migration is directed mainly to neighbouring spatial units where living is attractive (for
example, because of low housing density). It is also possible to take international
migration into consideration, but in this case estimates from exogenous sources need to
be used.
Once the population dynamics is known, the consequences for landuse can be
estimated through a set of time-dependent land-use coefficients. These coefficients
depend on income levels and social variables. Income drives the demand for living
space, which increases as income increases. Social changes, such as individualisation,
lead to smaller households and consequently to a greater use of space per capita.

343
 In this model, the use of land for housing is driven solely by demand. There is no
feedback mechanism which includes the supply of space. It is also assumed that other
landuse types disappear ifthere is a demand for housing in a certain area.

3.1.2. Model for industry, services and trade

The model is based on the total production level of these sectors. In its simplest form,
the total production in a spatial unit r at time t depends on the production in rand on
the total production in all other regions. A spatial interaction model expresses the
degree of the correlation between regions. The top-down character of the approach is
evident from the time-transition mechanism, based on exogenous growth rates
established at macro level. Different growth rates, however, can be selected for
different regions. The structure of the model for the total production of industry, service
and trade is the following:

Y.,I = [a· Y.,t-I + (1- a)· Les>:'.I]· GR (1)

where Yr.1 is the total production for region r at time t, a is the share of total production
which depends on r alone (0 ::; a ::; 1), es is the coefficient of trade relationship between
region s and region r, and GR is the exogenous growth rate.
The es coefficients are mainly influenced by transport costs. Improvements in the
infrastructure of certain regions are expressed by a uniform decrease of transport costs
in relation to those regions. Trade barriers, or the removal of these barriers, can be
included in the model with a similar mechanism.
This approach serves to calculate the production level of all spatial units
simultaneously. From the computational perspective, this can slow down calculations a
great deal. To speed-up calculations, two solutions can be used: (I) the production level
is calculated for smalI, relatively free-standing regions; or (2) Ys. l • j is approximated with
1'.;, in which case all calculation problems vanish, since the new production values for
each spatial unit can be calculated separately.
After the calculation of the production values, the consequences for land use can be
estimated through a set of time-dependent land-use coefficients (similar to the model
for housing). If the model is extended into a multisectoral approach, landuse intensities
can be calculated per sector.

3.1.3. Model for other land uses

The models for housing, industry, services and trade produce spatial claims that must be
incorporated in the remaining landuse categories. The rest of these landuse categories
show their own dynamics. for example linked up with development processes in
agriculture. The government can also direct these developments. for example through
subsidies or constraints for certain types of land use. Because of the large number of
spatial units included in the spatial database. and to limit the computation time, the

344
remammg land use types are analysed separately for each spatial unit. Correction
mechanisms are included to make sure that regional or areal balances of land-use types
are respected.
The suggestion is to use a demand function that has the following structure:
• The cost of preparing the soil for a given land use is paid for by the user of the land.
This implies that the demand curve for a given land-use type decreases as the land-
use share increases.
• The demand function is raised or lowered when subsidies or taxes are applied.
• These demand functions are aggregated per spatial unit. The balance between supply
and demand can be found by comparing the aggregated demand with the amount of
available ground in the spatial unit (from which the space for housing, industry,
services and trade has already been subtracted).

This approach leads to a elear pattern of landuse changes per period and per spatial
unit. For certain types of landuse there might be restrictions at regional level (for
example, restrictions which apply to the agricultural production of a certain type in a
certain area). This leads to an iterative process, where the landuse in all areas is
recalculated, but now with adjusted values for parameters of the demand function,
which guarantee that the results are elose to the aggregated ultimate values.

4. Landuse: from present to future land use

The starting point for calculating future land use is the current land use. Reliable estimates
of the current situation are thus needed. The parameters of the housing, industry, services
and trade models also need to be estimated before running the model.
The definition for the types ofland use are based on the standard land-use database of the
Dutch Institute for Public Health and Environmental Protection (RIVM: Velde et al., 1994).
The following eight types ofland use are considered in SES:
• Farmland and pastures for agriculturalland use:
- Non-permanent crops, such as farmland under continuous use; fields temporarily
unexploited on at least 50% of the surface; cornfields.
- Improved pastures, dry as weIl as wet pastures, including rangeland or land used
intensively for fodder.
• Permanent crops:
- Vineyards, orchards and pastures, ineluding vineyards altemated with grass strips or
other plants;
- Olive groves, orchards and pastures;
- Soft fruits, citrus fruits, stone-fruits, nuts, mixed orchards and berries.
• Deciduous forests:
- All forests and wooded areas with mainly deciduous trees; productive as weIl as non-
productive forests and wooded areas, including production forests as weIl as (serni)
natural recreation forests.

345
• Evergreen forests and mixed forests:
- All forests and wooded areas with mainly evergreens or mixed forests, productive as
weil as non-productive forests and wooded areas.
• Inland waterways:
- For inland waterways and lakes the WDB2 (World DataBase) is used, where only large
inland waterways in Europe are present. The data about rivers are not used due to their
limited size.
• Urban areas:
- This class originates from the DCW (Digital Chart of the World) and contains only
large urban areas in Europe. Airports, large trafiic jWlctions, industrial areas outside cities
and other built-up areas are not included.
• Others:
- Abandoned pastures, dry as weil as wet pasture;
- Non-cultivable land: rocks, glaciers, sand dWles, swarnps and peat areas;
- Areas with an extremely low productivity: scattered pastures, (reindeer) moss, rocks
and low shrubs, permanent snow cover (>6 months), areas that predominantly consist of
swarnps and peat bogs.

In the above classification, the category ''nature'' rnay be estimated in an indirect fashion.
For forested areas, for instance, it can be estimated from the ratio between productive and
non-productive forest.
In this classification there is a clear emphasis on rural land use. Rural land use raises
several issues (cf. Whitby and Ollerenshaw, 1988), besides the amoWlt of land allocated to a
specific type of use. In particular:
• the intensity of each type of landuse;
• the location of land belonging to each type of landuse;
• the general level of economic activity in rural areas.

The type of classification chosen does not allow for a systematic comparison of land
uses based on "intensity" of uses. In the database there are percentages of all various
types of landuse, but it is not possible to derive from these figures how intensive the use
iso From the earlier discussion on the development processes, however, it is clear that
the trend is from intensive to attenuated land use.
The focus in this study is primarily on used space, that is on areas characterised by
location and surface to which fWlctions are attached (which is not necessarily the same
as actual users of the space).
In order to specify future land-use images for the whole Europe, it is important to
consider land use in relation to location and spatial scale: the same current land use in
two different areas does not necessarily imply the same land use in the future, even if
the same prediction models apply to the two areas. A form of regional correction is
necessary to take into account factors like "length of the coast", "distance from the
capital", "density of the transport network", "elevation", and "number of telephone

346
connections" (which gives an indication of the centrality or peripherality of an area).
The Eurostat data that underlies this regional classification make it possible to come to
a meaningful regional correction. Moreover, other variables are included in the
determination of regional indices, such as suitability of soil for certain crops, landscape
structures (such as corridors or protected areas) and productivity of the land (as an
indication of economic development). Furthermore, in the design of a Scenario
Evaluation System, attention is paid to sector-oriented economic activities that in some
way find their expression spatially. That means that besides the focus on the agricultural
sector, information is gathered about, for example, forestry, nature and landscape
management (for example, LNV, 1993) as well as recreation (cf. Klundert et al., 1994).

