DANEL P.

L0U0K5 and
EEL00 VAN BEEK
WATER
RE50UR0E5
5Y5TEM5
PLANNN0
AND MANA0EMENT
An ntroduction to Methods,
Models and Applications
with contributions Irom
Jery R. 5tedinger
Jos P.M. Di|kman
Monique T. Villars
5
t
u
d
i
e
s

a
n
d

R
e
p
o
r
t
s

i
n

H
y
d
r
o
l
o
g
y
Water Resources Systems Planning
and Management
An Introduction to Methods,
Models and Applications
wrm_fm.qxd 8/31/2005 12:06 PM Page i
Water Resources Systems
Planning and Management
An Introduction to Methods, Models
and Applications
Daniel P. Loucks and Eelco van Beek
with contributions from
Jery R. Stedinger
Jozef P.M. Dijkman
Monique T. Villars
Studies and Reports in Hydrology
UNESCO PUBLISHING
wrm_fm.qxd 8/31/2005 12:06 PM Page iii
The designations employed and the presentation of material
throughout this publication do not imply the expression of any
opinion whatsoever on the part of UNESCO concerning the legal
status of any country, territory, city or area or of its authorities,
or concerning the delimitation of its frontiers or boundaries.
The authors are responsible for the choice and the presentation
of the facts contained in this book and for the opinions
expressed therein, which are not necessarily those of UNESCO
and do not commit the Organization.
Published in 2005 by the United Nations Educational,
Scientific and Cultural Organization
7, place de Fontenoy F-75352 Paris 07 SP
Typeset by SR Nova Pvt. Ltd, Bangalore, India
Printed by Ages Arti Grafiche, Turin
ISBN 92-3-103998-9
All rights reserved
UNESCO 2005
Printed in Italy
103998_Prelims.qxd 17/11/05 12:07 Page iv
Within the Netherlands, as in much
of the world, the quality of our lives
is directly related to the quality of our
natural environment our air, land
and water resources. We consider a
quality environment crucial to human
health and economic and social
development as well as for ecosystem
preservation and diversity. How well
we manage our natural resources today will determine
just how well these resources will serve our descendants
and us. Hence, we care much about the management of
these resources, especially our water resources.
Many of us in the Netherlands are living in areas that
exist only because of the successful efforts of our past
water engineers, planners and managers. Managing water
in ways that best meet all our diverse needs for water and
its protection, including the needs of natural ecosystems,
is absolutely essential. But in spite of our knowledge and
experience, we Dutch, as others throughout the world,
still experience droughts, floods and water pollution.
These adverse impacts are not unique to us here in
Europe. In too many other regions of this world the need
for improved water management is much more critical
and much more urgent. Too many people, especially
children, suffer each day because of the lack of it.
As we take pride in our abilities to manage water, we
also take pride in our abilities to help others manage
Foreword
water. Institutions such as WL | Delft Hydraulics have
been doing this throughout its seventy-five years of exis-
tence. This book was written and published, in part, to
celebrate its seventy-fifth anniversary.
This book was written by individuals who have
simultaneously served as university professors as well as
consulting engineers throughout much of their profes-
sional careers. They have provided an introduction to
practical ways of modeling and analysing water resources
systems.
Whether you are studying at a university or working
in a developed or developing region, the methods and
advice presented in this book can help you develop your
skills in the use of quantitative methods of identifying
and evaluating effective water resources management plans
and policies. It can serve as a guide on ways of obtaining
the information you and your organization need when
deciding how to best manage these important resources.
This book, introducing an integrated systems
approach to water management, can serve many students,
teachers, and practising water resource engineers and
planners in the years to come.
His Royal Highness the Prince of Orange
The Netherlands
wrm_fm.qxd 8/31/2005 12:06 PM Page v
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Throughout history much of the world has witnessed
ever-greater demands for reliable, high-quality and
inexpensive water supplies for domestic consumption,
agriculture and industry. In recent decades there have
also been increasing demands for hydrological regimes
that support healthy and diverse ecosystems, provide for
water-based recreational activities, reduce if not prevent
floods and droughts, and in some cases, provide for the
production of hydropower and ensure water levels ade-
quate for ship navigation. Water managers are challenged
to meet these multiple and often conflicting demands. At
the same time, public stakeholder interest groups have
shown an increasing desire to take part in the water
resources development and management decision-
making process. Added to all these management chal-
lenges are the uncertainties of natural water supplies and
demands due to changes in our climate, changes in
peoples standards of living, changes in watershed land
uses and changes in technology. How can managers
develop, or redevelop and restore, and then manage water
resources systems systems ranging from small water-
sheds to those encompassing large river basins and coastal
zones in a way that meets societys changing objectives
and goals? In other words, how can water resources
systems become more integrated and sustainable?
Before engineering projects can be undertaken to
address water management problems or to take advantage
of opportunities for increased economic, ecological, envi-
ronmental and social benefits, they must first be planned.
This involves identifying various alternatives for address-
ing the problems or opportunities. Next, the various
impacts of each proposed alternative need to be estimated
and evaluated. A variety of optimization and simulation
models and modelling approaches have been developed
to assist water planners and managers in identifying and
evaluating plans. This book introduces the science and
art of modelling in support of water resources planning
and management. Its main emphasis is on the practice of
developing and using models to address specific water
resources planning and management problems. This must
be done in ways that provide relevant, objective and
meaningful information to those who are responsible for
making informed decisions about specific issues in
specific watersheds or river basins.
Readers of this book are not likely to learn this art of
modelling unless they actually employ it. The informa-
tion, examples and case studies contained in this book,
together with the accompanying exercises, we believe, will
facilitate the process of becoming a skilled water resources
systems modeller, analyst and planner. This has been our
profession, and we can highly recommend it to others.
Planning and management modelling is a multi-
disciplinary activity that is an essential part of almost all proj-
ects designed to increase the benefits, however measured,
Preface
wrm_fm.qxd 8/31/2005 12:06 PM Page vi
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Preface vii
from available water and related land resources. The
modelling and analysis of water resources systems
involves science and it also involves people and politics.
It is a challenge, but it is also fulfilling.
This book builds on a text titled Water Resources
Systems Planning and Analysis by Loucks, Stedinger and
Haith published by Prentice Hall in 1981. The present
work updates much of what was in that text, introduces
some new modelling methods that are proving to be
useful, and contains considerably more case studies. It
benefits considerably from the experiences of WL | Delft
Hydraulics, one of the many firms involved around
the world using the approaches and methods discussed
in this book.
Developments in graphics-based menu-driven interac-
tive computer programs and computer technology during
the last quarter of a century have had a significant and
beneficial impact on the use of modelling in the practice
of water resources engineering, planning and manage-
ment. All the models discussed in this book are designed
for use on micro-computers. The software we use to illus-
trate the solutions to various problems can be obtained
from the Internet free of charge. Commonly available
spreadsheet software can also be used. None of this was
available in 1981.
Although we have attempted to incorporate into each
chapter current approaches to water resources systems
planning and analysis, this book does not pretend to be a
review of the state-of-the-art of water resources systems
analysis found in the literature. Rather it is intended to
introduce readers to some of the more commonly used
models and modelling approaches applied to the plan-
ning and managing of water resources systems. We have
tried to organize our discussion of these topics in a way
useful for teaching and self-study. The contents reflect
our belief that the most appropriate methods for planning
and management are often the simpler ones, chiefly
because they are easier to understand and explain, require
less input data and time, and are easier to apply to specific
issues or problems. This does not imply that more sophis-
ticated and complex models are less useful. Sometimes
their use is the only way one can provide the needed
information. In this book, we attempt to give readers the
knowledge to make appropriate choices regarding model
complexity. These choices will depend in part on factors
such as the issues being addressed and the information
needed, the level of accuracy desired, the availability of
data and their cost, and the time required and available
to carry out the analysis. While many analysts have their
favourite modelling approach, the choice of model
should be based on a knowledge of various modelling
approaches and their advantages and limitations.
This book assumes readers have had some mathemat-
ical training in algebra, calculus, geometry and the use of
vectors and matrices. Readers of Chapters 7 through 9
will benefit from some background in probability and
statistics. Similarly, some exposure to micro-economic
theory and welfare economics will be useful for readers of
Chapter 10. Some knowledge of hydrology, hydraulics
and environmental engineering will also be beneficial, but
not absolutely essential. Readers wanting an overview of
some of natural processes that take place in watersheds,
river basins, estuaries and coastal zones can refer to
Appendix A. An introductory course in optimization and
simulation methods, typically provided in either an oper-
ations research or an economic theory course, can also
benefit the reader, but again it is not essential.
Chapter 1 introduces water resources systems plan-
ning and management and describes some examples of
water resources systems projects in which modelling has
had a critical role. These example projects also serve to
identify some of the current issues facing water managers
in different parts of the world. Chapter 2 defines the mod-
elling approach in general and the role of models in water
resources planning and management projects. Chapter 3
begins the discussion of optimization and simulation
modelling methods and how they are applied and used
in practice. It also discusses how modelling activities in
water resources development, planning and/or manage-
ment projects should be managed.
Chapter 4 is devoted to optimization modelling. This
relatively large chapter focuses on the use of various
optimization methods for the preliminary definition of
infrastructure design and operating policies. These
preliminary results define alternatives that usually need to
be further analysed and improved using simulation
methods. The advantages and limitations of different
optimization approaches are presented and illustrated
using some simple water allocation, reservoir operation
and water quality management problems. Chapter 5
extends this discussion of optimization to problems char-
acterized by fuzzy (more qualitative) objectives.
wrm_fm.qxd 8/31/2005 12:06 PM Page vii
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Chapter 6 introduces some of the more recently devel-
oped methods of statistical modelling, including artificial
neural networks and evolutionary search methods includ-
ing genetic algorithms. This chapter expects interested
readers will refer to other books, many of which are solely
devoted to just these topics, for more detail.
Chapters 7 through 9 are devoted to probabilistic
models, uncertainty and sensitivity analyses. These meth-
ods are useful not only for identifying more realistic
infrastructure designs and operating policies given hydro-
logical variability and uncertain parameter values and
objectives, but also for estimating some of the major
uncertainties associated with model predictions. Such
probabilistic and stochastic models can also help identify
just what model input data are needed and how accurate
those data need be with respect to their influence on the
decisions being considered.
Water resources planning and management today
inevitably involve multiple goals or objectives, many of
which may be conflicting. It is difficult, if not impossible,
to please all stakeholders all the time. Models containing
multiple objectives can be used to identify the tradeoffs
among conflicting objectives. This is information useful
to decision-makers who must decide what the best
tradeoffs should be, both among conflicting objectives
and among conflicting stakeholder interest groups. Multi-
objective modelling, Chapter 10, identifies various types
of economic, environmental and physical objectives, and
some commonly used ways of including multiple objec-
tives in optimization and simulation models.
Chapter 11 is devoted to various approaches for mod-
elling the hydrological processes in river basins. The focus
is on water quantity prediction and management. This is
followed by Chapter 12 on the prediction and manage-
ment of water quality processes in river basins and
Chapter 13 on the prediction and management of water
quantity and quality in storm water runoff, water supply
distribution and treatment, and wastewater collection and
treatment in urban areas. The final Chapter (Chapter 14)
provides a synopsis, reviewing again the main role of
models, introducing measures that can be used to evalu-
ate their usefulness in particular projects, and presenting
some more case studies showing the application of mod-
els to water resources management issues and problems.
Following these fourteen chapters are five appendices.
They contain descriptions of A) natural hydrological and
ecological processes in river basins, estuaries and coastal
zones, B) monitoring and adaptive management, C) drought
management, D) flood management, and E) a framework
for assessing, developing and managing water resources
systems as practiced by WL | Delft Hydraulics.
We believe Chapters 1 through 4 are useful prerequi-
sites to most of the remaining chapters. For university
teachers, the contents of this book represent more than
can normally be covered in a single quarter or semester
course. A first course can include Chapters 1 through 4,
and possibly Chapters 10, 11 or 13 in addition to
Chapter 14, depending on the background of the partic-
ipants in the class. A second course could include
Chapters 7 through 9 and/or any combination of
Chapters 5, 6, 12, 13 or 14, as desired. Clearly much
depends on the course objectives and on the background
knowledge of the course participants. Some exercises for
each chapter are included in the attached CD.
(Instructors may write to the authors to obtain suggested
solutions to these exercises.)
The writing of this book began at WL | Delft
Hydraulics as a contribution to its seventy-fifth anniver-
sary. We are most grateful for the companys support,
both financial and intellectual. While this book is not
intended to be a testimony to Delft Hydraulics contribu-
tions to the development and application of models to
water resources planning and management projects, it
does reflect the approaches taken, and modelling tools
used by them and other such firms and organizations that
engage in water resources planning, development and
management projects worldwide.
Many have helped us prepare this book. Jery Stedinger
wrote much of Chapters 7, 8 and 9, Nicki Villars helped
substantially with Chapter 12, and Jozef Dijkman con-
tributed a major portion of Appendix D. Vladam Babovic,
Henk van den Boogaard, Tony Minns, and Arthur Mynett
contributed material for Chapter 6. Roland Price provided
material for Chapter 13. Others who offered advice and
who helped review earlier chapter drafts include Martin
Baptist, Herman Breusers, Harm Duel, Herman Gerritsen,
Peter Gijsbers, Jos van Gils, Simon Groot, Karel Heynert,
Joost Icke, Hans Los, Marcel Marchand, Erik Mosselman,
Erik Ruijgh, Johannes Smits, Mindert de Vries and Micha
Werner. Ruud Ridderhof and Engelbert Vennix created
the figures and tables in this book. We thank all these
individuals and others, including our students, who
viii Preface
wrm_fm.qxd 8/31/2005 12:06 PM Page viii
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Preface ix
Most importantly we wish to acknowledge and thank
all our teachers, students and colleagues throughout the
world who have taught us all we know and added to the
quality of our professional and personal lives. We have
tried our best to make this book error free, but inevitably
somewhere there will be flaws. For that we apologize and
take responsibility for any errors of fact, judgment or
science that may be contained in this book. We will be
most grateful if you let us know of any or have other
suggestions for improving this book.
Daniel. P. Loucks,
Cornell University, Ithaca, N.Y., USA
Eelco van Beek,
WL | Delft Hydraulics, Delft,
the Netherlands
November 2004
provided assistance and support on various aspects
during the entire time this book was being prepared. We
have also benefited from the comments of Professors
Jan-Tai Kuo at National Taiwan University in Taipei, Jay
Lund at the University of California at Davis, Daene
McKinney of the University of Texas in Austin, Peter
Rogers at Harvard University in Cambridge,
Massachusetts, Tineke Ruijgh at TU-Delft, and Robert
Traver at Villanova University in Philadelphia, all of
whom have used earlier drafts of this book in their
classes. Finally we acknowledge with thanks the support
of Andras Szllsi-Nagy and the publishing staff at
UNESCO for publishing and distributing this book as a
part of their International Hydrological Programme. We
have written this book for an international audience, and
hence we are especially grateful for, and pleased to have,
this connection to and support from UNESCO.
wrm_fm.qxd 8/31/2005 12:06 PM Page ix
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Introduction xix
1. Water Resources Planning and Management:
An Overview 3
1. Introduction 3
2. Planning and Management Issues: Some Case
Studies 4
2.1. Kurds Seek Land, Turks Want Water 4
2.2. Sharing the Water of the Jordan River Basin: Is
There a Way? 6
2.3. Mending the Mighty and Muddy Missouri 7
2.4. The Endangered Salmon 7
2.5. The Yellow River: How to Keep the Water
Flowing 9
2.6. Lake Source Cooling: Aid to Environment or
Threat to Lake? 10
2.7. Managing Water in the Florida Everglades 11
2.8. Restoration of Europes Rivers and Seas 13
2.8.1. The Rhine 13
2.8.2. The Danube 14
2.8.3. The North and Baltic Seas 15
2.9. Egypt and the Nile: Limits to Agricultural
Growth 16
2.10. Damming the Mekong 16
3. So, Why Plan, Why Manage? 18
3.1. Too Little Water 20
3.2. Too Much Water 20
3.3. Polluted Water 21
3.4. Degradation of Aquatic and Riparian
Ecosystems 21
3.5. Other Planning and Management Issues 21
4. System Components, Planning Scales and
Sustainability 22
4.1. Spatial Scales for Planning and
Management 22
4.2. Temporal Scales for Planning and
Management 23
4.3. Sustainability 23
5. Planning and Management 24
5.1. Approaches 24
5.1.1. Top-Down Planning and
Management 25
5.1.2. Bottom-Up Planning and
Management 25
5.1.3. Integrated Water Resources
Management 26
5.2. Planning and Management Aspects 26
5.2.1. Technical Aspects 26
5.2.2. Economic and Financial Aspects 27
5.2.3. Institutional Aspects 28
5.3. Analyses for Planning and Management 28
5.4. Models for Impact Prediction and
Evaluation 30
5.5. Shared-Vision Modelling 31
5.6. Adaptive Integrated Policies 31
5.7. Post-Planning and Management Issues 32
Contents
wrm_fm.qxd 8/31/2005 12:06 PM Page x
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Contents xi
6. Meeting the Planning and Management Challenges:
A Summary 32
7. References 34
2. Water Resource Systems Modelling: Its Role in
Planning and Management 39
1. Introduction 39
2. Modelling of Water Resources Systems 41
2.1. An Example Modelling Approach 41
2.2. Characteristics of Problems to be Modelled 41
3. Challenges in Water Resources Systems
Modelling 43
3.1. Challenges of Planners and Managers 43
3.2. Challenges of Modelling 44
3.3. Challenges of Applying Models in Practice 45
4. Developments in Modelling 46
4.1. Modelling Technology 46
4.2. Decision Support Systems 47
4.2.1. Shared-Vision Modelling 49
4.2.2. Open Modelling Systems 51
4.2.3. Example of a DSS for River Flood
Management 51
5. Conclusions 54
6. References 55
3. Modelling Methods for Evaluating
Alternatives 59
1. Introduction 59
1.2. Model Components 60
2. Plan Formulation and Selection 61
2.1. Plan Formulation 61
2.2. Plan Selection 63
3. Modelling Methods: Simulation or Optimization 64
3.1. A Simple Planning Example 65
3.2. Simulation Modelling Approach 66
3.3. Optimization Modelling Approach 66
3.4. Simulation Versus Optimization 67
3.5. Types of Models 69
3.5.1. Types of Simulation Models 69
3.5.2. Types of Optimization Models 70
4. Model Development 71
5. Managing Modelling Projects 72
5.1. Creating a Model Journal 72
5.2. Initiating the Modelling Project 72
5.3. Selecting the Model 73
5.4. Analysing the Model 74
5.5. Using the Model 74
5.6. Interpreting Model Results 75
5.7. Reporting Model Results 75
6. Issues of Scale 75
6.1. Process Scale 75
6.2. Information Scale 76
6.3. Model Scale 76
6.4. Sampling Scale 76
6.5. Selecting the Right Scales 76
7. Conclusions 77
8. References 77
4. Optimization Methods 81
1. Introduction 81
2. Comparing Time Streams of Economic Benefits
and Costs 81
2.1. Interest Rates 82
2.2. Equivalent Present Value 82
2.3. Equivalent Annual Value 82
3. Non-linear Optimization Models and Solution
Procedures 83
3.1. Solution Using Calculus 84
3.2. Solution Using Hill Climbing 84
3.3. Solution Using Lagrange Multipliers 86
3.3.1. Approach 86
3.3.2. Meaning of the Lagrange
Multiplier 88
4. Dynamic Programming 90
4.1. Dynamic Programming Networks and Recursive
Equations 90
4.2. Backward-Moving Solution Procedure 92
4.3. Forward-Moving Solution Procedure 95
4.4. Numerical Solutions 96
4.5. Dimensionality 97
4.6. Principle of Optimality 97
4.7. Additional Applications 97
4.7.1. Capacity Expansion 98
4.7.2. Reservoir Operation 102
4.8. General Comments on Dynamic
Programming 112
5. Linear Programming 113
5.1. Reservoir Storage CapacityYield Models 114
5.2. A Water Quality Management Problem 117
5.2.1. Model Calibration 118
5.2.2. Management Model 119
5.3. A Groundwater Supply Example 124
5.3.1. A Simplified Model 125
5.3.2. A More Detailed Model 126
wrm_fm.qxd 8/31/2005 12:06 PM Page xi
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
5.3.3. An Extended Model 127
5.3.4. Piecewise Linearization Methods 128
5.4. A Review of Linearization Methods 129
6. A Brief Review 132
7. References 132
5. Fuzzy Optimization 135
1. Fuzziness: An Introduction 135
1.1. Fuzzy Membership Functions 135
1.2. Membership Function Operations 136
2. Optimization in Fuzzy Environments 136
3. Fuzzy Sets for Water Allocation 138
4. Fuzzy Sets for Reservoir Storage and
Release Targets 139
5. Fuzzy Sets for Water Quality Management 140
6. Summary 144
7. Additional References (Further Reading) 144
6. Data-Based Models 147
1. Introduction 147
2. Artificial Neural Networks 148
2.1. The Approach 148
2.2. An Example 151
2.3. Recurrent Neural Networks for the Modelling of
Dynamic Hydrological Systems 153
2.4. Some Applications 153
2.4.1. RNN Emulation of a Sewerage System in
the Netherlands 154
2.4.2. Water Balance in Lake IJsselmeer 155
3. Genetic Algorithms 156
3.1. The Approach 156
3.2. Example Iterations 158
4. Genetic Programming 159
5. Data Mining 163
5.1. Data Mining Methods 163
6. Conclusions 164
7. References 165
7. Concepts in Probability, Statistics and
Stochastic Modelling 169
1. Introduction 169
2. Probability Concepts and Methods 170
2.1. Random Variables and Distributions 170
2.2. Expectation 173
2.3. Quantiles, Moments and Their Estimators 173
2.4. L-Moments and Their Estimators 176
3. Distributions of Random Events 179
3.1. Parameter Estimation 179
3.2. Model Adequacy 182
3.3. Normal and Lognormal Distributions 186
3.4. Gamma Distributions 187
3.5. Log-Pearson Type 3 Distribution 189
3.6. Gumbel and GEV Distributions 190
3.7. L-Moment Diagrams 192
4. Analysis of Censored Data 193
5. Regionalization and Index-Flood Method 195
6. Partial Duration Series 196
7. Stochastic Processes and Time Series 197
7.1. Describing Stochastic Processes 198
7.2. Markov Processes and Markov Chains 198
7.3. Properties of Time-Series Statistics 201
8. Synthetic Streamflow Generation 203
8.1. Introduction 203
8.2. Streamflow Generation Models 205
8.3. A Simple Autoregressive Model 206
8.4. Reproducing the Marginal Distribution 208
8.5. Multivariate Models 209
8.6. Multi-Season, Multi-Site Models 211
8.6.1. Disaggregation Models 211
8.6.2. Aggregation Models 213
9. Stochastic Simulation 214
9.1. Generating Random Variables 214
9.2. River Basin Simulation. 215
9.3. The Simulation Model 216
9.4. Simulation of the Basin 216
9.5. Interpreting Simulation Output 217
10. Conclusions 223
11. References 223
8. Modelling Uncertainty 231
1. Introduction 231
2. Generating Values From Known Probability
Distributions 231
3. Monte Carlo Simulation 233
4. Chance Constrained Models 235
5. Markov Processes and Transition
Probabilities 236
6. Stochastic Optimization 239
6.1. Probabilities of Decisions 243
6.2. A Numerical Example 244
7. Conclusions 251
8. References 251
xii Contents
wrm_fm.qxd 8/31/2005 12:06 PM Page xii
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Contents xiii
9. Model Sensitivity and Uncertainty
Analysis 255
1. Introduction 255
2. Issues, Concerns and Terminology 256
3. Variability and Uncertainty In Model Output 258
3.1. Natural Variability 259
3.2. Knowledge Uncertainty 260
3.2.1. Parameter Value Uncertainty 260
3.2.2. Model Structural and Computational
Errors 260
3.3. Decision Uncertainty 260
4. Sensitivity and Uncertainty Analyses 261
4.1. Uncertainty Analyses 261
4.1.1. Model and Model Parameter
Uncertainties 262
4.1.2. What Uncertainty Analysis Can
Provide 265
4.2. Sensitivity Analyses 265
4.2.1. Sensitivity Coefficients 267
4.2.2. A Simple Deterministic Sensitivity
Analysis Procedure 267
4.2.3. Multiple Errors and Interactions 269
4.2.4. First-Order Sensitivity Analysis 270
4.2.5. Fractional Factorial Design
Method 272
4.2.6. Monte Carlo Sampling Methods 273
5. Performance Indicator Uncertainties 278
5.1. Performance Measure Target Uncertainty 278
5.2. Distinguishing Differences Between Performance
Indicator Distributions 281
6. Communicating Model Output Uncertainty 283
7. Conclusions 285
8. References 287
10. Performance Criteria 293
1. Introduction 293
2. Informed Decision-Making 294
3. Performance Criteria and General Alternatives 295
3.1. Constraints On Decisions 296
3.2. Tradeoffs 296
4. Quantifying Performance Criteria 297
4.1. Economic Criteria 298
4.1.1. Benefit and Cost Estimation 299
4.1.2. A Note Concerning Costs 302
4.1.3. Long and Short-Run Benefit
Functions 303
4.2. Environmental Criteria 305
4.3. Ecological Criteria 306
4.4. Social Criteria 308
5. Multi-Criteria Analyses 309
5.1. Dominance 310
5.2. The Weighting Method 311
5.3. The Constraint Method 312
5.4. Satisficing 313
5.5. Lexicography 313
5.6. Indifference Analysis 313
5.7. Goal Attainment 314
5.8. Goal-Programming 315
5.9. Interactive Methods 315
5.10. Plan Simulation and Evaluation 316
6. Statistical Summaries of Performance Criteria 320
6.1. Reliability 321
6.2. Resilience 321
6.3. Vulnerability 321
7. Conclusions 321
8. References 322
11. River Basin Planning Models 325
1. Introduction 325
1.1. Scales of River Basin Processes 326
1.2. Model Time Periods 327
1.3. Modelling Approaches for River Basin
Management 328
2. Modelling the Natural Resources System and Related
Infrastructure 328
2.1. Watershed Hydrological Models 328
2.1.1. Classification of Hydrological Models 329
2.1.2. Hydrological Processes: Surface Water 329
2.1.3. Hydrological Processes:
Groundwater 333
2.1.4. Modelling Groundwater: Surface Water
Interactions 336
2.1.5. Streamflow Estimation 339
2.1.6. Streamflow Routing 341
2.2. Lakes and Reservoirs 342
2.2.1. Estimating Active Storage Capacity 343
2.2.2. Reservoir StorageYield Functions 344
2.2.3. Evaporation Losses 346
2.2.4. Over and Within-Year Reservoir Storage
and Yields 347
2.2.5. Estimation of Active Reservoir Storage
Capacities for Specified Yields 348
2.3. Wetlands and Swamps 354
2.4. Water Quality and Ecology 354
wrm_fm.qxd 8/31/2005 12:06 PM Page xiii
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
3. Modelling the Socio-Economic Functions In a River
Basin 355
3.1. Withdrawals and Diversions 355
3.2. Domestic, Municipal and Industrial Water
Demand 356
3.3. Agricultural Water Demand 357
3.4. Hydroelectric Power Production 357
3.5. Flood Risk Reduction 359
3.5.1. Reservoir Flood Storage Capacity 360
3.5.2. Channel Capacity 362
3.6. Lake-Based Recreation 362
4. River Basin Analysis 363
4.1. Model Synthesis 363
4.2. Modelling Approach Using Optimization 364
4.3. Modelling Approach Using Simulation 365
4.4. Optimization and/or Simulation 368
4.5. Project Scheduling 368
5. Conclusions 371
6. References 371
12. Water Quality Modelling and Prediction 377
1. Introduction 377
2. Establishing Ambient Water Quality Standards 378
2.1. Water-Use Criteria 379
3. Water Quality Model Use 379
3.1. Model Selection Criteria 380
3.2. Model Chains 381
3.3. Model Data 382
4. Water Quality Model Processes 383
4.1. Mass-Balance Principles 384
4.1.1. Advective Transport 385
4.1.2. Dispersive Transport 385
4.1.3. Mass Transport by Advection and
Dispersion 385
4.2. Steady-State Models 386
4.3. Design Streamflows for Water Quality 388
4.4. Temperature 389
4.5. Sources and Sinks 390
4.6. First-Order Constituents 390
4.7. Dissolved Oxygen 390
4.8. Nutrients and Eutrophication 393
4.9. Toxic Chemicals 396
4.9.1. Adsorbed and Dissolved Pollutants 396
4.9.2. Heavy Metals 398
4.9.3. Organic Micro-pollutants 399
4.9.4. Radioactive Substances 400
4.10. Sediments 400
4.10.1. Processes and Modelling
Assumptions 401
4.10.2. Sedimentation 401
4.10.3. Resuspension 401
4.10.4. Burial 402
4.10.5. Bed Shear Stress 402
4.11. Lakes and Reservoirs 403
4.11.1. Downstream Characteristics 405
4.11.2. Lake Quality Models 406
4.11.3. Stratified Impoundments 407
5. An Algal Biomass Prediction Model 408
5.1. Nutrient Cycling 408
5.2. Mineralization of Detritus 408
5.3. Settling of Detritus and Inorganic Particulate
Phosphorus 409
5.4. Resuspension of Detritus and Inorganic
Particulate Phosphorus 409
5.5. The Nitrogen Cycle 409
5.5.1. Nitrification and Denitrification 409
5.5.2. Inorganic Nitrogen 410
5.6. Phosphorus Cycle 410
5.7. Silica Cycle 411
5.8. Summary of Nutrient Cycles 411
5.9. Algae Modelling 412
5.9.1. Algae Species Concentrations 412
5.9.2. Nutrient Recycling 413
5.9.3. Energy Limitation 413
5.9.4. Growth Limits 414
5.9.5. Mortality Limits 414
5.9.6. Oxygen-Related Processes 415
6. Simulation Methods 416
6.1. Numerical Accuracy 416
6.2. Traditional Approach 417
6.3. Backtracking Approach 418
6.4. Model Uncertainty 420
7. Conclusions: Implementing a Water Quality
Management Policy 421
8. References 422
13. Urban Water Systems 427
1. Introduction 427
2. Drinking Water 428
2.1. Water Demand 428
2.2. Water Treatment 428
2.3. Water Distribution 430
2.3.1. Open Channel Networks 432
xiv Contents
wrm_fm.qxd 8/31/2005 12:06 PM Page xiv
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Contents xv
2.3.2. Pressure Pipe Networks 432
2.3.3. Water Quality 434
3. Wastewater 434
3.1. Wastewater Production 434
3.2. Sewer Networks 434
3.3. Wastewater Treatment 435
4. Urban Drainage 437
4.1. Rainfall 437
4.1.1. Time Series Versus Design Storms 437
4.1.2. Spatial-Temporal Distributions 438
4.1.3. Synthetic Rainfall 438
4.1.4. Design Rainfall 438
4.2. Runoff 439
4.2.1. Runoff Modelling 439
4.2.2. The Horton Infiltration Model 441
4.2.3. The US Soil Conservation Method (SCS)
Model 442
4.2.4. Other RainfallRunoff Models 444
4.3. Surface Pollutant Loading and Washoff 445
4.3.1. Surface Loading 446
4.3.2. Surface Washoff 446
4.3.3. Stormwater Sewer and Pipe
Flow 447
4.3.4. Sediment Transport 448
4.3.5. Structures and Special Flow
Characteristics 448
4.4. Water Quality Impacts 448
4.4.1. Slime 448
4.4.2. Sediment 448
4.4.3. Pollution Impact on the
Environment 448
4.4.4. Bacteriological and Pathogenic
Factors 451
4.4.5. Oil and Toxic Contaminants 451
4.4.6. Suspended Solids 452
5. Urban Water System Modelling 452
5.1. Model Selection 452
5.2. Optimization 453
5.3. Simulation 455
6. Conclusions 456
7. References 457
14. A Synopsis 461
1. Meeting the Challenge 461
2. The Systems Approach to Planning and
Management 461
2.1. Institutional Decision-Making 462
2.2. The Water Resources Systems 464
2.3. Planning and Management Modelling:
A Review 465
3. Evaluating Modelling Success 466
4. Some Case Studies 467
4.1. Development of a Water Resources Management
Strategy for Trinidad and Tobago 468
4.2. Transboundary Water Quality Management in
the Danube Basin 470
4.3. South Yunnan Lakes Integrated Environmental
Master Planning Project 473
4.4. River Basin Management and Institutional
Support for Poland 475
4.5. Stormwater Management in The Hague in the
Netherlands 476
5. Summary 478
6. References 478
Appendix A: Natural System Processes and
Interactions 480
1. Introduction 483
2. Rivers 483
2.1. River Corridor 484
2.1.1. Stream Channel Structure
Equilibrium 485
2.1.2. Lateral Structure of Stream or River
Corridors 486
2.1.3. Longitudinal Structure of Stream or River
Corridors 487
2.2. Drainage Patterns 488
2.2.1. Sinuosity 489
2.2.2. Pools and Riffles 489
2.3. Vegetation in the Stream and River
Corridors 489
2.4. The River Continuum Concept 490
2.5. Ecological Impacts of Flow 490
2.6. Geomorphology 490
2.6.1. Channel Classification 491
2.6.2. Channel Sediment Transport and
Deposition 491
2.6.3. Channel Geometry 493
2.6.4. Channel Cross sections and Flow
Velocities 494
2.6.5. Channel Bed Forms 495
2.6.6. Channel Planforms 495
2.6.7. Anthropogenic Factors 496
2.7. Water Quality 497
2.8. Aquatic Vegetation and Fauna 498
2.9. Ecological Connectivity and Width 500
wrm_fm.qxd 8/31/2005 12:06 PM Page xv
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
2.10. Dynamic Equilibrium 501
2.11. Restoring Degraded Aquatic Systems 501
3. Lakes and Reservoirs 504
3.1. Natural Lakes 504
3.2. Constructed Reservoirs 505
3.3. Physical Characteristics 505
3.3.1. Shape and Morphometry 505
3.3.2. Water Quality 506
3.3.3. Downstream Characteristics 507
3.4. Management of Lakes and Reservoirs 508
3.5. Future Reservoir Development 510
4. Wetlands 510
4.1. Characteristics of Wetlands 511
4.1.1. Landscape Position 512
4.1.2. Soil Saturation and Fibre Content 512
4.1.3. Vegetation Density and Type 512
4.1.4. Interaction with Groundwater 513
4.1.5. OxidationReduction 513
4.1.6. Hydrological Flux and Life
Support 513
4.2. Biogeochemical Cycling and Storage 513
4.2.1. Nitrogen (N) 514
4.2.2. Phosphorus (P) 514
4.2.3. Carbon (C) 514
4.2.4. Sulphur (S) 514
4.2.5. Suspended Solids 514
4.2.6. Metals 515
4.3. Wetland Ecology 515
4.4. Wetland Functions 515
4.4.1. Water Quality and Hydrology 515
4.4.2. Flood Protection 516
4.4.3. Shoreline Erosion 516
4.4.4. Fish and Wildlife Habitat 516
4.4.5. Natural Products 516
4.4.6. Recreation and Aesthetics 516
5. Estuaries 516
5.1. Types of Estuaries 517
5.2. Boundaries of an Estuary 518
5.3. Upstream Catchment Areas 519
5.4. Water Movement 519
5.4.1. Ebb and Flood Tides 519
5.4.2. Tidal Excursion 520
5.4.3. Tidal Prism 520
5.4.4. Tidal Pumping 520
5.4.5. Gravitational Circulation 520
5.4.6. Wind-Driven Currents 521
5.5. Mixing Processes 521
5.5.1. Advection and Dispersion 522
5.5.2. Mixing 522
5.6. Salinity Movement 523
5.6.1. Mixing of Salt- and Freshwaters 523
5.6.2. Salinity Regimes 523
5.6.3. Variations due to Freshwater Flow 523
5.7. Sediment Movement 524
5.7.1. Sources of Sediment 524
5.7.2. Factors Affecting Sediment
Movement 524
5.7.3. Wind Effects 525
5.7.4. Ocean Waves and Entrance Effects 525
5.7.5. Movement of Muds 526
5.7.6. Estuarine Turbidity Maximum 527
5.7.7. Biological Effects 527
5.8. Surface Pollutant Movement 528
5.9. Estuarine Food Webs and Habitats 528
5.9.1. Habitat Zones 529
5.10. Estuarine Services 531
5.11. Estuary Protection 531
5.12. Estuarine Restoration 533
5.13. Estuarine Management 533
5.13.1. Engineering Infrastructure 534
5.13.2. Nutrient Overloading 534
5.13.3. Pathogens 534
5.13.4. Toxic Chemicals 534
5.13.5. Habitat Loss and Degradation 534
5.13.6. Introduced Species 535
5.13.7. Alteration of Natural Flow Regimes 535
5.13.8. Declines in Fish and Wildlife
Populations 535
6. Coasts 535
6.1. Coastal Zone Features and Processes 535
6.1.1. Water Waves 536
6.1.2. Tides and Water Levels 538
6.1.3. Coastal Sediment Transport 538
6.1.4. Barrier Islands 538
6.1.5. Tidal Deltas and Inlets 538
6.1.6. Beaches 538
6.1.7. Dunes 539
6.1.8. Longshore Currents 540
6.2. Coasts Under Stress 540
6.3. Management Issues 540
6.3.1. Beaches or Buildings 542
6.3.2. Groundwater 542
xvi Contents
wrm_fm.qxd 8/31/2005 12:06 PM Page xvi
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Contents xvii
6.3.3. Sea Level Rise 542
6.3.4. Subsidence 543
6.3.5. Wastewater 544
6.3.6. Other Pollutants 544
6.3.7. Mining of Beach Materials 544
6.4. Management Measures 545
6.4.1. Conforming Use 546
6.4.2. Structures 546
6.4.3. Artificial Beach Nourishment 547
7. Conclusion 548
8. References 549
Appendix B: Monitoring and Adaptive
Management 559
1. Introduction 559
2. System Status 561
2.1. System Status Indicators 562
3. Information Needs 562
3.1. Information Objectives and Priorities 563
4. Monitoring Plans 563
5. Adaptive Monitoring 564
5.1. Risk Assessments For Monitoring 564
5.2. Use of Models 565
6. Network Design 565
6.1. Site Selection 566
6.2. Sampling/Measurement Frequencies 566
6.3. Quality Control 566
6.4. Water Quantity Monitoring 567
6.5. Water Quality Monitoring 568
6.6. Ecological Monitoring 569
6.7. Early-Warning Stations 569
6.8. Effluent Monitoring 570
7. Data Sampling, Collection and Storage 570
7.1. Overview 570
7.2. Remote Sensing 571
7.2.1. Optical Remote Sensing for Water
Quality 571
7.2.2. Applications in the North Sea 572
8. Data Analyses 572
9. Reporting Results 573
9.1. Trend Plots 573
9.2. Comparison Plots 573
9.3. Map Plots 576
10. Information Use: Adaptive Management 576
11. Summary 578
12. References 578
Appendix C: Drought Management 581
1. Introduction 581
2. Drought Impacts 581
3. Defining Droughts 584
4. Causes of Droughts 585
4.1. Global Patterns 586
4.2. Teleconnections 588
4.3. Climate Change 588
4.4. Land Use 590
5. Drought Indices 590
5.1. Percent of Normal Indices 590
5.2. Standardized Precipitation Index 590
5.3. Palmer Drought Severity Index 591
5.4. Crop Moisture Index 592
5.5. Surface Water Supply Index 592
5.6. Reclamation Drought Index 593
5.7. Deciles 594
5.8. Method of Truncation 594
5.9. Water Availability Index 594
5.10. Days of Supply Remaining 595
6. Drought Triggers 596
7. Virtual Drought Exercises 596
8. Conclusion 598
9. References 599
Appendix D: Flood Management 603
1. Introduction 603
2. Managing Floods in the Netherlands 605
2.1. Flood Frequency and Protection 605
2.2. The Rhine River Basin 605
2.3. Problems and Solutions 609
2.4. Managing Risk 609
2.4.1. Storage 610
2.4.2. Discharge-Increasing Measures 612
2.4.3. Green Rivers 614
2.4.4. Use of Existing Water Courses 615
2.4.5. The Overall Picture 615
2.5. Dealing With Uncertainties 615
2.6. Summary 617
3. Flood Management on the Mississippi 617
3.1. General History 619
3.2. Other Considerations 623
3.3. Interactions Among User Groups 624
3.4. Creating a Flood Management Strategy 626
3.5. The Role of the Government and
NGOs 626
wrm_fm.qxd 8/31/2005 12:06 PM Page xvii
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
3. Analytical Framework: Phases of Analysis 652
4. Inception Phase 654
4.1. Initial Analysis 655
4.1.1. Inventory of Characteristics,
Developments and Policies 655
4.1.2. Problem Analysis 655
4.1.3. Objectives and Criteria 656
4.1.4. Data Availability 656
4.2. Specification of the Approach 657
4.2.1. Analysis Steps 657
4.2.2. Delineation of System 657
4.2.3. Computational Framework 658
4.2.4. Analysis Conditions 659
4.2.5. Work Plan 660
4.3. Inception Report 660
4.4. Communication with Decision-Makers and
Stakeholders 661
5. Development Phase 661
5.1. Model Development and Data Collection 661
5.1.1. Analysis of the Natural Resources System
(NRS) 661
5.1.2. Analysis of the Socio-Economic System
(SES) 664
5.1.3. Analysis of the Administrative and
Institutional System (AIS) 666
5.1.4. Integration into a Computational
Framework 667
5.2. Preliminary Analysis 668
5.2.1. Base Case Analysis 669
5.2.2. Bottleneck (Reference Case)
Analysis 669
5.2.3. Identification and Screening of
Measures 669
5.2.4. Finalization of the Computational
Framework 669
6. Selection Phase 670
6.1. Strategy Design and Impact
Assessment 670
6.2. Evaluation of Alternative Strategies 671
6.3. Scenario and Sensitivity Analysis 672
6.4. Presentation of Results 672
7. Conclusions 672
Index 677
xviii Contents
4. Flood Risk Reduction 627
4.1. Reservoir Flood Storage Capacity 627
4.2. Channel Capacity 630
4.3. Estimating Risk of Levee Failures 631
4.4. Annual Expected Damage From Levee
Failure 633
4.4.1. Risk-Based Analyses 634
5. Decision Support and Prediction 635
5.1. Floodplain Modelling 636
5.2. Integrated 1D2D Modelling 637
6. Conclusions 638
7. References 640
Appendix E: Project Planning and Analysis:
Putting it All Together 644
1. Basic Concepts and Definitions 645
1.1. The Water Resources System 645
1.2. Functions of the Water Resources System 646
1.2.1. Subsistence Functions 646
1.2.2. Commercial Functions 646
1.2.3. Environmental Functions 647
1.2.4. Ecological Values 647
1.3. Policies, Strategies, Measures and
Scenarios 647
1.4. Systems Analysis 648
2. Analytical Description of WRS 649
2.1. System Characteristics of the Natural Resources
System 650
2.1.1. System Boundaries 650
2.1.2. Physical, Chemical and Biological
Characteristics 650
2.1.3. Control Variables: Possible
Measures 651
2.2. System Characteristics of the Socio-Economic
System 651
2.2.1. System Boundaries 651
2.2.2. System Elements and System
Parameters 651
2.2.3. Control Variables: Possible Measures 652
2.3. System Characteristics of the Administrative and
Institutional System 652
2.3.1. System Elements 652
2.3.2. Control Variables: Possible
Measures 652
wrm_fm.qxd 8/31/2005 12:06 PM Page xviii
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water resources are special. In their natural states they are
beautiful. People like to live and vacation near rivers, lakes
and coasts. Water is also powerful. Water can erode rock,
alter existing landscapes and form new ones. Life on this
planet depends on water. Most of our economic activities
consume water. All of the food we grow, process and eat
requires water. Much of our waste is transported and
assimilated by water. The importance of water to our well-
being is beyond question. Our dependence on water will
last forever.
So, what is the problem? The answer is simply that
water, although plentiful, is not distributed as we might
wish. There is often too much or too little, or what exists
is too polluted or too expensive. A further problem is that
the overall water situation is likely to further deteriorate
as a result of global changes. This is a result not only of
climatic change but also of other global change drivers
such as population growth, land use changes, urbaniza-
tion and migration from rural to urban areas, all of which
will pose challenges never before seen. Water obviously
connects all these areas and any change in these drivers
has an impact on it. Water has its own dynamics that are
fairly non-linear. For example, while population growth
in the twentieth century increased three-fold from 1.8
billion to 6 billion people water withdrawal during the
same period increased six-fold! That is clearly unsustain-
able. Freshwater, although a renewable resource, is finite
Introduction
and is very vulnerable. If one considers all the water on
Earth, 97.5% is located in the seas and oceans and what
is available in rivers, lakes and reservoirs for immediate
human consumption comprises no more than a mere
0.007 per cent of the total. This is indeed very limited and
on average is roughly equivalent to 42,000 cubic kilo-
metres per year.
If one looks at the past thirty years only in terms of
reduction in per capita water availability in a year the pic-
ture is even more disturbing. While in 1975 availability
stood at around 13,000 cubic metres per person per year,
it has now dropped to 6,000 cubic metres; meanwhile
water quality has also severely deteriorated. While this
cannot be extrapolated in any meaningful manner, it
nevertheless indicates the seriousness of the situation.
This will likely be further exacerbated by the expected
impacts of climate change. Although as yet unproven to
the required rigorous standards of scientific accuracy,
increasing empirical evidence indicates that the hydro-
logical cycle is accelerating while the amount of water
at a given moment in time is remains the same. If this
acceleration hypothesis is true then it will cause an
increase in the frequency and magnitude of flooding. At
the other end of the spectrum, the prevailing laws of
continuity mean that the severity and duration of drought
will also increase. These increased risks are likely to have
serious regional implications. Early simulation studies,
wrm_fm.qxd 8/31/2005 12:06 PM Page xix
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
xx Introduction
carried out by IHP, suggest that wet areas will become
even more humid while dry areas will become increas-
ingly arid. This will not occur overnight; similarly, appro-
priate countermeasures will need time to establish
policies that integrate the technical and social issues in a
way that takes appropriate consideration of the cultural
context.
Tremendous efforts and political will are needed to
achieve the two water related Millennium Development
Goals (MDGs), that is, to halve the number of human
beings who have no access to safe drinking water and
adequate sanitation facilities respectively, by 2015. In the
case of drinking water, we have 1.2 billion fellow human
beings that have no access to safe drinking water, while in
the case of sanitation, the figure is 2.4 billion.
The substantial growth of human populations
especially as half of humanity already lives in urban areas
and the consequent expansion of agricultural and
industrial activities with a high water demand, have only
served to increase problems of water availability, quality
and in many regions waterborne disease. There is now
an increasing urgency in the UN system to protect water
resources through better management. Data on the scale
of deforestation with subsequent land use conversion, soil
erosion, desertification, urban sprawl, loss of genetic
diversity, climate change and the precariousness of food
production through irrigation, all reveal the growing
seriousness of the problem. We have been forced to
recognize that societys activities can no longer continue
unchecked without causing serious damage to the very
environment and ecosystems we depend upon for our
survival. This is especially critical in water scarce
regions, many of which are found in the developing
world and are dependent on water from aquifers that
are not being recharged as fast as their water is being
withdrawn and consumed. Such practices are clearly not
sustainable.
Proper water resources management requires consid-
eration of both supply and demand. The mismatch of
supply and demand over time and space has motivated
the development of much of the water resources
infrastructure that is in place today. Some parts of the
globe witness regular flooding as a result of monsoons
and torrential downpours, while other areas suffer from
the worsening of already chronic water shortages. These
conditions are often aggravated by the increasing
discharge of pollutants resulting in a severe decline in
water quality.
The goal of sustainable water management is to pro-
mote water use in such a way that societys needs are
both met to the extent possible now and in the future. This
involves protecting and conserving water resources
that will be needed for future generations. UNESCOs
International Hydrological Programme (IHP) addresses
these short- and long-term goals by advancing our under-
standing of the physical and social processes affecting
the globes water resources and integrating this knowledge
into water resources management. This book describes
the kinds of problems water managers can and do face and
the types of models and methods one can use to define
and evaluate alternative development plans and manage-
ment policies. The information derived from these models
and methods can help inform stakeholders and decision-
makers alike in their search for sustainable solutions to
water management problems. The successful application
of these tools requires collaboration among natural and
social scientists and those in the affected regions, taking
into account not only the water-related problems but also
the social, cultural and environmental values.
On behalf of UNESCO it gives me great pleasure to
introduce this book. It provides a thorough introduction
to the many aspects and dimensions of water resources
management and presents practical approaches for
analysing problems and identifying ways of developing
and managing water resources systems in a changing and
uncertain world. Given the practical and academic expe-
rience of the authors and the contributions they have
made to our profession, I am confident that this book
will become a valuable asset to those involved in water
resources planning and management. I wish to extend our
deepest thanks to Professors Pete Loucks and Eelco van
Beek for offering their time, efforts and outstanding expe-
rience, which is summarized in this book for the benefit
of the growing community of water professionals.
Andrs Szllsi-Nagy
Deputy Assistant Director-General, UNESCO
Secretary, International Hydrological Programme
wrm_fm.qxd 8/31/2005 12:06 PM Page xx
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Water Resources Planning and Management: An Overview
1. Introduction 3
2. Planning and Management Issues: Some Case Studies 4
2.1. Kurds Seek Land, Turks Want Water 4
2.2. Sharing the Water of the Jordan River Basin: Is There a Way? 6
2.3 Mending the Mighty and Muddy Missouri 7
2.4. The Endangered Salmon 7
2.5. The Yellow River: How to Keep the Water Flowing 9
2.6. Lake Source Cooling: Aid to Environment or Threat to Lake? 10
2.7. Managing Water in the Florida Everglades 11
2.8. Restoration of Europes Rivers and Seas 13
2.8.1. The Rhine 13
2.8.2. The Danube 14
2.8.3. The North and Baltic Seas 15
2.9. Egypt and the Nile: Limits to Agricultural Growth 16
2.10. Damming the Mekong 16
3. So, Why Plan, Why Manage? 18
3.1. Too Little Water 20
3.2. Too Much Water 20
3.3. Polluted Water 21
3.4. Degradation of Aquatic and Riparian Ecosystems 21
3.5. Other Planning and Management Issues 21
4. System Components, Planning Scales and Sustainability 22
4.1. Spatial Scales for Planning and Management 22
4.2. Temporal Scales for Planning and Management 23
4.3. Sustainability 23
5. Planning and Management 24
5.1. Approaches 24
5.1.1. Top-Down Planning and Management 25
5.1.2. Bottom-Up Planning and Management 25
5.1.3. Integrated Water Resources Management 26
5.2. Planning and Management Aspects 26
5.2.1. Technical Aspects 26
5.2.2. Economic and Financial Aspects 27
5.2.3. Institutional Aspects 28
5.3. Analysis for Planning and Management 28
5.4. Models for Impact Prediction and Evaluation 30
5.5. Shared-Vision Modelling 31
5.6. Adaptive Integrated Policies 31
5.7. Post-Planning and Management Issues 32
6. Meeting the Planning and Management Challenges: A Summary 32
7. References 34
wrm_ch01.qxd 8/31/2005 11:48 AM Page 2
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Over the centuries, surface and ground waters have been
a source of water supplies for agricultural, municipal and
industrial consumers. Rivers have provided hydroelectric
energy and inexpensive ways of transporting bulk cargo
between different ports along their banks, as well as
water-based recreational opportunities, and have been a
source of water for wildlife and its habitat. They have also
served as a means of transporting and transforming waste
products that are discharged into them. The quantity and
quality regimes of streams and rivers have been a major
factor in governing the type, health and biodiversity of
riparian and aquatic ecosystems. Floodplains have pro-
vided fertile lands for agricultural production and rela-
tively flat lands for roads, railways and commercial and
industrial complexes. In addition to the economic bene-
fits that can be derived from rivers and their floodplains,
3
Water Resources Planning and
Management: An Overview
Water resource systems have benefited both people and their economies for many
centuries. The services provided by such systems are multiple. Yet in many regions,
water resource systems are not able to meet the demands, or even the basic needs, for
clean fresh water, nor can they support and maintain resilient biodiverse ecosystems.
Typical causes of such failures include degraded infrastructures, excessive withdrawals
of river flows, pollution from industrial and agricultural activities, eutrophication from
excessive nutrient loads, salinization from irrigation return flows, infestations of exotic
plants and animals, excessive fish harvesting, floodplain and habitat alteration from
development activities, and changes in water and sediment flow regimes. Inadequate
water resource systems reflect failures in planning, management and decision-
making and at levels broader than water. Planning, developing and managing water
resource systems to ensure adequate, inexpensive and sustainable supplies and
qualities of water for both humans and natural ecosystems can only be successful if
such activities address the causal socio-economic factors, such as inadequate
education, population pressures and poverty.
1
the aesthetic beauty of most natural rivers has made lands
adjacent to them attractive sites for residential and recre-
ational development. Rivers and their floodplains have
generated and, if managed properly, can continue to
generate substantial economic, environmental and social
benefits for their inhabitants.
Human activities undertaken to increase the benefits
obtained from rivers and their floodplains may also
increase the potential for costs and damage when the river
is experiencing rare or extreme flow conditions, such as
during periods of drought, floods and heavy pollution.
These costs and impacts are economic, environmental
and social in nature and result from a mismatch between
what humans expect or demand, and what nature (and
occasionally our own activities) offers or supplies. Human
activities tend to be based on the usual or normal range
of river flow conditions. Rare or extreme flow or water
quality conditions outside these normal ranges will
wrm_ch01.qxd 8/31/2005 11:48 AM Page 3
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
continue to occur, and possibly with increasing frequency
as climate change experts suggest. River-dependent,
human activities that cannot adjust to these occasional
extreme conditions will incur losses.
The planning of human activities involving rivers and
their floodplains must consider certain hydrological facts.
One of these facts is that flows and storage volumes vary
over space and time. They are also finite. There are limits
to the amounts of water that can be withdrawn from sur-
face and groundwater bodies. There are also limits to the
amounts of potential pollutants that can be discharged
into them without causing damage. Once these limits are
exceeded, the concentrations of pollutants in these waters
may reduce or even eliminate the benefits that could be
obtained from other uses of the resource.
Water resources professionals have learned how to
plan, design, build and operate structures that, together
with non-structural measures, increase the benefits
people can obtain from the water resources contained in
rivers and their drainage basins. However, there is a limit
to the services one can expect from these resources.
Rivers, estuaries and coastal zones under stress from over-
development and overuse cannot reliably meet the expec-
tations of those depending on them. How can these
renewable yet finite resources best be managed and used?
How can this be accomplished in an environment of
uncertain supplies and uncertain and increasing
demands, and consequently of increasing conflicts among
individuals having different interests in the management
of a river and its basin? The central purpose of water
resources planning and management activities is to
address and, if possible, answer these questions. These
issues have scientific, technical, political (institutional)
and social dimensions and thus, so must water resources
planning processes and their products.
River basin, estuarine and coastal zone managers
those responsible for managing the resources in those
areas are expected to manage them effectively and effi-
ciently, meeting the demands or expectations of all users
and reconciling divergent needs. This is no small task,
especially as demands increase, as the variability of hydro-
logical and hydraulic processes becomes more pronounced,
and as stakeholder measures of system performance
increase in number and complexity. The focus or goal is
no longer simply to maximize net economic benefits while
ensuring the equitable distribution of those benefits. There
4 Water Resources Systems Planning and Management
are also environmental and ecological goals to consider.
Rarely are management questions one-dimensional, such
as: How can we provide more high-quality water to irri-
gation areas in the basin at acceptable costs? Now added
to that question is how those withdrawals would affect
the downstream water quantity and quality regimes, and
in turn the riparian and aquatic ecosystems. To address
such what if questions requires the integration of a
variety of sciences and technologies with people and
their institutions.
Problems and opportunities change over time. Just as
the goals of managing and using water change over time,
so do the processes of planning to meet these changing
goals. Planning processes evolve not only to meet
new demands, expectations and objectives, but also in
response to new perceptions of how to plan more
effectively.
This book is about how quantitative analysis, and in
particular computer models, can support and improve
water resources planning and management. This first
chapter attempts to review some of the issues involved. It
provides the context and motivation for the chapters that
follow, which describe in more detail our understanding
of how to plan and how to manage and how computer-
based programs and models can assist those involved
in these activities. Additional information is available in
many of the references listed at the end of each chapter.
2. Planning and Management
Issues: Some Case Studies
Managing water resources certainly requires knowledge
of the relevant physical sciences and technology. But at
least as important, if not more so, are the multiple insti-
tutional, social or political issues confronting water
resources planners and managers. The following brief
descriptions of some water resources planning and man-
agement studies at various geographic scales illustrate
some of these issues.
2.1. Kurds Seek Land, Turks Want Water
The Tigris and Euphrates Rivers (Figure 1.1) in the
Middle East created the Fertile Crescent where some
of the first civilizations emerged. Today their waters are
wrm_ch01.qxd 8/31/2005 11:48 AM Page 4
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 5
critical resources, politically as well as geographically. In
one of the worlds largest public works undertakings,
Turkey is spending over $30 billion in what is called the
Great Anatolia Project (GAP), a complex of 22 reservoirs
and 19 hydroelectric plants. Its centrepiece, the Ataturk
Dam (Figure 1.2) on the Euphrates River, is already
completed. In the lake formed behind the dam, sailing
and swimming competitions are being held on a spot
where, for centuries, there was little more than desert
(Figure 1.3).
When the project is completed it is expected to
increase the amount of irrigated land in Turkey by 40%
and provide up to a quarter of the countrys electric
power needs. Planners hope this can improve the stan-
dard of living of six million of Turkeys poorest people,
most of them Kurds, and thus undercut the appeal of
revolutionary separatism. It will also reduce the amount
of water Syria and Iraq believe they need water that
Turkey fears might ultimately be used for anti-Turkish
causes.
The region of Turkey where Kurds predominate is
more or less the same region covered by the Great
Anatolia Project, encompassing an area about the size of
Austria. Giving that region autonomy by placing it under
Kurdish self-rule could weaken the Central Governments
control over the water resources that it recognizes as a
keystone of its future power.
In other ways also, Turkish leaders are using their
water as a tool of foreign as well as domestic policy.
Among their most ambitious projects considered is a fifty-
mile undersea pipeline to carry water from Turkey to
the parched Turkish enclave on northern Cyprus. The
pipeline, if actually built, will carry more water than
northern Cyprus can use. Foreign mediators, frustrated
by their inability to break the political deadlock on
Cyprus, are hoping that the excess water can be sold to
the ethnic Greek republic on the southern part of the
island as a way of promoting peace.
Turkey
Euphrates
Tigris
Syria
Damascus
Baghdad
Jordan
Saudi Arabia
Iraq
Iran
Caspian
Sea
Persian
Gulf
Mediterranean
Sea
E
0
1
1
2
1
7a
Ataturk dam
Figure 1.1. The Tigris and Euphrates Rivers in Turkey,
northern Syria and Iraq.
Figure 1.2. Ataturk Dam on the Euphrates River in Turkey (DSI).
Figure 1.3. Water sports on the Ataturk Reservoir on the
Euphrates River in Turkey (DSI).
wrm_ch01.qxd 8/31/2005 11:48 AM Page 5
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
2.2. Sharing the Water of the Jordan River
Basin: Is There a Way?
A growing population approximately 12 million people
and intense economic development in the Jordan River
Basin (Figure 1.4) are placing heavy demands on its
scarce freshwater resources. Though the largely arid
region receives less than 250 millimetres of rainfall each
year, total water use for agricultural and economic activi-
ties has been steadily increasing. This and encroaching
urban development have degraded many sources of high-
quality water in the region.
The combined diversions by the riparian water users
have changed the river in its lower course into little bet-
ter than a sewage ditch. Of the 1.3 billion cubic metres
(mcm or 10
6
m
3
) of water that flowed into the Dead Sea
in the 1950s, only a small fraction remains at present. In
normal years the flow downstream from Lake Tiberias
(also called the Sea of Galilee or Lake Kinneret) is some
60 mcm about 10% of the natural discharge in this
section. It mostly consists of saline springs and sewage
water. These flows are then joined by what remains
of the Yarmouk, by some irrigation return flows and by
6 Water Resources Systems Planning and Management
winter runoff, adding up to an annual total of 200
to 300 mcm. This water is unsuitable for irrigation in
both quantity and quality, nor does it sufficiently supply
natural systems. The salinity of the Jordan River reaches
up to 2,000 parts per million (ppm) in the lowest section,
which renders it unfit for crop irrigation. Only in flood
years is fresh water released into the lower Jordan Valley.
One result of this increased pressure on freshwater
resources is the deterioration of the regions wetlands,
which are important for water purification and flood and
erosion control. As agricultural activity expands, wetlands
are being drained, and rivers, aquifers, lakes and streams
are being polluted with runoff containing fertilizers and
pesticides. Reversing these trends by preserving natural
ecosystems is essential to the future availability of fresh
water in the region.
To ensure that an adequate supply of fresh, high-quality
water is available for future generations, Israel, Jordan and
the Palestinian Authority will have to work together to
preserve aquatic ecosystems (National Research Council,
1999). Without these natural ecosystems, it will be difficult
and expensive to sustain high-quality water supplies. The
role of ecosystems in sustaining water resources has largely
been overlooked in the context of the regions water provi-
sion. Vegetation controls storm runoff, filters polluted
water and reduces erosion and the amount of sediment
that makes its way into water supplies. Streams assimilate
wastewater, lakes store clean water, and surface waters
provide habitats for many plants and animals.
The Jordan River Basin, like most river basins, should
be evaluated and managed as a whole to permit the com-
prehensive assessment of the effects of water management
options on wetlands, lakes, the lower river and the Dead
Sea coasts. Damage to ecosystems and loss of animal and
plant species should be weighed against the potential ben-
efits of developing land and creating new water resources.
For example, large river-management projects that divert
water to dry areas have promoted intensive year-round
farming and urban development, but available river water
is declining and becoming increasingly polluted. Attempting
to meet current demands sol el y by wi thdrawing more
ground and surface water could result in widespread
environmental degradation and depletion of freshwater
resources.
There are policies that, if implemented, could help
preserve the capacity of the Jordan River to meet future Figure 1.4. The Jordan River between Israel and Jordan.
E
0
1
1
2
1
7
b
0 25 km
Golan
West
Bank
Gaza Strip
occupied territories
Negev Desert
Damascus
Mediterranean Sea
Lebanon
Syria
Egypt
Jordan
Haifa
Tel Aviv
Jerusalem
Lake Tiberias
Dead Sea
Litani
Yarmuk
Jordan
Israel
J
o
r
d
a
n
wrm_ch01.qxd 8/31/2005 11:48 AM Page 6
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 7
demands. Most of the options relate to improving the effi-
ciency of water use: that is, they involve conservation and
better use of proven technologies. Also being considered
are policies that emphasize economic efficiency and reduce
overall water use. Charging higher rates for water use in
peak periods and surcharges for excessive use, would
encourage conservation. In addition, new sources of fresh
water can be obtained by capturing rainfall through rooftop
cisterns, catchment systems and storage ponds.
Thus there are alternatives to a steady deterioration of
the water resources of the Jordan Basin. They will require
coordination and cooperation among all those living in
the basin. Will this be possible?
2.3. Mending the Mighty and Muddy Missouri
Nearly two centuries after an epic expedition through the
western United States in search of a northwest river pas-
sage to the Pacific Ocean, there is little enchantment left to
the Missouri River. Shown in Figure 1.5, it has been
dammed, dyked and dredged since the 1930s to control
floods and float cargo barges. The river nicknamed the
Mighty Missouri and the Big Muddy by its explorers is
today neither mighty nor very muddy. The conservation
group American Rivers perennially lists the Missouri
among the United States ten most endangered rivers.
Its wilder upper reaches are losing their cottonwood
trees to dam operations and cattle that trample seedlings
along the rivers banks. In its vast middle are multiple
dams that hold back floods, generate power and provide
pools for boats and anglers.
Its lower third is a narrow canal sometimes called the
Ditch that is deep enough for commercial tow-boats.
Some of the rivers banks are armoured with rock and con-
crete retaining walls that protect half a million acres of
farm fields from flooding. Once those floods produced and
maintained marshlands and side streams habitats for a
wide range of wildlife. Without these habitats, many wild
species are unable to thrive, or in some cases even survive.
Changes to restore at least some of the Missouri to a
more natural state are being implemented. Protection of
fish and wildlife habitat has been added to the list of
objectives to be achieved by the government agencies
managing the Missouri. The needs of wildlife are now
seen to be as important as other competing interests on
the river, including navigation and flood control. This is
in reaction, in part, to the booming $115 million-a-year
outdoor recreation industry. Just how much more
emphasis will be given to these back-to-nature goals
depends on whether the Missouri River Basin Association,
an organization representing eight states and twenty-eight
Native American tribes, can reach a compromise with the
traditional downstream uses of the river.
2.4. The Endangered Salmon
Greater Seattle in the northwestern US state of
Washington may be best known around the world for its
software and aviation industry, but residents know it for
something less flashy: its dwindling stock of wild salmon
(see Figure 1.6). The Federal Government has placed
seven types of salmon and two types of trout on its
list of threatened or endangered species. Saving the fish
from extinction will require sacrifices and could slow
development in one of the fastest-growing regions of the
United States.
Before the Columbia River and its tributaries in the
northwestern United States were blocked with dozens of
dams, about 10 to 16 million salmon made the annual
run back up to their spawning grounds. In 1996, a little
less than a million did. But the economy of the Northwest
depends on the dams and locks that have been built in the
Columbia to provide cheap hydropower production and
navigation.
Snake
Columbia
Colorado
Rio Grande
Missouri
Arkansas
Red
Ohio
Tennessee
E
0
1
1
2
1
7
c
Mississippi
Figure 1.5. Major rivers in the continental United States.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 7
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
8 Water Resources Systems Planning and Management
For a long time, engineers tried to jury-rig the system
so that fish passage would be possible. It has not worked
all that well. Still too many young fish enter the
hydropower turbines on their way down the river. Now,
as the debate over whether or not to remove some dams
takes place, fish are caught and carried by truck around
the turbines. The costs of keeping these salmon alive, if
not completely happy, are enormous.
Over a dozen national and regional environmental
organizations have joined together to bring back salmon
and steelhead by modifying or partially dismantling five
federal dams on the Columbia and Snake Rivers. Partial
removal of the four dams on the lower Snake River in
Washington State and lowering the reservoir behind
John Day Dam on the Columbia bordering Oregon and
Washington (see Figure 1.7) should help restore over
300 km of vital river habitat. Running the rivers in a more
Figure 1.6. A salmon swimming upstream (US Fish and
Wildlife Service, Pacific Region).
John
Day Dam
Ice
Harbour
Dam
Little
Goose
Dam
Lower
Granite
Dam
Oregon
Idaho
Washington
Alberta
Montana
Nevada California Utah
Lower
Monumental
Dam
British Columbia
P
a
c
i
f
i
c