5. Future land-use images

The SES system produces simulated land-use images for Europe, individual states or
smaller areas, at different times in the future. The results are presented in maps. Figure
3 shows a sarnple of the resuIts of the SES system for the year 2010 under the four
scenarios. As can be seen, the amount and pattern of different land uses are significantly
different for different scenario assumptions. These future images raise important
questions concerning the consequences for the future, the desirability of the outcomes,
the possibilities and opportunities for changing policies, and the co-operation between
govemments at local as weil as national levels. A key issue in this respect, besides the
visualisation of land-use changes, is the evaluation of these land use images so that
better support to policy makers can be provided,

5.1. Evaluation offuture land use

The concept of spatial quality is an interesting not ion developed by spatial planners.
This concept regards space, society, and the co-ordination between the two, so that an
evaluation scheme for future land use can be identified (Petrus, 1989). Within spatial
planning, there is a growing emphasis on the adjustment of quality rather than
expansion of quantity. This emphasis on adjustment, management and maintenance of
space is undoubtedly related to the (perceived) pressure on space, negative
environmental effects, and the people's concerns and criticisms about the current spatial
organisation.
AIthough the idea and the operationalisation of this concept is shaped by the Dutch
experience, where the spatial pressures of urbanisation and agricuIture are very high,
the concept of spatial quality can be adapted for other areas in Europe. Spatial quality
has two dimensions: an objective and a subjective dimension. In the first case, spatial
quality can be related to the land-use composition in rural areas, which leads to a
productivity or an employment-value of land.

347
 Current situation Legend
•
c::::J Intensive agriculture
c::::J Extensive agriculture
_ Permanent crops
_ Forests
_ Urban areas

Growth scenario Distribution scenario

Sustainable deve lopment scenario Crisis scenario

~
~~ •..... 1

Figure 3. Example of outputs ofthe Scenario Evaluation System. The first map at the upper lejt corner
shows the current land use (1990) . The other four maps show the predicted land use for the year 20 J0
under thefour scenarios.

The subjective interpretation is based on a combination of the exploitation,

enjoyment and future value of land use. The exploitation value refers primarily to the
social meaning of the use of space. An increase in the exploitation value can be

348
obtained by combining functions and concentrating them to increase the efficiency of
land use. The enjoyment value refers to the cultural and landscape identity of an area, as
perceived by an ob server. An increase in the enjoyment value can be achieved, for
instance, by creating diversity in landuse and landscape. The future value is mainly
related to the economic and environmental relevance of a certain land-use type in the
future. The future value depends on sustainability, in the economic and environmental
sense. Figure 4 shows how spatial quality is measured in SES. The spatial quality is a
function ofthe environmental pressure, productivity value and employment value.

Do minant (derault configuration) Most numerous By land use type

thematic view ofvisually thematic view ofmost thematic view ofthe percentage .
corrected most numerous land-use numerous IlI1d-use types in a of one land-use type in a
types in a grid cell. grid cell. gridcell

i Group land-use types and value for each group I

Value land-use types

Environmental pressure
Produetivity Employment
thematic view of the surface
thematic view of the surface thematic view of the surface
value for the environmental
value for productivity for each value for employment for each
pressure for each land-use type
hmd-use type in a grid cell land-use type in a grid cell
in a grid cell

Enjoyment value
thematic value ofthe diversity (10
• •
Exploitation value
thematic value of the average score of
productivity and employment
which extent certain groups of
IlI1d-use types occur. where every
group has a weight value).
Spatial qualiry
I average score of enjoyment
value. exploitation value and t+--
environmental pressure

Figure 4. The SES scheme Jor evaluating spatial qua/ity.

6. Further developments of GIS and land-use planning

This paper has shown how spatial analysis and GIS technology can be used to explore
future land use in Europe. In the social sciences in general, and in quantitative geography
and regional economics in particular, a wide range of spatial models has been developed in
the past decades to describe, analyse, forecast and appraise the economic developments
within a set of localities or regions. Such models deal with the internal structures of regions

349
and the relationships within one region, as weil as with inter-regional relationships. The
tradition of models in quantitative geography and regional science has shown a strong
orientation towards location-allocation spatial problems (Baumann et al., 1988; Haynes and
Fotheringham, 1988). These model focus on the flows ofpeople, commodities and resources
between regions. Most of these models are static in nature and consider space as a set of
discrete points (areas, grids) rather than as a continuum. Problems of ecological
inconsistencies and spatial autocorrelation caused by the arbitrary spatial demarcation have
led to various misspecifications in these models (see Baxter, 1987).
The development of regional and inter-regional economic models corresponded with
growing interest in national input-output models. This led to the specification of regional
labour and housing markets, land use and transportation models (see, for instance, Nijkamp,
1986).
In general, spatial models can be used for three different purposes: forecasting and
scenario generation, policy impact analysis and policy generation, and design. A major
obstacle for the practical application of these models and for the interpretation of their
results is the specificity of each situation which is modelIed. The spatial arrangement, size,
shape and organisation of the basic spatial units for which data are gathered affect the spatial
processes involved, the parameters of which depend on the spatial framework used. All
these models have fimctions which commercial GIS systems almost totally lack, but which
are important for the type of questions policy decision-makers are interested in. However,
very fast progress has been observed in the last few years (e.g. Birkin et al., 1996; Douven,
1997). These new approaches are based on the full integration of spatial models and GIS-
fimctionality.
As a closing remark, it is worthwhile bearing in mind the risks involved in the use of
such large scale models. Lee (1973) pointed out that the relationship between theory
and the models is often weak and unsatisfactory. After twenty five years, and many
scientific and technologie al improvements, the issue of the credibility and robustness of
these models is still a key issue. However, it was the same Lee who also said that when
the models are expressly designed for policy analysis, the pragmatic criterion must be
the primary basis for evaluation. With all the necessary caution, it is clear that models
of this type and the results which they provide are an essential part in an informed
policy analysis and evaluation.

Acknowledgements
The colleagues at Geodan, especially Maarten Hilferink, are gratefully acknowledged
for their scientific contribution. The research was supported by the National Physical
Planning Agency ofthe Netherlands. We would like to thank Jan Oroen ofthe Agency
for his supportive role in the project. The system is available on CD-ROM.

350
References
Baumann, 1., M.M. Fischer, U. Schubert (1988) A choice-theoretical labour-market model:
Empirical tests at the mesolevel, Environment and Planning A, 20: 1085-1102.
Baxter, M.J.( 1987) Testing for misspecification in models of spatial flow, Environment and
PlanningA, 19: 1153:1160.
Birkin, M., G. Clarke, M. Clarke, A. Wilson (1996) Intelligent GIS. Loeation decisions and
strategie planning, GeoInformation International, Cambridge.
Douven, W. (1997) Improving the aeeessibility 01 spatial information lor environmental
management: an applieation to pesticide risk management, PhD thesis, Vrije Universiteit,
Amsterdam.
Geodan (1996) Ruimtegebruik in Europa: simulatie, evaluatie en presentatie van europees
grondgebruik, Geodan, Amsterdam.
Haynes, K.E., A.S. Fotheringham (1988) Gravity and spatial interaction models, Sage
Publications, Beverley Hills.
Klundert, A.F. van, A.G.J. Dietvorst, 1. van Os (1994) Back to the Future, nieuwefuneties van
landelijke gebieden in Europa, Staring Centrum, DLO, Wageningen.
Lee, D.B. (1973) Requiem for large-scale models, Journal 01 the Ameriean Institute 01
Planners, May 1973: 163-178.
LNV: Ministerie van Landbouw, Natuurbeheer en Visserij (1993) Bosbeleidsplan,
Regeringsbeslissing, Den Haag.
Nijkamp, P. (1986, Ed.) Handbook 01 Regional and Urban Eeonomies, North-Holland,
Amsterdam.
Petrus, P.W.F. (1989) Ruimtelijke kwaliteit en de praktijk van de ruimtelijke ordening,
Stedebouwen Volkshuisvesting, 70: 4-8.
Velde, R.J. van de, W.S. Faber, V.F. van Katwijk, J.C.I. Kuylenstierna, H.J. Scholten, T.J.M.
Thewessen, M. Verspuij (1994) The preparation 01 a European Landuse Database, RIVM,
report 712401001, Bilthoven.
Whitby, M., 1. Ollerenshaw (1988) Landuse and the European environment, Belhaven Press,
London.

351
long term comprehensive evaluation strategies for
spatial planning, design and management

Hubert N. van Lier l and Pat D. Taylor2

I Laboratory for Spatial Analysis, Planning and Design; Department of Environmental

Sciences, Wageningen Agricultural University, The Netherlands

2The Graduate Program in Landscape Architecture
The University ofTexas at Arlington, USA

Abstract
The planning, design and management of the countryside increasingly demand a careful balancing
between two extremes. One extreme is the aim for fuH economic development of rural areas,
predominantly by maximizing food production and minimizing costs. The other extreme is fuH
conservation ofthe natural resource base, including habitats and landscapes. While the first extreme
affects the environment through use of soH, water and air, as well as through the loss of native
ecosystems and landscapes, the second condition puts pressure on socio-economic values and other
competing social demands. The result is that societies face the choice of paying early on to sustain
both conditions, or to postpone costs by choosing one condition over the other. Severa\ quantitative
and qualitative approaches are being used to find the right balance between the two extremes. In open
space planning and design (including their related management systems) the development and use of
spatial concepts have increased over the last quarter century. Examples of such spatial concepts -
often measured quantitatively - are integration versus segregation, and the framework or spatial
network concepts such as ecological networking or ecologica\ infrastructure development. The
application of such concepts, among other things, demands an evaluation system capable of being
widely used. This evaluation system should originate at the outset ofthe planning process, when the
idea of a spatial plan first germinates. To do so can make sensitive, advanced thinking an integral
part of planning, design and management. Wide and early use of such a system can enable the
monitoring of planning objectives for an area and can assure the on-going evaluation of conditions
after a plan is executed. Quantitative attempts are made through so called evaluative design
procedures (such as the HELP method or the BEL method). This chapter assesses the efficacy of
these strategies and concepts, and urges a better balancing ofhuman needs with those ofthe physical
environment.

Keywords: sustainability; rural developments; conservation;" hwnan-nature paradigm;

abiotic factors; biotic factors; quantitative and qualitative approaches; spatial concepts;
investment-effect; evaluation design; HELP-method; BEL-method.

1. Spatial planning, design and management

As with many activities and developments in today's society the planning, design and
management of space are related to the concept of sustainability. The tenn is mostly used in
relation to the protection of natural resources, stemming from its original use in production

353
agriculture. Sustainability has several dimensions of which Bryden (1994) distinguishes
three: husbandry, interdependence and ethical obligations.
Husbandry refers to the continuity, durability and exploitation of our natural resources
over very long time-horizons; for example, through crop-rotation systems, the application of
fallowing or the restoration of soil and water systems. Interdependence refers to such
aspects as fragmentation and the relations between different land uses (for example, between
farmed areas and semi-natural vegetative areas). Ethical obligations pertain to future
generations. This last dimension deals with observed losses and depletions of a society's
natural resources related to the expected increase in population; for example, what are the
impacts of population on fossil fuels, forests, soillosses, water and air pollution? Thus, the
planning of future land uses, a10ng with the design for and management of these land uses,
is closely related today to these severaI dimensions of sustainability. Moreover, the process
of planning, design and management has to deaI with both conservation and development.
"Land-use planning is the systematic assessment of land and water potential, alternatives
for land use and economic and social conditions in order to select and adapt the best land-
use options" (F.A.O., 1993). This definition embraces the systematic approach of
possibilities for different land uses in the future, as weil as the (feIt) need for changes and the
willingness to execute aplan.
All present land-use planning is caught up between two seemingly contradictory dimensi-
ons: ecological conservation and economic existence. In a way or another, both dimensions
are related to sustainability. The first dimension refers to conservation; i.e., to the
conservation ofnatural resources, including clean water, air and soils and to the preservation
ofplailts"'and aIiimals (biodiversity for purposes of gene sources) etc. In many cases it even
goes further than just conservation: it seeks a re-creation of lost values. Examples are the
creation of nature areas from farmland or the reforestation of land Another example is the
restoration ofhigh water tables in formerly drained lands in order to create low intensive use
of meadows and thus increase meadow bird areas. Other examples can be given for many
parts of the world Generally, this approach is a clear one, especiallY in terms of spatial
consequences. Sustainability in terms of conservation means either halting certain
autonomous developments, retracking past developments or a combination of both.
Although this approach can come into conflict with the second dimension of sustainability
(interdependence) it does not necessarily do so.
The second dimension of landscape planning - that of a durable socio-economic existence
- is often argued as an important goal for creating a sustainable rural system. In many global
locations, local economies are under strong pressure, most notably farming. Surplus
productions, low qualities, worsening production conditions due to a lack of water or other
important resources, and rising production costs make it even more difficult for people to
attain a reasonable standard ofliving. These conditions result in such things as outmigration.
Moreover, this dimension of rural sustainability is generally feit when activities conceming
land-use planning and management are at stake. A very important task for land use planners
was a1ways, and still is, to improve the socio-economic situation of the rural population,
including such practice measures as creating more farmland, improving farmland, increasing
production, relocating farm buildings, routing and re-routing transportation systems or

354
creating villages, to name a few. The conflicting problem of economic development and
ecological coilservation is often mentioned: "We cannot save the environment without
development, and we cannot continue to develop just anywhere unless we save the
environment" (Science Council ofCanada, 1989). There also is a corresponding belief in the
possibility of bringing conservation and development together: "Unless development
conforms to conservation principles, it is not sustainable and human welfare is not served.
At the same time, unless there is adequate development - which must be ecologically sound
- conservation will be undermined by the subsistence requirements of the increasing
populations needing development assistance" (Talbot, 1984).
This last point is often expressed this way: "Tbe economy and its participants exist within
the environment, not outside it: we cannot expect to maintain economic prosperity unIess we
protect the environment and our resource base, the building blocks of development"
(National Task Force on Environment and Economics, 1987,) or by saying: "Just as long-
term economic growth depends on a healthy environment. so the maintenance of a healthy
environment requires continued economic development" (Aitken, 1988).
It is the Brundtland Commission itself, fmally, that addresses the sustainability concept as
a task for everybody (inc\uding land use planners) by stressing that: " ... environmental
protection and sustainable development must be an integral part of the mandates of all
agencies. These must be made responsible and accountable for ensuring that their policies,
programs and budgets encourage and support activities that are economically and
ecologically sustainable both in the short and longer terms" (Brundtland Commission,
1987).
In theory a wide spectrum of land-use planning goals can be defmed. as Figure 1 makes
c\ear albeit in a simple way.

I III 11

Full conservation of Full development of

rural areas: maximise rural areas: maximise
protection of nature production

room for creative

solutions to combine
extremes

Figure I. The land-use planning spectrum.

The conservation attitude strives to restore and conserve. In the extreme, it means fencing
everything that is beautiful or that can be made beautiful so as to protect it trom human
beings. Tbe development attitude strives to create and improve land for more and higher
production. In its most extreme sense, the countryside is seen solely as a fOod-producing
machine.