O
c
e
a
n
Columbia
R
i
v
e
r
C
olum
bia
K
o
o
t
e
n
a
y
Salmon
R
iv
e
r
S
n
a
k
e
R
i
v
e
r
River
Snake
River
E
0
1
1
2
1
7
e
Figure 1.7. The Snake and
Columbia River reservoirs
identified by the Columbia and
Snake Rivers Campaign for
modification or dismantling to
permit salmon passage.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 8
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 9
natural way may return salmon and steelhead to the har-
vestable levels of the 1960s before the dams were built.
Dismantling part of the four Lower Snake dams will
leave most of each dam whole. Only the dirt bank con-
necting the dam to the riverbank will be removed. The
concrete portion of the dam will remain in place, allow-
ing the river to flow around it. The process is reversible
and, the Columbia and Snake Rivers Campaign argues, it
will actually save taxpayers money in planned dam main-
tenance by eliminating subsidies to shipping industries
and agribusinesses, and by ending current salmon recov-
ery measures that are costly. Only partially removing the
four Lower Snake River dams and modifying John Day
Dam will restore rivers, save salmon and return balance to
the Northwests major rivers.
2.5. The Yellow River: How to Keep the Water
Flowing
The Yellow River is one of the most challenging in the
world from the point of view of water and sediment
management. Under conditions of normal and low flow,
the water is used for irrigation, drinking and industry to
such an extent that the lower reach runs dry during many
days each year. Under high-flow conditions, the river is
heavily laden with very fine sediment originating from the
Lss Plateau, to the extent that a hyperconcentrated flow
occurs. Through the ages the high sediment load has
resulted in the building-out of a large delta in the Bohai
Sea and a systematic increase of the large-scale river
slope. Both have led to what is now called the suspended
river: the riverbed of the lower reach is at points some
10 metres above the adjacent land, with dramatic effects
if dyke breaching were to occur.
The Yellow River basin is already a very water-scarce
region. The rapid socio-economic development in China
is putting the basin under even more pressure.
Agricultural, industrial and population growth will fur-
ther increase the demand for water. Pollution has reached
threatening levels. The Chinese government, in particular
the Yellow River Conservancy Commission (YRCC), has
embarked on an ambitious program to control the river
and regulate the flows. Their most recent accomplishment
is the construction of the Xiaolangdi Dam, which will
0 500 km 100 200 300 400
E
0
4
0
2
1
8
b
Xi' an
Bohai
Sea
Zhaling
Lake
Lanzhou
Huhehaote
Zhengzhou
San-
menxia
Q
i
n
Xiaolangdi
Longyangxia
Wei
F
e
n
Eling
Lake
B
a
i
H
e
i
Tao
K
u
y
i
e
L
u
o
J
i
n
g D
a
w
e
n
Xining
Yinchuan
Taiyuan
Jinan
Huayuankou
Liujiaxia
Qingtongxia
Wanjiazhai
H
u
a
n
g
s
h
u
i
Lss Plateau
main reservoir
Beijing
Tianjin
main irrigation area
Yellow River (Huang He)
Y
ilu
o
Wuding
Figure 1.8. The Yellow River
Basin.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 9
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
control water and sediment just before the river enters the
flat lower reach. This controlling includes a concentrated
release of high volumes of water to flush the sediment out
to sea.
In the delta of the Yellow River, fresh water wetlands
have developed with a dynamic and unique ecosystem
of valuable plant species and (transmigratory) birds. The
decreased and sometimes zero flow in the river is threat-
ening this ecosystem. To protect it, the YRCC has started
to release additional water from the Xiaolangdi dam to
supply these wetlands with water during dry periods.
The water demand of the wetlands is in direct competi-
tion with the agricultural and industrial demands
upstream, and there have been massive complaints about
this waste of valuable water. Solving this issue and
agreeing upon an acceptable distribution over users and
regions is a nearly impossible task, considering also that
the river crosses nine rather autonomous provinces.
How can water be kept flowing in the Yellow River
basin? Under high-flow conditions the sediment has to be
flushed out of the basin to prevent further build-up of the
suspended river. Under low-flow conditions water has to
be supplied to the wetlands. In both cases the water is seen
as lost for what many consider to be its main function: to
support the socio-economic development of the region.
2.6. Lake Source Cooling: Aid to Environment
or Threat to Lake?
It seems an environmentalists dream: a cost-effective system
that can cool some 10 million square feet of high school
and university buildings simply by pumping cold water
from the depths of a nearby lake (Figure 1.9), without the
emission of chlorofluorocarbons (the refrigerants that can
destroy protective ozone in the atmosphere) and at a cost
substantially smaller than for conventional air conditioners.
The water is returned to the lake, with a few added calories.
However, a group of local opponents insists that
Cornell Universitys $55-million lake-source-cooling
plan, which has replaced its aging air conditioners, is
actually an environmental threat. They believe it could
foster algal blooms. Pointing to five years of studies,
thousands of pages of data, and more than a dozen permits
from local and state agencies, Cornells consultants say the
system could actually improve conditions in the lake. Yet
another benefit, they say, is that the system would reduce
10 Water Resources Systems Planning and Management
Cornells contribution to global warming by reducing the
need to burn coal to generate electricity.
For the most part, government officials agree. But a
small determined coalition of critics from the local com-
munity argue over the expected environmental impacts,
and over the process of getting the required local, state
and federal permits approved. This is in spite of the fact
that the planning process, which took over five years,
requested and involved the participation of all interested
stakeholders from the very beginning. Even the local
chapter of the Sierra Club and biology professors at
other universities have endorsed the project. However, in
almost every project where the environmental impacts are
uncertain, there will be debates among scientists as well
as among stakeholders. In addition, a significant segment
of society distrusts scientists anyway. This is a major
societal problem, wrote a professor and expert in the
dynamics of lakes. A scientist says X and someone else
says Y and youve got chaos. In reality, we are the prob-
lem. Every time we flush our toilets, fertilize our lawns,
gardens and fields, or wash our cars, we contribute to the
nutrient loading of the lake.
The project has now been operating for over five years,
and so far no adverse environmental effects have been
noticed at any of the many monitoring sites.
Figure 1.9. The cold deep waters of Lake Cayuga are being
used to cool the buildings of a local school and university
(Ithaca City Environmental Laboratory).
wrm_ch01.qxd 8/31/2005 11:48 AM Page 10
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 11
2.7. Managing Water in the Florida Everglades
The Florida Everglades (Figure 1.10) is the largest single
wetland in the continental United States. In the mid-
1800s it covered a little over 3.6 million ha, but since that
time the historical Everglades has been drained and half
of the area is now devoted to agriculture and urban devel-
opment. The remaining wetland areas have been altered
by human disturbances both around and within them.
Water has been diverted for human uses, flows have been
lowered to protect against floods, nutrient supplies to the
wetlands from runoff from agricultural fields and urban
areas have increased, and invasions of non-native or
otherwise uncommon plants and animals have out-com-
peted native species. Populations of wading birds (includ-
ing some endangered species) have declined by 85 to
90% in the last half-century, and many species of South
Floridas mammals, birds, reptiles, amphibians and plants
are either threatened or endangered.
The present management system of canals, pumps,
and levees (Figure 1.11) will not be able to provide
adequate water supplies or sufficient flood protection to
agricultural and urban areas, let alone support the natural
(but damaged) ecosystems in the remaining wetlands.
The system is not sustainable. Problems in the greater
Everglades ecosystem relate to both water quality and
quantity, including the spatial and temporal distri-
bution of water depths, flows and flooding durations
(called hydroperiods). Issues arise due to variations in
Figure 1.10. Scenes of the Everglades in southern Florida
(South Florida Water Management District).
Figure 1.11. Pump station on a drainage canal in southern
Florida (South Florida Water Management District).
wrm_ch01.qxd 8/31/2005 11:48 AM Page 11
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the natural/historical hydrological regime, degraded water
quality and the sprawl from fast-growing urban areas.
To meet the needs of the burgeoning population and
increasing agricultural demands for water, and to begin
the restoration of the Everglades aquatic ecosystem to a
more natural state, an ambitious plan has been developed
by the US Army Corps of Engineers (USACE) and its local
sponsor, the South Florida Water Management District.
The proposed Corps plan is estimated to cost over $8 bil-
lion. The plan and its Environmental Impact Statement
(EIS) have received input from many government agen-
cies and non-governmental organizations, as well as from
the public at large.
The plan to restore the Everglades is ambitious and
comprehensive, involving the change of the current
12 Water Resources Systems Planning and Management
hydrological regime in the remnant of the Everglades to one
that resembles a more natural one, the re-establishment of
marshes and wetlands, the implementation of agricultural
best-management practices, the enhancement of wildlife
and recreation areas, and the distribution of provisions for
water supply and flood control to the urban population,
agriculture and industry.
Planning for and implementing the restoration effort
requires application of state-of-the-art large systems
analysis concepts, hydrological and hydroecological
data and models incorporated within decision support
systems, integration of social sciences, and monitoring
for planning and evaluation of performance in an adaptive
management context. These large, complex challenges
of the greater Everglades restoration effort demand the
Atlantic
Ocean
Norwegian
Sea
Arctic Ocean
North Sea
Black
Sea
B
a
lt
ic
S
e
a
Bay of
Biscay
A
d
r
ia
t
ic
S
e
a
Mediterranean Sea
S
k
a
g
e
rra
k
S
e
a
I
r
i
s
h
B
o
t
h
n
i
a
o
f
G
u
l
f
K
a
t
t
e
g
a
t
EnglishChannel
Tyrrhenian
Sea
Str. of Gibraltar
L
ig
u
r
ia
n
Sea
Ionian
Sea
Sea
AEgean
Thames
Rhine
Seine
Meuse
Loire
Danube
Po
Elbe
Vistula
Volga
Dnepr
0 500 km
Ebro
Tagus
Guadiana
Rhone
Odra
E
0
1
1
2
1
7
f
Figure 1.12. Europes major rivers and seas.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 12
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 13
most advanced, interdisciplinary and scientifically-sound
analysis capabilities available. They also require the
political will to make compromises and to put up with
lawsuits by anyone who may be disadvantaged by some
restoration measure.
Who pays for all this? Both the taxpayers of Florida,
and the taxpayers of the United States.
2.8. Restoration of Europes Rivers and Seas
2.8.1. The Rhine
The map of Figure 1.13 shows the areas of the nine
countries that are part of river Rhine basin. In the Dutch
area of the Rhine basin, water is partly routed northward
through the Ijssel and westward through the highly
interconnected river systems of the Rhine, Meuse and
Waal. About 55 million people live in the Rhine River
basin and about 20 million of those people drink the river
water.
In the mid-1970s, some called the Rhine the most
romantic sewer in Europe. In November 1986, a chemical
spill degraded much of the upper Rhines aquatic ecosys-
tem. This damaging event was reported worldwide. The
Rhine was again world news in the first two months of
1995 when its water level reached a height that occurs
on average once in a century. In the Netherlands, some
200,000 people, 1,400,000 pigs and cows and 1,000,000
chickens had to be evacuated. During the last two months
of the same year there was hardly enough water in the
Aare
Alpenrhein
Hochrhein
Moezel
Neckar
Main
Oberrhein
Lahn
Sieg
Mittelrhein
Wupper
Ruhr
Lippe
Niederrhein
Nahe
Thur
Bodensee
Lake
IJssel
IJssel
Waal
Lek
Nederrijn
The
Netherlands
Germany
Luxembourg
France
Austria
Belgium
Switzerland
E
0
1
1
2
1
7
g
Figure 1.13. The Rhine River
basin of western Europe and its
extent in the Netherlands.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 13
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Rhine for navigation. It is fair to say these events have
focused increased attention on what needs to be done to
restore and protect the Rhine.
To address just how to restore the Rhine, it is useful to
look at what has been happening to the river during the past
150 years. The Rhine, the only river connecting the Alps
with the North Sea, was originally a natural watercourse. To
obtain greater economic benefits from the river, it was
engineered for navigation, hydropower, water supply and
flood protection. Floodplains, now protected from floods,
provided increased land areas suitable for development.
The main stream of the Rhine is now considerably shorter,
narrower and deeper than it was originally.
From an economic development point of view, the
engineering works implemented in the river and its basin
worked. The Rhine basin is now one of the most industri-
alized regions in the world and is characterized by intensive
industrial and agricultural activities: it contains some 20%
of the worlds chemical industry. The river is reportedly the
busiest shipping waterway in the world, containing long
canals with regulated water levels, connecting the Rhine
and its tributaries with the rivers of almost all the sur-
rounding river basins, including the Danube. This provides
water transport to and from the North and Black Seas.
From an environmental and ecological viewpoint, and
from the viewpoint of flood control as well, the economic
14 Water Resources Systems Planning and Management
development that has taken place over the past two
centuries has not worked perfectly. The concerns aroused
by the recent toxic spill and floods, and from a generally
increasing interest by the inhabitants of the basin in envi-
ronmental and ecosystem restoration and the preservation
of natural beauty, have resulted in basinwide efforts to
rehabilitate the basin to a more living sustainable entity.
A Rhine Action Programme has been created to revive
the ecosystem. The goal of that program is the revival
of the main stream as the backbone of the ecosystem,
particularly for migratory fish, and the protection, main-
tenance and revival of ecologically important areas along
the Rhine. Implemented in the 1990s, the plan was given
the name Salmon 2000, since the return of salmon to the
Rhine is seen as a symbol of ecological revival. A healthy
salmon population will need to swim throughout the river
length. This will be a challenge, as no one pretends that
the engineering works that provide navigation and
hydropower benefits, but which also inhibit fish passage,
are no longer needed or desired.
2.8.2. The Danube
The Danube River (shown in Figure 1.14) is in the heart-
land of central Europe. Its basin includes large parts of
the territories of thirteen countries. It additionally receives
Figure 1.14. The Danube River
in central Europe.
Munich
Vienna
Budapest
Belgrade
Bucarest
E
0
4
0
2
1
8
a
Zagreb
Germany
Czech
Republic
Slovak Republic Ukraine
Moldava
Romania
Hungary
Austria
Slovenia
Croatia
Bosnia and
Herzegovina
Yugoslavia
Bulgaria
Inn
Mura
Drova
Sava
Morava
Morava
Tisza
Oltul
Danube
wrm_ch01.qxd 8/31/2005 11:48 AM Page 14
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 15
runoff from small catchments located in five other coun-
tries. About 85 million people live in the basin. This river
encompasses greater political, economic and social varia-
tions than arguably any other river basin in Europe.
The river discharges into the Black Sea. The Danube
delta and the banks of the Black Sea have been designated
a biosphere reserve by UNESCO. Over half of the delta has
been declared a wet zone of international significance.
Throughout its length the Danube provides a vital resource
for drainage, communications, transport, power genera-
tion, fishing, recreation and tourism. It is considered to be
an ecosystem of irreplaceable environmental value.
More than forty dams and large barrages, plus over
500 smaller reservoirs have been constructed on the main
Danube River and its tributaries. Flood-control dykes
confine most of the length of the main stem of the Danube
River and the major tributaries. Over the last fifty years
natural alluvial floodplain areas have declined from about
26,000 km
2
to about 6,000 km
2
.
There are also significant reaches with river training
works and river diversion structures. These structures trap
nutrients and sediment in the reservoirs, which causes
changes in downstream flow and sediment transport
regimes that reduce the ecosystems habitats both
longitudinally and transversely and decrease the effi-
ciency of natural purification processes. Thus, while these
engineered facilities provide important opportunities for
the control and use of the rivers resources, they also
illustrate the difficulties of balancing these important
economic activities with environmentally sound and sus-
tainable management.
The environmental quality of the Danube River is also
under intense pressure from a diverse range of human
activities, including point source and non-point source
agricultural, industrial and municipal wastes. Because of
the poor water quality (sometimes affecting human
health), the riparian countries of the basin have been
participating in environmental management activities
on regional, national and local levels for several decades.
All Danube countries signed a formal Convention on
Cooperation for the Protection and Sustainable Use of the
Danube River in June 1994. The countries have agreed to
take all appropriate legal, administrative and technical
measures to improve the current environmental and
water quality conditions of the Danube River and of the
waters in its catchment area, and to prevent and reduce as
much as possible the adverse impacts and changes occur-
ring or likely to be caused.
2.8.3. The North and Baltic Seas
The North and Baltic Seas (shown in Figure 1.12) are the
most densely navigated seas in the world. Besides ship-
ping, military and recreational uses, there is an offshore
oil and gas industry, and telephone cables cover the
seabed. The seas are rich and productive, with resources
that include not only fish but also crucial minerals (in
addition to oil) such as gas, sand and gravel. These
resources and activities play major roles in the economies
of the surrounding countries.
Since the seas are so intensively exploited and are sur-
rounded by advanced industrialized countries, pollution
problems are serious. The main pollution sources include
rivers and other outfalls, dumping by ships (of dredged
materials, sewage sludge and chemical wastes) and opera-
tional discharges from offshore installations and ships.
Deposition of atmospheric pollutants is an additional major
source of pollution.
Those parts of the seas at greatest risk from pollution
are where the sediments come to rest, where the water
replacement is slowest, and where nutrient concentra-
tions and biological productivity are highest. A number of
warning signals have occurred.
Algal populations have changed in number and species.
There have been algal blooms, caused by excessive nutrient
discharge from land and atmospheric sources. Species
changes show a tendency toward more short-lived species
of the opportunistic type and a reduction, sometimes to the
point of disappearance, of some mammal, fish and sea grass
species. Decreases of ray, mackerel, sand eel and echino-
derms due to eutrophication have resulted in reduced
plaice, cod, haddock and dab, mollusk and scoter. The
impact of fishing activities is also considerable. Sea mam-
mals, sea birds and Baltic fish species have been particularly
affected by the widespread release of toxins and pollutants
that accumulate in the sediments and in the food web.
Some species, such as the grey seal and the sea eagle, are
threatened with extinction.
Particular concern has been expressed about the
Wadden Sea, which serves as a nursery for many North
Sea species. Toxic PCB contamination, for example,
almost caused the disappearance of seals in the 1970s.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 15
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The 1988 massive seal mortality in the North and
Wadden Seas, although caused by a viral disease, is still
thought by many to have a link with marine pollution.
Although the North Sea needs radical and lengthy
treatment, it is probably not a terminal case. Actions are
being taken by bordering countries to reduce the dis-
charge of wastes into the sea. A major factor leading to
agreements to reduce discharges of wastewaters has been
the verification of predictive pollutant circulation models
of the sea that identify the impacts of discharges from
various sites along the sea boundary.
2.9. Egypt and the Nile: Limits to Agricultural
Growth
Egypt, located in a belt of extreme aridity, is nearly
completely dependent on the River Nile (Figure 1.15)
for its water resources. Therefore, it is no wonder that
most of Egypts population lives close to the Nile. In rela-
tion to arable land and water, Egypts population density
is among the highest in the world: of its population of
63 million in 2000, 97% lives on 5% of land in the small
strip along the Nile and in the Delta where water is abun-
dant. The population density continues to increase as a
result of a population growth of about 2% per year.
To relieve the population pressure in the Nile Delta and
Nile Valley, the government has embarked on an ambitious
programme to increase the inhabited area in Egypt from
the present 5% to about 25% in the future. The agricultural
area is to be enlarged by horizontal expansion, which
should increase the agricultural area from 3.4 million ha
in 1997 to 4.1 million in 2017. New industrial areas are
planned in the desert, to be supplied by Nile water. Most
of these new agricultural and industrial developments are
based on publicprivate partnerships, requiring the gov-
ernment to give guarantees for the availability of water. The
Toskha project in the south and the El-Salaam scheme in
the Sinai are examples of this kind of development.
However, the availability of Nile water remains the
same. Under the present agreement with Sudan, Egypt is
allowed to use 55.5 billion m
3
of Nile water each year.
That water is nearly completely used already and a further
increase in demand will result in a lower availability of
water per hectare. Additional measures can and will be
taken to increase the efficiency of water use in Egypt,
but that will not be sufficient. It is no wonder that Egypt
is looking into possibilities to increase the supply by tak-
ing measures upstream in Sudan and Ethiopia. Examples
are the construction of reservoirs on the Blue Nile in
Ethiopia and the Jonglei Canal in Sudan that will partly
drain the swamps in the Sudd and decrease the evapora-
tion from them. Cooperation with the other (nine) coun-
tries in the Nile basin is essential to enable those
developments (see Figure 1.15). Hence, Egypt is a strong
supporter of the work of the Nile Basin Initiative that pro-
vides a framework for this cooperation. Other countries
in the basin are challenging the claim of Egypt for addi-
tional water. If Egypt is unable to increase its supply, it
will be forced to lower its ambitions on horizontal expan-
sion of agriculture in the desert and to provide other
means of livelihood for its growing population.
2.10. Damming the Mekong
The Mekong River (Figures 1.16 and 1.17) flows some
4,200 km through Southeast Asia to the South China
Sea through Tibet, Myanmar (Burma), Vietnam, Laos,
Thailand and Cambodia. Its development has been
restricted over the past several decades because of
regional conflicts, indeed those that have altered the
history of the world. Now that these conflicts are reduced,
investment capital is becoming available to develop the
Mekongs resources for improved fishing, irrigation, flood
control, hydroelectric power, tourism, recreation and
navigation. The potential benefits are substantial, but so
are the environmental and ecological risks.
During some months of the year the lack of rainfall
causes the Mekong to fall dramatically. Salt water may pen-
etrate as much as 500 km inland. In other months the flow
can be up to thirty times the low flows, causing the water
in the river to back up into wetlands and flood some
12,000 km
2
of forests and paddy fields in the Vietnamese
delta region alone. The ecology of a major lake, Tonle Sap
in Cambodia, depends on these backed-up waters.
While flooding imposes risks on some 50 million inhab-
itants of the Mekong floodplain, there are also distinct
advantages. High waters deposit nutrient-rich silts on the
low-lying farmlands, thus sparing the farmers from having
to transport and spread fertilizers on their fields. Also,
shallow lakes and submerged lands provide spawning
habitats for about 90% of the fish in the Mekong Basin.
Fish yield totals over half a million tons annually.
16 Water Resources Systems Planning and Management
wrm_ch01.qxd 8/31/2005 11:48 AM Page 16
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 17
Figure 1.15. The Nile Basin.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 17
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
What will happen to the social fabric and to the natural
environment if the schemes to build big dams across the
mainstream of the Mekong are implemented? Depending
on their operation, they could disrupt the current fertility
cycles, habitats and habits of the fish in the river. Increased
erosion downstream from major reservoirs is also a threat.
Add to these the possible adverse impacts the need to evac-
uate and resettle thousands of people displaced by the lake
behind the dams. How will they be resettled? And how
long will it take them to adjust to new farming conditions?
There have been suggestions that a proposed dam in
Laos could cause deforestation in a wilderness area of
18 Water Resources Systems Planning and Management
some 3,000 km
2
. Much of the wildlife, including ele-
phants, big cats and other rare animals, would have to be
protected if they are not to become endangered. Malaria-
carrying mosquitoes, liver fluke and other disease bearers
might find ideal breeding grounds in the mud flats of the
shallow reservoir. These are the types of issues that need
to be considered now that increased development seems
possible, and even likely.
Consider, for example, the impacts of a dam con-
structed on the Nam Pong River in northeast Thailand.
The Nam Pong project was to provide hydroelectric power
and irrigation water, the avowed purposes of many reser-
voir projects throughout the world. Considerable attention
was paid to the social aspects of this project, but not to
the environmental impacts. The project had a number of
unexpected consequences, both beneficial and adverse.
Because the reservoir was acting as a bioreactor for
most of the year, the fish population became so large that
a major fishery industry has developed around the reser-
voir. The economic benefits of fish production exceeded
those derived from hydropower. However, lack of ade-
quate planning for this development resulted in less than
ideal living and economic conditions for the migrating
fishermen who came to this region.
Despite the availability of irrigation water, most farm-
ers were still practising single-crop agriculture after the
dam was built, and still growing traditional crops in their
traditional ways. No training was provided for them to
adapt their skills to the new conditions and opportunities.
In addition, while farming income did not decrease, the
general welfare and health of the population seems to
have deteriorated. Again, little attention was given to diet
and hygiene under these new conditions.
The reservoir itself had some adverse impacts along with
the beneficial ones. These included increased erosion of the
stream banks, silting up of the channel and a large increase
in aquatic vegetation that clogged hydraulic machinery and
reduced transport capacity.
3. So, Why Plan, Why Manage?
Water resources planning and management activities are
usually motivated, as they were in each of the previous
sections case examples, by the realization that there
are both problems to solve and opportunities to obtain
increased benefits from the use of water and related land
Figure 1.16. The Mekong River is one of the few rivers that is
still in equilibrium with surrounding life.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 18
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 19
resources. These benefits can be measured in many dif-
ferent ways. Inevitably, it is not easy to agree on the best
way to do so, and whatever is proposed may provoke
conflict. Hence there is the need for careful study and
research, as well as full stakeholder involvement, in the
search for a shared vision of the best compromised plan
or management policy.
Reducing the frequency and/or severity of the adverse
consequences of droughts, floods and excessive pollution
are common goals of many planning and management
exercises. Other goals include the identification and eval-
uation of alternative measures that may increase the avail-
able water supplies or hydropower, improve recreation
and/or navigation, and enhance the quality of water and
Vietnam
China
Laos
Myanmar
Cambodia
Thailand
0 50 100 150 200 250 km
E
0
1
1
2
1
7
k
Mekong
Chi
Mun Mekong
Se Kong
Se San
Lake
Tonle Sap
Hanoi
Vientiane
Bangkok
Phnom Peng
Ho Chi Minh City
Figure 1.17. The Lower Mekong
River Basin.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 19
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
aquatic ecosystems. Quantitative system performance cri-
teria can help one judge the relative net benefits, however
measured, of alternative plans and management policies.
System performance criteria of interest have evolved
over time. They have developed from being primarily
focused on safe drinking water just a century ago, to
multipurpose economic development a half-century ago, to
goals that now include environmental and ecosystem
restoration and protection, aesthetic and recreational expe-
riences, and more recently, sustainability (ASCE, 1998).
Some of the multiple purposes served by a river can be
conflicting. A reservoir used solely for hydropower or
water supply is better able to meet its objectives when it
is full of water, rather than when it is empty. On the other
hand, a reservoir used solely for downstream flood con-
trol is best left empty, until the flood comes of course.
A single reservoir serving all three purposes introduces
conflicts over how much water to store in it and how it
should be operated. In basins where diversion demands
exceed the available supplies, conflicts will exist over
water allocations. Finding the best way to manage, if not
resolve, these conflicts that occur over time and space are
other reasons for planning.
3.1. Too Little Water
Issues involving inadequate supplies to meet demands
can result from conflicts or concerns over land and water
use. They can result from growing urbanization, the
development of additional water supplies, the need to
meet instream flow requirements, and conflicts over pri-
vate property and public rights regarding water alloca-
tions. Other issues can involve trans-basin water transfers
and markets, objectives of economic efficiency versus the
desire to keep non-efficient activities viable, and demand
management measures, including incentives for water
reuse and water reuse financing.
Measures to reduce the demand for water in times of
supply scarcity should be identified and agreed upon before
everyone has to cope with an actual water scarcity. The
institutional authority to implement drought measures
when their designated triggers such as decreasing storage
volumes in reservoirs have been met should be established
before the measures are needed. Such management
responses may include increased groundwater abstractions
to supplement low surface-water flows and storage volumes.
20 Water Resources Systems Planning and Management
Conjunctive use of ground and surface waters can be sus-
tainable as long as the groundwater aquifers are recharged
during conditions of high flow and storage volumes.
3.2. Too Much Water
Damage due to flooding is a direct result of floodplain
development that is vulnerable to floods. This is a risk
many take, and indeed on average it may result in posi-
tive private net benefits, especially when public agencies
subsidize these private risk takers in times of flooding.
In many river basins of developed regions, the level of
annual expected flood damage is increasing over time, in
spite of increased expenditures on flood damage reduc-
tion measures. This is mainly due to increased economic
development on river floodplains, not to increased fre-
quencies or magnitudes of floods.
The increased economic value of the development on
floodplains often justifies increased expenditure on flood
damage reduction measures. Flood protection works
decrease the risks of flooding and consequent damage,
creating an incentive for increased economic develop-
ment. Then when a flood exceeding the capacity of exist-
ing flood protection works occurs, and it will, even more
damage results. This cycle of increasing flood damage and
cost of protection is a natural result of the increasing
values of floodplain development.
Just what is the appropriate level of risk? It may
depend, as Figure 1.18 illustrates, on the level of flood
insurance or subsidy provided when flooding occurs.
Flood damage will decrease only if restrictions are
placed on floodplain development. Analyses carried out
during planning can help identify the appropriate level of
e
x
p
e
c
t
e
d