355
 This last viewpoint is less extreme than it seems. In a recent study for the European
Union the viewpoint was investigated within the framework of agricultural potential; that is,
if agricultural activities are performed in the best technical modem way and the production
volume is restricted to the expected consumption level (see van Latesteyn and Rabbinge,
1994). The study was based on four scenarios: free market and free trade; regional develop..
ment; nature and landscape; and environmental protection. The study made clear that the
total amount of land in agricultural production under such conditions could decline by at
least 40010 and at most by 80%. Associated labor also could decline by 50% to 75%. Accom-
plishinga drastic reduction in nitrogen and pesticide use could also be technically feasible
(see von Meyer, 1994).

2. Long term negative effects on spatial developments

2.1. Thefacts ofphysical condition

The increase of intensely used lands, both in urban and rural areas, has led to many negative
effects on the quality and quantity of natural resources and the environment. For the
countryside, probably the most accountable cause of these negative effects is farming.
Farming has developed along such lines as enIargement of scale, intensification, specia-
lization, and mechanization. Many factors contributed to this line of development, not in the
least socio-economic and technical ones. It is little wonder that such developments have led
to severe..cproblems in the countryside. such .PS the misuse, overuse, damage and loss of
natural systems. The following problems are known:

Abiotic factors:
• desertification (see Postei, 1989);
• soil erosion (wind and water erosion; see Brown, 1984 and Postel, 1989);
• soil degradation and soil compaction (see Boels. 1982);
• water pollution (see de Haan et al., 1986);
• air pollution (see Harrington, 1988 and Posch and Kauppi, 1991).

Biotic factors:
• loss ofnature: plants and animals; in fact whole ecosystems (see Bamey, 1980).

2.2. The human and/or nature paradigm

To some scientists., it is not so much that these problems exist, but rather a question ofhow
we look at them. For example, it is fashionable to think of spatial development as negative
and non-development as positive. This model sterns from the faulty environmental paradigm
that views humankind and nature as separate actors on the planetary stage. The weakness in
this view is that it fails to account for human actions as natural, even acceptable behaviors
within and upon the physical environment. Therefore, the question is not so much one of
searching for the negative effects of human action, but rather of accepting human actions
356
and their occurrence as something as natural as moming dew. With this understanding in
miml, the issue becomes one of carrying out spatial development and of maximizing as
much as possible the pre-development conditions which attracted human activity in the first
place.

Spotial development in the Dallas/Ft. Worth (Texas, US.A.) metropolitan areas serves as an
example. This region provides a contemporary laboratory for examining our philosophical
point. The north American prairie upon which these huge city complexes sit is a developer 's
paradise.
First, there are no geophysical ba"iers to municipal growth. No seashores, mountain
ranges, or river canyons impede, restrict or direct growth. Therefore, growth occurs in a 360-
circ/e, limited only by political or social barriers. This condition al/ows developers, from
farmers to builders, to respond primarily to market conditions in making decisions about the
conversion of prairie land to urban uses (van den Brink, 1996). Second, on the north
American prairie the dominant or c1imax plant communities are grasses, with canopy
vegetation normal/y found along the many small streams, creeks and rivers which drain the
countryside. Some canopies can be found creeping up the cooler northern slopes of hi/lsides,
but the gentleness ofthe prairie's topography quickly exposes trees to hotter, drier conditions
and their spread onto the grass lands is restricted naturally; that is, until human nurturing
enables them to adapt, safoly in their new developments on the prairie crown. Third, the
gentle topography requires almost effortless preporation for constructing new development.
Heavy earthwork, except for the giant excavations requiredfor under ground parking garages
and skyscrapers, is held to aminimum. Therefore, construction costs are more affordable.
It is nat much of a stretch to conclude that the human response (that is, the Dal/as/Ft.
Worth metroplex) that has flourished upon the north Texas prairie, is as natural as the cotton
farming that dominated the economy until the mid-twentieth century. Urbanization, or more
accurately suburbanization, has replaced farming, and has become as vital an economic
entity as agriculture once was. The 10.000 square mi/es that make-up the mass of cities and
suburbs support over 4.000.000 people, nearly Jour times os many as the area did in 1940. In
effect, the prairie is grow{ng houses, people and businesses as efficiently as it grew crops. It
can be said that this new human activity on the prairie is a natural response to pre-
development conditions, and in accordance with democratic principles, it primarily is market
driven, just like farming was.
There are negative effects, but many of them are seen as negative because their
naturalness is overlooked or because they are emotional/y charged, or both. These include
sprawl, job displacements, loss of habitat and the fauna it supported, pollution /rom
construction and urban runoff, and increased automobile emissions, to name a few. None of
these goes unchecked or unattended in America or in other first world countries. However,
success in solving or stemming these problems is intermittent, and some problems receive
more attention than others, depending on the political and social priorities of the moment,
and a little upon the latest scientific knowledge.

Again, what cannot be ignored about spatial development in the case of DallasIFt. Worth,
as weil as in much of North America and other places in the world, is that it very much is
market driven; and, increasingly, decision makers, developers and consumers are finding
they have shared goals, particularly where environmental quality is concemed. The result is

357
that local govemment standards for such amenities as landscaping, tree preservation and
open space can be exceeded by deve\opers who respond to consumer demands that surpass
govemmental minimums. Consumers in this prairie renaissance demonstrate their preference
for outdoor amenities by their willingness to pay for landscaped transportation corridors,
preserved streams and rivers, or native tree stands. On occasion public agencies which
regulate development are challenged to strengthen regulations in order to retlect the
planning and design standards already being practiced. Public agencies (most of which are
local in the U.S.A. since state and federal physical planning is seldom practiced) can also be
the mediators of change over developers who resist the inclusion of adequate public
amenities even when there is a willing buyer (the consumer).
In either case, either by law, or just because it makes good business sense, fewer environ-
mental decisions are being made without the serious solicitation of public priorities. lt is this
solicitation and the techniques used to execute it that mark a shift in the evaluation paradigm
for spatial planning, design and management in America and many other pi aces.
So, while there are negative effects resulting from spatial development, a paradigm
which is shifting from evaluating public projects late in the planning and design process to
evaluating them earlier, makes it possible to focus public attention on these effects. And, in a
democratic society there exists the belief that the better informed the public is, the wiser
public decisions will be. Fortunately, there are evaluation processes available which not only
measure but add to public awareness of the effects of development. These procedures are
introduced in the following section.