r
e
t
u
r
n
flood risk
unacceptable risk level
without / with
insurance or subsidy
E
0
1
1
2
1
7
m
Figure 1.18. The lowest risk of flooding on a floodplain does
not always mean the best risk, and what risk is acceptable
may depend on the amount of insurance or subsidy provided
when flood damage occurs.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 20
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 21
development and flood damage protection works, on the
basis of both the beneficial and the adverse economic,
environmental and ecological consequences of floodplain
development. People are increasingly recognizing the eco-
nomic as well as environmental and ecological benefits of
allowing floodplains to do what they were formed to do:
store flood waters when floods occur.
3.3. Polluted Water
The discharges of wastewater by industry and households
can have considerable detrimental effects on water quality
and, hence, often on public and ecosystem health.
Planning and management activities should pay attention
to these possible negative consequences of industrial
development, population growth and the intensive use of
pesticides and fertilizers in urban as well as in agricul-
tural areas. Issues regarding the environment and water
quality include:
upstream versus downstream conflicts on meeting
water quality standards
threats from aquatic nuisance species
threats from the chemical, physical and biological
water quality of the watersheds aquatic resources
quality standards for recycled water
non-point source pollution discharges, including sed-
iment from erosion
inadequate groundwater protection compacts and
concerned institutions.
We still know too little about the environmental and
health impacts of many of the wastewater constituents
found in river waters. As more is learned about, for exam-
ple, the harmful effects of heavy metals and dioxins, our
plans and management policies should be adjusted
accordingly. Major fish losses and algae blooms point to
the need to manage water quality as well as quantity.
3.4. Degradation of Aquatic and Riparian
Ecosystems
Aquatic and riparian ecosystems may be subject to a
number of threats. The most important include habitat
loss due to river training and reclamation of floodplains
and wetlands for urban and industrial development, poor
water quality due to discharges of pesticides, fertilizers
and wastewater effluents, and the infestation of aquatic
nuisance species.
Exotic aquatic nuisance species can be major threats to
the chemical, physical and biological water quality of a
rivers aquatic resources, and a major interference with
other uses. The destruction and/or loss of the biological
integrity of aquatic habitats caused by introduced exotic
species is considered by many ecologists to be among
the most important problems facing natural aquatic and
terrestrial ecosystems. The biological integrity of natural
ecosystems is controlled by habitat quality, water flows
or discharges, water quality and biological interactions
including those involving exotic species.
Once exotic species are established, they are usually
difficult to manage and nearly impossible to eliminate.
This creates a costly burden for current and future gener-
ations. The invasion in North America of non-indigenous
aquatic nuisance species such as the sea lamprey, zebra
mussel, purple loosestrife, European green crab and
various aquatic plant species, for example, has had
pronounced economic and ecological consequences for
all who use or otherwise benefit from aquatic ecosystems.
Environmental and ecological effectiveness as well as
economic efficiency should be a guiding principle in eval-
uating alternative solutions to problems caused by aquatic
nuisance organisms. Funds spent on proper prevention
and early detection and eradication of aquatic nuisance
species may reduce the need to spend considerably
greater funds on management and control once such
species are well established.
3.5. Other Planning and Management Issues
Navigation
Industrial and related port development may result in the
demand for deeper rivers to allow the operation of larger-
draught cargo vessels in the river. River channel improve-
ment cannot be detached from functions such as water
supply and flood control. Narrowing the river for ship-
ping purposes may increase floodwater levels.
River Bank Erosion
Bank erosion can be a serious problem where people are
living close to morphologically active (eroding) rivers.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 21
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Bangladesh, where bank erosion is considered to be a much
more urgent problem than the well-known floods of that
country, is an example of this. Predictions of changes in
river courses due to bank erosion and bank accretion are
important inputs to land use planning in river valleys and
the choice of locations for bridges and hydraulic structures.
Reservoir Related Issues
Degradation of the riverbed upstream of reservoirs may
increase the risks of flooding in those areas. Reservoir
construction inevitably results in loss of land and forces
the evacuation of residents due to impoundment. Dams
can be ecological barriers for migrating fish species such
as salmon. The water and sediment quality in the reser-
voir may deteriorate and the in-flowing sediment may
accumulate, reducing the active (useful) capacity of the
reservoir. Other potential problems may include those
stemming from stratification, water related diseases, algae
growth and abrasion of hydropower turbines.
Environmental and morphological impacts down-
stream of the dam are often due to a changed river hydro-
graph and decreased sediment load in the water released
from the reservoir. Lower sediment loads result in higher
scouring of downstream riverbeds and consequently a
lowering of their elevations. Economic as well as social
impacts include the risk of dams breaking. Environmental
impacts may result from sedimentation control measures
(e.g., sediment flushing) and reduced oxygen content of
the out-flowing water.
The ecological, environmental and economic impacts
of dams and reservoirs are heavily debated among plan-
ners and environmentalists. In creating a new framework
for decision-making, the World Commission on Dams
compiled and considered the arguments of all sides of this
debate (WCD, 2000).
4. System Components, Planning
Scales and Sustainability
Water resources management involves influencing and
improving the interaction of three interdependent
subsystems:
the natural river subsystem in which the physical,
chemical and biological processes take place
22 Water Resources Systems Planning and Management
the socio-economic subsystem, which includes the
human activities related to the use of the natural river
system
the administrative and institutional subsystem of
administration, legislation and regulation, where the
decision and planning and management processes
take place.
Figure 1.19 illustrates the interaction between these sub-
systems, all three of which should be included in any
analysis performed for water resource systems planning
and management. Inadequate attention to one can destroy
the value of any work done to improve the performance
of the others.
Appendix A describes the major components of the
natural system and their processes and interactions.
4.1. Spatial Scales for Planning and
Management
Watersheds or river basins are usually considered logical
regions for water resources planning and management.
This makes sense if the impacts of decisions regarding
water resources management are contained within the
Figure 1.19. Interactions among subsystems and between
them and their environment.
l
e
g
is
la
tion / organ
iz
a
t
io
n
NRS SES
impacts
resources use
AIS
legislation,
regulation
infrastructure
s
o
c
i
o
-
e
c
o
n
o
m
i
c
d
e
v
e
l
o
p
m
e
n
t
p
l
a
n
s
n
a
t
u
r
a
l
b
o
u
n
d
a
r
y
c
o
n
d
i
t
i
o
n
s
E
0
4
0
2
2
3
a
wrm_ch01.qxd 8/31/2005 11:48 AM Page 22
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 23
watershed or basin. How land and water are managed in
one part of a river basin can affect the land and water in
other parts of the basin. For example, the discharge of
pollutants or the clearing of forests in the upstream por-
tion of the basin may degrade the quality and increase the
variability of the flows and sedimentation downstream.
The construction of a dam or weir in the downstream part
of a river may prevent vessels and fish from travelling
upstream. To maximize the economic and social benefits
obtained from the entire basin, and to ensure that these
benefits and accompanying costs are equitably distrib-
uted, planning and management is often undertaken on a
basin scale.
While basin boundaries make sense from a hydrologi-
cal point of view, they may be inadequate for addressing
particular water resources problems that are caused by
events taking place outside the basin. What is desired is
the highest level of performance, however defined, of the
entire physical, socio-economic and administrative water
resource system. To the extent that the applicable prob-
lems, stakeholders and administrative boundaries extend
outside the river basin, the physically based river basin
focus of planning and management should be expanded
to include the entire applicable problem-shed. Hence,
consider the term river basin used in this book to mean
problem-shed when appropriate.
4.2. Temporal Scales for Planning and
Management
Water resources planning requires looking into the
future. Decisions recommended for the immediate future
should take account of their long-term future impacts.
These impacts may also depend on economic, demo-
graphic and physical conditions now and on into some
distant future. The question of just how far into the future
one need look, and try to forecast, is directly dependent
on the influence that future forecast has on the present
decisions. What is most important now is what decision
to make now. Decisions that are to be made later can be
based on updated forecasts, then-current information and
planning and management objectives. Planning is a con-
tinuing sequential process. Water resources plans need to
be periodically updated and adapted to new information,
new objectives, and updated forecasts of future supplies,
demands, costs and benefits.
The number and duration of within-year time peri-
ods explicitly considered in the planning process will be
dependent in part on the need to consider the variabil-
ity of the supplies and demands for water resources and
on the purposes to be served by the water resources
within the basin. Irrigation planning and summer-
season water recreation planning may require a greater
number of within-year periods during the summer
growing and recreation season than might be the case if
one were considering only municipal water supply plan-
ning, for example. Assessing the impacts of alternatives
for conjunctive surface and groundwater management,
or for water quantity and quality management, require
attention to processes that take place on different spatial
and temporal scales.
4.3. Sustainability
Sustainable water resources systems are those designed
and managed to best serve people living today and in the
future. The actions that we as a society take now to satisfy
our own needs and desires should depend not only on
what those actions will do for us but also on how they will
affect our descendants. This consideration of the long-
term impacts on future generations of actions taken now is
the essence of sustainable development. While the word
sustainability can mean different things to different peo-
ple, it always includes a consideration of the welfare of
those living in the future. While the debate over a more
precise definition of sustainability will continue, and ques-
tions over just what it is that should be sustained may
remain unanswered, this should not delay progress toward
achieving more sustainable water resources systems.
The concept of environmental and ecological sustain-
ability has largely resulted from a growing concern about
the long-run health of our planet. There is increasing evi-
dence that our present resource use and management
activities and actions, even at local levels, can significantly
affect the welfare of those living within much larger
regions in the future. Water resource management prob-
lems at a river basin level are rarely purely technical and
of interest only to those living within the individual river
basins where those problems exist. They are increasingly
related to broader societal structures, demands and goals.
What would future generations like us to do for them?
We dont know, but we can guess. As uncertain as these
wrm_ch01.qxd 8/31/2005 11:48 AM Page 23
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
guesses will be, we should take them into account as we
act to satisfy our own immediate needs, demands and
desires. There may be tradeoffs between what we wish to
do for ourselves in our current generation versus what we
think future generations might wish us to do for them.
These tradeoffs, if any, between what present and future
generations would like should be considered and debated
in the political arena. There is no scientific theory to help
us identify which tradeoffs, if any, are optimal.
The inclusion of sustainability criteria along with the
more common economic, environmental, ecological and
social criteria used to evaluate alternative water resources
development and management strategies may identify a
need to change how we commonly develop and use our
water resources. We need to consider the impacts of change
itself. Change over time is certain just what it will be is
uncertain. These changes will affect the physical, biologi-
cal and social dimensions of water resource systems. An
essential aspect in the planning, design and management
of sustainable systems is the anticipation of change. This
includes change due to geomorphologic processes, the
aging of infrastructure, shifts in demands or desires of a
changing society, and even increased variability of water
supplies, possibly because of a changing climate. Change
is an essential feature of sustainable water resources
development and management.
Sustainable water resources systems are those designed
and operated in ways that make them more adaptive,
robust and resilient to an uncertain and changing future.
They must be capable of functioning effectively under con-
ditions of changing supplies, management objectives and
demands. Sustainable systems, like any others, may fail,
but when they fail they must be capable of recovering and
operating properly without undue costs.
In the face of certain changes, but with uncertain
impacts, an evolving and adaptive strategy for water
resources development, management and use is a neces-
sary condition of sustainable development. Conversely,
inflexibility in the face of new information, objectives
and social and political environments is an indication of
reduced system sustainability. Adaptive management is a
process of adjusting management actions and directions,
as appropriate, in the light of new information on the
current and likely future condition of our total
environment and on our progress toward meeting our
24 Water Resources Systems Planning and Management
goals and objectives. Water resources development and
management decisions can be viewed as experiments,
subject to modification, but with goals clearly in mind.
Adaptive management recognizes the limitations of cur-
rent knowledge and experience as well as those that we
learn by experimenting. It helps us move toward meeting
our changing goals over time in the face of this incomplete
knowledge and uncertainty. It accepts the fact that there
is a continual need to review and revise management
approaches because of the changing, as well as uncertain,
nature of our socio-economic and natural environments.
Changing the social and institutional components of
water resources systems is often the most challenging
task, because it involves changing the way individuals
think and act. Any process involving change will require
that we change our institutions the rules under which
we as a society function. Individuals are primarily respon-
sible for, and adaptive to, changing political and social
situations. Sustainability requires that public and private
institutions also change over time in ways that are respon-
sive to the needs of individuals and society.
Given the uncertainty of what future generations will
want, and the economic, environmental and ecological
problems they will face, a guiding principle for the
achievement of sustainable water resource systems is to
provide options to future generations. One of the best
ways to do this is to interfere as little as possible with the
proper functioning of natural life cycles within river
basins, estuaries and coastal zones. Throughout the water
resources system planning and management process,
it is important to identify all the beneficial and adverse
ecological, economic, environmental and social effects
especially the long-term effects associated with any
proposed project.
5. Planning and Management
5.1. Approaches
There are two general approaches to planning and man-
agement. One is from the top down, often called com-
mand and control. The other is from the bottom up, often
called a grass-roots approach. Both approaches can lead
to an integrated plan and management policy.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 24
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 25
5.1.1. Top-Down Planning and Management
Over much of the past half century, water resources
professionals have been engaged in preparing integrated,
multipurpose master development plans for many of the
worlds river basins. These plans typically consist of a series
of reports, complete with numerous appendices, describing
all aspects of water resources management and use. In these
documents alternative structural and non-structural man-
agement options are identified and evaluated. On the basis
of these evaluations, the preferred plan is presented.
This master planning exercise has typically been a top-
down approach that professionals have dominated. Using
this approach there is usually little if any active participa-
tion by interested stakeholders. The approach assumes
that one or more institutions have the ability and author-
ity to develop and implement the plan, in other words,
that will oversee and manage the coordinated develop-
ment and operation of the basins activities that affect the
surface and ground waters of the basin. In todays envi-
ronment, where publics are calling for less governmental
oversight, regulation and control, and increasing partici-
pation in planning and management activities, top-down
approaches are becoming less desirable or acceptable.
5.1.2. Bottom-Up Planning and Management
Within the past decade water resources planning and
management processes have increasingly involved the
active participation of interested stakeholders those
affected in any way by the management of the water and
land resources. Plans are being created from the bottom
up rather than top down. Concerned citizens and non-
governmental organizations, as well as professionals in
governmental agencies, are increasingly working together
towards the creation of adaptive comprehensive water
management programs, policies and plans.
Experiences of trying to implement plans developed pri-
marily by professionals without significant citizen involve-
ment have shown that, even if such plans are technically
flawless, they have little chance of success if they do not take
into consideration the concerns of affected local stakehold-
ers and do not have their support. To gain this, concerned
stakeholders must be included in the decision-making
process as early as possible. They must become part of that
process, not merely as spectators or advisors to it. This will
help gain their cooperation and commitment to the plans
adopted. Participating stakeholders will have a sense of
ownership, and as such will strive to make the plans work.
Such plans, if they are to be successfully implemented,
must also fit within existing legislative, permitting, enforce-
ment and monitoring programmes. Stakeholder participa-
tion improves the chance that the system being managed
will be sustainable.
Successful planning and management involves moti-
vating all potential stakeholders and sponsors to join in
the water resources planning and management process,
determining their respective roles and establishing how
to achieve consensus on goals and objectives. Ideally this
should occur before addressing conflicting issues so that
all involved know each other and are able to work
together more effectively. Agreements on goals and objec-
tives and on the organization (or group formed from
multiple organizations) that will lead and coordinate the
water resources planning and management process
should be reached before stakeholders bring their
individual priorities or problems to the table. Once
the inevitable conflicts become identified, the settling of
administrative matters doesnt get any easier.
Bottom-up planning must strive to achieve a common
or shared vision of goals and priorities among all stake-
holders. It must be aware of and comply with all applica-
ble laws and regulations. It should strive to identify and
evaluate multiple alternatives and performance criteria
including sustainability criteria and yet keep the process
from producing a wish-list of everything each stakeholder
wants. In other words, it must identify tradeoffs among
conflicting goals or measures of performance, and priori-
tize appropriate strategies. It must value and compare,
somehow, the intangible and non-monetary impacts of
environmental and ecosystem protection and restoration
with other activities whose benefits and costs can be
expressed in monetary units. In doing so, planners should
use modern information technology to improve both the
process and product. This technology, however, will not
eliminate the need to reach conclusions and make deci-
sions on the basis of incomplete and uncertain data and
scientific knowledge.
These process issues focus on the need to make water
resources planning and management as efficient and
wrm_ch01.qxd 8/31/2005 11:48 AM Page 25
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
effective as possible. Many issues will arise in terms of
evaluating alternatives and establishing performance cri-
teria (prioritizing issues and possible actions), performing
incremental cost analysis, and valuing monetary and non-
monetary benefits. Questions must be answered as to how
much data must be collected and with what precision,
and what types of modern information technology (e.g.,
geographic information systems (GIS), remote sensing,
Internet, decision support systems, etc.) can be benefi-
cially used for both analyses and communication.
5.1.3. Integrated Water Resources Management
The concept of integrated water resources management
(IWRM) has been developing since the beginning of the
eighties. IWRM is the response to the growing pressure
on our water resources systems caused by growing popu-
lation and socio-economic developments. Water short-
ages and deteriorating water quality have forced many
countries in the world to reconsider their options with
respect to the management of their water resources. As
a result water resources management (WRM) has been
undergoing a change worldwide, moving from a mainly
supply-oriented, engineering-biased approach towards a
demand-oriented, multi-sectoral approach, often labelled
integrated water resources management.
In international meetings, opinions are converging to a
consensus about the implications of IWRM. This is best
reflected in the Dublin Principles of 1992 (see Box 1.1),
which have been universally accepted as the base for
IWRM. The concept of IWRM makes us move away from
top-down water master planning (see Section 5.1.1),
which focuses on water availability and development,
towards comprehensive water policy planning which
addresses the interaction between different sub-sectors,
seeks to establish priorities, considers institutional require-
ments and deals with the building of management capacity.
IWRM considers the use of the resources in relation
to social and economic activities and functions. These
also determine the need for laws and regulations for the
sustainable use of the water resources. Infrastructure
made available, in relation to regulatory measures and
mechanisms, will allow for effective use of the resource,
taking due account of the environmental carrying
capacity (Box 1.2).
26 Water Resources Systems Planning and Management
5.2. Planning and Management Aspects
5.2.1. Technical Aspects
Technical aspects of planning include hydrological assess-
ments. These identify and characterize the properties of,
and interactions among, the resources in the basin or
region, including the land, the rainfall, the runoff, the
stream and river flows and the groundwater.
Existing watershed land use and land cover, and future
changes in this use and cover, result in part from existing
and future changes in regional population and economy.
Planning involves predicting changes in land use/covers
and economic activities at watershed and river basin levels.
These will influence the amount of runoff, and the concen-
trations of sediment and other quality constituents (organic
wastes, nutrients, pesticides, etc.) it contains as a result of
any given pattern of rainfall over the land area. These pre-
dictions will help planners estimate the quantities and
qualities of flows and their constituents throughout a
watershed or basin, associated with any land use and water
management policy. This in turn provides the basis for
Box 1.1. The Dublin Principles
1. Water is a finite, vulnerable and essential resource,
essential to sustain life, development and the
environment.
2. Water resources development and management
should be based on a participatory approach,
involving users, planners and policy makers at all
levels.
3. Women play a central role in the provision, man-
agement and safeguarding of water.
4. Water has an economic value in all its competing
uses and should be recognized as an economic
good.
Box 1.2. Definition of IWRM
IWRM is a process which promotes the coordinated
development and management of water, land and
related resources, in order to maximize the resultant
economic and social welfare in an equitable manner
without compromising the sustainability of vital
ecosystems.
(GWP, 2000)
wrm_ch01.qxd 8/31/2005 11:48 AM Page 26
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 27
predicting the type and health of terrestrial and aquatic
ecosystems in the basin. All of this may affect the economic
development of the region, which in part determines the
future demands for changes in land use and land cover.
Technical aspects also include the estimation of the
costs and benefits of any measures taken to manage the
basins water resources. These measures might include:
engineering structures for making better use of scarce
water
canals and water-lifting devices
dams and storage reservoirs that can retain excess
water from periods of high-flow for use during the
periods of low-flow (and may reduce flood damage
below the reservoir by storing floodwater)
open channels that may take the form of a canal,
flume, tunnel or partly filled pipe
pressure conduits
diversion structures, ditches, pipes, checks, flow
dividers and other engineering facilities necessary
for the effective operation of irrigation and drainage
systems
municipal and industrial water intakes, including
water purification plants and transmission facilities
sewerage and industrial wastewater treatment plants,
including waste collection and ultimate disposal
facilities
hydroelectric power storage, run-of-river or pumped
storage plants,
river channel regulation works, bank stabilization,
navigation dams and barrages, navigation locks and
other engineering facilities for improving a river for
navigation
levees and floodwalls for confinement of the flow
within a predetermined channel.
Not only must the planning process identify and evaluate
alternative management strategies involving structural and
non-structural measures that will incur costs and bring
benefits, but it must also identify and evaluate alternative
time schedules for implementing those measures. The
planning of development over time involving interdepen-
dent projects, uncertain future supplies and demands as
well as costs, benefits and interest (discount) rates is part of
all water resources planning and management processes.
With increasing emphasis placed on ecosystem preser-
vation and enhancement, planning must include ecologic
impact assessments. The mix of soil types and depths and
land covers together with the hydrological quantity and
quality flow and storage regimes in rivers, lakes, wetlands
and aquifers affect the riparian and aquatic ecology of the
basin. Water managers are being asked to consider ways
of improving or restoring ecosystems by, for example,
reducing:
the destruction and/or loss of the biological integrity of
aquatic habitats caused by introduced exotic species
the decline in number and extent of wetlands and the
adverse impacts on wetlands of proposed land and
water development projects
the conflicts between the needs of people for water
supply, recreation, energy, flood control, and naviga-
tion infrastructure and the needs of ecological com-
munities, including endangered species.
And indeed there are and will continue to be conflicts
among alternative objectives and purposes of water man-
agement. Planners and managers must identify the trade-
offs among environmental, ecologic, economic and social
impacts, however measured, and the management alterna-
tives that can balance these often-conflicting interests.
5.2.2. Economic and Financial Aspects
The fourth Dublin principle states that water has an eco-
nomic value in all its competing uses and should be
recognized as an economic good. This principle addresses
the need to extract the maximum benefits from a limited
resource as well as the need to generate funds to recover
the costs of the investments and of the operation and
maintenance of the system.
The maximization of benefits is based on a common
economic market approach. Many past failures in water
resources management are attributable to the fact that
water has been and still is viewed as a free good. Prices
of water for irrigation and drinking water are in many
countries well below the full cost of the infrastructure and
personnel needed to provide that water, which comprises
the capital charges involved, the operation and mainte-
nance (O&M) costs, the opportunity cost, economic
externalities and environmental externalities (see GWP,
2000). Charging for water at less than full cost means that
the government, society and/or environment subsidizes
water use and leads to sub-optimal use of the resource.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 27
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Recognizing water as an economic good does not always
mean that full costs should be charged. Poor people have
the right to safe water and this should be taken into
account. For that reason the fourth Dublin principle is
often referred to as water being an economic and social
good.
Cost recovery is the second reason for the fourth
Dublin principle. The overriding financial component
of any planning process is to make sure that the
recommended plans and projects are able to pay for them-
selves. Revenues are needed to recover construction costs,
if any, and to maintain, repair and operate any infrastruc-
ture designed to manage the basins water resources. This
may require cost-recovery policies that involve pricing the
outputs of projects. Beneficiaries should be expected to
pay at least something, and in some way, for the added
benefits they get. Planning must identify equitable cost
and risk-sharing policies and improved approaches to
risk/cost management. In many developing countries a
distinction is made between cost recovery of investments
and cost recovery of O&M costs. Cost recovery of O&M
costs is a minimum condition for a sustainable project.
Without that, it is likely that the performance of the
project will deteriorate seriously over time.
In most WRM studies, financial viability is viewed as a
constraint that must be satisfied. It is not viewed as an
objective whose maximization could result in a reduction
in economic efficiency, equity or other non-monetary
objectives.
5.2.3. Institutional Aspects
The first condition for successful project implementation
is to have an enabling environment. There must exist
national, provincial and local policies, legislation and
institutions that make it possible for the right decisions to
be taken and implemented correctly. The role of the gov-
ernment is crucial. The reasons for governmental involve-
ment are manifold:
Water is a resource beyond property rights: it cannot
be owned by private persons. Water rights can be
given to persons or companies, but only the rights to
use the water and not to own it. Conflicts between
users automatically turn up at the table of the final
owner of the resource the government.
28 Water Resources Systems Planning and Management
Water is a resource that often requires large invest-
ment to develop. Many water resources development
projects are very expensive and have many beneficiar-
ies. Examples are multipurpose reservoirs and the
construction of dykes along coasts and rivers. The
required investments need large financial commit-
ments which only can be made by the government or
state-owned companies.
Water is a medium that can easily transfer external
effects. The use of water by one person often has
negative effects on others (externalities). The obvious
example is the discharge of waste into a river that may
have negative effects on downstream users.
Only the government can address these issues and good
governance is necessary for good water management.
An insufficient institutional setting and the lack of a
sound economic base are the main causes of water resources
development project failure, not technical inadequacy of
design and construction. This is also the reason why at
present much attention is given to institutional develop-
ments in the sector, in both developed and developing
countries. In Europe, various types of water agencies are
operational (e.g., the Agence de lEau in France and the
water companies in England), each having advantages
and disadvantages. The Water Framework Directive of the
European Union requires that water management be
carried out at the scale of a river basin, particularly when
this involves transboundary management. It is very likely
that this will result in a shift in responsibilities of the
institutions involved and the establishment of new
institutions. In other parts of the world experiments are
being carried out with various types of river basin organiza-
tions, combining local, regional and sometimes national
governments.
5.3. Analyses for Planning and Management
Analyses for water resources planning and management
generally comprise several stages. The explicit description
of these stages is referred to as the analytical (or concep-
tual) framework. Within this framework, a set of coherent
models for the quantitative analysis of measures and
strategies is used. This set of models and related databases
will be referred to as the computational framework. This
book is mainly about the computational framework.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 28
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 29
situations change (political, international, societal
developments).
As an example, the analytical framework that is used
by Delft Hydraulics for WRM studies is depicted in
Figure 1.20. The three elementary phases of that frame-
work are:
inception
development
selection.
During each phase the processes have a cyclic component
(comprehensive cycle). Interaction with the decision-
makers, or their representatives, is essential throughout
the process. Regular reporting through inception and interim
reports will improve the effectiveness of the communication.
The first phase of the process is the inception phase.
Here the subject of the analysis (what is analysed under
initial analysis
approach
analysis conditions
workplan
inception
inception report interim reports
t
r
i
g
g
e
r
s
decision-makers / stakeholder representatives
detailed analysis
selection
characteristics of WRS
activities & developments
policies & institutions
problems & measures
objectives and criteria
data availability
design of alternative
strategies
preliminary analysis
base case analysis
bottleneck analysis
screening of measures
final report
data collection and
modelling
development
NRS
SES
AIS
E
0
4
0
2
2
3
b
natural resource system
socio-economic system
administrative-
institutional system
impact assessment
evaluation of alternatives
sensitivity and scenario
analysis
analysis steps
delineation system
presentation of results
computational
framework
Figure 1.20. Typical analytical
framework for water resources
studies.
The purpose of the analyses is to prepare and support
planning and management decisions. The main phases
of the analytical framework therefore correspond to the
phases of the decision process. Such a decision process is
not a simple, one-line sequence of steps. Inherent in a
decision-making process are factors causing the decision-
makers to return to earlier steps of the process. Part of the
process is thus cyclic. A distinction is made between
comprehension cycles and feedback cycles. A comprehen-
sion cycle improves the decision-makers understanding
of a complex problem by cycling within or between steps.
Feedback cycles imply returning to earlier phases of the
process. They are needed when:
solutions fail to meet criteria.
new insights change the perception of the problem and
its solutions (e.g., due to more/better information).
essential system links have been overlooked.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 29
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
what conditions) and its object (the desired results of
the analysis) are specified. Based on this initial analysis,
during which intensive communication with (representa-
tives of the) decision-makers is essential, the approach for
the policy analysis is specified. The results of the incep-
tion phase are presented in the inception report, which
includes the work plan for the other phases of the analy-
sis process (project).
In the development phase tools are developed for
analysing and identifying possible solutions to the WRM
problems. The main block of activities is usually related to
data collection and modelling. Various preliminary analy-
ses will be made to ensure that the tools developed for the
purpose are appropriate for solving the WRM problems.
Individual measures will be developed and screened in this
phase, and preliminary attempts will be made to combine
promising measures into management strategies. The devel-
opment phase is characterized by an increased under-
standing of the functioning of the water resources system,
starting with limited data sets and simplified tools and end-
ing at the levels of detail deemed necessary in the inception
phase. Scanning of possible measures should also start as
soon as possible during this phase. The desired level of
detail in the data collection and modelling strongly
depends on what is required to distinguish among the vari-
ous measures being considered. Interactions with decision-
makers are facilitated through the presentation of interim
results in interim reports.
The purpose of the selection phase is to prepare a
limited number of promising strategies based on a
detailed analysis of their effects on the evaluation crite-
ria, and to present them to the decision-makers, who
will make the final selection. Important activities in this
phase are strategy design, evaluation of strategies and
presentation. The results of this phase are included in
the final report
Although it is clear that analyses are made to support
the decision-making process, it is not always clear who
will make the final decision, or who is the decision-
maker. If analyses are contracted to a consultant, careful
selection of the appropriate coordinating agency is instru-
mental to the successful implementation of the project. It
is always advantageous to use existing line agencies as
much as possible. Interactions with the decision-makers
usually take place through steering commissions (with
an interdepartmental forum) and technical advisory
30 Water Resources Systems Planning and Management
committees. Appendix E of this book describes this ana-
lytical framework in more detail.
5.4. Models for Impact Prediction and
Evaluation
The process of planning has undergone a significant
transformation over the past several decades, mainly due
to the continuing development of improved computa-
tional technology, and various water resource simulation
and optimization models together with their associated
databases and user-friendly interactive interfaces.
Planning today is heavily dependent on the use of
computer-based impact prediction models. Such models
are used to assist in the identification and evaluation of
alternative ways of meeting various planning and man-
agement objectives. They provide an efficient way of
analysing spatial and temporal data in an effort to predict
the interaction and impacts, over space and time, of
various river basin components under alternative designs
and operating policies.
Many of the systems analysis approaches and models
discussed in the accompanying chapters of this book have
been, and continue to be, central to the planning and
management process. Their usefulness is directly depend-
ent on the quality of the data and models being used.
Models can assist planning and management at different
levels of detail. Some are used for preliminary screening
of alternative plans and policies, and as such do not
require major data collection efforts. Screening models
can also be used to estimate how significant certain data
and assumptions are for the decisions being considered,
and hence can help guide additional data collection activ-
ities. At the other end of the planning and management
spectrum, much more detailed models can be used for
engineering design. These more complex models are
more data demanding, and typically require higher levels
of expertise for their proper use.
The integration of modelling technology into the social
and political components of the planning and manage-
ment processes in a way that enhances those processes
continues to be the main challenge of those who develop
planning and management models. Efforts to build and
apply interactive generic modelling programs or shells,
on which interested stakeholders can draw in their sys-
tem, enter their data and operating rules at the level of
wrm_ch01.qxd 8/31/2005 11:48 AM Page 30
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 31
detail desired, run simulations, and discover the effect of
alternative assumptions and operating rules, has in many
cases helped to create a common or shared understanding
among these stakeholders. Getting stakeholders involved
in developing and experimenting with their own interac-
tive data-driven models has been an effective way of
building a consensusa shared vision.
5.5. Shared-Vision Modelling
Participatory planning inevitably involves conflict man-
agement. Each stakeholder or interest group has its
objectives, interests and agendas. Some of these may be
in conflict with others. The planning and management
process is one of negotiation and compromise. This takes
time but, from it can come decisions that have the best
chance of being considered right and fair or equitable by
most participants. Models can assist in this process of
reaching a common understanding and agreement among
different stakeholders. This has a greater chance of hap-
pening if the stakeholders themselves are involved in the
modelling and analysis process.
Involving stakeholders in model-building accom-
plishes a number of things. It gives them a feeling of own-
ership. They will have a much better understanding of
just what their model can do and what it cannot. If they
are involved in model-building, they will know the
assumptions built into their model.
Being involved in a joint modelling exercise is a way to
understand better the impacts of various assumptions.
While there may be no agreement on the best of various
assumptions to make, stakeholders can learn which of
those assumptions matter and which do not. In addition,
the process of model development by numerous stake-
holders will itself create discussions that will lead toward a
better understanding of everyones interests and concerns.
Though such a model-building exercise, it is possible
those involved will reach not only a better understanding
of everyones concerns, but also a common or shared
vision of at least how their system (as represented by their
model, of course) works.
5.6. Adaptive Integrated Policies
One of the first issues to address when considering water
resources planning and management activities is the
product desired. If it is to be a report, what should the
report contain? If it is to be a model or a decision support
system, what should be its capabilities?
Clearly a portion of any report should contain a discus-
sion of the water resources management issues and options.
Another part of the report might include a prioritized list of
strategies for addressing existing problems and available
development or management opportunities in the basin.
Recent emphasis has shifted from structural engineering
solutions to more non-structural alternatives, especially for
environmental and ecosystem restoration. Part of this shift
reflects the desire to keep more options open for future
generations. It reflects the desire to be adaptive to new
information and to respond to surprises impacts not fore-
casted. As we learn more about how river basins, estuaries
and coastal zones work, and how humans can better man-
age those resources, we do not want to regret what we have
done in the past that may preclude this adaptation.
In some situations it may be desirable to create a
rolling plan one that can be updated at any time. This
permits responses to resource management and regula-
tory questions when they are asked, not just at times
when new planning and management exercises take
place. While this appears to be desirable, will planning
and management organizations have the financing and
support to maintain and update the modelling software
used to estimate various impacts, collect and analyse new
data, and maintain the expertise, all of which are neces-
sary for continuous planning (rolling plans)?
Consideration also needs to be given to improving the
quality of the water resources planning and management
review process, and focusing on outcomes themselves
rather than output measures. One of the outcomes should
be an increased understanding of some of the relation-
ships between various human activities and the hydrology
and ecology of the basin, estuary or coastal zone. Models
developed for predicting the economic as well as ecologic
interactions and impacts due to changes in land and water
management and use could be used to address questions
such as:
What are the hydrological, ecological and economic
consequences of clustering or dispersing human land
uses such as urban and commercial developments and
large residential areas? Similarly, what are the conse-
quences of concentrated versus dispersed patterns of
reserve lands, stream buffers and forestland?
wrm_ch01.qxd 8/31/2005 11:48 AM Page 31
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
What are the costs and ecological benefits of a conser-
vation strategy based on near-stream measures (e.g.,
riparian buffers) versus near-source (e.g., upland/site-
edge) measures? What is the relative cost of forgone
upland development versus forgone valley or riparian
development? Do costs strongly limit the use of stream
buffer zones for mitigating agriculture, residential and
urban developments?
Should large intensive developments be best located in
upland or valley areas? Does the answer differ depend-
ing on economic, environmental or aquatic ecosystem
perspectives? From the same perspectives, is the most
efficient and desirable landscape highly fragmented or
highly zoned with centres of economic activity?
To what extent can riparian conservation and
enhancement mitigate upland human land use
effects? How do the costs of upland controls compare
with the costs of riparian mitigation measures?
What are the economic and environmental quality
tradeoffs associated with different areas of various
classes of land use such as commercial/urban, residen-
tial, agriculture and forest?
Can adverse effects on hydrology, aquatic ecology and
water quality of urban areas be better mitigated
through upstream or downstream management
approaches? Can land controls like stream buffers be
used at reasonable cost within urban areas, and if so,
how effective are they?
Is there a threshold size for residential/commercial
areas that yield marked ecological effects?
What are the ecological states at the landscape scale
that, once attained, become irreversible with reason-
able mitigation measures? For example, once stream
segments in an urban setting become highly altered by
direct and indirect effects (e.g., channel bank protec-
tion and straightening and urban runoff ), can they be
restored with feasible changes in urban land use or
mitigation measures?
Mitigating flood risk by minimizing floodplain devel-
opments coincides with conservation of aquatic life in
streams. What are the economic costs of this type of
risk avoidance?
What are the economic limitations and ecological ben-
efits of having light residential zones between water-
ways and commercial, urban or agricultural lands?
What are the economic development decisions that are
irreversible on the landscape? For example, once land
32 Water Resources Systems Planning and Management
is used for commercial development, it is normally too
costly to return it to agriculture. This would identify
limits on planning and management for conservation
and development.
What are the associated ecological and economic
impacts of the trend in residential, commercial and
forest lands replacing agricultural lands?
The answers to these and similar questions may well dif-
fer in different regions. However, if we can address them
on a regional scale in multiple river basins we might
begin to understand and predict better the interactions
among economy, environment and ecology as a function
of how we manage and use its land and water. This in
turn may help us better manage and use our land and
water resources for the betterment of all now and in
the future.
5.7. Post-Planning and Management Issues
Once a plan or strategy is produced, common imple-
mentation issues include seeing that the plan is
followed, and modified, as appropriate, over time. What
incentives need to be created to ensure compliance? How
are the impacts resulting from the implementation of any
decision going to be monitored, assessed and modified as
required and desired? Who is going to be responsible?
Who is going to pay, and how much? Who will keep the
stakeholders informed? Who will keep the plan current?
How often should plans and their databases be updated?
How can new projects be operated in ways that increase
the efficiencies and effectiveness of joint operation of
multiple projects in watersheds or river basins rather
than each project being operated independently of the
others? These questions should be asked and answered,
at least in general terms, before the water resources plan-
ning and management process begins. The questions
should be revisited as decisions are made and when
answers to them can be much more specific.
6. Meeting the Planning and
Management Challenges:
A Summary
Planning (the formulation of development and manage-
ment plans and policies) is an important and often
indispensable means to support and improve operational
wrm_ch01.qxd 8/31/2005 11:48 AM Page 32
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 33
management. It provides an opportunity to:
assess the current state of the water resources and the
conflicts and priorities over their use, formulate
visions, set goals and targets, and thus orient opera-
tional management
provide a framework for organizing policy relevant
research and public participation
increase the legitimacy, public acceptance of (or even
support for) the way the resources are to be allocated
or controlled, especially in times of stress
facilitate the interaction, discussion and coordination
among managers and stakeholders, and generate a com-
mon point of reference a management plan or policy.
Many of the concerns and issues being addressed by
water resources planners and managers today are similar
to those faced by planners and managers in the past. But
some are different. Most of the new ones are the result
of two trends: first, a growing concern for the sustain-
ability of natural ecosystems and second, an increased
recognition of the need of a bottom-up grass-roots par-
ticipatory approach to planning, managing and decision-
making.
Today planners work for economic development and
prosperity as they did in the past, keeping in mind
environmental impacts and goals as they did then, but now
recognizing ecological impacts and values as well. Water
resources management may still be focused on controlling
and mitigating the adverse impacts of floods and droughts
and water pollution, on producing hydropower, on devel-
oping irrigation, on controlling erosion and sediment, and
on promoting navigation, but only as these and similar
activities are compatible with healthy ecosystems. Natural
ecosystems generally benefit from the variability of natural
hydrological regimes. Other uses prefer less variability.
Much of our engineering infrastructure is operated so as to
reduce hydrological variability. Today water resource sys-
tems are increasingly required to provide rather than
reduce hydrological (and accompanying sediment load)
variability. Reservoir operators, for example, can modify
their water release policies to increase this variability.
Farmers and land-use developers must minimize rather than
encourage land-disturbing activities. Floodplains may need
to get wet occasionally. Rivers and streams may need to
meander and fish species that require habitats along the full
length of rivers to complete their life cycles must have access
to those river reaches. Clearly these ecological objectives,
added to all the other economic and environmental ones,
can only compound the conflicts and issues with respect to
land and water management and use.
So, how can we manage all this conflict and
uncertainty? We know that water resources planning and
management should be founded on sound science,
efficient public programme administration and the broad
participation of stakeholders. Yet obtaining each of these
three conditions is a challenge. While the natural and
social sciences can help us predict the economic, envi-
ronmental and ecological impacts of alternative decisions,
those predictions are never certain. In addition, these sci-
ences offer no help in determining the best decision to
make in the face of multiple conflicting goals held by mul-
tiple stakeholders goals that have changed, and no
doubt will continue to change. Water resources planning
and management and decision-making is not as easy as
we professionals can tell you what to do, all you need is
the will to do it. Very often it is not clear what should be
done. Professionals administering the science, often from
public agencies, non-governmental organizations, or even
from universities, are merely among all the stakeholders
having an interest in and contributing to the management
of water.
Each governmental agency, consulting firm, environ-
mental interest group and citizen typically has particular
limitations, authorities, expertise and conflicts with other
people, agencies and organizations, all tending to detract
from achieving a fully integrated approach to water
resources planning and management. But precisely
because of this, the participation and contributions of all
these stakeholders are needed. They must come together
in a partnership if indeed an integrated approach to water
resources planning and management is to be achieved
and sustained. All views must be heard, considered
and acted upon by all involved in the water resources
planning and management process.
Water resources planning and management is not sim-
ply the application and implementation of science. It is
creating a social environment that brings in all of us who
should be involved, from the beginning, in a continuing
planning process. This process is one of:
educating ourselves about how our systems work and
function
identifying existing or potential options and opportu-
nities for enhancement and resource development
and use
wrm_ch01.qxd 8/31/2005 11:48 AM Page 33
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
resolving the inevitable problems and conflicts that will
result over who gets what and when, and who pays
who for what and when and how much
making and implementing decisions, and finally of
monitoring the impacts of those decisions.
This process is repeated as surprises or new opportunities
or new knowledge dictates.
Successful water resources planning and management
requires the active participation of all community institu-
tions involved in economic development and resource
management. How can this begin at the local stakeholder
level? How does anyone get others interested in prevent-
ing problems before those problems are apparent, and
especially before unacceptable solutions are offered to
deal with them? And how do you deal with the inevitable
group or groups of stakeholders who see it in their best
interest not to participate in the planning process, but
simply to criticize it from the outside? Who is in a
position at the local level to provide the leadership and
financial support needed? In some regions, non-govern-
mental institutions have been instrumental in initiating
and coordinating this process at local grass-root levels.
Water resources planning and management processes
should identify a vision that guides development and
operational activities in the affected region. Planning and
management processes should:
recognize and address the goals and expectations of
the regions stakeholders
identify and respond to the regions water-related
problems
function effectively within the regions legal/institu-
tional frameworks
accommodate both short and long-term issues
generate a diverse menu of alternatives
integrate the biotic and abiotic parts of the basin
take into account the allocation of water for all needs,
including those of natural systems
be stakeholder driven
take a global perspective
be flexible and adaptable
drive regulatory processes, not be driven by them
be the basis for policy making
foster coordination among planning partners and
consistency among related plans
be accommodating of multiple objectives
34 Water Resources Systems Planning and Management
be a synthesizer, recognize and deal with conflicts
produce recommendations that can be implemented.
All too often integrated planning processes are hampered
by the separation of planning, management and imple-
menting authorities, turf-protection attitudes, shortsighted
focusing of efforts, lack of objectivity on the part of
planners, and inadequate funding. These deficiencies need
addressing if integrated holistic planning and management
is to be more than just something to write about.
Effective water resources planning and management is
a challenge today, and will be an increasing challenge
into the foreseeable future. This book introduces some
of the tools that are being used to meet these challenges.
We consider it only a step towards becoming an accom-
plished planner or manager.
7. References
ASCE (AMERICAN SOCIETY OF CIVIL ENGINEERS).
1998. Sustainability criteria for water resource systems.
Reston, Va., ASCE.
GWP (GLOBAL WATER PARTNERSHIP). 2000.
Integrated water resources management. Tac Background
Papers No. 4. Stockholm, Sweden, GWP.
NRC (NATIONAL RESEARCH COUNCIL). 2001.
Compensating for wetland losses under the clean water act.
Committee on Mitigating Wetland Losses, Board on
Environmental Studies and Toxicology, Water Science
and Technology Board.
NATIONAL RESEARCH COUNCIL. 1999. Water for the
future: the West Bank and Gaza Strip, Israel, and Jordan.
Water Science and Technology Board and the Board on
Environmental Studies and Toxicology, National Research
Council, National Academy Press, Washington, DC.
WCD. 2000. Dams and Developments A new framework
for decision-making: the report of the World Commission on
Dams. UK, Earthscan.
Additional References (Further Reading)
ABU-ZEID, M.A. and BISWAS, A.K. (eds.). 1996. River
basin planning and management. Oxford University Press,
Calcutta.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 34
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resources Planning and Management: An Overview 35
BARROW, C.J. 1998. River basin development planning
and management: a critical review. World-Development,
Vol. 26, No. 1, pp. 17186.
BISWAS, A.K. (ed.). 1997. Water resources: environmental
planning, management, and development. New York,
McGraw-Hill.
COOPER, A.B. and BOTTCHER, A.B. 1993. Basin-scale
modelling as a tool for water-resource planning. Journal of
Water Resources Planning and Management (ASCE), Vol.
119, No. 3, pp. 30623.
DIAMANTINI, C. and ZANON, B. 1996. River basin
planning in Italy: resource and risk management.
European-Environment, Vol. 6, No. 4, pp. 11925.
ECKSTEIN, O. 1958. Water resource development: the
economics of project evaluation. Cambridge, Mass., Harvard
University Press.
GLOBAL WATER PARTNERSHIP (GWP). 2000. Water as
a social and economic good: how to put the principle into
practice. Stockholm, Sweden, GWP, Tac Background
Papers No. 2.
GLOBAL WATER PARTNERSHIP (GWP). 2000. Effective
water governance. Stockholm, Sweden, GWP, Tac
Background Papers No. 7.
GOULTER, I.C. 1985. Equity issues in the implementation
of river basin planning. In: J. Lundqvist (ed.), Strategies
for river basin management: environmental integration of
land and water in a river basin, pp. 28792. Dordrecht,
Holland, D. Reidel.
HOWE, C.W. 1996. Water resources planning in a fed-
eration of states: equity versus efficiency. Natural
Resources Journal, Vol. 36, No. 1, pp. 2936.
HUAICHENG, G. and BEANLANDS, G. 1994. A compar-
ative study on Canadian and Chinese river basin plan-
ning. Journal of Environmental Science China, Vol. 6, No. 2,
pp. 22433.
KARAMOUZ, M.; ZINSSER, W.K. and SZIDAROVSZKY, F.
2003. Water resources systems analysis. Boca Raton, Fla.,
Lewis.
KRUTILLA, J.V. and ECKSTEIN, O. 1958. Multiple
purpose river development. Baltimore, Md., Johns
Hopkins Press.
KULSHRESHTHA, S. 1998. A global outlook for water
resources to the year 2025. Water Resources Management,
Vol. 12, No. 3, June, pp. 16784.
LEE, D.J. and DINAR, A. 1996. An integrated model of
river basin planning, development, and management.
Water International, Vol. 21, No. 4, pp. 21322. Also see
1995. Review of integrated approaches to river basin planning,
development and management, Washington, D.C., World
Bank, Agriculture and Natural Resources Department.
LINS, H.F.; WOLOCK, D.M. and MCCABE, G.J. 1997.
Scale and modelling issues in water resources planning.
Climate Change, Vol. 37, No. 1, pp. 6388.
LOUCKS, D.P. (ed.). 1998. Restoration of degraded rivers:
challenges, issues and experiences. Dordrecht, Holland,
Kluwer Academic.
LOUCKS, D.P. and DA COSTA, J.R. (eds.). 1991. Decision
support systems: water resources planning and research.
Berlin, Springer-Verlag.
LOUCKS, D.P.; STEDINGER, J.R. and HAITH, D.A.
1981. Water resources systems planning and analysis.
Englewood Cliffs, N.J., Prentice-Hall.
MAASS, A.; HUFSCHMIDT, M.M.; DORFMAN, R.;
THOMAS, H.A. Jr.; MARGLIN, S.A. and FAIR, G.M.
1962. Design of water resource systems. Cambridge, Mass.,
Harvard University Press.
MAIDMENT, D.R. (ed.). 1993. Handbook of hydrology.
New York, McGraw-Hill.
MAJOR, D.C. and LENTON, R.L. 1979. Applied water
resource systems planning. Englewood Cliffs, N.J.,
Prentice-Hall.
MAYS, L.W. (ed.). 1996. Water resources handbook.
New York, McGraw-Hill.
MCMILLAN, T. 1990. Water resource planning in
Canada. Journal of Soil and Water Conservation, Vol. 45,
No. 6, November/December, pp. 61416.
MITCHELL, B. 1983. Comprehensive river basin plan-
ning in Canada: problems and opportunities. Water
International, Vol. 8, No. 4, pp. 14653.
ORIORDAN, J. 1981. New strategies for water resource
planning in British Columbia. Canadian Water Resources
Journal, Vol. 6, No. 4, pp. 1343.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 35
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
RAZAVIAN, D.; BLEED, A.S.; SUPALLA, R.J. and
GOLLEHON, N.R. 1990. Multistage screening process for
river basin planning. Journal of Water Resources Planning
and Management (ASCE), Vol. 116, No. 3, May/June ,
pp. 32334.
REITSMA, R.F. and CARRON, J.C. 1997. Object-oriented
simulation and evaluation of river basin operations.
Journal of Geographic Information and Decision Analysis,
Vol. 1, No. 1, pp. 924.
REYNOLDS, P.J. Ecosystem approaches to river basin
planning strategies for river basin management. In:
J. Lundqvist (ed.), 1985. Environmental integration of
land and water in a river basin, pp. 418. Dordrecht,
Holland, D. Reidel.
SAHA, S.K. and BARROW, C.J. (eds.). 1981. River basin
planning: theory and practice. Chichester, UK, Wiley
Interscience.
SAVENIJE, H.H.G. and VAN DER ZAAG, P. (eds.). 1998.
The Management of Shared River Basins. Neda, The Hague,
Ministry of Foreign Affairs.
SCHRAMM, G. 1980. Integrated river basin planning in
a holistic universe. Natural Resources Journal, Vol. 20,
No. 4, October, pp. 787806.
36 Water Resources Systems Planning and Management
SMITH, S.C. and CASTLE E.N. (eds.). 1964. Economics
and public policy in water resources development. Ames,
Iowa, Iowa University Press.
SOMLYODY, L. Use of optimization models in river basin
water quality planning. In: M.B. Beck and P. Lessard (eds.).
1997. WATERMATEX 97: systems analysis and computing
in water quality management. Towards a New Agenda,
pp. 7387. London, International Water Association.
STOUT, G.E. 1998. Sustainable development requires the
full cooperation of water users. Water International,
Vol. 23, No. 1, March, pp. 37.
THANH, N.C. and BISWAS, A.K. (eds.). Environmentally-
sound water management. Delhi, Oxford University Press.
THIESSEN, E.M.; LOUCKS, D.P. and STEDINGER, J.R.
1998. Computer-assisted negotiations of water resources
conflicts. Group Decision and Negotiation, Vol. 7, No. 2,
pp. 10929.
VIESSMAN, W. 1996. Integrated water management.
Water Resources Update, No. 106, Winter, pp. 212.
VIESSMAN, W. 1998. Water Policies for the Future.
Water Resources Update, No. 111, Spring, pp. 47,
10410.
wrm_ch01.qxd 8/31/2005 11:48 AM Page 36
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
2. Water Resource Systems Modelling: Its Role in Planning and Management
1. Introduction 39
2. Modelling of Water Resources Systems 41
2.1. An Example Modelling Approach 41
2.2. Characteristics of Problems to be Modelled 41
3. Challenges in Water Resources Systems Modelling 43
3.1. Challenges of Planners and Managers 43
3.2. Challenges of Modelling 44
3.3. Challenges of Applying Models in Practice 45
4. Developments in Modelling 46
4.1. Modelling Technology 46
4.2. Decision Support Systems 47
4.2.1. Shared-Vision Modelling 49
4.2.2. Open Modelling Systems 51
4.2.3. Example of a DSS for River Flood Management 51
5. Conclusions 54
6. References 55
wrm_ch02.qxd 8/31/2005 11:49 AM Page 38
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
When design and management decisions are made about
environmental and water resources systems, they are
based on what the decision-makers believe, or perhaps
hope, will take place as a result of their decisions. These
predictions are either based on very qualitative informa-
tion and beliefs in peoples heads or crystal balls
(Figure 2.1) or, at least in part, on quantitative infor-
mation provided by mathematical or computer-based
models (Figure 2.2). Today computer-based modelling is
used to enhance mental models. These quantitative
mathematical models are considered essential for carry-
ing out environmental impact assessments. Mathematical
simulation and optimization models packaged within
interactive computer programs provide a common way
for planners and managers to predict the behaviour of
any proposed water resources system design or manage-
ment policy before it is implemented.
39
Water Resource Systems
Modelling: Its Role in
Planning and Management
Planning, designing and managing water resources systems today inevitably
involve impact prediction. Impact prediction involves modelling. While
acknowledging the increasingly important role of modelling in water resources
planning and management, we also acknowledge the inherent limitation of models
as representations of any real system. Model structure, input data, objectives and
other assumptions related to how the real system functions or will behave under
alternative infrastructure designs and management policies or practices may be
controversial or uncertain. Future events are always unknown and of course any
assumptions about them may affect model outputs, that is, their predictions. As
useful as they may or may not be, the results of any quantitative analysis are
always only a part, but an important part, of the information that should be
considered by those involved in the overall planning and management
decision-making process.
2
Modelling provides a way, perhaps the principal way,
of predicting the behaviour of proposed infrastructural
designs or management policies. The past thirty years have
witnessed major advances in our abilities to model the
engineering, economic, ecological, hydrological and some-
times even the institutional or political impacts of large,
complex, multipurpose water resources systems. Appli-
cations of models to real systems have improved our under-
standing, and hence have often contributed to improved
system design, management and operation. They have also
taught us how limited our modelling skills remain.
Water resources systems are far more complex than
anything analysts have been, or perhaps ever will be,
able to model and solve. The reason is not simply any
computational limit on the number of model variables,
constraints, subroutines or executable statements in
those subroutines. Rather it is because we do not under-
stand sufficiently the multiple interdependent physical,
biochemical, ecological, social, legal and political
wrm_ch02.qxd 8/31/2005 11:49 AM Page 39
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
(human) processes that govern the behaviour of
water resources systems. These processes are affected by
uncertainties in things we can measure, such as water
supply and water demands. They are also affected by
the unpredictable actions of multiple individuals and
institutions that are affected by what they get or do not
get from the management and operation of such
systems, as well as by other events having nothing
directly to do with water.
The development and application of models in other
words, the art, science and practice of modelling, as
will be discussed in the following chapters should be
preceded by the recognition of what can and cannot be
40 Water Resources Systems Planning and Management
Figure 2.1. Using mental models for prediction.
Figure 2.2. Using computer models for prediction.
achieved from the use of models. Models of real-world
systems are always simplified representations. What
features of the actual system are represented in a model,
and what features are not, will depend in part on what the
modeller thinks is important with respect to the issues
being discussed or the questions being asked. How well
this is done will depend on the skill of the modeller, the
time and money available, and, perhaps most impor-
tantly, the modellers understanding of the real system
and decision-making process.
Developing models is an art. It requires knowledge of
the system being modelled, the clients objectives, goals
and information needs, and some analytical and pro-
gramming skills. Models are always based on numerous
assumptions or approximations, and some of these may
be at issue. Applying these approximations of reality in
ways that improve understanding and eventually lead to a
good decision clearly requires not only modelling skills
but also the ability to communicate effectively.
Models produce information. They do not produce
decisions. Water resources planners and managers must
accept the fact that decisions may not be influenced by their
planning and management model results. To know, for
example, that cloud seeding may, on average, reduce the
strength of hurricanes over a large region does not mean
that such cloud-seeding activities will or should be under-
taken. Managers or operators may know that not everyone
will benefit from what they would like to do, and those
who lose will likely scream louder than those who gain.
In addition, decision-makers may feel safer in inaction
than action (Shapiro, 1990; Simon, 1988). There is a
strong feeling in many cultures and legal systems that
failure to act (nonfeasance) is more acceptable than
acts that fail (misfeasance or malfeasance). We all feel
greater responsibility for what we do than for what we do
not do. Yet our aversion to risk should not deter us from
addressing sensitive issues in our models. Modelling
efforts should be driven by the need for information
and improved understanding. It is that improved
understanding (not improved models per se) that may
eventually lead to improved system design, management
and/or operation. Models used to aid water resources
planners and managers are not intended to be, and rarely
are (if ever), adequate to replace their judgement. This
we have learned, if nothing else, in over forty years of
modelling experience.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 40
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 41
This brief chapter serves as an overview of modelling
and its applications. The emphasis is on application. This
chapter is about modelling in practice more than in
theory. It is based on the considerable experience and
literature pertaining to how well, or how poorly, profes-
sional practitioners and researchers have done over the
past four decades or more in applying various modelling
approaches or tools to real problems with real clients (also
see, for example, Austin, 1986; Gass, 1990; Kindler, 1987,
1988; Loucks et al., 1985; Reynolds, 1987 and Rogers
and Fiering, 1986).
In attempting to understand how modelling can better
support planners and managers, it may be useful to
examine just what planners and managers of complex
water resources systems do. What they do governs to
some extent what they need to know. And what they need
to know governs to a large extent what modellers or ana-
lysts should be trying to provide. In this book the terms
analysts or modellers, planners, and managers can refer to
the same person or group of individuals. The terms are
used to distinguish the activities of individuals, not
necessarily the individuals themselves.
First, a brief example is presented to demonstrate
the value of modelling. Then we offer some general
thoughts on the major challenges facing water resources
systems planners and managers, the information they
need to meet these challenges, and the role analysts have
in helping to provide this information. Finally, we argue
why we think the practice of modelling is in a state of
transition, and how current research and development
in modelling and computing technology are affecting
that transition. New computer technology has had and
will continue to have a significant impact in the devel-
opment and use of models for water resources planning
and management.
2. Modelling of Water Resources
Systems
2.1. An Example Modelling Approach
Consider for example the sequence or chain of models
required for the prediction of fish and shellfish survival
as a function of nutrient loadings into an estuary. The
condition of the fish and shellfish are important to
the stakeholders. One way to maintain healthy stocks is
to maintain sufficient levels of oxygen in the estuary. The
way to do this is to control algae blooms. This in turn
requires limiting the nutrient loadings to the estuary that
can cause algae blooms and subsequent dissolved oxygen
deficits. The modelling challenge is to link nutrient
loading to fish and shellfish survival. In other words,
can some quantitative relationship be defined relating the
amount of nutrient loading to the amount of fish and
shellfish survival?
The negative effects of excessive nutrients (e.g., nitro-
gen) in an estuary are shown in Figure 2.3. Nutrients
stimulate the growth of algae. Algae die and accumulate
on the bottom where bacteria consume them. Under calm
wind conditions density stratification occurs. Oxygen is
depleted in the bottom water. Fish and shellfish may die
or become weakened and more vulnerable to disease.
A sequence of deterministic or better, of probabilistic
models, each providing input data to the next model,
can be defined (Chapter 12) to predict shellfish and fish
abundance in the estuary based on upstream nutrient
loadings. These models, for each link shown in Figure 2.4,
can be a mix of judgemental, mechanistic and/or statisti-
cal ones. Statistical models could range from simple regres-
sions to complex artificial neural networks (Chapter 6).
Any type of model selected will have its advantages as
well as its limitations, and its appropriateness may largely
depend on the amount and precision of the data available
for model calibration and verification.
The biological endpoints shell-fish abundance and
number of fish-kills are meaningful indicators to stake-
holders and can easily be related to designated water
body use.
2.2. Characteristics of Problems to be
Modelled
Problems motivating modelling and analyses exhibit a
number of common characteristics. These are reviewed
here because they provide insight into whether a
modelling study of a particular problem may be worth-
while. If the planners objectives are very unclear, if few
alternative courses of action exist, or if there is little
scientific understanding of the issues involved, then
mathematical modelling and sophisticated methodologies
are likely to be of little use.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 41
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Successful applications of modelling are often charac-
terized by:
A systems focus or orientation. In such situations atten-
tion needs to be devoted to the interdependencies and
interactions of elements within the system as a whole,
as well as to the elements themselves.
The use of interdisciplinary teams. In many complex and
non-traditional problems it is not at all clear from the
start what disciplinary viewpoints will turn out to be
most appropriate or acceptable. It is essential that
participants in such work coming from different
established disciplines become familiar with the
techniques, vocabulary and concepts of the other
disciplines involved. Participation in interdisciplinary
modelling often requires a willingness to make mis-
takes at the fringes of ones technical competence and
to accept less than the latest advances in ones own
discipline.
The use of formal mathematics. Most analysts prefer
to use mathematical models to assist in system
42 Water Resources Systems Planning and Management
description and the identification and evaluation of
efficient tradeoffs among conflicting objectives, and to
provide an unambiguous record of the assumptions
and data used in the analysis.
Not all water resources planning and management prob-
lems are suitable candidates for study using modelling
methods. Modelling is most appropriate when:
The planning and management objectives are reason-
ably well defined and organizations and individuals
can be identified who can benefit from understanding
the model results.
There are many alternative decisions that may satisfy
the stated objectives, and the best decision is not
obvious.
The water resources system and the objectives being
analysed are describable by reasonably tractable math-
ematical representations.
The information needed, such as the hydrological,
economic, environmental and ecological impacts
under calm wind conditions,
density stratification
oxygen is depleted in the bottom water
fish and shellfish may die or become weakened and
vulnerable to disease
algae die
and accumulate
on the bottom and
are consumed by bacteria
nutrients
stimulate
the growth
of algae
a
b
c
e
d
a
b
c
d
e
E
0
2
0
7
3
0
b
Figure 2.3. The impacts of
excessive nutrients in an
estuary (Borsuk et al., 2001).
wrm_ch02.qxd 8/31/2005 11:49 AM Page 42
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 43
resulting from any decision, can be better estimated
through the use of models.
The parameters of these models are estimable from
readily obtainable data.
3. Challenges in Water Resources
Systems Modelling
3.1. Challenges of Planners and Managers
Planners and managers of water resources systems are
the people responsible for solving particular water-related
problems or meeting special water resources needs. When
they fail, they hear about it. The public lets them know.
What makes their job particularly challenging is that the
public has differing needs and expectations. Furthermore,
institutions where water resources planners and managers
work (or hire consultants to work for them) are like most
institutions these days: they must do what they can with
limited financial and human resources. Their clients are
all of us who use water, or at least all of us who are
affected by the decisions they make.
The overall objective of these planners and managers
and their institutions is to provide a service, such as a
reliable and inexpensive supply of water, an assurance
of water quality, the production of hydropower, protection
from floods, the provision of commercial navigation and
recreational opportunities, the preservation of wildlife
and enhancement of ecosystems, or some combination of
these or other purposes. Furthermore they are expected
to do this at a cost no greater than what people are willing
to pay. Meeting these goals (i.e., keeping everyone happy)
is not always easy or even possible.
Simple technical measures or procedures are rarely
able to ensure a successful solution to any particular set
of water resources management problems. Furthermore,
everyone who has had any exposure to water resources
planning and management knows that one cannot design
or operate a water resources system without making
compromises. These compromises are over competing
purposes (such as hydropower and flood control) or
competing objectives (such as who benefits and who
pays, and how much and where and when). After
analysts, using their models of course, identify possible
ways of achieving various goals and objectives and
provide estimates of associated economic, environmental,
ecological and social impacts, it is the planners and
managers who have the more difficult job. They must
work with and influence everyone who will be affected by
whatever decision they make.
Planning and managing involves not only decision-
making, but also developing among all interested and
influential individuals an understanding and consensus
that legitimizes the decisions and enhances their successful
implementation. Planning and managing are processes that
take place in a social or political environment. They involve
leadership and communication among people and institu-
tions, and the skills required are learned from experience of
working with people, not with computers or models.
Moving an organization or institution into action to
achieve specific goals involves a number of activities,
including goal-setting, debating, coordinating, motivating,
deciding, implementing and monitoring. Many of these
activities must be done simultaneously and continuously,
especially as conditions (goals and objectives, water
supplies, water demands, financial budgets) change over
E
0
2
0
7
3
0
c
nutrient
inputs
carbon
production
chlorophyll
violations
harmfull
algal
blooms
sediment
oxygen
demand
number
of
fishkills
shellfish
abundance
algal
density
river
flow
frequency
of
hupoxia
duration
of
stratification
fish
health
Figure 2.4. Cause and effect diagram for estuary
eutrophication due to excessive nutrient loadings
(Borsuk et al., 2001).
wrm_ch02.qxd 8/31/2005 11:49 AM Page 43
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
time. These activities create a number of challenges that
are relevant to modellers or analysts. They include how to:
identify creative alternatives for solving problems
find out what each interest group wants to know in
order to reach an understanding of the issues and a
consensus on what to do
develop and use models and present their results
so that everyone can reach a common or shared
understanding and agreement that is consistent with
their individual values
make decisions and implement them, given differences
in opinions, social values and objectives.
In addressing these needs or challenges, planners and
managers must consider the relevant:
legal rules and regulations
history of previous decisions
preferences of important actors and interest groups
probable reactions of those affected by any decision
relative importance of various issues being addressed
applicable science, engineering and economics the
technical aspects of their work.
We mention these technical aspects lastly not to suggest
that they are the least important factor to be considered.
We do so to emphasize that they are only one set among
many factors and probably, in the eyes of planners and
managers, not the most decisive or influential (Ahearne,
1988; Carey, 1988; Pool, 1990 and Walker, 1987).
So, does the scientific, technical, systematic approach
to modelling for planning and management really mat-
ter? We believe it can if it addresses the issues of concern
to the modellers clients: the planners and the managers.
Analysts need to be prepared to interact with the
political or social structure of the institutions they are
attempting to assist, as well as with the public and the
press. Analysts should also be prepared to have their
work ignored. Even if they are presenting facts based on
the current state of the sciences, these sciences may
not be considered relevant. Fortunately for scientists and
engineers, this is not always the case. The challenge of
modellers or analysts interested in having an impact on
the practice of water resources systems planning and
management is to become a part of the largely political
planning and management process and to contribute
towards its improvement.
44 Water Resources Systems Planning and Management
3.2. Challenges of Modelling
To engage in a successful water resources systems
study, the modeller must possess not only the requisite
mathematical and systems methodology skills, but also
an understanding of the environmental engineering,
economic, political, cultural and social aspects of water
resources planning problems. Consider, for example, the
study of a large land-development plan. The planner
should be able to predict how the proposed development
will affect the quantity and quality of the surface and
subsurface runoff, and the impact this will have on the
quantity and quality of surface and ground waters
and their ecosystems. These impacts, in turn, might
affect the planned development itself, or other land uses
downstream. To do this the analysts must have an under-
standing of the biological, chemical, physical and even
social processes that are involved in water resources
management.
A reasonable knowledge of economic theory, law,
regional planning and political science can be just as
important as an understanding of hydraulic, hydrogeo-
logic, hydrological, ecologic and environmental engineer-
ing disciplines. It is obvious that the results of most water
resources management decisions have a direct impact on
people and their relationships. Hence inputs from those
having a knowledge of those other disciplines are also
needed during the comprehensive planning of water
resources systems, especially during the development and
evaluation of the results of various planning models.
Some of the early water resources systems studies were
undertaken with a nave view of the appropriate role and
impact of models and modellers in the policy-making
process. The policy-maker could foresee the need to make
a decision. He or she would ask the systems group
to study the problem. They would then model it,
identify feasible solutions and their consequences, and
recommend one or at most a few alternative solutions.
The policy-maker, after waiting patiently for these recom-
mendations, would then make a yes or no decision.
However, experience to date suggests the following:
A final solution to a water resources planning problem
rarely exists: plans and projects are dynamic. They
evolve over time as facilities are added and modified to
adapt to changes in management objectives and in the
demands placed on the facilities.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 44
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 45
For every major decision there are many minor
decisions, made by different agencies or management
organizations responsible for different aspects of a
project.
The time normally available to study particular water
resources problems is shorter than the time needed;
if there is sufficient time, the objectives of the original
study will probably have shifted significantly by the
time the study is completed.
This experience emphasizes some of the limitations and
difficulties that any water resources systems study may
encounter, but more importantly, it underscores the need
for constant communication among the analysts, system
planners, managers and operators, and policy-makers.
The success or failure of many past water resource stud-
ies is due largely to the efforts expended or not expended
in ensuring adequate, timely and meaningful communi-
cation communication among systems analysts, plan-
ners, those responsible for system operation and design,
and public officials responsible for major decisions and
setting general policies. Decision-makers who need the
information that can be derived from various models and
analyses, need it at particular times and in a form useful
and meaningful to them. Once their window of opportu-
nity for decision-making has passed, such information, no
matter how well presented, is often useless.
At the beginning of any study, objectives are usually
poorly defined. As more is learned about what can be
achieved, stakeholders are better able to identify what
they want to do. Close communication among analysts
and all interested stakeholders and decision-makers
throughout the modelling process is essential if systems
studies are to make their greatest contribution to the
planning process. Objectives as stated at the beginning of
a study are rarely the objectives as understood at its end.
Furthermore, those who will use models, and present
the information derived from models to those responsible
for making decisions, must be intimately involved with
model development, solution and analysis. It is only then
can they appreciate the assumptions upon which any par-
ticular model is based, and hence adequately evaluate the
reliability of the results. A water resources systems study
that involves only outside consultants, and has minimal
communication between consultants and planners
within a responsible management agency or involved
stakeholders, is unlikely to have a significant impact
on the planning process. Models that are useful are
constantly being modified and applied by those involved
in plan preparation, evaluation and implementation.
The interaction described above is illustrated in
Figure 1.20 of the previous chapter. Models are devel-
oped and applied during the second and third phase of
this analytical framework. A continuous communication
with the decision-makers and stakeholder representa-
tives should ensure the models and results will indeed
serve their purpose.
3.3. Challenges of Applying Models in Practice
As already mentioned, the clients of modellers or analysts
are typically planners and managers who have problems
to solve and who could benefit from a better understand-
ing of what options they have and what impacts may
result. They want advice on what to do and why, what
will happen as a result of what they do, and who will care
and how much. The aim of analysts is to provide planners
and managers with meaningful (understandable), useful,
accurate and timely information. This information serves
to help them better understand their system, its problems,
and alternative ways to address them. The purpose of
water resources systems planning and management mod-
elling, stated once again, is to provide useful and timely
information to those involved in managing such systems.
Modelling is a process or procedure intended to focus
and force clearer thinking and to promote more informed
decision-making. The approach involves problem recog-
nition, system definition and bounding, identification of
various goals or objectives, identification and evaluation
of various alternatives, and very importantly, effective
communication of this information to those who need
to know.
The focus of most books and articles on water resource
systems modelling is on modelling methods. This book is
no different. But what all of us should also be interested
in, and discuss more than we do, is the use of these tools
in the processes of planning and management. If we did,
we could learn much from each other about what tools
are needed and how they can be better applied in
practice. We could extend the thoughts of those who,
in a more general way, addressed these issues over two
decades ago (Majoni and Quade, 1980; Miser, 1980;
Stokey and Zeckhauser, 1977 and Tomlison, 1980).
wrm_ch02.qxd 8/31/2005 11:49 AM Page 45
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
There is always a gap between what researchers in
water resources systems modelling produce and publish,
and what the practitioner finds useful and uses. Those
involved in research are naturally interested in developing
new and improved tools and methods for studying, iden-
tifying and evaluating alternative water resources system
designs and management and operation policies. If there
were no gap between what is being developed or
advocated by researchers and that which is actually used
by practitioners, either the research community would be
very ineffective in developing new technology or the
practitioners would be incredibly skilled in reading,
assimilating, evaluating and adapting this research to
meet their needs. Evaluation, testing and inevitable
modifications take time. Not all published research is
ready or suited for implementation. Some research results
are useful, some are not. It is a work in progress.
How can modellers help reduce the time it takes for
new ideas and approaches to be used in practice? Clearly,
practitioners are not likely to accept a new modelling
approach, or even modelling itself, unless it is obvious
that it will improve the performance of their work as well
as help them address problems they are trying to solve.
Will some new model or computer program make it
easier for practitioners to carry out their responsibilities?
If it will, there is a good chance that the model or com-
puter program might be successfully used, eventually.
Successful use of the information derived from models or
programs is, after all, the ultimate test of the value of
those tools. Peer review and publication is only one, and
perhaps not even a necessary, step towards that ultimate
test or measure of value of a particular model or model-
ling approach.
4. Developments in Modelling
4.1. Modelling Technology
The increasing developments in computer technology
from microcomputers and workstations to supercomputers
have motivated the concurrent development of an
impressive set of new models and computer software.
This software is aimed at facilitating model use and, more
importantly, interaction and communication between the
analysts or modellers and their clients. It includes:
46 Water Resources Systems Planning and Management
interactive approaches to model operation that put
users more in control of their computers, models, and
data
computer graphics that facilitate data input, editing,
display and comprehension
geographic information systems that provide
improved spatial analysis and display capabilities
expert systems that can help the user understand
better how complex decision problems might be
solved, and at the same time explain to the users why
one particular decision may be better than another
electronic mail and the Internet, which let analysts,
planners and managers communicate and share data
and information with others worldwide, and to run
models that are located and maintained at distant sites
multimedia systems that permit the use of sound and
video animation in analyses, all aimed at improving
communication and understanding.
These and other software developments are giving planners
and managers improved opportunities for increasing their
understanding of their water resources systems. Such
developments in technology should continue to aid all of
us in converting model output data to information; in other
words, it should provide us with a clearer knowledge and
understanding of the alternatives, issues and impacts asso-
ciated with potential solutions to water resources systems
problems. But once again, this improved information and
understanding will only be a part of everything planners
and managers must consider.
Will all the potential benefits of new technology
actually occur? Will analysts be able to develop and apply
these continual improvements in new technology wisely?
Will we avoid another case of oversell or unfulfilled
promises? Will we avoid the temptation of generating
fancy animated, full-colour computer displays just
because we are easily able to produce them, rather than
working on the methods that will add to improved under-
standing of how to solve problems more effectively? Will
we provide the safeguards needed to ensure the correct
use and interpretation of the information derived from
increasingly user-friendly computer programs? Will we
keep a problem-solving focus, and continue to work
towards increasing our understanding of how to improve
the development and management of our water resources,
whether or not our planning models are incorporated into
wrm_ch02.qxd 8/31/2005 11:49 AM Page 46
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 47
some sort of interactive computer-aided support system?
We can, but it will take discipline.
As modellers or researchers, we must discipline our-
selves to work more closely with our clients: the planners,
managers and other specialists who are responsible for
the development and operation of our water resources
systems. We must study their systems and their problems,
and we must identify their information needs. We must
develop better tools that they themselves and other inter-
ested stakeholders can use to model their water resource
systems and obtain an improved understanding a
shared vision of how their system functions and of their
available management options and associated impacts or
consequences. We must be willing to be multidisciplinary
and capable of including all relevant data in our analyses.
We must appreciate and see the perspectives of the agron-
omists, ecologists, economists, engineers, hydrologists,
lawyers or political and regional scientists as appropriate.
Viewing a water resources system from a single-discipline
perspective is rarely sufficient for todays water resource
systems planning.
Even if we have successfully incorporated all relevant
disciplines and data in our analyses, we should have a
healthy scepticism about our resulting information. We
must admit that this information, especially concerning
what might happen in the future, is uncertain. If we are
looking into the future (whether using crystal balls as
shown in Figure 2.1 or models as in Figure 2.2), we must
admit that many of our assumptions, such as parameter
values, cannot even be calibrated, let alone verified. Our
conclusions or estimates can be very sensitive to those
assumptions. One of our major challenges is to commu-
nicate this uncertainty in understandable ways to those
who ask for our predictions.
4.2. Decision Support Systems
Water resources planners and managers today must con-
sider the interests and goals of numerous stakeholders.
The planning, managing and decision-making processes
involve negotiation and compromise among these numer-
ous stakeholders, like those shown in Figure 2.5, who
typically have different interests, objectives and opinions
about how their water resources system should be man-
aged. How do we model to meet the information needs of
all these different stakeholders? How can we get them to
believe in and accept these models and their results? How
do we help them reach a common shared vision? How
can we help create a shared vision among all stakeholders
of at least how their system works and functions, if not
how they would like it to?
Today we know how to build some rather impressive
models of environmental systems. We know how to
incorporate within our models the essential biology,
chemistry and physics that govern how the environmen-
tal system works. We have also learned a little about how
to include the relevant economics, ecology and engineer-
ing into these models. Why do we do this? We do all this
modelling simply to be able to estimate, or identify, and
compare and evaluate the multiple impacts resulting from
different design and management decisions we might
make. Such information, we assume, should be of value
to those responsible for choosing the best decision.
If our goal is to help prevent, or contribute to the
solution of, water resources problems, then simply having
information from the worlds best models and technology,
as judged by our peers, is not a guarantee of success. To
be useful in the political decision-making process, the
Figure 2.5. Stakeholders involved in river basin planning and
management, each having different goals and information
needs (Engineering News Record, 20 September 1993, with
permission).
wrm_ch02.qxd 8/31/2005 11:49 AM Page 47
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
information we generate with all our models and com-
puter technology must be understandable, credible and
timely. It must be just what is needed when it is needed.
It must be not too little nor too much.
The optimal format and level of detail and precision of
any information generated from models should depend
on the needs and backgrounds of each individual
involved in the decision-making process. The value of
such information, even if the format and content are opti-
mal, will also depend on when it is available. Information
on an issue is only of value if it is available during the time
when the issue is being considered that is, when there
is an interest in that issue and a decision concerning what
to do about it has not yet been made. That is the window
of opportunity when information can have an impact.
Information is of no value after the decision is made,
unless of course it results in opening up another window
of opportunity.
If there is truth in the expression decision-makers
dont know what they want until they know what they
can get, how do modellers know what decision-makers
will need before even they do? How will modellers know
what is the right amount of information, especially if
48 Water Resources Systems Planning and Management
they are to have that information available, and in the
proper form, before or at the time, not after, it is needed?
Obviously modellers cannot know this. However, over
the last two decades or so this challenge has been
addressed by developing and implementing decision sup-
port systems (DSSs) (Fedra, 1992; Georgakakos and
Martin, 1996; Loucks and da Costa, 1991). These inter-
active modelling and display technologies can, within
limits, adapt to the level of information needed and can
give decision-makers some control over data input, model
operation and data output. But will each decision-maker,
each stakeholder, trust the model output? How can they
develop any confidence in the models contained in a DSS?
How can they modify those models within a DSS to address
issues the DSS developer may not have considered? An
answer to these questions has been the idea of involving the
decision-makers themselves not only in interactive model
use, but in interactive model building as well.
Figure 2.6 gives a general view of the components of
many decision support systems. The essential feature is
the interactive interface that permits easy and meaningful
data entry and display, and control of model (or com-
puter) operations.
simple
diagnostic
tools
system
development
tools
strategy
development
tools
evaluation
tools
surveys in-situ
measurements
remote sensing
rule and
knowledge bases
data bases geographic
information system
natural
system models
user-
function models
economic models
u
s
e
r
u
s
e
r