3. Future strategies: quantitative and qualitative approaches

As noted, the negative effects of spatial development are simple to identify, but their degree
of negativity is often judged according to the evaluator's preference for the human or the
nature side of the equation. If the evaluation is made with a balanced man-nature paradigm
in mind, then the negative effects of spatial development can be seen as minimal, because
the effected environment can partially be preserved. Determining whether this balance is to
be a goal in spatial planning is a policy decision. However, policy decisions drawn from
soundly determined public input have the promise of longevity and acceptance because
neither the public nor the planner is apt to encounter surprises in thoroughly researched
outcomes.
Fortunately, there are some research techniques being used which both retlect and effect
this balanced man-nature paradigm. Landscape architects are among those planners and
designers having to learn proper scientific techniques for soliciting public input. For those
practitioners without the training to implement questionnaires and analyze quantitative data,
there are related professional services available. Survey analysts, statisticians and others
frequently join planning and design teams to create surveys, test and implement them, and
evaluate their results.
However, important data do not always wait to appear during survey time. The planning
and design milieu is rife with data that emerge during contract negotiations, informal
discussions, pre\iminary meetings, board and commission meetings, public hearings, and
358
generally throughout the daily Iife of a community, as welI as during formal data colIection
periods. It is during these activities that some ofthe most significant needs and preferences
regarding spatial planning become elear, and qualitative methods of data colIection are best
suited to testing out these needs and preferences (Taylor and Bogden, 1984; Henderson
1991, 1995).
Qualitative methods alIow investigators, ineluding planners and designers, to
empathicaUy see the world of the participant through the use of such techniques as
observations, group and face-to-face interviews, and document content analyses, to narne a
few (Spradley, 1979, 1980). Qualitative techniques possess the additional characteristics of
being humane, meaning that actors who contribute data through qualitative inquiry are apt to
disengage from the process with a genuine sense of participation (Taylor and Bogden,
1984). This characteristic can engender public projects with the additional benefit of
increased citizen support, although it does not assure the adoption or implementation of an
environmental plan or design. Researchers have found, however, a elose connection between
participation and subsequent support for an idea or product (Rogers and Shoemaker, 1971;
Rogers and Rogers, 1976). Testing ofthis connection with environmental projects is an area
of research needing urgent attention because even with better resource-based planning,
advances in knowledge and technology are indicating that what is being done,
environmentally speaking, is not enough.
A case in point, for example, involves the adaptation of environmental corridors or
greenways in American cities. Lewis' (1964) caU to preserve our communities' pre-
development infrastructure (natural drainage systems to be exact) only received wide
adoption in the late 1970s and 1980s. As a case in point, new fast growing satellite cities
around DallasIFt. Worth have preserved local creeks, streams, and flood plains which in turn
have contributed to increased real estate values and to the formation of community images
shaped by open space. Such communities are ribboned with greenways, vestiges of the
original prairie habitat, though more manicured than their wild pre-development condition.
Certain wildlife have adaptoo to these corridors retaining the people's exposure to flora and
fauna that they would seldom see otherwise.
Other native wildlife and their associated habitats, however, are not enhanced by these
"progressive" environmental greenways. Recent discoveries in the biosciences suggest the
narrow configuration of greenways over-exposes songbirds to birds-of-prey or competitive
species (Neill, 1996). Certain desirable birds, it seems, require a thicker, deeper canopy,
suggesting that at least in places greenways need to widen-out with let-grow zones
containing several hundred acres. If this requirement is ineluded in the formulae used to
develop the prairie, the costs of development can be expected to rise, and the form and shape
of American cities can be expected to change.
Here is a case, then, where advancing knowledge can change our colIective behavior, at
least where spatial planning is concemed. Two questions arise at this point. First, are people
wiUing to re-configure the open space system of existing cities; and second, are planners,
developers and consumers willing to change the standards and raise the costs of develop-
ment to greatly increase urban open space? There is, of course, a third question: what if new

359
 scientific knowledge proves to be errant over time, and meanwhile large amounts of
financial resources are invested based on ephemeral data?
Answering these questions requires more than surveying the affected constituents. It
likely requires a longitudinal approach in which researchers both educate and measure these
constituents over time. Fortunately, qualitative research embraces techniques well-suited to
this purpose.
Unfortunately, the adoption of such techniques is in the innovative stage, so a break-out
of qualitative data collection is yet to happen. Widespread use of qualitative methods in
spatial planning, design and management is on the horizon however, because the number of
planners and designers exposed to their use is increasing. In landscape architecture at the
University of Texas at Arlington, for example, graduate students have been exposed to
quantitative and qualitative research methods - and their application in practice - since 1990.
This exposure is reinforced With the charge to "think theoretically and act practically", both
in their academie work and in their practice. More important, the closely allied field of parks
and recreation has embraced an array of qualitative data collection techniques, making it
only a matter of time before planners, designers and managers who come in contact with
recreation professionals will become aware of the value of qualitative approaches
(Henderson, 1991).
Several new attitudes are required on the part of planners, politicians, officials and the
public in order for qualitative techniques to work. These attitudes, and their associated
behavioral ehanges, can be demonstrated by professionals who understand them, and only
work when modeled early-on and throughout the planning, design and management
processes.
One such attitude involves the one-up one-down model of clientlconsultant relationships
(Block, 1981). Traditionally, clients are seen as persons in need of information, and
eonsultants are seen as those who hold information for whieh a client is willing to expend
resourees. In reality, all relationships are dynamie, with elientlconsultant roles being
reversed from time-to-time, regardless of the organizational or social strueture of the
relationship. In fact, landscape arehitecture students in the University of Texas at Arlington
graduate program are taught specifieally that at the beginning of eontractual relationships -
even before a eontract is developed - they as professional planners and designers usually are
the clients, and not the eonsultants. Understanding this role reversal is difficult for both
eontraetual parties. Clients are anxious to attain aresolution to their problems. Consultants
are anxious to provide it. In reality, neither is prepared for aresolution at the beginning of
the relationship.
Therefore, planners and designers must force elients to see themselves as the early
holders of important data. Only they know their own problems, their own needs and wants,
and their own proposed solutions, even if sueh solutions are impraetical. Foreing this role
reversal also forces the aspiring eonsultant at first to appear ignorant or uninformed because
he or she is forced to ask questions rather than to give answers. The risk is that the client
temporarily (permanently, if the client breaks the relationship ) sees the consultant as being
intellectually impotent, and the consultant must be patient not to see himlherself the same
way.

360
 However, by quietly staying the course, that is converting the dient into the consultant
role through serious, empathetic questioning, the consultant begins to accumulate enough
deep background to carefuHy relinquish bits ofthe knowledge he or she can ultimately apply
to the project. Tbe result is that the dient, because of being seriously heard, becomes
rewarded as a fuH and respected participant in the process. Furthermore, this feeling of
importance can carry over not only as support for the consultant but for the project, because
the dient knows that he or she has contributed to its fmal outcome.
Tbis early life of a project is where evaluation strategies need attention. However,
planners, designers and managers need to understand that to use these techniques they are
injecting greater democracy (and therefore reducing autocracy) into their processes. Once
these strategies are embraced there is no easy or painless way to terminate them. Tbey attract
the possibility of fuH participation until the project's completion because participants are
inteHectually and emotionaHy vested in it.
Tbe techniques used to achieve these outcomes can be used by public agencies, private
developers, individuals, and even educators., not just for specific projects but for long-term
evaluation ofthe perspectives and viewpoints of clients and client groups. Fortunately, these
techniques can be learned through training from practitioners in qualitative research.
AdditionaHy, they can serve as validations or justifications for the use of quantitative data
collection when appropriate (Henderson, 1991).
Land Redevelopment Projects (LRPs) in the Netherlands, however, have a long standing
tradition of using a combination of quantitative and qualitative evaluation methods, as
shown in Section 5. Tbe strategies apply mainly to specific spatial planning projects,
whereas what is needed is the long-term, on-going use of qualitative techniques to monitor
the trends, identify the themes and assess the dynarnic moods of human communities and
constituent groups over time. lt is one thing to know how the public feels about a particular
project after it has been exposed to the project's particulars. It is something else to have the
public actua1ly generate (or thwart) a project because it is genuinely in tune with the
resource needs and capabilities of its native space.

As an example, the suburban city 01 Plano, Texas voted in August 011996 to retain its
participation in the Dallas Area Rapid Transit (DART) system, after several years 01JUli
participation. By state law, member eities in regional transportation systems eould vote at
least every six years to withdraw or stay in. By nearly /wo to one, Plano's citizens voted to
stay in, and they partieipated in an eleetion that eould have been unneeessary had city council
members "had the pulse 01 the eommunity" before authorizing the vote. The eouncil had
submitted to avocal but inaecurate minority olopponents to regional transit in ealling the
eleetion. Had the council been assessing the eommunity's attitude through the sustained use 01
longitudinal research methods, it is likely the eost and angst 01 a publie eleetion eould have
been avoided.