f
r
i
e
n
d
l
y

i
n
t
e
r
f
a
c
e
g
e
o
g
r
a
p
h
i
c
a
l

p
r
e
s
e
n
t
a
t
i
o
n
measurement system
information system
model system
analysis system
E
0
4
0
4
0
5
g
Figure 2.6. Common components of
many decision support systems.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 48
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 49
Depending on the particular issue at hand, and more
importantly the particular individuals and institutions
involved, a decision support system in the broadest sense
can range from minimal if any computer model use
where the decision-makers provide all the data and
analyses, make the decision, and they or their institutions
implement those decisions to decision support systems
that are fully automated and where no human involve-
ment is present. The latter are rare, but they do exist. The
automatic closing of the flood gates in Rotterdam harbour
is an example of this. These extremes, and various levels
of DSS in between are outlined in Figure 2.7.
4.2.1. Shared-Vision Modelling
Involving stakeholders in model building gives them a
feeling of ownership. They will have a much better
understanding of just what their model can do and
what it cannot. If they are involved in model-building,
they will know the assumptions built into their model.
Being involved in a joint modelling exercise is a way to
better understand the impacts of various assumptions.
While there may be no agreement on the best of various
assumptions to make, stakeholders can learn which
of those assumptions matter and which do not. In
addition, just the process of model development by
numerous stakeholders will create discussions that can
lead toward a better understanding of everyones inter-
ests and concerns. Through such model-building
exercises, it is just possible those involved will reach
not only a better understanding of everyones concerns,
but also a common or shared vision of at least how
their water resources system works (as represented by
their model, of course). Experience in stakeholder
involvement in model-building suggests such model-
building exercises can also help multiple stakeholders
reach a consensus on how their real system should be
developed and managed.
In the United States, one of the major advocates of
shared vision modelling is the Institute for Water
Resources of the US Army Corps of Engineers. They
have applied their interactive general-purpose model-
building platform in a number of exercises where con-
flicts existed over the design and operation of water
systems (Hamlet, et al., 1996a, 1996b, 1996c; Palmer,
Keys and Fisher, 1993; Werick, Whipple and Lund,
1996). Each of these model-building shared-vision
exercises included numerous stakeholders together with
experts in the use of the software. Bill Werick of the
Corps writes:
Because experts and stakeholders can build these models
together, including elements that interest each group,
they gain a consensus view of how the water system
decision
implemented
by
decision
selection
by
approach
to decision-
making
options
generated
by
data
analysed
by
data
provided
by
completely
unsupported
1
decision-
maker
information
supported
2
decision-
maker
systematic
analysis
3
decision-
maker
sys. analysis
alternatives
4
decision-
maker
system with
over-ride
5
decision-
maker
automated 6
GIS / DB
GIS / DB
GIS / DB
GIS / DB
GIS / DB
MODEL
MODEL
MODEL
MODEL
E040325c
Various Phases of Decision Support Systems
Figure 2.7. Various types of
computer-aided decision support
systems (based on OCallaghan, 1996).
wrm_ch02.qxd 8/31/2005 11:49 AM Page 49
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
works as a whole, and how it affects stakeholders and
the environment. Without adding new bureaucracies or
reassigning decision-making authority, the shared vision
model and the act of developing it create a connectedness
among problem solvers that resembles the natural
integration of the conditions they study.
Now the question is how to get all the stakeholders, many
of whom may not really want to work together, involved
in a model-building exercise. This is our challenge!
One step in that direction is the development of
improved technologies that will facilitate model develop-
ment and use by stakeholders with various backgrounds
and interests. We need better tools for building DSSs, not
just better DSSs themselves. We need to develop better
modelling environments that people can use to make their
own models. Researchers need to be building the model
building blocks, as opposed to the models themselves, and
to focus on improving those building blocks that can
be used by others to build their own models. Clearly if
stakeholders are going to be involved in model-building
exercises, it will have to be an activity that is enjoyable and
require minimal training and programming skills.
Traditional modelling experiences seem to suggest that
there are five steps in the modelling process. The first is
to identify the information the model is to provide. This
includes criteria or measures of system performance that
are of interest to stakeholders. These criteria or measures
are defined as functions of the behaviour or state of the
system being modelled. Next, this behaviour needs to be
modelled so the state of the system associated with any
external inputs can be predicted. This requires model-
ling the physical, chemical, biological, economic, ecolog-
ical and social processes that take place, as applicable, in
the represented system. Thirdly, these two parts are put
together, along with a means of entering the external
inputs and obtaining in meaningful ways the outputs.
Next, the model must be calibrated and verified or
validated, to the extent it can. Only now can the model
be used to produce the information desired.
This traditional modelling process is clearly not going
to work for those who are not especially trained or expe-
rienced (or even interested) in these modelling activities.
They need a model-building environment where they can
easily create models that:
they understand
are compatible with available data
50 Water Resources Systems Planning and Management
work and provide the level and amount of information
needed
are easily calibrated and verified when possible
give them the interactive control over data input,
editing, model operation and output display that they
can understand and that they need in order to make
informed decisions.
The challenge in creating such model-building environ-
ments is to make them sufficiently useful and attractive
that multiple stakeholders will want to use them. They
will have to be understandable. They will have to be rela-
tively easy and transparent, and even fun, to build. They
must be capable of simulating and producing different
levels of detail with regard to natural, engineering,
economic and ecological processes that take place at
different spatial and temporal scales. And they must
require no programming and debugging by the users. Just
how can this be done?
One approach is to develop interactive modelling
shells specifically suited to modelling environmental
problems. Modelling shells are data-driven programs that
become models once sufficient data have been entered
into them.
There are a number of such generic modelling shells
for simulating water resources systems. AQUATOOL
(Andreu et al., 1991), RIBASIM (Delft Hydraulics, 2004),
MIKE-BASIN (Danish Hydraulic Institute, 1997) and
WEAP (Raskin et al., 2001) (Shown in Figure 2.8) are
representative of interactive riveraquifer simulation shells
that require the system to be represented by, and drawn in
as, a network of nodes and links. Each node and link
requires data, and these data depend on what that node
or link represents, as well as what the user wants to get
from the output. If what is of interest is the time series of
quantities of water flowing, or stored, within the system
as a result of reservoir operation and/or water allocation
policies, then water quality data need not be entered, even
though there is the capacity to model water quality. If
water quality outputs are desired, then the user can choose
the desired various water quality constituents. Obviously,
the more different types of information desired or the
greater spatial or temporal resolution desired in the model
output, the more input data required.
Interactive shells provide an interactive and adaptive
way to define models and their input data. Once a model
wrm_ch02.qxd 8/31/2005 11:49 AM Page 50
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 51
is defined, the shell provides the interface for input data
entry and editing, model operation and output data
display.
To effectively use such shells, some training is useful
in the use of the shell and what it can and cannot do. The
developers of such shells have removed the need to
worry about database management, solving systems of
equations, developing an interactive interface, preserving
mass balances and continuity of flow, and the like. Any
assumptions built into the shell should be readily trans-
parent and acceptable to all before it is used in any
shared-vision exercises.
4.2.2. Open Modelling Systems
The next step in shared-vision modelling will be to create
a modelling environment that will enable all stakeholders
to include their own models in the overall system descrip-
tion. Stakeholders tend to believe their own models more
than those provided by governmental agencies or research
institutes. Their own models include the data they trust,
and are based on their own assumptions and views on
how the system works. For example, in transboundary
water resources issues, different countries may want to
include their own hydrodynamic models for the river
reaches in their country.
Various developments on open modelling systems
are taking place in Europe and the United States,
although most of them are still in a research phase. The
implementation of the Water Framework Directive in
Europe has stimulated the development of OpenMI
(European Open Modelling Interface and Environment).
OpenMI will simplify the linking of water-related mod-
els that will be used in the strategic planning required by
the Water Framework Directive (Gijsbers et al., 2002).
An initiative in the United States aims to establish a sim-
ilar framework for Environmental Models (Whelan and
Nicholson, 2002).
4.2.3. Example of a DSS for River Flood Management
In the Netherlands the flood management policy on the
Rhine Branches is aimed at reducing flood stages, since
further raising of the dyke system is judged as unsustain-
able in the long term. Possible measures to reduce flood
stages include the removal of hydraulic obstacles, lower-
ing of groins, widening of the low-flow channel, lowering
of floodplains, setting back of dykes, construction of side
WEAP: Weaping River Basin
Area Edit View Schematic General Help
River
Diversion
Reservoir
Groundwater
Other Local Supply
Demand Site
Transmission Link
Wastewater Treatment PI
Return Flow
Run of River Hydro
Flow Requirement
Streamflow Gauge
Rivers
Counties
Schematic
Data
Results
Overviews
Notes
Area: Weaping River Basin Schematic View Registered to: Daniel P. Loucks, Cornell University
North
Aquifer
Industry
North (2)
Agriculture
North (3) North
Reservoir
(1)
Weaping
River
Central
Reservoir
West
Aquifer
West
City (1)
(1)
(1)
Agriculture
West (3)
West
WW TP
South City
WW TP
South
City (1)
Industry
East (2)
Grey
River
Grey Hydro
Blue
River
Figure 2.8. The main interface of the
WEAP program, which is typical of a
variety of generic river basin models
that are able to simulate any river
system drawn into the computer and
displayed on the computer terminal,
as shown.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 51
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
channels, detention basins and other measures (as
described in more detail in Appendix D). These options
are illustrated in Figure 2.9.
Determining which set of river improvement measures
to implement involves a complex process of public
decision-making that includes many stakeholders.
Exploratory investigations along the river have identified
over 600 possible improvement measures. Which of
these alternatives should be chosen? A decision needs
to be made, and it needs to be acceptable to at least the
majority of stakeholders. As this is being written this
decision-making process is taking place. It is benefiting
from the use of online decision support that provides
information on the flood levels resulting from combina-
tions of measures along the river. This relatively simple
and user-friendly decision support system is called a
Planning Kit. (This kit is available from Delft Hydraulics.)
The preliminary design phase of this scheme, called
Room for the Rhine Branches, consists to a large extent
of bottom-up public decision-making processes.
Starting from the notion that multiple usage of space
located between the river dykes (i.e., the area between
the main embankments) should be possible. On the
basis of a number of exploratory studies that identify
possible measures and their respective effects, stake-
holders and local authorities are to identify their
preferred plans. These are to be judged on a number of
criteria, such as the flood conveyance capacity of the
river, its navigability, and its impact on the landscape
and ecological infrastructure. The envisaged result of
this procedure is an outline of a coherent scheme of
river improvement.
52 Water Resources Systems Planning and Management
The Planning Kit is developed for online decision
support and to facilitate a public discussion as well as
one among professionals in the planning and prelimi-
nary design phases.
Because of the large number of options and stake-
holders, the selection process is complicated. However,
reaching a technical optimum is not the objective. All of
the rivers functions, including its impact on the basins
ecology and cultural heritage, have to be respected. Public
acceptance is an essential requirement. In the meantime,
a number of overall criteria have to be satisfied. Without
further support from a variety of models incorporated
within the Planning Kit, this decision-making process
would be much less directed or focused, and hence much
less effective.
Numerical models of the Rhine exist. They vary in
scale level (from basin-wide down to local scale) and in
sophistication (1-D cross-sectionally averaged, 2-D
depth-averaged, 2-D or 3-D eddy-resolving, etc.). In the
studies in the framework of Room for the Rhine
Branches, a 1-D model of the Lower Rhine is used for
the large-scale phenomena and morphological computa-
tions and a 2-D depth-averaged model for more detailed
local computations. Flood-level computations are
made with a 2-D depth-averaged model of the entire
Rhine Branches.
These models are being intensively used in the explo-
ration and design phases of the river improvement works.
They provide help in setting the target design water levels
in order to avoid dyke raising, in checking the safety of
the flood defences, and in assessing the hydraulic and
morphological effects of proposed measures.
high water
8
1
2
3
4
lowering groines
deepening low flow channel
removing hydraulic obstacles
lowering flood plains
E
0
2
1
0
0
1
d
-
-
-
-
5
6
7
8
setting back dykes locally
setting back dykes globally
using detention reservoirs
reducing lateral inflow
-
-
-
-
7
1
1
2
low water
5, 6
3
4
Figure 2.9. River improvement
measures (see Appendix D).
wrm_ch02.qxd 8/31/2005 11:49 AM Page 52
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 53
The flood level computations with the 2-D model
for the Rhine Branches are too time-consuming to be
performed online during a meeting or a brainstorming
session. Therefore, model runs for each of the suggested
measures have been performed beforehand, and the water-
level effects of individual measures have been stored in a
database. This database is the core of the Planning Kit.
The basic assumption underlying the Planning Kit
approach is that, as a first-order approximation, the
water-level effects of each individual measure can be sep-
arated from those of every other measure, and the effects
of combinations of measures can be obtained by super-
position of their individual effects. At first sight, this may
seem a disputable approach, since the hydrodynamic
equations are essentially non-linear. Indeed, the total
water-level effect of two combined measures may easily
be 50% higher or lower than the sum of the two individ-
ual effects. But for large sets of measures (say, more than
twenty-five over a 100-km river stretch) this approach
has proven to be quite acceptable. As exemplified by
Figure 2.10, a fully non-linear 2-D hydrodynamic model
computation for a combination of forty measures along
the Rhine Branch Waal results in water levels which are at
most 10 cm lower than the results of the Planning Kit,
in which water-level effects of individual measures are
simply totalled.
The database approach allows for an online presentation
of effects of measures. With the super-position principle, the
design water-level effects of any combination of measures
can be composed from those of each individual measure.
For each of the proposed measures, the database con-
tains a situation sketch or an aerial photograph, ground
photographs, the effects on the longitudinal water surface
profile under the design flood conditions, the area
covered, a cost estimate, the length of dyke to be rebuilt,
ecological effects, amounts of material to be excavated for
various soil types and so on. Thus, when analysing and
synthesizing a set of measures, the aspects relevant to the
decision-making can be immediately provided.
The central element in the presentation of this
information (Figure 2.11) is a diagram showing the water-
surface profile as far as it exceeds the desired profile
according to the Room for the River principle (no dyke
reinforcement). By clicking on a measure, it is activated
and the water surface profile is adjusted accordingly.
Thus, one can see immediately how much more is needed
in order to reach the desired situation. By opening
windows connected to the name of the measure, one can
display photographs and any other available information.
The Planning Kit can easily be installed and run on a
PC or a laptop computer; hence, it can conveniently be
used online during meetings and hearings.
0.1
0.0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
-0.7
960 950 940 930 920 910 900 880 870 860 850 840 890
river kilometre (km)
w
a
t
e
r

l
e
v
e
l

e
f
f
e
c
t

(
m
)
E
0
4
0
4
0
7
b
2D hydrodynamic calculation
simulating measures simultaneously
Planning Kit method: adding up the
water level effect of individual measures
Figure 2.10. Water-level effect
of forty measures along the
Rhine Branch Waal under
design flood conditions as
calculated with a 2-D model
simulating all measures
(blue line) and as a result of
the Planning Kit method in
which results for individual
measures are simply summed
together.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 53
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
5. Conclusions
In our opinion, the most important aspect of model use
today is communication. Unless water resources planners
and managers can articulate well their needs for informa-
tion, it will be difficult for modellers to generate such
information. If the modellers cannot communicate effec-
tively their modelling assumptions and results, or how
others can use their tools to obtain their own results, then
little understanding will be gained from such models. Both
users and producers of modelling analyses must work
together to improve communication. This takes time,
patience and the willingness to understand what each has
to say has well as the real meaning behind what is said.
To expect everyone to communicate effectively and
to understand one another fully may be asking too
much. There is a story written in the Bible (Genesis;
Chapter 11, Verses 19) that tells us of a time when
everyone on the earth was together and spoke one
language. It seems these people decided to build a tower
whose top may reach into the heaven. Apparently
this activity came to the attention of the Lord, who for
some reason did not like this tower-building idea. So,
54 Water Resources Systems Planning and Management
according to the Bible, the Lord came down to earth
and confounded the peoples language so they could not
understand one another. They could no longer work
together to build their tower.
Is it any wonder we have to work so hard to commu-
nicate more effectively with one another, even in our
single, but multidisciplinary, field of water resources
planning and management? Let all of us modellers or
analysts, planners and managers work together to build a
new tower of understanding. To do this we need to
control our jargon and take the time to listen, com-
municate and learn from each other and from all of our
experiences.
Those who are involved in the development of water
resources systems modelling methodology know that
the use of these models cannot guarantee development
of optimal plans for water resources development
and management. Given the competing and changing
objectives and priorities of different interest groups, the
concept of an optimal plan is not very realistic. What
modellers can do, however, is to help define and
evaluate, in a rather detailed manner, numerous alterna-
tives that represent various possible compromises among
Figure 2.11. Main screen of the
Planning Kit. The small
geometric shapes are
alternatives at various locations
on the selected and displayed
river reach. The plot shows the
water levels associated with
selected alternatives.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 54
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Resource Systems Modelling 55
conflicting groups, values and management objectives. A
rigorous and objective analysis should help to identify the
possible tradeoffs among quantifiable objectives so that
further debate and analysis can be more informed. The art
of modelling is to identify those issues and concerns that
are important and significant and to structure the analysis
to shed light on these issues.
Although water resources planning and management
processes are not restricted to mathematical modelling,
modelling is an important part of those processes.
Models can represent in a fairly structured and ordered
manner the important interdependencies and interac-
tions among the various control structures and users
of a water resources system. Models permit an evalua-
tion of the consequences of alternative engineering
structures, of various operating and allocating policies,
and of different assumptions regarding future supplies,
demands, technology, costs, and social and legal require-
ments. Although models cannot define the best objec-
tives or set of assumptions, they can help identify the
decisions that best meet any particular objective and
assumptions.
We should not expect, therefore, to have the precise
results of any quantitative systems study accepted and
implemented. A measure of the success of any systems
study resides in the answer to the following questions:
Did the study have a beneficial impact in the planning
and decision-making process? Did the results of such
studies lead to a more informed debate over the proper
choice of alternatives? Did it introduce competitive alter-
natives that otherwise would not have been considered?
There seems to be no end of challenging water
resources systems planning problems facing water
resources planners and managers. How one models any
specific water resource problem depends on: first, the
objectives of the analysis; second, the data required to
evaluate the projects; third, the time, data, money and
computational facilities available for the analysis; and
fourth, the modellers knowledge and skill. Model devel-
opment is an art; it requires judgement both in abstract-
ing from the real world the components that are
important to the decision to be made and that can be illu-
minated by quantitative methods, and also in expressing
those components and their inter-relationships mathe-
matically in the form of a model. This art is to be intro-
duced in the next chapter (Chapter 3).
6. References
AHEARNE, J.F. 1988. Addressing public concerns in
science. Physics Today, Vol. 41, No. 9, pp. 3642.
ANDREU, J.; CAPILLA, J. and SANCHIS, E. 1991. AQUA-
TOOL. A computer assisted support system for water
resources research management including conjunctive use.
In: D.P. Loucks and J.R. da Costa (eds.), Decision support
systems, Berlin, Springer-Verlag, pp. 33355.
AUSTIN, T.A. 1986. Utilization of models in water
resources. Water Resources Bulletin, Vol. 22, No. 1, pp. 4956.
BORSUK, M.E.; STOW, C.A.; HIGDON, D. and
RECKHOW, K.H. 2001. A Bayesian hierarchical model
to predict benthic oxygen demand from organic matter
loading in estuaries and coastal zones. Ecological
Modelling, No. 143, pp. 16581.
CAREY, W.D. 1988. Scientists and sandboxes: regions of
the mind. American Scientist, No. 76, pp. 1435.
DANISH HYDRAULIC INSTITUTE (DHI). 1997. MIKE-
BASIN: operating manual and description. Denmark,
Hrsholm.
DELFT HYDRAULICS. 2004. RIBASIM River basin simula-
tion program operating manual and description. Netherlands,
Delft.
FEDRA, K. 1992. Advanced computer applications. Options,
December.
GASS, S.I. 1990. Model world: 1-lave model, will travel.
Interfaces, Vol. 20, No. 2, pp. 6771.
GEORGAKAKOS, A.P. and MARTIN, Q.W. (eds.). 1996.
An international review of decision support systems in
river basin operation. Proceedings of the Fifth Water
Resources Operations Management Workshop, March
Arlington, Va., ASCE.
GIJSBERS, P.J.A.; MOORE, R.V. and TINDALL, C.I.
2002. Hamonil T: towards OMI, an open modelling interface
and environment to harmonise European developments in
water related simulation software. Delft, the Netherlands,
Delft Hydraulics (internal paper).
HAMLET, A. et al. 1996a. Simulating basinwide alternatives
using the ACT-ACF shared vision models. Mobil District,
Mobil, Al., USACE.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 55
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
HAMLET, A. et al. 1996b. A history of shared vision model-
ling in the ACT-ACF comprehensive study. Mobil District,
Mobil, Al., USACE.
HAMLET, A. et al. 1996c. Basic STELLA II users manual for
the ACT-ACF shared vision models. Mobil District, Mobil,
Al., USACE.
KINDLER, J. 1987. Systems analysis approach to water
resources management: state of the art. Symposium
on recent developments and perspectives in systems
analysis in water resources management. Perugia, Italy,
WARREDOC.
KINDLER, J. 1988. Systems analysis and water resources
planning. Fourth IFAC international symposium on systems
analysis applied to management of water resources. Rabat,
Morocco, IFAC, 11 to 13 October.
LOUCKS, D.P. 1995. Developing and implementing
decision support systems: a critique and a challenge.
Water Resources Bulletin, Journal of American
Water Resources Association, Vol. 31, No. 4, August, pp.
57182.
LOUCKS, D.P. and DA COSTA, J.R. (eds.). 1991. Decision
support systems. NATO Series G. Vol. 26. Berlin, Springer-
Verlag.
LOUCKS, D.P.; STEDINGER, J.R. and SHAMIR, U. 1985.
Modelling water resources systems: issues and experi-
ences. Civil Engineering Systems, No. 2, pp. 22331.
MAJONI, H. and QUADE, E.S. 1980. Pitfalls of analysis.
New York, John Wiley.
MISER, H.J. 1980. Operations research and systems
analysis. Science, No. 209, pp. 17482.
OCALLAGHAN, J.R. (ed.) 1996. The interaction of eco-
nomic, ecology and hydrology, Chapman & Hall, London.
PALMER, R.N.; KEYS, A.M. and FISHER, S. 1995.
Empowering stakeholders through simulation in water
resources planning. Proceedings of the 20th Conference on
56 Water Resources Systems Planning and Management
Water Resources Planning and Management Division. New
York, ASCE.
POOL, R. 1990. Struggling to do science for society. Sci.,
No. 248, pp. 67273.
RASKIN, P.; SIEBER, J. and HUBER-LEE, A. 2001. Water
evaluation and planning system: user guide for WEAP21.
Boston, Mass., Tellus Institute.
REYNOLDS, P.J. 1987. Future research needs in systems
analysis and to application to water resources manage-
ment. International symposium on recent developments and
perspectives in systems analysis in water resources manage-
ment. Perugia, Italy, WARREDOC.
ROGERS, P.P. and FIERING, M.B. 1986. Use of systems
analysis in water management. Water Resources Research,
Vol. 22, No. 9, pp. 14658S.
SHAPIRO, H.T. 1990. The willingness to risk failure.
Science, Vol. 250, No. 4981, p. 609.
SIMON, I-I.A. 1998. Prediction and prescription in system
modelling. 15th Anniversary of IIASA, International Institute
for Applied Systems Analysis. Laxenburg, Austria, IIASA.
STOKEY, E. and ZECKHAUSER, R. 1977. A primer for
policy analysis. New York, W.W. Norton.
TOMLISON, R. 1980. Doing something about the future.
Journal of the Operational Research Society, No. 31,
pp. 46776.
WALKER, W.E. 1987. Important hard problems in public
policy analysis. Report P-7282. Santa Monica, Calif., Rand
Corp.
WERICK, W.J.; WHIPPLE, W. Jr. and LUND, J. 1996. ACT-
ACF basinwide study. Mobil District, Mobil, Al., USACE.
WHELAN, G. and NICHOLSON, T.J. (eds.). 2002.
Proceedings of the environmental software systems compabil-
ity and linkage workshop, March 79, 2000. NRC report
NUREG/CP-0177. Richl and, Wash., Pacific Northwest
National Laboratory.
wrm_ch02.qxd 8/31/2005 11:49 AM Page 56
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
3. Modelling Methods for Evaluating Alternatives
1. Introduction 59
1.1. Model Components 60
2. Plan Formulation and Selection 61
2.1. Plan Formulation 61
2.2. Plan Selection 63
3. Modelling Methods: Simulation or Optimization 64
3.1. A Simple Planning Example 65
3.2. Simulation Modelling Approach 66
3.3. Optimization Modelling Approach 66
3.4. Simulation Versus Optimization 67
3.5. Types of Models 69
3.5.1. Types of Simulation Models 69
3.5.2. Types of Optimization Models 70
4. Model Development 71
5. Managing Modelling Projects 72
5.1. Create a Model Journal 72
5.2. Initiate the Modelling Project 72
5.3. Selecting the Model 73
5.4. Analysing the Model 74
5.5. Using the Model 74
5.6. Interpreting Model Results 75
5.7. Reporting Model Results 75
6. Issues of Scale 75
6.1. Process Scale 75
6.2. Information Scale 76
6.3. Model Scale 76
6.4. Sampling Scale 76
6.5. Selecting the Right Scales 76
7. Conclusions 77
8. References 77
wrm_ch03.qxd 8/31/2005 11:50 AM Page 58
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Water resources system planners must identify and
evaluate alternative water resources system designs or
management plans on the basis of their economic, eco-
logical, environmental, and social or political impacts.
One important criterion for plan identification and evalu-
ation is the economic benefit or cost a plan would entail
were it to be implemented. Other criteria can include the
extent to which any plan meets environmental, ecological
and social targets. Once planning or management per-
formance measures (objectives) and various general alter-
natives for achieving desired levels of these performance
measures have been identified, models can be developed
and used to help identify specific alternative plans that
best meet those objectives.
Some system performance objectives may be in
conflict, and in such cases models can help identify the
efficient tradeoffs among these conflicting measures of
system performance. These tradeoffs indicate what com-
binations of performance measure values can be obtained
from various system design and operating policy variable
values. If the objectives are the right ones (that is, they are
59
Modelling Methods
for Evaluating Alternatives
Water resources systems are characterized by multiple interdependent
components that together produce multiple economic, environmental, ecological
and social impacts. Planners and managers working to improve the performance
of these complex systems must identify and evaluate alternative designs and
operating policies, comparing their predicted performance with the desired goals
or objectives. These alternatives are defined by the values of numerous design,
target and operating policy variables. Constrained optimization together with
simulation modelling is the primary way we have of estimating the values of the
decision variables that will best achieve specified performance objectives. This
chapter introduces these optimization and simulation methods and describes what
is involved in developing and applying them in engineering projects.
3
what the stakeholders really care about), such quantita-
tive tradeoff information should be of value during the
debate over what decisions to make.
Regional water resources development plans designed
to achieve various objectives typically involve investments
in land and infrastructure. Achieving the desired eco-
nomic, environmental, ecological and social objective
values over time and space may require investments in
storage facilities, including surface or groundwater
reservoirs and storage tanks, pipes, canals, wells, pumps,
treatment plants, levees and hydroelectric generating
facilities, or in fact the removal of some of them.
Many capital investments can result in irreversible
economic and ecological impacts. Once the forest of a
valley is cleared and replaced by a lake behind a dam, it
is almost impossible to restore the site to its original
condition. In parts of the world where river basin or
coastal restoration activities require the removal of engi-
neering structures, water resources engineers are learning
just how difficult and expensive that effort can be.
The use of planning models is not going to eliminate
the possibility of making mistakes. These models can,
however, better inform. They can provide estimates of the
wrm_ch03.qxd 8/31/2005 11:50 AM Page 59
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
different impacts associated with, say, a natural unregu-
lated river system and a regulated river system. The
former can support a healthier ecosystem that provides a
host of flood protection and water quality enhancement
services. The latter can provide more reliable and cheaper
water supplies for off-stream users and increased
hydropower and some protection from at least small
floods for those living on flood-prone lands. In short,
models can help stakeholders assess the future conse-
quences, the benefits and costs, and a multitude of other
impacts associated with alternative plans or management
policies.
This chapter introduces some mathematical optimiza-
tion and simulation modelling approaches commonly
used to study and analyse water resources systems. The
modelling approaches are illustrated by their application
to some relatively simple water resources planning and
management problems. The purpose here is to introduce
and compare some commonly used methods of (or
approaches to) modelling. This is not a text on the state-
of-the-art of optimization or simulation modelling. In
subsequent chapters of this book, more details will be
given about optimization models and simulation meth-
ods. More realistic and more complex problems usually
require much bigger and more complex models than
those developed in this book, but these bigger and more
complex models are often based on the principles and
techniques introduced here.
Regardless of the problem complexity or size, the mod-
elling approaches are the same. Thus, the emphasis here is
on the art of model development: just how one goes about
constructing a model that will provide information needed
to solve a particular problem, and various ways models
might be solved. It is unlikely anyone will ever use any of
the specific models developed in this or other chapters,
simply because they will not be solving the specific exam-
ples used to illustrate the different approaches to model
development and solution. However, it is quite likely
that water resources managers and planners will use the
modelling approaches and solution methods presented
in this book to analyse similar types of problems. The
particular problems used here, or any others that could
have been used, can be the core of more complex models
addressing more complex problems in practice.
Water resources planning and management today is
dominated by the use of predictive optimization and
60 Water Resources Systems Planning and Management
simulation models. While computer software is becoming
increasingly available for solving various types of opti-
mization and simulation models, no software currently
exists that will build those models themselves. What and
what not to include and assume in models requires
judgement, experience and knowledge of the particular
problem being addressed, the system being modelled and
the decision-making environment. Understanding the
contents of, and performing the exercises for, this chapter
will be a first step towards gaining some judgement and
experience in model development.
1.1. Model Components
Mathematical models contain algebraic equations. These
equations include variables that are assumed to be known
and others that are unknown and to be determined.
Known variables are usually called parameters, and
unknown variables are called decision variables. Models
are developed for the primary purpose of identifying the
best values of the latter. These decision variables can
include design and operating policy variables of various
water resources system components.
Design variables can include the active and flood
storage capacities of reservoirs, the power generating
capacity of hydropower plants, the pumping capacity of
pumping stations, the efficiencies of wastewater treatment
plants, the dimensions or flow capacities of canals and
pipes, the heights of levees, the hectares of an irrigation
area, the targets for water supply allocations and so on.
Operating variables can include releases of water from
reservoirs or the allocations of water to various users over
space and time. Unknown decision variables can also
include measures of system performance, such as net
economic benefits, concentrations of pollutants, ecological
habitat suitability values or deviations from particular
ecological, economic or hydrological targets.
Models describe, in mathematical terms, the system
being analysed and the conditions that the system has
to satisfy. These conditions are often called constraints.
Consider, for example, a reservoir serving various water
supply users downstream. The conditions included in a
model of this reservoir would include the assumption that
water will flow in the direction of lower heads (that is,
downstream unless it is pumped upstream), and the
volume of water stored in a reservoir cannot exceed the
wrm_ch03.qxd 8/31/2005 11:50 AM Page 60
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 61
reservoirs storage capacity. Both the storage volume over
time and the reservoir capacity might be unknown. If
the capacity is known or assumed, then it is among the
known model parameters.
Model parameter values, while assumed to be known,
can often be uncertain. The relationships between
various decision variables and assumed known model
parameters (i.e., the model itself) may be uncertain. In
these cases the models can be solved for a variety of
assumed conditions and parameter values. This provides
an estimate of just how important uncertain parameter
values or uncertain model structures are with respect to
the output of the model. This is called sensitivity analysis.
Sensitivity analyses will be discussed in Chapter 9 in
much more detail.
Solving a model means finding values of its unknown
decision variables. The values of these decision variables
can define a plan or policy. They can also define the costs
and benefits or other measures of system performance
associated with that particular management plan or
policy.
While the components of optimization and simulation
models include system performance indicators, model
parameters and constraints, the process of model devel-
opment and use includes people. The drawing shown in
Figure 3.1 illustrates some interested stakeholders busy
studying their river basin, in this case perhaps with the
use of a physical model. Whether a mathematical model
or physical model is being used, one important consider-
ation is that if the modelling exercise is to be of any value,
it must provide the information desired and in a form that
the interested stakeholders can understand.
2. Plan Formulation and Selection
Plan formulation can be thought of as assigning particular
values to each of the relevant decision variables. Plan
selection is the process of evaluating alternative plans and
selecting the one that best satisfies a particular objective
or set of objectives. The processes of plan formulation and
selection involve modelling and communication among
all interested stakeholders, as the picture in Figure 3.1
suggests.
The planning and management issues being discussed
by the stakeholders in the basin pictured in Figure 3.1
could well include surface and groundwater water alloca-
tions, reservoir operation, water quality management and
infrastructure capacity expansion.
2.1. Plan Formulation
Model building for defining alternative plans or policies
involves a number of steps. The first is to clearly specify
the issue or problem or decision(s) to be made. What are
the fundamental objectives and possible alternatives?
Such alternatives might require defining allocations of
water to various water users, the level of wastewater treat-
ment, the capacities and operating rules of multipurpose
reservoirs and hydropower plants, and the extent and
reliability of floodplain protection from levees. Each of
these decisions may affect system performance criteria or
objectives. Often these objectives include economic meas-
ures of performance, such as costs and benefits. They may
also include environmental and social measures not
Figure 3.1. These stakeholders have an interest in how their
watershed or river basin is managed. Here they are using a
physical model to help them visualize and address planning
and management issues. Mathematical models often replace
physical models, especially for planning and management
studies. (Reprinted with permission from Engineering
News-Record, copyright The McGraw-Hill Companies, Inc.,
September 20, 1993. All rights reserved.).
wrm_ch03.qxd 8/31/2005 11:50 AM Page 61
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
expressed in monetary units. (More detail on performance
criteria is contained in Chapter 10.)
To illustrate this plan formulation process, consider
the problem of designing a tank to hold a specific amount
of water. The criterion to be used to compare different
feasible designs is cost. The goal in this example is to find
the least-cost shape and dimensions of a tank that will
hold a specified volume, say V, of water.
The model of this problem must somehow relate
the unknown design variable values to the cost of the
tank. Assume, for example, a rectangular tank shape. The
design variables are the length, L, width, W, and height,
H, of the tank. These are the unknown decision variables.
The objective is to find the combination of L, W, and H
values that minimizes the total cost of providing a tank
capacity of at least V units of water. This volume V will be
one of the model parameters. Its value is assumed known
even though in fact it may be unknown and dependent in
part on the cost.
The cost of the tank will be the sum of the costs of
the base, the sides and the top. These costs will depend
on the area of the base, sides and top. The costs per
unit area may vary depending on the values of L, W and
H; however, even if those cost values depend on the
values of those decision variables, given any specific
values for L, W and H, one can define an average
cost-per-unit area. Here we will assume these average
costs per unit area are known. They can be adjusted if
they turn out to be incorrect for the derived values of L,
W and H.
These average unit costs of the base, sides and top will
probably differ. They can be denoted as C
base
, C
side
and
C
top
respectively. These unit costs together with the tanks
volume, V, are the parameters of the model. If L, W, and
H are measured in metres, then the areas will be
expressed in units of square metres and the volume will
be expressed in units of cubic metres. The average unit
costs will be expressed in monetary units per square
metre.
The final step of model building is to specify all the
relations among the objective (cost), function and deci-
sion variables and parameters, including all the condi-
tions that must be satisfied. It is often wise to first state
these relationships in words. The result is a word model.
Once that is written, mathematical notation can be
defined and used to construct a mathematical model.
62 Water Resources Systems Planning and Management
The word model for this tank design problem is to
minimize total cost where:
Total cost equals the sum of the costs of the base, the
sides and the top.
Cost of the sides is the cost-per-unit area of the sides
times the total side area.
Cost of the base is the cost-per-unit area of the base
times the total base area.
Cost of the top is the cost-per-unit area of the top
times the total top area.
The volume of the tank must at least equal some spec-
ified volume capacity.
The volume of the tank is the product of the length,
width and height of the tank.
Using the notation already defined, and combining some
of the above conditions, a mathematical model can be
written as:
Minimize Cost (3.1)
Subject to:
Cost (C
base
C
top
)(LW) 2(C
side
) (LH WH) (3.2)
LWH V (3.3)
Equation 3.3 permits the tanks volume to be larger than
that required. While this is allowed, it will cost more if the
tanks capacity is larger than V, and hence the least-cost
solution of this model will surely show that LWH will
equal V. In practice, however, there may be practical,
legal and/or safety reasons why the decisions with respect
to L, W and H may result in a capacity that exceeds the
required volume, V.
This model can be solved a number of ways, which
will be discussed later in this and the next chapters. The
least-cost solution is
W L [2C
side
V/(C
base
C
top
)]
1/3
(3.4)
and H V/[2C
side
V/(C
base
C
top
)]
2/3
(3.5)
or H V
1/3
[(C
base
C
top
)/2C
side
]
2/3
(3.6)
The modelling exercise should not end here. If there is
any doubt about the value of any of the parameters, a sen-
sitivity analyses should be performed on those uncertain
parameters or assumptions. In general these assumptions
could include the values of the cost parameters (e.g., the
costs-per-unit area) as well as the relationships expressed
in the model (that is, the model itself). How much does
wrm_ch03.qxd 8/31/2005 11:50 AM Page 62
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 63
the total cost change with respect to a change in any of the
cost parameters or with the required volume V? How
much does any decision-variable change with respect
to changes in those parameter values? What is the percent
change in a decision-variable value given a unit percent
change in some parameter value (what economists call
elasticity)?
If indeed the decision-variable values do not change
significantly with respect to a change in the value of an
uncertain parameter value, there is no need to devote
more effort to reducing that uncertainty. Any time and
money available for further study should be directed
toward those parameters or assumptions that substan-
tially influence the models decision-variable values.
This capability of models to help identify what data are
important and what data are not can guide monitoring
and data collection efforts. This is a beneficial attribute of
modelling often overlooked.
Continuing with the tank example, after determining,
or estimating, all the values of the model parameters and
then solving the model to obtain the cost-effective values
of L, W and H, we now have a design. It is just one of a
number of designs that could be proposed. Another
design might be for a cylindrical tank having a radius and
height as decision-variables. For the same volume V and
unit area costs, we would find that the total cost is less,
simply because the areas of the base, side and top are less.
We could go one step further and consider the possibility
of a truncated cone, having different bottom and top
radii. In this case both radii and the height would be the
decision-variables. But whatever the final outcome of our
modelling efforts, there might be other considerations or
criteria that are not expressed or included in the model
that might be important to those responsible for plan
(tank design) selection.
2.2. Plan Selection
Assume P alternative plans (e.g., tank designs) have
been defined, each designated by the index p. For each
plan, there exist n
p
decision variables x
j
p
indexed with
the letter j. Together these variables and their values,
expressed by the vector X
p
, define the specifics of the
pth plan. The index j distinguishes one decision-variable
from another, and the index p distinguishes one plan
from another. The task at hand, in this case, may be to
find the particular plan p, defined by the known values
of each decision-variable in the vector X
p
, that maximizes
the present value of net benefits, B(X
p
), derived from
the plan.
Assume for now that an overall performance objective
can be expressed mathematically as:
maximize B(X
p
) (3.7)
The values of each decision-variable in the vector X
p
that
meet this objective must be feasible; in other words, they
must meet all the physical, legal, social and institutional
constraints.
X
p
feasible for all plans p. (3.8)
There are various approaches to finding the best plan or
best set of decision-variable values. By trial and error, one
could identify alternative plans p, evaluate the net bene-
fits derived from each plan, and select the particular plan
whose net benefits are a maximum. This process could
include a systematic simulation of a range of possible
solutions in a search for the best. When there is a large
number of feasible alternatives that is, many decision-
variables and many possible values for each of them it
may no longer be practical to identify and simulate all
feasible combinations of decision-variable values, or even
a small percentage of them. It would simply take too long.
In this case it is often convenient to use an optimization
procedure.
Equations 3.7 and 3.8 represent a discrete optimiza-
tion problem. There are a finite set of discrete alternatives.
The set could be large, but it is finite. The tank problem
example is a continuous optimization problem having,
at least mathematically, an infinite number of feasible
solutions. In this case optimization involves finding
feasible values of each decision-variable x
j
in the set of
decision variables X that maximize (or minimize) some
performance measure, B(X). Again, feasible values are
those that satisfy all the model constraints. A continuous
constrained optimization problem can be written as:
maximize B(X) (3.9)
X feasible (3.10)
While maximization of Equation 3.7 requires a comparison
of B(X
p
) for every discrete plan p, the maximization of
Equation 3.9, subject to the feasibility conditions required
in Equation 3.10, by complete enumeration is impossible.
If there exists a feasible solution in other words, at least
wrm_ch03.qxd 8/31/2005 11:50 AM Page 63
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
one that satisfies all the constraints mathematically there
are likely to be an infinite number of possible feasible
solutions or plans represented by various values of the
decision-variables in the vector X.
Finding by trial and error the values of the vector X
that maximizes the objective Equation 3.9 and at the same
time meet all the constraints is often difficult. Some type
of optimization procedure, or algorithm, is useful in such
cases. Mathematical optimization methods are designed
to make this search for the best solution (or better solu-
tions) more efficient. Optimization methods are used to
identify those values of the decision-variables that satisfy
specified objectives and constraints without requiring
complete enumeration.
While optimization models might help identify the
decision-variable values that will produce the best plan
directly, they are based on all the assumptions incorpo-
rated in the model. Often these assumptions are limiting.
In these cases the solutions resulting from optimization
models should be analysed in more detail, perhaps
through simulation methods, to improve the values of the
decision-variables and to provide more accurate estimates
of the impacts associated with those decision-variable
values. In these situations, optimization models are used
for screening out the clearly inferior solutions, not for
finding the very best one. Just how screening is performed
using optimization models will be discussed in the next
chapter.
The values that the decision-variables may assume are
rarely unrestricted. Usually various functional relation-
ships among these variables must be satisfied. This is what
is expressed in constraint Equations 3.8 and 3.10. For
example, the tank had to contain a given amount of water.
In a water-allocation problem, any water allocated to and
completely consumed by one user cannot simultaneously
or subsequently be allocated to another user. Storage
reservoirs cannot store more water than their maximum
capacity. Technological restrictions may limit the capaci-
ties and sizes of pipes, generators and pumps to those
commercially available. Water quality concentrations
should not exceed those specified by water quality
standards or regulations. There may be limited funds
available to spend on water resources development
projects. These are a few examples of physical, legal and
financial conditions or constraints that may restrict the
ranges of variable values in the solution of a model.
64 Water Resources Systems Planning and Management
Equations or inequalities can generally express any
physical, economic, legal or social restrictions on the
values of the decision-variables. Constraints can also
simply define relationships among decision-variables.
For example, Equation 3.2 above defines a new decision-
variable called Cost as a function of other decision-
variables and model parameters.
In general, constraints describe in mathematical terms
the system being analysed. They define the system com-
ponents and their inter-relationships, and the permissible
ranges of values of the decision-variables, either directly
or indirectly.
Typically, there exist many more decision-variables
than constraints, and hence, if any feasible solution exists,
there may be many such solutions that satisfy all the con-
straints. The existence of many feasible alternative plans is
a characteristic of most water resources systems planning
problems. Indeed it is a characteristic of most engineering
design and operation problems. The particular feasible
solution or plan that satisfies the objective function that
is, that maximizes or minimizes it is called optimal. It is
the optimal solution of the mathematical model, but it
may not necessarily be considered optimal by any decision-
maker. What is optimal with respect to some model may
not be optimal with respect to those involved in a plan-
ning or decision-making process. To repeat what was
written in Chapter 2, models are used to provide infor-
mation (useful information, one hopes), to the decision-
making process. Model solutions are not replacements for
individuals involved in the decision-making process.
3. Modelling Methods: Simulation
or Optimization
The modelling approach discussed in the previous section
focused on the use of optimization methods to identify
the preferred design of a tank. Similar methods can be
used to identify preferred design-variable values and oper-
ating policies for multiple reservoir systems, for example.
Once these preferred designs and operating policies have
been identified, unless there is reason to believe that a
particular alternative is really the best and needs no fur-
ther analysis, each of these preferred alternatives can be
further evaluated with the aid of more detailed and robust
simulation models. Simulation models address what if
wrm_ch03.qxd 8/31/2005 11:50 AM Page 64
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 65
questions: What will likely happen over time and at one
or more specific places if a particular design and/or
operating policy is implemented?
Simulation models are not limited by many of the
assumptions incorporated into optimization models. For
example, the inputs to simulation models can include a
much longer time series of hydrological, economic and
environmental data such as rainfall or streamflows,
water supply demands, pollutant loadings and so on
than would likely be included in an optimization model.
The resulting outputs can better identify the variations
of multiple system performance indicator values:
that is, the multiple hydrological, ecological, economic
and environmental impacts that might be observed over
time, given any particular system design and operating
policy.
Simulating multiple sets of values defining the designs
and operating policies of a water resources system can
take a long time. Consider, for example, only 30 decision-
variables whose best values are to be determined. Even if
only two values are assumed for each of the 30 variables,
the number of combinations that could be simulated
amounts to 2
30
or in excess of 10
9
. Simulating and com-
paring even 1% of these billion at a minute per simulation
amounts to over twenty years, continuously and without
sleeping. Most simulation models of water resources
systems contain many more variables and are much more
complex than this simple 30-binary-variable example. In
reality there could be an infinite combination of feasible
values for each of the decision-variables.
Simulation works when there are only a relatively few
alternatives to be evaluated, not when there are a large
number of them. The trial and error process of simulation
can be time consuming. An important role of optimiza-
tion methods is to reduce the number of alternatives
for simulation analyses. However, if only one method
of analysis is to be used to evaluate a complex water
resources system, simulation together with human judge-
ment concerning which alternatives to simulate is often,
and rightly so, the method of choice.
Simulation can be based on either discrete events or
discrete time periods. Most simulation models of water
resources systems are designed to simulate a sequence of
discrete time periods. In each discrete time period, the
simulation model converts all the initial conditions and
inputs to outputs. The duration of each period depends in
part on the particular system being simulated and the
questions being addressed.
3.1. A Simple Planning Example
Consider the case of a potential reservoir releasing water
to downstream users. A reservoir and its operating policy
can increase the benefits each user receives over time by
providing increased flows during periods of otherwise
low flows relative to the user demands. Of interest is
whether or not the increased benefits the water users
obtain from an increased flow and more reliable
downstream flow conditions will offset the costs of the
reservoir. This water resources system is illustrated in
Figure 3.2.
Before this system can be simulated, one has to define
the active storage capacity of the reservoir and how much
water is to be released depending on the storage volume
and time period; in other words, one has to define the
reservoir operating policy. In addition, one must also
define the allocation policy: how much water to allocate
to each user and to the river downstream of the users
given any particular reservoir release.
For this simple illustration assume these operating and
allocation policies are as shown in Figure 3.3. Also for
simplicity assume they apply to each discrete time period.
The reservoir operating policy, shown as a red line in
Figure 3.3, attempts to meet a release target. If insufficient
water is available, all the water will be released in the time
period. If the inflow exceeds the target flow and the reser-
voir is at capacity, a spill will occur. This operating policy
is sometimes called the standard operating policy. It is
not usually followed in practice. Most operators, as
indeed specified by most reservoir operating policies, will
reduce releases in times of drought in an attempt to save
r
iv
e
r
Q
t
X
1t
user 1
user 2
user 3
X
2t X
3t
reservoir
E
0
2
0
1
0
8
w
Figure 3.2. Reservoir-water allocation system to be simulated.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 65
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
some water in the reservoir for future releases in case
of an extended period of low inflows. This is called a
hedging policy. Any reservoir release policy, including a
hedging policy, can be defined within the blue portion of
the release policy plot shown in Figure 3.3. The dashdot
line in Figure 3.3 is one such hedging function.
Once defined, any reservoir operating policy can be
simulated.
3.2. Simulation Modelling Approach
The simulation process for the three-user system shown
in Figure 3.2 proceeds from one time period to the next.
The reservoir inflow, obtained from a database, is added
to the existing storage volume, and a release is determined
from the release policy. Once the release is known, the
final storage volume is computed and this becomes the
66 Water Resources Systems Planning and Management
initial volume for the next simulation time period. The
reservoir release is then allocated to the three downstream
users and to the river downstream of those users as
defined by the allocation policy. The resulting benefits
can be calculated and stored in an output database.
Additional data pertaining to storage volumes, releases
and the allocations themselves can also be stored in the
output database, as desired. The process continues for the
duration of the simulation run. Then the output data can
be summarized for later comparison with other simula-
tion results based on other reservoir capacities and oper-
ation policies and other allocation policies. Figure 3.4
illustrates this simulation process.
It would not be too difficult to write a computer program
to carry out this simulation. In fact, it can be done on a
spreadsheet. However easy that might be for anyone familiar
with computer programming or spreadsheets, one cannot
expect it to be easy for many practicing water resources
planners and managers who are not doing this type of work
on a regular basis. Yet they might wish to perform a simula-
tion of their particular system, and to do it in a way that
facilitates changes in many of its assumptions.
Computer programs capable of simulating a wide vari-
ety of water resources systems are becoming increasingly
available. Simulation programs together with their inter-
faces that facilitate the input and editing of data and the
display of output data are typically called decision support
systems. Their input data define the components of the
water resources system and their configuration. Inputs
include hydrological data and design and operating policy
data. These generic simulation programs are now becom-
ing capable of simulating surface and groundwater water
flows, storage volumes and qualities under a variety of
system infrastructure designs and operating policies.
3.3. Optimization Modelling Approach
The simple reservoir-release and water-allocation plan-
ning example of Section 3.1 can also be described as an
optimization model. The objective remains that of maxi-
mizing the total benefits that the three users obtain from
the water that is allocated to them. Denoting each users
benefit as B
it
(i 1, 2, 3) for each of T time periods t, this
objective, expressed symbolically is to:
maximize total benefits (3.11) { }
B B B
t t t
t
T
1 2 3
+ +