This example suggests that outcomes are more predictable when longitudinal research is
on-going. lt also suggests that outcomes can be different depending upon which types of
data coHection methods are used. Tbere is little evidence to reinforce this latter hypothesis,
but there is need for further research into the types of data that are unveiled depending upon

36\
the techniques used. Quantitative methods tend to seek verification of the presence or
absence ofpreviously identified things. Qualitative methods, on the other hand, tend to seek
data on whatever the experience is of those being surveyed. While qualitative methods are
new to spatial planning, both methods have their place. It is the depth of data from
qualitative procedures, however, that excites researchers the most.

4. Spatial concepts

Spatial concepts are a land-use-planner's methods for addressing the problems of conserva-
tion and (or versa) development, or the man-nature paradigrn, as related to the sustainability
concept. Following is abrief description of three spatial concepts: the integration versus
segregation concept; the framework concept; and the ecological network concept (for a
more extensive description see van Lier, 1996).
The integration versus segregation concept tries to restore the many relationships
between the severalland uses in the countryside. In this concept, integration seeks to correct
upon segregation, which is a dramatic change from planning strategies in years past. The
integration concept puts restrictions on the developments of a certain land use because of a
desired or chosen development of another. The concept is actually the basis of landscape
ecology in which land use types (such as farming, outdoor recreation, infrastructure, and
trafik) have to be planned and developed in such a way that the basic ecosystems ofa rural
area are kept to function.
The framework-concept is fundamentally based upon a distinction between a low
dynamic l part of the rural system (nature, ecosystem) and a highly dynamic one (farming,
housing, recreation and transport). Extensive research (see Kerkstra and Vrijlandt, 1990; van
Buuren, 1991 and van Buuren and Kerkstra, 1993) on the changes in Dutch rural systems
showed that:
• the natural variation in moisture content (wetldry) and in nutrition levels
(eutrophie/oligotrophic) of soils diminished;
• this disappearance of variation in area properties resulted in a sharp decrease in number
and species of native vegetation and wildlife; and
• the visual character of the landscape became more uniform, meaning that contrasts
between largely open areas and small enclosed areas (with small scale farmlands)
disappeared.
The framework concept seeks a correction through a spatial segregation of intensely used
lands requiring a flexible lay-out and use, and extensively used lands, requiring stability.
This involves the planning of a durable and stable framework, an interconnected pattern of
zones in which nature management, forestry, outdoor recreation and water management are

1 "Dynamic" refers to the speed in which changes in rural areas take place. A rapid change in fanning systems
and in transportation (in society in general) for example. is considered to be "highly" dynamic. whereas
changes taking place in ecosystems demand long time horizons and are therefore considered to be "Iow"
dynamic.

362
concentrated. Such a framework enve\ops large open areas in which optimal farming is
possible.
The ecological network-concept is more or less based upon the island theory of Mac
Arthur and Wilson (1967) and the application of this theory to mainland situations. In this
concept the fragmentation of nature in rural systems creates nature islands, and the
increasing isolation of these islands determines whether certain populations of plants and
animals can survive over time. The concept of an ecological network is a reaction to this
fragmentation and isolation. An ecological network is a constellation of landscape elements
that is functional for the dispersion of a species in a landscape.
Several studies have demonstrated the specific properties on which ecological networks
are developed or are adapted as criteria for development of the concept (see examples in
Cook and van Lier, 1994). Connectivity, indicator species, stepping stones, core regions, and
metapopulations are all important issues considered in the network approach. Such a
network consists of core regions, nature development regions and connecting areas. The
core regions are areas of existing ecological values that are of international or national
significance. Nature development regions are areas that offer opportunities for the
development of future nature values of (inter)national significance or for a significant
increase in present nature values. The objective of the connecting zones is to improve and
develop the possibilities for migration within and between the core regions for some species.
It is assurned that providing for the needs ofthese species can be ofpositive significance for
other species in the core areas.
The main connections between nature areas can be formed through corridors of sufficient
length and width, sufficient variations in soils, heights, cover, waters and the like. In practice
such connections can be achieved by combining a corridor with existing watercourses, wet
areas, small hedgerows or other landscape elements, and in specific cases with main
drainage systems, rura\ roads, and other corridors. The small sideways connections in their
turn can follow field borders, small water courses, or unpaved rural roads to create a
connection with smaller nature preserves. In intensively farmed regions the concept is a
challenge to be developed in such a way that it achieves its own goals and that it makes
profitable farming possible. If the land is used for other functions, such as forestry or
outdoor recreation, the challenge is there as weil. The combination of enjoying the rura\
countryside, as weil as growing and harvesting timber combined with the concept of
ecologica\ networks, is another achievable connection, spiritually ifnot physically.

5. Evaluative design approach

The evaluative design approach is based upon the assumption that the evaluation of a plan is
not a purpose in itself, but a means to improve the planning process. Evaluating plans or
several plan alternatives can be of great help to either improve and or select plans. The
method can be intrinsic or extrinsic. Intrinsic means that the planner or the planning team
judges its own different steps in the planning process by trying to balance the benefits and
costs(positive and negative effects oftheir plans) in both qualitative and quantitative ways.

363
In the extrinsic process, prescribed methods are used to make a more objective comparison
between plans or parts of plan.
An example ofthis is the system of evaluative design evolved over the centuries in Dutch
Land Redevelopment Plans (LRP's). In the 1950's and 1960's a type ofbenefit-cost method,
called the investment-effect (i.e.) was introduced. This method calculated the benefits that
farmers were to receive from aland redevelopment plan, at that time known as aland
reaIlocation plan. The i.e. was defined as:

fanning benefits increase in income + decrease in maintainance costs (1)

i.e. =
investments costs for improving land, water, road, and the like

The method bad severaI shortcomings. The two most important ones were (1.) that it took
only farming into consideration, whereas the plans had increasingly more benefits and
related costs than for farming alone; and (2.) that it compared two situations - before and
after the project - thus ignoring the autonomous changes that would bave taken place.
Furthermore, there was a need for using evaluation systems during rather than after the
planning process in order to improve the plans. The result was the so-called evaluative
design approach, which in a general way can be visualized in Figure 2.

Tentative Pre·plans
plans

~
-
o ,_ ~ .<

Initiative
Inventory Ultimate
to make a
plan
E---+ phase 1-- ~ plan

~
~

!
Figure 2. The evaluative design approach in land use planning.

In the Dutch LRPs a new evaluation method was introduced caIIed the HELP method,
the acronym which (in Dutch) refers to 'Renewing Evaluation Land Redevelopment Plans.'
The following basic conditions were formulated for this method:
• Land redevelopment is a project based approach for improving the rural countryside for
existing and new purposes.

364
• Costs occur to execute the plan from the use of production means (such as labor and
machines) which are withdrawn from other employment, thus making it less or not
possible to realize other objectives.
• Positive benefits occur ifthe plan leads to set objectives.
• Negative benefits occur if the plan leads to worsening conditions with regard to set
objectives.
• Benefits and costs bave to be determined within theprinciple of the with-without
approach. The situation after the execution ofthe plan is compared with the situation that
would bave existed had the plan not been executed (a comparison of two processes rather
than of two moments). This principle forces the planner to "predict" the autonomous
situation by determining how the region most probably would have developed bad no
plan been designed and executed.
• Costs inc1ude different types of costs, inc1uding those costs made by government, and
those additional investments made by private persons, such as farmers.
• Benefits inc1ude the indirect ones such as those from increased business by frrrns which
deliver or process the increased goods from the project.
• Several plan alternatives bave to be generated.