initial storage and inflow

E
0
2
0
1
0
8
x
r
e
l
e
a
s
e
release target
storage capacity
0
0
s
p
i
l
l

=

r
e
l
e
a
s
e

o
n

a
r
e
a
s
o
f

t
a
r
g
e
t

r
e
l
e
a
s
e
zone of feasible release
h
e
d
g
in
g
reservoir release flow
E
0
2
0
1
0
8
y
a
l
l
o
c
a
t
i
o
n
10
8
6
4
2
0
0 2 4 6 8 10 12 14 16 18 20
user
river
user
user
3
Q
1
2
Figure 3.3. Policy defining the reservoir release to be made
as a function of the current storage volume and current inflow
and the reservoir release allocation policy for the river flow
downstream of the reservoir. The blue zone in the reservoir
release policy indicates the zone of feasible releases. It is
physically impossible to make releases represented by points
outside that blue zone.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 66
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 67
The benefit, B
it
, for each user i in each time period t
depends on the amount of water, X
it
, allocated to it.
These benefit functions, B
it
B
it
(X
it
), need to be
known and expressed in a form suitable for solution
using the particular optimization solution algorithm
selected. The unknown variables include the allocations,
X
it
, and associated reservoir releases R
t
for all periods
t 1, 2, 3, . . . , T. Assuming there is no significant incre-
mental runoff between the upstream reservoir and the
sites where water is diverted from the river, the amounts
allocated to all users, the sum of all X
it
in each period t,
cannot exceed the amount of water released from the
reservoir, R
t
, in the period. This is one of the optimiza-
tion model constraints:
R
t
X
1t
X
2t
X
3t
(3.12)
The remaining necessary constraints apply to the reser-
voir. A mass balance of water storage is needed, along
with constraints limiting initial storage volumes, S
t
, to the
capacity, K, of the reservoir. Assuming a known time-
series record of reservoir inflows, I
t
, in each of the time
periods being considered, the mass-balance or continuity
equations for storage changes in each period t can be
written:
S
t
I
t
R
t
S
t1
for t 1, 2, . . . , T;
If t T, then T1 1. (3.13)
The capacity constraints simply limit the unknown initial
storage volume, S
t
, to be no greater than the reservoir
capacity, K.
S
t
K for t 1, 2, . . . , T. (3.14)
A one-year analysis period with T12 time periods of one
month each in combination with three allocation vari-
ables, X
it
, a storage variable S
t
, and a release R
t
, variable in
each period t, includes a total of sixty unknown decision-
variables. The job of the optimization solution procedure
is to find the values of these sixty variables that will satisfy
the objective, Equation 3.11, that is to say, maximize total
benefits, and at the same time satisfy all of the thirty-six
constraint equations and inequalities as well.
In this example the reservoir inflows, I
t
, its storage capac-
ity, K, and each users benefit functions, B
it
, are assumed
known. In some cases such information is not known. Nor
were other purposes, such as hydropower, flood control,
water quality or recreation considered in this example, to
mention only a few possible extensions. Such conditions
and extensions will be considered in later chapters.
3.4. Simulation Versus Optimization
Unlike simulation models, the solutions of optimization
models are based on objective functions of unknown
decision-variables that are to be maximized or minimized.
E
0
2
0
1
0
8
z
stop end
data
storage
compute (t),
i
(t),
i
(t) and (t)
R
X B Q
file
t
S S
= 0
( ) = o 1
start
yes no
read (t)
S
S
(t+ ) =
(t) +
1
(t) (t) R
t t = +1
(t)
S(t)
R(t)
i
(t) X
Q(t)
B
i
(t)
- inflow
- initial storage volume
- reservoir release
- allocation to user i
- allocation to stream
- benefit for user i
- time period t
Figure 3.4. Flow diagram of the
reservoir-allocation system
simulation process. The
simulation terminates after
some predefined number of
simulation time steps.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 67
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The constraints of an optimization model contain
decision-variables that are unknown and parameters
whose values are assumed known. Constraints are
expressed as equations and inequalities. The tank model
(Equations 3.1, 3.2 and 3.3) is an example of an opti-
mization model. So is the reservoir water-allocation
model, Equations 3.11 to 3.14. The solution of an opti-
mization model, if one exists, contains the values of all
of the unknown decision-variables. It is mathematically
optimal in that the values of the decision-variables satisfy
all the constraints and maximize or minimize an objective
function. This optimal solution is of course based on the
assumed values of the model parameters, the chosen
objective function and the structure of the model itself.
The procedure (or algorithm) most appropriate for
solving any particular optimization model depends in
part on the particular mathematical structure of the
model. There is no single universal solution procedure
that will efficiently solve all optimization models. Hence,
model builders tend to model water resources systems
by using mathematical expressions that are of a form
compatible with one or more known solution procedures.
Approximations of reality, made to permit model solution
by a chosen optimization solution procedure (algorithm),
may justify a more detailed simulation to check and
improve on any solution obtained from that optimization.
Simulation models are not restricted to any particular form
of mathematics, and can define many relations including
those not easily incorporated into optimization models.
One of many challenges in the use of optimization
modelling is our inability to quantify and express
68 Water Resources Systems Planning and Management
mathematically all the planning objectives, the technical,
economic, and political constraints and uncertainties, and
other important considerations that will influence the
decision-making process. Hence at best a mathematical-
model of a complex water resources system is only an
approximate description of the real system. The optimal
solution of any model is optimal only with respect to the
particular model, not necessarily with respect to the real
system. It is important to realize this limited meaning of the
word optimal, a term commonly used by water resources
and other systems analysts, planners and engineers.
Figure 3.5 illustrates the broad differences between
simulation and optimization. Optimization models need
explicit expressions of objectives. Simulation models do not.
Simulation simply addresses what-if sceanarios what may
happen if a particular scenario is assumed or if a particular
decision is made. Users must specify the values of design
and operating decision-variables before a simulation can be
performed. Once these values of all decision-variables are
defined, simulation can help us estimate more precisely the
impacts that may result from those decisions. The difficulty
with using simulation alone for analysing multiple alterna-
tives occurs when there are many alternative, and potentially
attractive, feasible solutions or plans and not enough time or
resources to simulate them all. Even when combined with
efficient techniques for selecting the values of each decision-
variable, an enormous computational effort may still lead to
a solution that is still far from the best possible.
For water resources planning and management, it
is often advantageous to use both optimization and
simulation modelling. While optimization will tell us
water resources system
water resources system
simulation
optimization
system design and operating policy
system design and operating policy
system inputs
system inputs
system outputs
system outputs
E
0
1
1
2
1
7
o
Figure 3.5. Distinguishing
between simulation and
optimization modelling.
Simulation addresses what if
questions; optimization can
address what should be
questions. Both types of models
are often needed in water
resources panning and
management studies.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 68
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 69
what we should do what the best decision is that
solution is often based on many limiting assumptions.
Because of this, we need to use optimization not as a way
to find the best solution, but to define a relatively small
number of good alternatives that can later be tested,
evaluated and improved by means of simulation. This
process of using optimization to reduce the large number
of plans and policies to a few that can then be simulated
and better evaluated is often called preliminary screening.
3.5. Types of Models
3.5.1. Types of Simulation Models
Simulation models can be statistical or process oriented,
or a mixture of both. Pure statistical models are based
solely on data (field measurements). Pure process-
oriented models are based on knowledge of the funda-
mental processes that are taking place. The example
simulation model just discussed is a process-oriented
model. It incorporated and simulated the physical
processes taking place in the system. Many simulation
models combine features of both of these extremes.
The range of various simulation modelling approaches
applied to water resources systems is illustrated in
Figure 3.6.
Regressions, such as that resulting from a least-squares
analysis, and artificial neural networks are examples of
purely statistical models. As discussed in Chapter 6, a
relationship is derived between input data (cause) and
output data (effect), based on measured and observed
data. The relationship between the input and the output
variable values is derived by calibrating a black-box or
statistical model with a predefined structure unrelated to
the actual natural processes taking place. Once calibrated,
the model can be used to estimate the output variable
values as long as the input variable values are within the
range of those used to calibrate the model.
Hybrid models incorporate some process relationships
into regression models or neural networks. These rela-
tionships supplement the knowledge contained in the
calibrated parameter values based on measured data.
Most simulation models containing process relationships
include parameters whose values need to be estimated. This
is called model calibration. Calibration requires measured
field data. These data can be used for initial calibration and
verification, and in the case of ongoing simulations, for
continual calibration and uncertainty reduction. The latter
is sometimes referred to as data assimilation.
Other simulation model classifications are possible.
Simulation models can be classified based on what the
regression
and
neural
networks
hybrid
models
process
models
& data
assimilation
process
based
simulation
models
fully process oriented fully data oriented
process
knowledge
measured
data
E
0
2
0
1
1
0
s
Figure 3.6. Range of simulation
models types based on the
extent to which measured field
data and descriptions of system
processes are included in the
model.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 69
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
model simulates: on the domain of application. Today
one can obtain, or develop, computer programs written to
simulate a wide variety of water resources system compo-
nents or events. Some of these include:
water quantity and/or quality of rivers, bays, estuaries
or coastal zones
reservoir operation for quantity and/or quality
saturated and/or unsaturated zone groundwater quan-
tity and/or quality
precipitation runoff, including erosion and chemicals
water system demands, supply distribution and treat-
ment
high-water forecasting and control
hazardous material spills
morphological changes
wastewater collection systems
wastewater purification facilities
irrigation operations
hydropower production
ecological habitats of wetlands, lakes, reservoirs and
flood plains
economic benefits and costs.
Simulation models of water resources systems generally
have both spatial and temporal dimensions. These dimen-
sions may be influenced by the numerical methods used,
if any, in the simulation, but otherwise they are usually
set, within the limits desired by the user. Spatial resolu-
tions can range from 0 to 3 dimensions. Models are some-
times referred to as quasi 2- or 3-dimensional models.
These are 1 or 2-dimensional models set up in a way that
approximates what takes place in 2- or 3-dimensional
space, respectively. A quasi-3D system of a reservoir
could consist of a series of coupled 2D horizontal layers,
for example.
Simulation models can be used to study what might
occur during a given time period, say a year, sometime in
the future, or what might occur from now to that given
time in the future. Models that simulate some particular
time in the future, where future conditions such as
demands and infrastructure design and operation are
fixed, are called stationary or static models. Models that
simulate developments over time are called dynamic
models. Static models are those in which the external
environment of the system being simulated is not chang-
ing. Water demands, soil conditions, economic benefit
70 Water Resources Systems Planning and Management
and cost functions, populations and other factors do not
change from one year to the next. Static models provide a
snapshot or a picture at a point in time. Multiple years of
input data may be simulated, but from the output statis-
tical summaries can be made to identify what the values
of all the impact variables could be, together with their
probabilities, at that future time period.
Dynamic simulation models are those in which the
external environment is also changing over time.
Reservoir storage capacities could be decreasing due to
sediment load deposition, costs could be increasing due
to inflation, wastewater effluent discharges could be
changing due to changes in populations and/or waste-
water treatment capacities, and so on.
Simulation models can also vary in the way they are
solved. Some use purely analytical methods while others
require numerical ones. Many use both methods, as
appropriate.
Finally, models can also be distinguished according to
the questions being asked and the level of information
desired. The type of information desired can range from
data of interest to policy-makers and planners (requiring
relatively simple models and broader in scope) to that of
interest to researchers desiring a better understanding
of the complex natural, economic and social processes
taking place (requiring much more detailed models and
narrower in scope). Water management and operational
models (for real-time operations of structures, for exam-
ple) and event-based calamity models are somewhere
between these two extremes with respect to model detail.
The scope and level of detail of any modelling study will
also depend on the time, money and data available for
that study (see Chapter 2).
3.5.2. Types of Optimization Models
There are many ways to classify various types of con-
strained optimization models. Optimization models can
be deterministic or probabilistic, or a mixture of both.
They can be static or dynamic with respect to time. Many
water resources planning and management models are
static, but include multiple time periods to obtain a
statistical snapshot of various impacts in some planning
period. Optimization models can be linear or non-linear.
They can consist of continuous variables or discrete or
integer variables, or a combination of both. But whatever
wrm_ch03.qxd 8/31/2005 11:50 AM Page 70
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 71
type they are, they have in common the fact that they are
describing situations where there exist multiple solutions
that satisfy all the constraints, and hence, there is the
desire to find the best solution, or at least a set of very
good solutions.
Regardless of the type of optimization model, they all
include an objective function. The objective function of
an optimization model (such as Equation 3.11 in the
example problem above) is used to evaluate multiple
possible solutions. Often multiple objective functions
may be identified (as will be discussed in Chapter 10). But
at least one objective must be identified in all optimization
models. Identifying the best objective function is often a
challenging task.
Optimization models can be based on the particular
type of application, such as reservoir sizing and/or opera-
tion, water quality management, or irrigation develop-
ment or operation. Optimization models can also be
classified into different types depending on the algorithm
to be used to solve the model. Constrained optimization
algorithms are numerous. Some guarantee to find the
best model solution and others can only guarantee locally
optimum solutions. Some include algebraic mathematical
programming methods and others include deterministic
or random trial-and-error search techniques. Mathematical
programming techniques include Lagrange multipliers,
linear programming, non-linear programming, dynamic
programming, quadratic programming, fractional pro-
gramming and geometric programming, to mention a
few. The applicability of each of these as well as other
constrained optimization procedures is highly dependent
on the mathematical structure of the model. The follow-
ing Chapter 4 illustrates the application of some of
the most commonly used constrained optimization tech-
niques in water resources planning and management.
These include classical constrained optimization using
calculus-based Lagrange multipliers, discrete dynamic
programming, and linear and non-linear programming.
Hybrid models usually include multiple solution
methods. Many generic multi-period simulation models
are driven by optimization methods within each time
period. (The CALSIM II model used by the State of
California and the US Bureau of Reclamation to allocate
water in central California is one such model.)
Each of a variety of optimization modelling types
and solution approaches will be discussed and illustrated
in more detail in subsequent chapters. In some cases, we
can use available computer programs to solve optimiza-
tion models. In other cases, we may have to write our
own software. To make effective use of optimization, and
even simulation, models one has to learn some model
solution methods, since those methods often dictate the
type of model most appropriate for analysing a particular
planning or management problem or issue.
To date, no single model type or solution procedure
has been judged best for all the different types of issues
and problems encountered in water resources planning
and management. Each method has its advantages and
limitations. One will experience these advantages and
limitations as one practices the art of model development
and application.
4. Model Development
Prior to the selection or development of a quantitative
simulation model, it is often useful to develop a concep-
tual one. Conceptual models are non-quantitative repre-
sentations of a system. The overall system structure is
defined but not all its elements and functional relation-
ships.
Figure 3.7 illustrates a conceptual model, without
indicating what each box represents, defining relation-
ships between what land and water managers can do and
the eventual ecological impacts of those actions.
Once a conceptual model has been quantified
(expressed in mathematical terms), it becomes a mathe-
matical model. The models equations typically include
state and auxiliary variables, parameters and other model
components.
The values of the models parameters need to be deter-
mined. Model calibration involves finding the best values
for these parameters. It is based on comparisons of the
model results with field measurements. Optimization
methods can sometimes be used to identify the values of
all model parameters. This is called model calibration or
identification. (Illustrations of the use of optimization for
estimating model parameter values are contained in
Chapters 4 and 6).
Sensitivity analysis (Chapter 9) may serve to identify
the impacts of uncertain parameter values and show
which parameter values substantially influence the
wrm_ch03.qxd 8/31/2005 11:50 AM Page 71
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
models results. Following calibration, the remaining
uncertainties in the model predictions may be quantified
in an uncertainty analysis as discussed in Chapter 9.
In addition to being calibrated, simulation models
should also be validated or verified. In the validation or
verification process the model results are compared with
an independent set of measured observations that were
not used in calibration. This comparison is made to verify
whether or not the model describes the system behaviour
sufficiently correctly.
5. Managing Modelling Projects
There are some steps that, if followed in modelling
projects, can help reduce potential problems and lead to
more effective outcomes. These steps are illustrated in
Figure 3.8 (Scholten et al., 2000).
Some of the steps illustrated in Figure 3.8 may not
be relevant in particular modelling projects. If so, these
parts of the process can be skipped. Each of these
modelling project steps is discussed in the next several
sections.
72 Water Resources Systems Planning and Management
5.1. Creating a Model Journal
One common problem of model studies once they are
underway occurs when one wishes to go back over a
series of simulation results to see what was changed, why
a particular simulation was made or what was learned. It
is also commonly difficult if not impossible for third
parties to continue from the point at which any previous
project terminated. These problems are caused by a lack
of information on how the study was carried out. What
was the pattern of thought that took place? Which actions
and activities were carried out? Who carried out what
work and why? What choices were made? How reliable
are the end results? These questions should be answerable
if a model journal is kept. Just like computer-programming
documentation, this study documentation is often
neglected under the pressure of time and perhaps because
it is not as interesting as running the models themselves.
5.2. Initiating the Modelling Project
Project initiation involves defining the problem to be
modelled and the objectives that are to be accomplished.
land and water management practices and policies
hydrological consequences
ecosystem habitat & function impacts
impacts on specific species
E
0
2
0
1
1
0
t
Figure 3.7. An example of a
conceptual model without its
detail, showing the cause and
effect links between
management decisions and
specific system impacts.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 72
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 73
There can be major differences in perceptions between
those who need information and those who are going to
provide it. The problem as stated is often not the prob-
lem as understood by either the client or the modeller. In
addition, problem perceptions and modelling objectives
can change over the duration of a modelling project.
The appropriate spatial and time scales also need to be
identified. The essential natural system processes must be
identified and described. One should ask and answer
the question of whether or not a particular modelling
approach, or even modelling in general, is the best way to
obtain the needed information. What are the alternatives
to modelling or a particular modelling approach?
The objective of any modelling project should be
clearly understood with respect to the domain and the
problem area, the reason for using a particular model,
the questions to be answered by the model, and the
scenarios to be modelled. Throughout the project these
objective components should be checked to see if any
have changed and if they are being met.
The use of a model nearly always takes place within a
broader context. The model itself can also be part of a
larger whole, such as a network of models in which many
are using the outputs of other models. These conditions
may impose constraints on the modelling project.
Proposed modelling activities may have to be justified
and agreements made where applicable. Any client at any
time may wish for some justification of the modelling
project activities. Agreement should be reached on how
this justification will take place. Are intermediate reports
required, have conditions been defined that will indicate
an official completion of the modelling project, is verifi-
cation by third parties required, and so on? It is particu-
larly important to record beforehand the events or times
when the client must approve the simulation results.
Finally, it is also sensible to reach agreements with respect
to quality requirements and how they are determined
or defined, as well as the format, scope and contents of
modelling project outputs (data files) and reports.
5.3. Selecting the Model
The selection of an existing model to be used in any proj-
ect depends in part on the processes that will be modelled
(perhaps as defined by the conceptual model), the data
available and the data required by the model. The avail-
able data should include system observations for compar-
ison of the model results. They should also include
estimates of the degree of uncertainty associated with
each of the model parameters. At a minimum this might
only be estimates of the ranges of all uncertain parameter
values. At best it could include statistical distributions of
them. In this step of the process it is sufficient to know
what data are available, their quality and completeness,
and what to do about missing or outlier data.
Determining the boundaries of the model is an essen-
tial consideration in model selection. This defines what is
to be included in a model and what is not. Any model
selected will contain a number of assumptions. These
assumptions should be identified and justified, and later
tested.
Project-based matters such as the computers to be
used, the available time and expertise, the modellers per-
sonal preferences, and the clients wishes or requirements
may also influence model choice. An important practical
criterion is whether there is an accessible manual for
operating the model program and a help desk available to
address any possible problems.
report model results
step 7
interpret model results
implement the model
analyse the model
select the model
initiate modeling project
establish project journal
step 6
step 5
step 4
step 3
step 2
step 1
E
0
2
0
1
1
0
u
Figure 3.8. The modelling project process is an iterative
procedure involving specific steps or tasks.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 73
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The decision to use a model, and which model to use,
is an important part of water resources plan formulation.
Even though there are no clear rules on how to select
the right model to use, a few simple guidelines can be
stated:
Define the problem and determine what information is
needed and what questions need to be answered.
Use the simplest method that will yield adequate accu-
racy and provide the answer to your questions.
Select a model that fits the problem rather than trying
to fit the problem to a model.
Question whether increased accuracy is worth the
increased effort and increased cost of data collection.
(With the advances in computer technology, compu-
tational cost is rarely an issue except perhaps for some
groundwater management problems.)
Do not forget the assumptions underlying the model
used and do not read more significance into the simu-
lation results than is actually there.
5.4. Analysing the Model
Once a modelling approach or a particular model has
been selected, its strengths and limitations should be
studied in more detail. The first step is to set up a plan for
testing and evaluating the model. These tests can include
mass (and energy) balance checks and parameter sensitivity
analyses (see Chapter 9). The model can be run under
extreme input data conditions to see if the results are as
expected.
Once a model is tested satisfactorily, it can be cali-
brated. Calibration focuses on the comparison between
model results and field observations. An important prin-
ciple is: the smaller the deviation between the calculated
model results and the field observations, the better the
model. This is indeed the case to a certain extent, as
the deviations in a perfect model are only due to measure-
ment errors. In practice, however, a good fit is by no
means a guarantee of a good model.
The deviations between the model results and the field
observations can be due to a number of factors. These
include possible software errors, inappropriate modelling
assumptions such as the (conscious) simplification of
complex structures, neglect of certain processes, errors in
the mathematical description or in the numerical method
applied, inappropriate parameter values, errors in input
74 Water Resources Systems Planning and Management
data and boundary conditions, and measurement errors
in the field observations.
To determine whether or not a calibrated model is
good, it should be validated or verified. Calibrated
models should be able to reproduce field observations not
used in calibration. Validation can be carried out for
calibrated models as long as an independent data set has
been kept aside for this purpose. If all available data are
used in the calibration process in order to arrive at
the best possible results, validation will not be possible.
The decision to leave out validation is often a justifiable
one especially when data are limited.
Philosophically, it is impossible to know if a model of
a complex system is correct. There is no way to prove it.
Experimenting with a model, by carrying out multiple
validation tests, can increase ones confidence in that
model. After a sufficient number of successful tests, one
might be willing to state that the model is good enough,
based on the modelling project requirements. The model
can then be regarded as having been validated, at least for
the ranges of input data and field observations used in the
validation.
If model predictions are to be made for situations or
conditions for which the model has been validated, one
may have a degree of confidence in the reliability of those
predictions. Yet one cannot be certain. Much less confi-
dence can be placed on model predictions for conditions
outside the range for which the model was validated.
While a model should not be used for extrapolations
as commonly applied in predictions and in scenario
analyses, this is often exactly the reason for the modelling
project. What is likely to happen given events we have not
yet experienced? A models answer to this question
should also include the uncertainties attached to these
predictions. Beck (1987) summarized this dilemma in the
following statement: using scientifically based models,
you will often predict an incorrect future with great accu-
racy, and when using complex, non-identifiable models,
you may be capable of predicting the correct future with
great uncertainty.
5.5. Using the Model
Once the model has been judged good enough, it may be
used to obtain the information desired. One should
develop a plan on how the model is to be used, identifying
wrm_ch03.qxd 8/31/2005 11:50 AM Page 74
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 75
the input to be used, the time period(s) to be simulated,
and the quality of the results to be expected. Again, close
communication between the client and the modeller is
essential, both in setting up this plan and throughout its
implementation, to avoid any unnecessary misunderstand-
ings about what information is wanted and the assump-
tions on which that information is to be based.
Before the end of this model-use step, one should
determine whether all the necessary model runs have
been performed and whether they have been performed
well. Questions to ask include:
Did the model fulfill its purpose?
Are the results valid?
Are the quality requirements met?
Was the discretization of space and time chosen well?
Was the choice of the model restrictions correct?
Was the correct model and/or model program chosen?
Was the numerical approach appropriate?
Was the implementation performed correctly?
Are the sensitive parameters (and other factors) clearly
identified?
Was an uncertainty analysis performed?
If any of the answers to these questions is no, then the
situation should be corrected. If it cannot be corrected,
then there should be a good reason for this.
5.6. Interpreting Model Results
Interpreting the information resulting from simulation
models is a crucial step in a modelling project, especially
in situations in which the client may only be interested in
those results and not the way they were obtained. The
model results can be compared to those of other similar
studies. Any unanticipated results should be discussed
and explained. The results should be judged with respect
to the modelling project objectives.
The results of any water resources modelling project
typically include large files of time-series data. Only the
most dedicated of clients will want to read those files, so
the data must be presented in a more concise form.
Statistical summaries should explicitly include any
restrictions and uncertainties in the results. They should
identify any gaps in the domain knowledge, thus generat-
ing new research questions or identifying the need for
more field observations and measurements.
5.7. Reporting Model Results
Although the results of a model should not be the sole basis
for policy decisions, modellers have a responsibility to
translate their model results into policy recommendations.
Policy-makers, managers, and indeed the participating
stakeholders often want simple, clear and unambiguous
answers to complex questions. The executive summary of a
report will typically omit much of the scientifically justified
discussion in its main body regarding, say, the uncertain-
ties associated with some of the data. This executive
summary is often the only part read by those responsible
for making decisions. Therefore, the conclusions of the
model study must not only be scientifically correct and
complete, but also concisely formulated, free of jargon,
and fully understandable by managers and policy-makers.
The report should provide a clear indication of the
validity, usability and any restrictions of the model results.
The use of visual aids, such as graphs and GIS, can be very
helpful.
The final report should also include sufficient detail to
allow others to reproduce the model study (including its
results) and/or to proceed from the point where this study
ended.
6. Issues of Scale
Scaling aspects play an important role in many modelling
projects. Four different types of scales can be distin-
guished: the process scale, the information scale, the
model scale and the sampling scale. Each of these is
discussed below.
6.1. Process Scale
Most hydrological processes vary over space and time.
The scale on which the process variations manifest them-
selves is referred to as the process scale. Spatially, process
scale can vary from the movement of small granules of
sediment, for example, to the flooding of large river basins
or coastal zones. All kinds of intermediary scale processes
can be found, such as drainage into ditches of runoff from
parcels of land, transport of sediment in brooks and flow
movements in aquifers.
Various temporal scales can also be distinguished,
varying from the intensity of rain in less than a minute to
wrm_ch03.qxd 8/31/2005 11:50 AM Page 75
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the change in landscape in geological time. Many process
descriptions require a spectrum of scales. Such is the case,
for example, in the simulation of interdependent surface
and groundwater quantity and quality processes taking
place in a watershed.
6.2. Information Scale
Information scale is the spatial and temporal scale of the
information required. Generally, a strategic water resources
manager (for example the local, regional or national
government) needs information on a scale relative to their
responsibilities and authorities. This level of information is
likely to differ from the level desired by operational water
managers dealing with day-to-day issues.
Information at scales smaller than what is needed is
seen as being noise. Information at scales larger than
what is needed is not relevant or helpful. For local
organizations (e.g., water boards) concerned with runoff,
for example, there is no need to collect information on
individual raindrops. The important spatial variances
are usually within a range varying from hundreds of
metres to hundreds of kilometres. Larger-scale variations
(differences between precipitation in the Netherlands
and Russia or between North and South America, for
example) are rarely if ever relevant. The information
scale depends on the task set for the water planner or
manager.
6.3. Model Scale
Model scales refer to their spatial and temporal discretiza-
tion. The model scales determine the required data inter-
polation and aggregation.
If the temporal and spatial scales of the problem have
not been defined clearly enough, this can affect the later
phases of the modelling process negatively. If the model
scale chosen is too large, this may result in too general a
schematization and relevant details might not be derived
from the results. If the chosen model scale is too small,
irrelevant small-scale variations can lead to non-optimal
calibrations for the large-scale variations.
In large, spatially distributed models in particular, it is
vital that the scale and the number of independent
parameters (degrees of freedom) are chosen on the basis
of the available data. If too many parameters are included
76 Water Resources Systems Planning and Management
in a model, there is a risk of it appearing to work well but
being unsuitable for interpolation or prediction. This can
actually only be determined if adequate measuring data
are available having a measurement frequency at least
equal to the chosen modelling time step. This must be
taken into account when selecting or constructing the
model, as there is otherwise the risk that the model
cannot be calibrated adequately.
6.4. Sampling Scale
Sampling scale is the scale at which samples are taken.
The sampling spatial scale can vary from point observa-
tions (for example, a temperature measurement at a
certain location at a certain time) to area observations
(using, for example, remote sensing images). The density
of the measuring network and the sampling or measuring
and recording frequency determine the sampling spatial
and temporal scales.
6.5. Selecting the Right Scales
Modellers must choose the model scale in such a manner
that the model provides information at the required
information scale, taking into account the process scales
present, in combination with the spatial and temporal
sampling scales. It is possible that situations will occur
that are impossible to model just because of these scale
issues.
The relationship between the types of scales is repre-
sented in Figure 3.9. The relative level of detail is given on
the horizontal axis, from considerable detail on the left to
much less detail on the right.
To show that the various types of scales may not be
mutually compatible, consider the following three
examples.
sampling
scale
small scale
process
scale
information
scale
large scale
E
0
2
0
1
1
0
v
Figure 3.9. Relationships among various scale types.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 76
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Methods for Evaluating Alternatives 77
Example 1: The information scale is different from the process
scale
Water resources planning studies are typically carried out
at a river basin or watershed scale. From a hydrological
(process) point of view, this makes sense because it enables
a comprehensive analysis (including upstream and down-
stream impacts), makes it easier to develop a water balance
for the study area, and reduces the amount of information
needed at the borders of the study area. However, most
decision-makers are not interested in results at river basin
or watershed scale; they want to know what these results
mean for their province, municipality or city. This conflict
of scales can be solved by a well-considered selection of the
(sub)watersheds that will be considered in the study and a
post-processor that translates the results at these process
scales into the required administrative scales.
Example 2: The information scale is smaller than the process
scale
Imagine that a water resources manager wants to evaluate
alternative anti-dehydration measures on the groundwater
level over a period of five years. Thus the required
information scale is a five-year period. However, the
groundwater level is characterized by a very slow response.
The relevant temporal groundwater-level process scale is
around fifteen to twenty years. Whatever management
alternative is implemented, it will take fifteen to twenty
years to determine its impact. Thus, regardless of the choice
of measuring frequency (sampling scale) and the model
scale, it is impossible within the five-year period of interest
to arrive at information on the groundwater-level changes as
a consequence of anti-dehydration measures.
Example 3: The sampling scale is larger than the process scale
and information scale
Assume it is necessary to estimate the change in concen-
trations of certain substances in the soil and groundwater
in an urban area. The information spatial scale is one to
two decimetres. This corresponds to the spatial variation
of cohesion processes that take place in the soil and
groundwater aquifer. However, logistic and budgetary
considerations make it impossible to increase the spatial
sampling density to less than a measurement site every
few hundred metres. As the spatial sampling scale is
much larger than the spatial process scale, useful inter-
polations cannot be made.
7. Conclusions
This chapter has reviewed some basic types of models and
presented guidelines for consideration when undertaking
a modelling project. Generic models for water resources
system analyses are increasingly becoming available, saving
many organizations that need model results from having to
develop their own individual models. While many readers of
this book may get involved in writing their own models,
most of those involved in water resources planning and man-
agement will be using existing models and analysing and
presenting their results. The information provided in this
chapter is intended to help model users plan and manage
effective modelling projects as well as improve the repro-
ducibility, transferability and usefulness of model results.
8. References
BECK, M.B. 1987. Water quality modeling: a review of
the analysis of uncertainty. Water Resources Research,
No. 23, pp. 13931442.
SCHOLTEN, H.; WAVEREN, R.H. VAN; GROOT, S.;
GEER, F.C. VAN; WSTEN, J.H.M.; KOEZE, R.D. and
NOORT, J.J. 2000. Good modelling practice in water manage-
ment, proceedings hydroinformatics 2000. Cedar Rapids, Iowa,
International Association for Hydraulic Research.
Additional References (Further Reading)
ABBOTT, M.B. 1999. Introducing hydroinformatics.
Journal of Hydroinformatics, Vol. 1, No. 1, pp. 320.
BISWAS, A.K. (ed.). 1997. Water resources: environmental
planning, management, and development. New York,
McGraw-Hill.
BOBBA, A., GHOSH, S.; VIJAY, P. and BENGTSSON, L.
2000. Application of environmental models to different
hydrological systems. Ecological Modelling, Vol. 125,
pp. 1549.
BOOCH, G. 1994. Object-oriented analysis with applica-
tions. Redwood City, Benjamin Cummings.
CHAN, H.W. and LAM, P.F. 1997. Visualizing Input and
output analysis for simulation. Simulation Practice and
Theory, Vol. 5, No. 5, pp. 42553.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 77
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
EDSALL, R.M.; HARROWER, M. and MENNIS, J.L. 2000.
Tools for visualizing properties of spatial and temporal
periodicities in geographic data. Computers & Geosciences,
Vol. 26, No. 1, pp. 10918.
FEDRA, K. and JAMIESON, D.G. 1996. An object-oriented
approach to model integration: a river basin information
system example. In: K. Kowar and H.P. Nachtnebel (eds.),
Proceedings of the HydroGIS96 Conference, No. 235,
Vienna, IAHS Press. pp. 66976.
GUARISO, G. and WERTHNER, H. 1989. Environmental
decision support systems. New York, Ellis Horwood-Wiley.
HO, J.K.K. and SCULLI, D. 1997. The scientific approach
to problem solving and decision support systems.
International Journal of Production Economics, Vol. 48,
No. 3, pp. 24957.
HUFSCHMIDT, M.M. and FIERING, M.B. 1966. Simulation
in the design of water resource systems. Cambridge, Mass.,
Harvard University Press.
LAM, D.C.L.; MAYFIELD, C.I. and SAWYNE, D.A. 1994.
A prototype information system for watershed manage-
ment and planning. Journal of Biological Systems, Vol. 2,
No. 4, pp. 499517.
LAM, D. and SWAYNE, D. 2001. Issues of EIS software
design: some lessons learned in the past decade. Environ-
mental Modeling and Software, Vol. 16, No. 5, pp. 41925.
LAW, A.M. and KELTON, W.D. 2000. Simulation modeling
and analysis, 3rd edn. Boston, Mass., McGraw-Hill.
LOUCKS, D.P.; STEDINGER, J.S. and HAITH, D.A. 1981.
Water resources systems planning and analysis. Englewood
Clifts, N.J., Prentice Hall.
MAASS, A.; HUFSCHMIDT, M.M.; DORFMAN, R.;
THOMAS, H.A.; MARGLIN, S.A. and FAIR, G.M. 1966.
Design of water resources systems. Cambridge, UK, Harvard
University Press.
REITSMA, R.F. and CARRON, J.C. 1997. Object-oriented
simulation and evaluation of river basin operations.
Journal of Geographic Information and Decision Analysis,
Vol. 1, No. 1, pp. 924.
RIZZOLI, A.E. and YOUNG, W.J. 1997. Delivering
environmental decision support systems: software tools
and techniques. Environmental Modeling & Software,
Vol. 12, No. 23, pp. 23749.
78 Water Resources Systems Planning and Management
RIZZOLI, A.E.; DAVIS, J.R. and ABEL, D.J. 1998. Model
and data integration and re-use in environmental decision
support systems. Decision support systems, Vol. 24, No. 2,
pp. 12744.
SARGENT, R.G. 1982. Verification and validation of
simulation models. In: F.E. Cellier (ed.), Progress in
modelling and simulation, pp. 15969. London and
elsewhere, Academic Press.
SARGENT, R.G. 1984a. A tutorial on verification and
validation of simulation models. In: S. Sheppard,
U. Pooch and D. Pegden (eds.), Proceedings of the 1984
winter simulation conference, pp. 11521. Piscataway, N.J.,
Institute of Electrical and Electronics Engineers.
SARGENT, R.G. 1984b. Simulation model validation. In:
T.I. ren, B.P. Zeigler and M.S. Elzas (eds.), Simulation and
model-based methodologies: an integrative view, pp. 53755.
Berlin and elsewhere, Springer-Verlag (No. 10 in the series:
NATO ASI Series F: Computer and Systems Sciences).
SINGH, V.P. (ed.). 1995. Computer models of watershed
hydrology. Littleton, Colo., Water Resources Publications.
SUN, L. and LIU, K. 2001. A method for interactive artic-
ulation of information requirements for strategic decision
support. Information and Software Technology, Vol. 43,
No. 4, pp. 24763.
VAN WAVEREN, R.H.; GROOT, S.; SCHOLTEN, H.;
VAN GEER, F.C.; WSTEN, J.H.M.; KOEZE, R.D. and
NOORT, J.J. 1999. Good modelling practice handbook: STOWA
report 9905. Lelystad, the Netherlands, Dutch Dept. of
Public Works, Institute for Inland Water Management and
Waste Water Treatment (report 99.036).
YOUNG, P. 1998. Data-based mechanistic modeling of
environmental, ecological, economic and engineering
systems. Environmental Modeling & Software, Vol. 13,
pp. 10522.
YOUNG, W.J.; LAM, D.C.L.; RESSEL, V. and WONG,
J.W. 2000. Development of an environmental flows
decision support system. Environmental Modeling &
Software, Vol. 15, pp. 25765.
ZEIGLER, B.P. 1976. Theory of modelling and simulation.
Malabar, Fla, Robert E. Krieger.
ZEILER, M. 1999. Modeling our world: the ESRI guide to
geodatabase design. Redlands, Calif., ESRI.
wrm_ch03.qxd 8/31/2005 11:50 AM Page 78
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
4. Optimization Methods
1. Introduction 81
2. Comparing Time Streams of Economic Benefits and Costs 81
2.1. Interest Rates 82
2.2. Equivalent Present Value 82
2.3. Equivalent Annual Value 82
3. Non-Linear Optimization Models and Solution Procedures 83
3.1. Solution Using Calculus 84
3.2. Solution Using Hill Climbing 84
3.3. Solution Using Lagrange Multipliers 86
3.3.1. Approach 86
3.3.2. Meaning of the Lagrange Multiplier 88
4. Dynamic Programming 90
4.1. Dynamic Programming Networks and Recursive Equations 90
4.2. Backward-Moving Solution Procedure 92
4.3. Forward-Moving Solution Procedure 95
4.4. Numerical Solutions 96
4.5. Dimensionality 97
4.6. Principle of Optimality 97
4.7. Additional Applications 97
4.7.1. Capacity Expansion 98
4.7.2. Reservoir Operation 102
4.8. General Comments on Dynamic Programming 112
5. Linear Programming 113
5.1. Reservoir Storage CapacityYield Models 114
5.2. A Water Quality Management Problem 117
5.2.1. Model Calibration 118
5.2.2. Management Model 119
5.3. A Groundwater Supply Example 124
5.3.1. A Simplified Model 125
5.3.2. A More Detailed Model 126
5.3.3. An Extended Model 127
5.3.4. Piecewise Linearization Methods 128
5.4. A Review of Linearization Methods 129
6. A Brief Review 132
7. Reference 132
wrm_ch04.qxd 8/31/2005 11:51 AM Page 80
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
All of us are optimizers. We all make decisions that
maximize our welfare in some way or another. Often the
welfare we are maximizing may come later in life. By opti-
mizing, it reflects our evaluation of future benefits versus
current costs or benefits forgone. In economics, the extent
to which we value future benefits today is reflected by
what is called a discount rate. While economic criteria are
only a part of everything we consider when making deci-
sions, they are often among those deemed very important.
Economic evaluation methods involving discount rates
can be used to consider and compare alternatives charac-
terized by various benefits and costs that occur over
time. This chapter begins with a quick and basic review
of what is often called applied engineering economics. Many
of the optimization methods used in practice incorporate
concepts from engineering economics.
Engineering economic methods typically identify a set of
mutually exclusive alternatives (only one alternative can be
selected) and then, using various methods involving the dis-
count rate, identify the best one. The values of the decision-
variables (e.g., the design and operating policy variables) are
known for each alternative. For example, consider again the
tank design problem presented in the previous chapter.
Alternative tank designs could be identified, and then each
81
Optimization Methods
Optimization methods are designed to provide the best values of system design
and operating policy variables values that will lead to the highest levels of
system performance. These methods, combined with more detailed and accurate
simulation methods, are the primary ways we have, short of actually building
physical models, of estimating the likely impacts of particular water resources
system designs and operating policies. This chapter introduces and illustrates
some of the art of optimization model development and use in analysing water
resources systems. The modelling methods introduced in this chapter are
extended in subsequent chapters.
4
could be evaluated, on the basis of cost and perhaps other
criteria as well. The best would be called the optimal one, at
least with respect to the objective criteria used.
The optimization methods introduced in this chapter
extend those engineering economics methods. Some are
discrete, some are continuous. Continuous optimization
methods can identify the best tank design, for example,
without having to identify numerous discrete, mutually
exclusive alternatives. Just how that can be done will be
discussed later in this chapter.
Before proceeding to a more detailed discussion of
optimization, a review of some methods of dealing with
time streams of economic incomes or costs (engineering
economics) may be useful.
2. Comparing Time Streams of
Economic Benefits and Costs
Alternative plans, p, may involve different benefits and
costs over time. These different plans need to be com-
pared somehow. Let the net benefit generated in time
period t by plan p be designated simply as B
t
p
. Each plan
is characterized by the time stream of net benefits it
generates over its planning period Tp.
B
1
p
, B
2
p
, B
3
p
, , B
Tp
p
(4.1)
wrm_ch04.qxd 8/31/2005 11:51 AM Page 81
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Clearly, if in any time period t the benefits exceed the costs,
then B
t
p
0; and if the costs exceed the benefits, B
t
p
0.
This section defines two ways of comparing different benefit,
cost or net-benefit time streams produced by different plans
perhaps having planning periods ending in different years Tp.
2.1. Interest Rates
Fundamental to the conversion of a time series of incomes
and costs to a single value is the concept of the time value
of money. From time to time, individuals, private corpo-
rations and governments need to borrow money to do
what they want to do. The amount paid back to the lender
has two components: (1) the amount borrowed and (2) an
additional amount called interest. The interest amount is
the cost of borrowing money, of having the money when
it is loaned compared to when it is paid back. In the
private sector the interest rate, the added fraction of the
amount loaned that equals the interest, is often identified
as the marginal rate of return on capital. Those who have
money, called capital, can either use it themselves or they
can lend it to others, including banks, and receive inter-
est. Assuming people with capital invest their money
where it yields the largest amount of interest, consistent
with the risk they are willing to take, most investors
should be receiving at least the prevailing interest rate as
the return on their capital.
The interest rate includes a number of considerations.
One is the time value of money (a willingness to pay
something to obtain money now rather than to obtain the
same amount later). Another is the risk of losing capital
(not getting the full amount of a loan or investment
returned at some future time). A third is the risk of reduced
purchasing capability (the expected inflation over time).
The greater the risks of losing capital or purchasing
power, the higher the interest rate will be compared to the
rate reflecting only the time value of money in a secure
and inflation-free environment.
2.2. Equivalent Present Value
To compare projects or plans involving different time
series of benefits and costs, it is often convenient to
express these time series as a single equivalent value. One
way to do this is to convert the time series to what it is
worth today, its present worth, that is, a single value at
82 Water Resources Systems Planning and Management
the present time. This present worth will depend on
the prevailing interest rate in each future time period.
Assuming an amount V
0
is invested at the beginning of
a time period, e.g., a year, in a project or a savings account
earning interest at a rate r per period, then at the end of
the period the value of that investment is (1 r)V
0
.
If one invests an amount V
0
at the beginning of period
t 1 and at the end of that period immediately reinvests
the total amount (the original investment plus interest
earned), and continues to do this for n consecutive
periods at the same period interest rate r, the value, V
n
, of
that investment at the end of n periods would be:
V
n
V
0
(1 r)
n
(4.2)
The initial amount V
0
is said to be equivalent to V
n
at the
end of n periods. Thus the present worth or present value,
V
0
, of an amount of money V
n
at the end of period n is:
V
0
V
n
/(1 r)
n
(4.3)
Equation 4.3 is the basic compound interest discounting
relation needed to determine the present value at the
beginning of period 1 of net benefits V
n
that accrue at the
end of n time periods.
The total present value of the net benefits generated by
plan p, denoted V
0
p
, is the sum of the values of the net
benefits V
t
p
accrued at the end of each time period t times
the discount factor for that period t. Assuming the inter-
est or discount rate r in the discount factor applies for the
duration of the planning period Tp,
(4.4)
The present value of the net benefits achieved by two or
more plans having the same economic planning horizons
Tp can be used as an economic basis for plan selection. If
the economic lives or planning horizons of projects differ,
then the present value of the plans may not be an appro-
priate measure for comparison and plan selection. A valid
comparison of alternative plans is possible if all plans
cover the same planning period.
2.3. Equivalent Annual Value
If the lives of various plans differ, but the same plans will
be repeated on into the future, then one need only compare
the equivalent constant annual net benefits of each plan.
Finding the average or equivalent annual amount V
p
is done
V V r
p p
t
t
t
Tp
0
1