By principle the benefits and costs are "measure" for the set objectives. For Dutch LRPs
these objectives are a high and stable income position (farming); socially acceptable living
and working conditions (farming and others); an attractive landscape for tourism and
outdoor recreation; the conservation and or creation of a differentiated ecosystem; the
improvement of the access and opening up of the region; and the creation or conservation of
an aesthetic and cultural-historic landscape.
The benefits for farming inc1ude higher yields, lower costs, less maintenance and more
free time. The benefits for outdoor recreation are determined on consumer surplus
approaches or the cost-effectiveness of provision. Benefits for nature are based on the
selection of indicator types and monitoring of increases or decreases. For trafik, changes in
speed (time-gains) and safety are taken into account.
The HELP-method has been successfully applied for about fifteen years during which it
played an important role in generating alternatives and in the design and selection of a final
plan. The combination of quantitative and qualitative analyses shows measurable
improvements in the planning process and in planning outcomes.
Nevertheless, a new system has been developed. Called the BEL system (the Dutch
acronym for "Decision-supporting Evaluation of Land use plans") it not only improves the
plan, but also the decision to execute the plan (go/no go) and the execution itself. The new
system supports decisions, places objectives as central, is selective, forms an integral part of
the planning process, uses criteria to check conditions, compares plan alternatives, is user-
friendly and is efficient.
WithinBEL:
• several plan alternatives are developed;
• the environment is studied carefully;

365
• four different land uses (farming, nature, landscape and tourismloutdoor recreation) are
studied;
• seven objectiveslintegral aspects (living conditions, identity, sustainability, social
acceptance, amount of land needed, execution aspects and integral rentability) are taken
into account;
• costs are included such as those for buying land (for new uses), the costs of executing the
plan, the costs for maintenance ofthe new situation and the distribution ofthe costs (who
pays what?)
The new system is supposed to be more flexible, but still has to prove its value.

6. Tightening-up the man-nature paradigm

Early in this chapter the man-nature paradigm was mentioned, in which the idea was set
forth that the effects of spatial development depended to some degree on the philosophy of
the evaluator. For example, ifthe evaluator saw mankind and nature as separate, competing
entities, the development activities ofhuman beings likely would be seen as negative.
Historically, the research techniques used in environmental planning have been born of
the natural sciences, such as biology; soil science and hydrology, to name three. Use ofthese
techniques has provided planners with data on how resources are used or affected under
human management, and thankfully, the world has benefited from development projects
such as land reallocation and land redevelopment plans in which more efficacious use of
resources has resulted in more satisfactory human use of the resource base (van Lier and
Taylor, 1982). In these projects the social sciences were brought into the process to help
scientists and citizens understand the human impact of reallocation.
The Dutch practice of reallocation (now more accurately referred to as redevelopment)
assumes that the larger social values gained from sustaining rural occupancy are as
important as the agricultural yields achieved by land reallocation. No such theoretical
connection exists between resource based cultural values and their transfer to urban citizens
in the types of spatial planning practiced in urban America. Instead, open space, spatial
relationships or architectural styles are determined more on the basis of availability, demand
or what is popular at the moment. Missing is a scientifically grounded understanding of the
connectedness between urban activities and the pre-development environment which drew
humans to our cities in the first place. Without this connection urban residents are insulated
from an awareness of how fragile and tough life - any life - can be. The struggle with
drought is blunted by confidence that there always will be water in the pipes. After all, there
always has been. The terrifying last moments of an animal trapped on a busy urban
thoroughfare are dismissed with the rationale that it is survival of the fittest at work here.
The shock from earthquakes, tornadoes, floods, severe heat or other natural disasters is
exacerbated by lack of understanding that these events have occurred on an occupied site
since long before today's cities were established there.
It is one of the benefits of qualitative research that participants in it can come to better
understand the subject being investigated (Rogers and Shoemaker, 1971). Couple this fact
with the notion that qualitative research relies on empathy to unveil the deep, rich

366
perspectives ofthe actors or participants, and there exists ample justification to employ these
research techniques in the search for human relationships with the physical environment.
Long-tenn assessment is needed of the urban population 's understanding of how human
behaviors are affected by these relationships. There is an assumption among physical
planners that human good comes from a elose association with the environment, but this
assumption is more an artiele of faith than a fact based on scientific evidence.
Kaplan and Kaplan (1982) and others have laid a framework for understanding and
measuring this good, but the unpredictable tenets of human behavior beg an on-going
search. Since qualitative research is longitudinal, it possesses the primary trait for long-tenn
data collection and analysis. What is needed is both basic and applied research from
practitioners and scientists in the environmental fields.
Thus, the call to incorporate new methodological approaches into spatial planning is also
a call to change collective thinking about the static nature of data collection. There is a meta-
planning element at work in these new approaches because in them the data collection
processes become a part of the planning process - not just a step in the process. The
qualitative strategies available to planners and designers contain the additional benefit of
elevating public knowledge about planning's value. Such an outcome promises to strengthen
the role of planning in society by sustaining public awareness of the aims and objectives of
visionaries, and by enabling the public to become more visionary itself.

References
Aitken, W.R.O. (1988) The environment and the Economy, Centre for Resource Studies Special
Paper, Queens University, Kingston.
Bamey, G.O. (1980) The global 2000 report to the President: entering the twenty-first century,
Pergamon, New York.
Block, P. (1981) Flawless Consulting, Leaming Concepts, Austin, Texas.
Boels, D. (1982) Physica\ soil degradation in the Netherlands, in D. Boels, D.S. Davies, A.E.
lohnston (Eds.) Soil degradation, Balkema, Den Haag.
Brink, A. van den (1996) Land Use Planning in the United States, Paper presented to the 24th
Symposium ofthe European Faculty on Land Use and Development. Vienna, Austria.
Brown, L.R. (1984) Conserving soils, in State olthe World 1984, A World Watch Institute Report
on Progress Toward a Sustainable Society: 53-73.
Brundtland Commission (1987) Gur Common Future, Oxford, New York.
Bryden, 1M. (1994) Some preliminary perspectives on sustainable rural communities, in 1M.
Bryden (Ed.) Towards Sustainable Rural Communities, The Guelph Seminar Series, The
Arkleton Trust and Univ. ofGuelph: 41-50.
Buuren, M. van (1991) A hydrological approach to landscape planning: The framework concept
elaborated from a hydrological perspective, Landscape and Urban Planning, 21: 91-107.
Buuren, M. van, K. Kerkstra (1993) The framework concept and the hydrological landscape
structure: A new perspective in the design of multifunctional landscapes, in e.e. Vos and P.
Opdam (Eds.) Landscape Ecology 01 a Stressed Environment, Chapman and Hall, London, 219-
243.
Cook, EA, H.N. van Lier (1994) Landscape planning and ecological networks, Isomul6F. Dev. in
Landsc. Plan. and Urb. Planning, Elsevier, Amsterdam.

367
F.A.O. (1993) Guide\ines for Land Use Planning, Pres. by Soil Res., Manag. and Service under guid.
ofthe lnter-Dept. Working Group on Land Use Planning, FAO, Rome.
Haan, F.A.M. de, Th. M. Lexmond, W.H. Riemsdijk (1986) Belasting van bodem en water door
mineraaloverschotten uit de intensieve veehouderij (Pollution of soils and water by minerals of
feed lots), Milieu, 2: 34-43.
Harrington, W. (1988) Breaking the deadlock on acid rain control, Resourcesfor the Future, 93: 1-4.
Henderson, K. A. (1991) Dimensions ofChoice: A Qualitative Approach to Recreation, Parks and
Leisure Research, State College PA: Venture Publishing, lnc.
Henderson, K. A. (1995) Evaluating Leisure Services: Making Enlightened Decisions, State College
PA, Venture Publishing, lnc.
Kaplan, R., S. Kaplan (1982) Cognition and Environment, Preager Press, New York.
Kerkstra, K., P. Vrijlandt (1990) Landscape Planning for lndustrial Agriculture: A proposed frame-
work for Rural Areas, Landscape and Urban Planning, 18: 275-282.
Latesteyn, H.e. van, R. Rabbinge (1994) Sustainable land use in the E.C.: an index ofpossibilities,
in H.N. van Lier, C.F. Jaarsma, C.R. Jurgens, A.J. de Buck (Eds.) Sustainable Land Use
Planning. ISOMUL 6E. Elsevier Amsterdam, 31-45.
Lewis, P.H. (1964) Qua1ity corridors in Wisconsin, Landscape Architecture Quarterly, 54(2):100-
108.
Lier, H.N. van (1996) Sustainable Rural Systems: Concepts from aland use planners perspective, in
H. Sasaki, I. Saito, A. Tabayashi, T. Morimoto (Eds.) Geographical perspectives on sustainable
rural systems, Kaisei Publishers, Tokyo, 14-25.
Lier, H.N. van, P.D. Taylor (1982) The New Dutch Landscape, Landscape Architecture magazine,
72(2):66-71.
Mac Arthur, R.H., E.O. Wilson (1967) The theory of biogeography, Princeton University Press,
Princeton.
Meyer, H. von (1994) Sustainable rura1land use from an European perspective, in J.M. Bryden (Ed.)
Towards Sustainable Rural Communities, The Guelph Seminar Series, The Arkleton Trust and
Univ. ofGuelph, 53-63.
Neill, R. (1996) An Interim Resource Management Plan for Eagle Mountain State Park, The
University ofTexas at Arlington, Center for Environmental Designs Research.
Posch, M., L. Kauppi (1991) Potential for acidification offorest soils in Europe, in F.M. Brouwer,
AJ. Thomas, MJ. Chadwick (Eds.) Land use changes in Europe, Kluwer, Dordrecht, 325-350.
Postei, S. (1989) Halting Land Degradation, in L.R. Brown (Ed.) State ofthe World, A World Watch
Institute Report on Progress Toward a Sustainable Society, Norton, New YorklLondon, 21-40.
Rogers, E.M., F.F. Shoemaker (1971) The Communication of Innavations, The Free Press, New
York.
Rogers, E.M., R.A. Rogers (1976) Communication in Organizations, The Free Press, New York.
Science Council of Canada (1989) Environmental Peace keepers: Science Technology and
Sustainable Development, Science Council, Ottawa
Spradley, J.P. (1979) The Ethnographie Interview, Holt, Rinehart and Winston, New York.
Spradley, J.P. (1980) Participant Observation, Holt, Rinehart and Winston, New York.
Talbot, L.M. (1984) The World Conservation Strategy, in F.R. Thibodeau and H.H. Fields (Eds.)
Sustaining Tomo"ow, Univ. Press ofNew England, London.
Taylor, S.J., Bogden (1984) An Introduction to Qualitative Research Methods, Wiley, New York.

368
list of authors

Jaques Antoine Food and Agriculture Organization (FAO)

Rome,ltaly

Carlos A. Bana e Costa Instituto Superior Tecnico-CESUR

Technical University 01 Lisbon
Lisbon, Portugal

Martin van Beek Faculty 01 Economics

Vrije Universiteit
Amsterdam, The Netherlands

Euro Beinat Institute lor Environmental Studies and Faculty 01 Economics

Vrije Universiteit
Amsterdam, The Netherlands

Robert Brinkman Food and Agriculture Organization (FAO)

Rome,ltaly

Arne J. Carisen Norwegian School 01 Management

Sandvika, Norway

Alessandro P. Costa Federal University olSanta Catarina - EPS

Florianopolis, Brazi/

Frankjohn J. Dent Food and Agriculture Organization (FAO)

Rome,ltaly

Michiel van Drunen Institute lor Environmental Studies

Vrije Universiteit
Amsterdam, The Netherlands

J. Ronald Eastman The Clark Labslor Cartographic Technology and Geographie

Analysis, Clark University
Worcester, Massachusetts, USA

Leonardo Ensslin Federal University olSanta Catarina - EPS

Florianopolis, Brazi/

Giulio Giannerini Centro di Teoria dei sistemi - CNR

Dipartimento di Elettronica e Inlormazione
Politecnico di Mi/ano
Mi/ano, Italy

369
Ca rio Giupponi Dipartimento Agronomia Ambientale e Produzioni Vegetali
Universita di Padova
Padova, Italy

Gerrit W. Heil Department 0/ Plant Ecology and Evolutionary Biology

Utrecht University
Utrecht, The Netherlands

Gerda Hermanides Faculty 0/ Economics

Vrije Universiteit
Amsterdam, The Netherlands

Marjan van Herwijnen Institute /or Environmental Studies

Vrije Universiteit
Amsterdam, The Netherlands

Ron Janssen Institute /or Environmental Studies

Vrije Universiteit
Amsterdam, The Netherlands

HongJiang The Clark Labs /or Cartographic Technology and Geographie

Analysis, Clark University
Worcester, Massachusetts, USA

A1ison Joubert Department 0/Statistical Sciences

University o/Cape Town
Cape Town, South Africa

Eliot Laniado Centro di Teoria dei sistemi - CNR

Dipartimento di Elettronica e Injormazione
Politecnico di Milano
Milano, Italy

Hubert N. van Lier Laboratory /or Spatial Analysis, Planning and Design
Department 0/ Environmental Sciences
Wageningen Agricultural University
Wageningen, The Netherlands

Francesco Marangon Dipartimento di Scienze Economiche

Universita di Udine
Udine, Italy

Giuseppe Munda Dep. Economia e Historia Economica

Universitat Autonoma de Barcelona
Barcelona, Spain

370
Paul Misdorp Geodan BV
Amsterdam, The Netherlands

Andrea Nardini Center /or Research on Environmental Sciences EULA-Chile

Universidad de Concepcion
Concepcion, Chile

Peter Nijkamp Faculty 0/ Economics

Vrije Universiteit
Amsterdam, The Netherlands

Massimo Parruccini European Commission

Joint Research Centre
Ispra, Italy

Andrea Patrono International Institute /or Aerospace Survey and Earth Sciences
(ITC)
Enschede, The Netherlands

Mark A. Ridgley Faculty o/Geography

University 0/ Hawaii
Honolulu, Hawaii, USA

Paolo Rosato Dipartimento Territorio e Sistemi Agro-Forestali

Universita di Padova
Padova, Italy

Giuseppe Rossi Universita di Catania

Catania, Italy

Henk J. Scholten Faculty 0/ Economics

Vrije Universiteit
Amsterdam, The Netherlands

Denis Sims Food and Agriculture Organization (FAO)

Rome, Italy

Giorgio Stagni Centro di Teoria dei sistemi - CNR

Dipartimento di Elettronica e Injormazione
Politecnico di Milano
Milano, ltaly

Theodor J. Stewart Department 0/ Statistical Sciences

University o/Cape Town
Cape Town, South Africa

371
Pat D. Taylor The Graduate Program in Landscape Architecture
The University 01 Texas at Arlington. USA

Tiziano Tempesta Dipartimento Territorio e S.A.F.

Universita di Padova
Padova. [taly

James Toledano The Clark Labs lor Cartographic Technology and Geographie
Analysis. Clark University
Worcester. Massachusetts. USA

Mark Weltz USDA-Agricultural Research Service

Southwest Watershed Research Center
Tueson. Arizona. USA

Fred Wenstop Norwegian School 01 Management

Sandvika. Norway

Diana S. Yakowitz USDA-Agricultural Research Service

Southwest Watershed Research Center; and
Department 01Agricultural and Biosystems Engineering
University 01Arizona
Tueson. Arizona. USA

372