/ ) (
wrm_ch04.qxd 8/31/2005 11:51 AM Page 82
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 83
in two steps. First, one can compute the present value, V
0
p
,
of the time stream of net benefits, using Equation 4.4. The
equivalent constant annual benefits, V
p
, all discounted to
the present must equal the present value, V
0
p
.
(4.5)
With a little algebra the average annual end-of-year
benefits V
p
of the project or plan p is:
V
p
V
0
p
[r(1 r)
Tp
]/[(1 r)
Tp
1] (4.6)
The capital recovery factor CRF
n
is the expression
[r(1r)
Tp
]/[(1r)
Tp
1] in Equation 4.6 that converts a
fixed payment or present value V
0
p
at the beginning of n
time periods into an equivalent fixed periodic payment V
p
at the end of each period. If the interest rate per period
is r and there are n periods involved, then the capital
recovery factor is:
CRF
n
[r(1 r)
n
]/[(1 r)
n
1] (4.7)
This factor is often used to compute the equivalent annual
end-of-year cost of engineering structures that have a
fixed initial construction cost C
0
and annual end-of-year
operation, maintenance, and repair (OMR) costs. The
equivalent uniform end-of-year total annual cost, TAC,
equals the initial cost times the capital recovery factor plus
the equivalent annual end-of-year uniform OMR costs.
TAC CRF
n
C
0
OMR (4.8)
For private investments requiring borrowed capital, interest
rates are usually established, and hence fixed, at the time of
borrowing. However, benefits may be affected by changing
interest rates, which are not easily predicted. It is common
practice in benefitcost analyses to assume constant interest
rates over time, for lack of any better assumption.
Interest rates available to private investors or borrow-
ers may not be the same rates that are used for analysing
public investment decisions. In an economic evaluation
of public-sector investments, the same relationships are
used even though government agencies are not generally
free to loan or borrow funds on private money markets.
In the case of public-sector investments, the interest rate
to be used in an economic analysis is a matter of public
policy; it is the rate at which the government is willing to
forego current benefits to its citizens in order to provide
V V V
p p
t
p
t
Tp
t
t
Tp
t
0
1 1

/( / ( r r ) )
benefits to those living in future time periods. It can be
viewed as the governments estimate of the time value of
public monies or the marginal rate of return to be
achieved by public investments.
These definitions and concepts of engineering eco-
nomics are applicable to many of the problems faced in
water resources planning and management. More detailed
discussions of the application of engineering economics
are contained in numerous texts on the subject.
3. Non-Linear Optimization Models
and Solution Procedures
Constrained optimization is also called mathematical
programming. Mathematical programming techniques include
Lagrange multipliers, linear and non-linear programming,
dynamic programming, quadratic programming, fractional
programming and geometric programming, to mention a few.
The applicability of each of these as well as other constrained
optimization procedures is highly dependent on the
mathematical structure of the model. The remainder of this
chapter introduces and illustrates the application of some
of the most commonly used constrained optimization
techniques in water resources planning and management.
These include classical constrained optimization using
calculus-based Lagrange multipliers, discrete dynamic
programming, and linear and non-linear programming.
Consider a river from which diversions are made to
three water-consuming firms that belong to the same
corporation, as illustrated in Figure 4.1. Each firm makes
a product. Water is needed in the process of making that
product, and is the critical resource. The three firms can
be denoted by the index j 1, 2 and 3 and their water
allocations by x
j
. Assume the problem is to determine the
allocations x
j
of water to each of three firms ( j 1, 2, 3)
that maximize the total net benefits,
j
NB
j
(x
j
), obtained
from all three firms. The total amount of water available
is constrained or limited to a quantity of Q.
Assume the net benefits, NB
j
(x
j
), derived from water x
j
allocated to each firm j, are defined by:
NB
1
(x
1
) 6x
1
x
1
2
(4.9)
NB
2
(x
2
) 7x
2
1.5x
2
2
(4.10)
NB
3
(x
3
) 8x
3
0.5x
3
2
(4.11)
wrm_ch04.qxd 8/31/2005 11:51 AM Page 83
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
These are concave functions exhibiting decreasing
marginal net benefits with increasing allocations. These
functions look like hills, as illustrated in Figure 4.2.
3.1. Solution Using Calculus
Calculus can be used to find the allocation to each firm
that maximizes its own net benefit, simply by finding
where the slope or derivative of the net benefit function
for the firm equals zero. The derivative, dNB(x
1
)/dx
1
, of
the net benefit function for Firm 1 is (6 2x
1
) and hence
the best allocation to Firm 1 would be 6/2 or 3. The best
allocations for Firms 2 and 3 are 2.33 and 8 respectively.
The total amount of water desired by all firms is the sum
of each firms desired allocation, or 13.33 flow units.
However, suppose only 6 units of flow are available for
all three firms. Introducing this constraint renders the
previous solution infeasible. In this case we want to
find the allocations that maximize the total net
benefit obtained from all firms subject to having only
6 flow units to allocate. Using simple calculus will not
suffice.
84 Water Resources Systems Planning and Management
3.2. Solution Using Hill Climbing
One approach to finding the particular allocations that
maximize the total net benefit derived from all firms in this
example is an incremental steepest-hill-climbing method.
This method divides the total available flow Q into incre-
ments and allocates each additional increment so as to get
the maximum additional net benefit from that incremental
amount of water. This procedure works in this example
because the functions are concave; in other words, the
marginal benefits decrease as the allocation increases. This
procedure is illustrated by the flow diagram in Figure 4.3.
Table 4.1 lists the results of applying the procedure
shown in Figure 4.3 to the problem of a) allocating 8 and
b) allocating 20 flow units to the three firms and the river.
Here a river flow of at least 2 is required and is to be
satisfied, when possible, before any allocations are made to
the firms.
The hill-climbing method illustrated in Figure 4.3 and
Table 4.1 assigns each incremental flow Q to the use
that yields the largest additional (marginal) net benefit.
An allocation is optimal for any total flow Q when the
Q
r
e
m
a
i
w
n
in
lo
g
f
r
iv
e
r
firm
producing
1
x
x
x
firm 3
producing
firm 2
producing

1
1
2
3
3
2
E
0
1
1
2
1
7
q
X
X X X
X X
NB ( ) X
dNB
/
dx
dNB
/
dx
dNB
/
dx
NB ( ) X NB ( ) X
E
0
1
1
2
1
7
r
1 1 2 2 3 3
3 3
3
3
2 2
2
2
1
1
1 1
* * *
Figure 4.1. Three water-using
firms obtain water from river
diversions. The amounts
allocated, x
j
, to each firm j will
depend on the amount of water
available, Q, in the river.
Figure 4.2. Concave net benefit
functions and their slopes at
allocations x
1
*, x
2
* and x
3
*.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 84
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 85
start set initial
conditions
set
Q = 0
Q Q
Q Q = +Q
max,
x , x , x , = 0
1 2 3
Q R <
stop Q Q
max
>
yes no
allocate
to firm j having
max. marginal
net benefits
d at : /dx x
j j j
NB
Q
x x
j j
Q + =
Store
Q and X
data
yes
no
E
0
1
1
2
1
7
s
E
0
2
0
8
2
0
g
1-3 0-2 0-2 0 0 0 6 7 8 3 2 0 0 1 7.5
4 3 2 0 0 1 6 7 7 4 2 0 0 2 14.0
5 4 2 0 0 2 6 7 6 5 2 0 1 2 19.5
6 5 2 0 1 2 6 4 6 6 2 0 1 3 25.0
7 6 2 0 1 3 6 4 5 7 2 1 1 3 30.0
8 7 2 1 1 3 4 4 5 8 2 1 1 4 34.5
9 8 2 1 1 4 4 4 4 -
R x R
1
x
2
x
3
6-2x
1
8- x
3
7-3x
2
x
1
x
2
x
3
x
1
NB
j a j
( )
iteration i Q
Q
i
allocations. R,
allocations. R,
x
x
j
j
marginal net
benefits
marginal net benefits
Q
i
+Q
new allocations total net
benefits
total net
benefits
- - - - ---
Q
Q
max
max
= 8;
= 20;
Q
i
= 0; Q
Q
= ;
0;
river flow R
river flow R
>
>
min
min
Q, 2
Q, 2
{
{
}
}
0-2 0-2 0 0 0 6 7 8 0.0
4 2 0 0.25 1.75 6.00 6.25 6.25 14.1
5 2 0.18 0.46 2.36 5.64 5.64 5.64 20.0
8 2 1.00 1.00 4.00 4.00 4.00 4.00 34.5
10 2 1.55 1.36 5.09 2.91 2.91 2.91 41.4
15 2 2.91 2.27 7.82 0.18 0.18 0.18 49.1
20 6.67 3.00 2.33 8.00 0 0 0 49.2
R x
1
x
2
x
3
6-2x
1
8- x
3
7-3x
2 1
NB
j j
( )
selected values of Q
1
x
Table 4.1. Hill-climbing iterations for
finding allocations that maximize total
net benefit given a flow of Qmax and a
required (minimum) streamflow of
R = 2.
Figure 4.3. Steepest hill-
climbing approach for finding
allocation of a flow Qmax to the
three firms, while meeting
minimum river flow
requirements R.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 85
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
marginal net benefits from each non-zero allocation are
equal, or as close to each other as possible given the size of
the increment Q. In this example, with a Q of 1 and
Qmax of 8, it just happens that the marginal net benefits are
all equal (to 4). The smaller the Q, the more optimal will
be the allocations in each iteration, as shown in the lower
portion of Table 4.1 where Q approaches 0.
Based on the allocations derived for various values of
available water Q, as shown in Table 4.1, an allocation pol-
icy can be defined. For this problem the allocation policy
that maximizes total net benefits is shown in Figure 4.4.
This hill-climbing approach leads to optimal
allocations only if all of the net benefit functions whose
sum is being maximized are concave: that is, the marginal
net benefits decrease as the allocation increases.
Otherwise, only a local optimum solution can be
guaranteed. This is true using any calculus-based
optimization procedure or algorithm.
3.3. Solution Using Lagrange Multipliers
3.3.1. Approach
As an alternative to hill-climbing methods, consider
a calculus-based method involving Lagrange multipliers.
To illustrate this approach, a slightly more complex
water-allocation example will be used. Assume that the
benefit, B
j
(x
j
), each water-using firm receives is determined,
in part, by the quantity of product it produces and the
price per unit of the product that is charged. As before,
these products require water and water is the limiting
resource. The amount of product produced, p
j
, by each
firm j is dependent on the amount of water, x
j
, allocated
to it.
86 Water Resources Systems Planning and Management
Let the function P
j
(x
j
) represent the maximum amount
of product, p
j
, that can be produced by firm j from an allo-
cation of water x
j
. These are called production functions.
They are typically concave: as x
j
increases the slope,
dP
j
(x
j
)/dx
j
, of the production function, P
j
(x
j
), decreases.
For this example assume the production functions for
the three water-using firms are:
p
1
0.4(x
1
)
0.9
(4.12)
p
2
0.5(x
2
)
0.8
(4.13)
p
3
0.6(x
3
)
0.7
(4.14)
Next consider the cost of production. Assume the asso-
ciated cost of production can be expressed by the
following convex functions:
C
1
3(p
1
)
1.3
(4.15)
C
2
5(p
2
)
1.2
(4.16)
C
3
6(p
3
)
1.15
(4.17)
Each firm produces a unique patented product, and
hence it can set and control the unit price of its product.
The lower the unit price, the greater the demand and thus
the more each firm can sell. Each firm has determined the
relationship between the amount that can be sold and the
unit price that is, the demand functions for that product.
These unit price or demand functions are shown in
Figure 4.5 where the p
j
s are the amounts of each product
produced. The vertical axis of each graph is the unit price.
To simplify the problem we are assuming linear demand
functions, but this assumption is not a necessary condition.
The optimization problem is to find the water alloca-
tions, the production levels and the unit prices that
together maximize the total net benefit obtained from all
firm 3
river R
firm 1
firm 2
0 2 4 6 8 10 12 14 16 18 20
10
8
a
l
l
o
c
a
t
i
o
n
Q
6
4
2
0
E
0
1
1
2
1
7
t
Figure 4.4. Water allocation policy
that maximizes total net benefits
derived from all three water-using
firms.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 86
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 87
three firms. The water allocations plus the amount that
must remain in the river, R, cannot exceed the total
amount of water Q available.
Constructing and solving a model of this problem for
various values of Q, the total amount of water available, will
define the three allocation policies as functions of Q. These
policies can be displayed as a graph, as in Figure 4.4,
showing the three best allocations given any value of Q.
This of course assumes the firms can adjust to varying
values of Q. In reality this may not be the case. (Chapter 10
examines this problem using more realistic benefit
functions that reflect the degree to which firms can adapt
to changing inputs over time.)
The model:
Maximize Net_benefit (4.18)
Subject to:
Definitional constraints:
Net_benefit Total_return Total_cost (4.19)
Total_return (12 p
1
)p
1
(20 1.5p
2
)p
2
(28 2.5p
3
)p
3
(4.20)
Total_cost 3(p
1
)
1.30
5(p
2
)
1.20
6(p
3
)
1.15
(4.21)
Production functions defining the relationship between
water allocations x
j
and production p
j
:
p
1
0.4(x
1
)
0.9
(4.22)
p
2
0.5(x
2
)
0.8
(4.23)
p
3
0.6(x
3
)
0.7
(4.24)
Water allocation restriction:
R x
1
x
2
x
3
Q (4.25)
One can first solve this model for the values of each p
j
that
maximize the total net benefits, assuming water is not
a limiting constraint. This is equivalent to finding each
individual firms maximum net benefits, assuming all the
water that is needed is available. Using calculus we can
equate the derivatives of the total net benefit function
with respect to each p
j
to 0 and solve each of the resulting
three independent equations:
Total Net_benefit [(12 p
1
)p
1
(20 1.5p
2
)p
2
(28 2.5p
3
)p
3
] [3(p
1
)
1.30
5(p
2
)
1.20
6(p
3
)
1.15
] (4.26)
Derivatives:
(Net_benefit)/p
1
0 12 2p
1
1.3(3)p
1
0.3
(4.27)
(Net_benefit)/p
2
0 20 3p
2
1.2(5)p
2
0.2
(4.28)
(Net_benefit)/p
3
0 28 5p
3
1.15(6)p
3
0.15
(4.29)
The result (rounded off ) is p
1
3.2, p
2
4.0, and
p
3
3.9 to be sold for unit prices of 8.77, 13.96, and
18.23 respectively, for a maximum net revenue of 155.75.
This would require water allocations x
1
10.2, x
2
13.6
and x
3
14.5, totaling 38.3 flow units. Any amount of
water less than 38.3 will restrict the allocation to, and
hence the production at, one or more of the three firms.
If the total available amount of water is less than that
desired, constraint Equation 4.25 can be written as an
equality, since all the water available, less any that must
remain in the river, R, will be allocated. If the available
water supplies are less than the desired 38.3 plus the
required streamflow R, then Equations 4.22 through 4.25
need to be added. These can be rewritten as equalities
since they will be binding.

32
28
24
20
16
12
8
4
0
u
n
i
t

p
r
i
c
e

(
$
)

32
28
24
20
16
12
8
4
0

32
28
24
20
16
12
8
4
0

12 -

20 - 1.5

28 - 2.5
E
0
1
1
2
1
7
u
1
1 2 3
3
2
Figure 4.5. Unit prices that
will guarantee the sale of the
specified amounts of products p
j
produced in each of the three
firms.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 87
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
p
1
0.4(x
1
)
0.9
0 (4.30)
p
2
0.5(x
2
)
0.8
0 (4.31)
p
3
0.6(x
3
)
0.7
0 (4.32)
R x
1
x
2
x
3
Q 0 (4.33)
The first three constraints, Equations 4.30, 4.31 and 4.32,
are the production functions specifying the relationships
between each water input x
j
and product output p
j
. The
fourth constraint Equation 4.33 is the restriction on the
total allocation of water. Since each of the four constraint
equations equals zero, each can be added to the net ben-
efit Equation 4.26 without changing its value. This is
done in Equation 4.34, which is equivalent to Equation
4.26. The variable L is the value of the Lagrange form of
the objective function:
L [(12 p
1
)p
1
(20 1.5p
2
)p
2
(28 2.5p
3
)p
3
]
[3(p
1
)
1.30
5(p
2
)
1.20
6(p
3
)
1.15
]

1
[p
1
0.4(x
1
)
0.9
]
2
[p
2
0.5(x
2
)
0.8
]

3
[p
3
0.6(x
3
)
0.7
]
4
[R x
1
x
2
x
3
Q]
(4.34)
Since each of the four constraint Equations 4.30 through
4.33 included in Equation 4.34 equals zero, each can be
multiplied by a variable
i
without changing the value of
Equation 4.34. These unknown variables
i
are called the
Lagrange multipliers of constraints i. The value of each
multiplier,
i
, is the marginal value of the original objec-
tive function, Equation 4.26, with respect to a change in
the value of the amount produced, p
i
, or in the case of
88 Water Resources Systems Planning and Management
constraint, Equation 4.33, the amount of water available,
Q. We will prove this shortly.
Differentiating Equation 4.34 with respect to each of
the ten unknowns and setting the resulting equations to
0 yields:
L/p
1
0 12 2p
1
1.3(3)p
1
0.3

1
(4.35)
L/p
2
0 20 3p
2
1.2(5)p
2
0.2

2
(4.36)
L/p
3
0 28 5p
3
1.15(6)p
3
0.15

3
(4.37)
L/x
1
0
1
0.9(0.4)(x
1
)
0.1

4
(4.38)
L/x
2
0
2
0.8(0.5)(x
2
)
0.2

4
(4.39)
L/x
3
0
3
0.7(0.6)(x
3
)
0.3

4
(4.40)
L/
1
0 p
1
0.4(x
1
)
0.9
(4.41)
L/
2
0 p
2
0.5(x
2
)
0.8
(4.42)
L/
3
0 p
3
0.6(x
3
)
0.7
(4.43)
L/
4
0 R x
1
x
2
x
3
Q (4.44)
These ten equations are the conditions necessary for an
optimal solution. They can be solved to obtain the values of
the ten unknown variables. The solutions to these equations
for various values of Q, (found in this case by using LINGO)
are shown in Table 4.2. (A demo version of LINGO is
available, together with its help files, at www.lindo.com.)
3.3.2. Meaning of the Lagrange Multiplier
In this example, Equation 4.34 is the objective function.
It is maximized (or minimized) by equating to zero each
E
0
2
0
8
2
0
h
Q-R x
1
x
2
x
3
available 1
10 1.2 3.7 5.1 0.46 1.44 1.88 8.0 9.2 11.0 2.8
20 4.2 7.3 8.5 1.46 2.45 2.68 4.7 5.5 6.6 1.5
30 7.5 10.7 11.7 2.46 3.34 3.37 2.0 2.3 2.9 0.6
38 10.1 13.5 14.4 3.20 4.00 3.89 0.1 0.1 0.1 0.0
38.3 10.2 13.6 14.5 3.22 4.02 3.91 0 0 0 0
p
1
p
2
p
3

1 2 3 4

marginal net benefits
water allocations
to firms
product
productions
Lagrange multipliers
2 3 1 2 3
Table 4.2. Solutions to Equations 4.35
through 4.44.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 88
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 89
of its partial derivatives with respect to each unknown
variable. Equation 4.34 consists of the original net benefit
function plus each constraint i multiplied by a weight
or multiplier
i
. This equation is expressed in monetary
units, such as dollars or euros. The added constraints are
expressed in other units: either the quantity of product
produced or the amount of water available. Thus the units
of the weights or multipliers
i
associated with these
constraints are expressed in monetary units per constraint
units. In this example the multipliers
1
,
2
and
3
represent the change in the total net benefit value of
the objective function (Equation 4.26) per unit change in
the products p
1
, p
2
and p
3
produced. The multiplier
4
represents the change in the total net benefit per unit
change in the water available for allocation, Q R.
Note in Table 4.2 that as the quantity of available
water increases, the marginal net benefits decrease. This is
reflected in the values of each of the multipliers,
i
. In
other words, the net revenue derived from a quantity of
product produced at each of the three firms, and from the
quantity of water available, are concave functions of those
quantities, as illustrated in Figure 4.2.
To review the general Lagrange multiplier approach
and derive the definition of the multipliers, consider the
general constrained optimization problem containing n
decision-variables x
j
and m constraint equations i.
Maximize (or minimize) F(X) (4.45)
subject to constraints
g
i
(X) b
i
i 1, 2, 3, , m (4.46)
where X is the vector of all x
j
. The Lagrange function L(X, )
is formed by combining Equations 4.46, each equalling
zero, with the objective function of Equation 4.45.
L(X, ) F(X)
i

i
(g
i
(X) b
i
) (4.47)
Solutions of the equations
L/x
j
0 for all decision-variables j
and
L/
i
0 for all constraints i (4.48)
are possible local optima.
There is no guarantee that a global optimum solution will
be found using calculus-based methods such as this one.
Boundary conditions need to be checked. Furthermore,
since there is no difference in the Lagrange multipliers
procedure for finding a minimum or a maximum
solution, one needs to check whether in fact a maximum
or minimum is being obtained. In this example, since
each net benefit function is concave, a maximum will
result.
The meaning of the values of the multipliers
i
at
the optimum solution can be derived by manipulation
of Equation 4.48. Taking the partial derivative of the
Lagrange function, Equation 4.47, with respect to an
unknown variable x
j
and setting it to zero results in:
L/x
j
0 F/x
j

i
(g
i
(X))/x
j
(4.49)
Multiplying each term by x
j
yields
F
i

i
(g
i
(X)) (4.50)
Dividing each term by b
k
associated with a particular
constraint, say k, defines the meaning of
k
.
F/b
k

i

i
(g
i
(X))/b
k

k
(4.51)
Equation 4.51 follows from the fact that ( g
i
(X))/b
k
0
for constraints i k and (g
i
(X))/b
k
1 for the
constraint i k. The latter is true since b
i
g
i
(X) and thus
(g
i
(X)) b
i
.
Thus from Equation 4.51, each multiplier
k
is the
marginal change in the original objective function F(X) with
respect to a change in the constant b
k
associated with the
constraint k. For non-linear problems it is the slope of the
objective function plotted against the value of b
k
.
Readers can work out a similar proof if a slack or
surplus variable, S
i
, is included in inequality constraints
to make them equations. For a less-than-or-equal
constraint g
i
(X) b
i
a squared slack variable S
i
2
can be
added to the left-hand side to make it an equation g
i
(X)
S
i
2
b
i
. For a greater-than-or-equal constraint g
i
(X) b
i
a
squared surplus variable S
i
2
can be subtracted from the
left hand side to make it an equation g
i
(X) S
i
2
b
i
.
These slack or surplus variables are squared to ensure
they are non-negative, and also to make them appear in
the differential equations.
L/S
i
0 2S
i

i
S
i

i
(4.52)
Equation 4.52 shows that either the slack or surplus
variable, S
i
, or the multiplier,
i
, will always be zero. If the
value of the slack or surplus variable S
i
is non-zero, the
constraint is redundant. The optimal solution will not be
affected by the constraint. Small changes in the values, b
i
,
of redundant constraints will not change the optimal value
wrm_ch04.qxd 8/31/2005 11:51 AM Page 89
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
of the objective function F(X). Conversely, if the constraint
is binding, the value of the slack or surplus variable S
i
will
be zero. The multiplier
i
can be non-zero if the value of
the function F(X) is sensitive to the constraint value b
i
.
The solution of the set of partial differential Equations
4.52 often involves a trial-and-error process, equating to
zero a
i
or a S
i
for each inequality constraint and solving the
remaining equations, if possible. This tedious procedure,
along with the need to check boundary solutions when non-
negativity conditions are imposed, detracts from the utility
of classical Lagrange multiplier methods for solving all but
relatively simple water resources planning problems.
4. Dynamic Programming
The water allocation problems in the previous section
considered a net-benefit function for each water-using
firm. In those examples they were continuous differen-
tiable functions, a convenient attribute if methods based
on calculus (such as hill-climbing or Lagrange multipliers)
are to be used to find the best solution. In many practical
situations these functions may not be so continuous, or
so conveniently concave for maximization or convex for
minimization, making calculus-based methods for their
solution difficult.
A possible solution method for constrained
optimization problems containing continuous and/or
discontinuous functions of any shape is called discrete
dynamic programming. Each decision-variable value can
assume one of a set of discrete values. For continuous
valued objective functions, the solution derived from
discrete dynamic programming may therefore be only an
approximation of the best one. For all practical purposes
this is not a significant limitation, especially if the intervals
between the discrete values of the decision-variables are
not too large and if simulation modelling is used to refine
the solutions identified using dynamic programming.
Dynamic programming is an approach that divides the
original optimization problem, with all of its variables,
into a set of smaller optimization problems, each of which
needs to be solved before the overall optimum solution to
the original problem can be identified. The water supply
allocation problem, for example, needs to be solved for a
range of water supplies available to each firm. Once this
is done the particular allocations that maximize the total
net benefit can be determined.
90 Water Resources Systems Planning and Management
4.1. Dynamic Programming Networks and
Recursive Equations
A network of nodes and links can represent each discrete
dynamic programming problem. Dynamic programming
methods find the best way to get to any node in that
network. The nodes represent possible discrete states that
can exist and the links represent the decisions one could
make to get from one state to another. Figure 4.6 illustrates
a portion of such a network for the three-firm allocation
problem shown in Figure 4.1. In this case the total amount
of water available, QR, to all three firms is 10.
Thus, dynamic programming models involve states,
stages and decisions. The relationships among states,
stages and decisions are represented by networks, such as
that shown in Figure 4.6. The states of the system are the
nodes and the values of the states are the numbers in the
nodes. Each node value in this example is the quantity of
water available to allocate to all remaining firms, that
is, to all connected links to the right of the node. These
state variable values typically represent some existing
condition either before making, or after having made, a
decision. The stages of the system are the separate
columns of linked nodes. The links in this example
represent possible allocation decisions for each of the
three different firms. Each stage is a separate firm.
E
0
1
1
2
1
7
w
firm 1 firm 2 firm 3
1
x
2
x
3
x
0
0
2
4
4
4
5
5
5
3
3
3
4
4
4
4
5
5
6
6
0
9
8
4
5
6
3 7
2
0
Figure 4.6. A network representing some of the possible
integer allocations of water to three water-consuming firms j.
The circles or nodes represent the discrete quantities of
water available, and the links represent feasible allocation
decisions x
j
.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 90
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 91
Each link connects two nodes, the left node value
indicating the state of a system before a decision is made,
and the right node value indicating the state of a system after
a decision is made. In this case the state of the system is the
amount of water available to allocate to the remaining firms.
In the example shown in Figure 4.6, the state and
decision-variables are represented by integer values an
admittedly fairly coarse discretization. The total amount
of water available, in addition to the amount that
must remain in the river, is 10. Note from the first row of
Table 4.2 the exact allocation solution is x
1
1.2,
x
2
3.7, and x
3
5.1. Normally we wouldnt know this
solution before solving for it using dynamic program-
ming, but since we do we can reduce the complexity of
the dynamic programming network so that the repetitive
process of finding the best solution is clearer. Thus
assume the range of x
1
is limited to integer values from
0 to 2, the range of x
2
is from 3 to 5, and the range of x
3
is from 4 to 6. These range limits are imposed here just to
reduce the size of the network. In this case, these assump-
tions will not affect or constrain the optimal solution. If
we did not make these assumptions the network would
have, after the first column of one node, three columns
of 11 nodes, one representing each integer value from 0
to 10. Finer (non-integer) discretizations would involve
even more nodes and connecting links.
The links of Figure 4.6 represent the water allocations.
Note that the link allocations, the numbers on the links,
cannot exceed the amount of water available, that is, the
number in the left node. The number in the right node is
the quantity of water remaining after an allocation has
been made. The value in the right node, state S
j1
, at the
beginning of stage j1, is equal to the value in the left
node, S
j
, less the amount of water, x
j
, allocated to firm j as
indicated on the link. Hence, beginning with a quantity of
water Q R that can be allocated to all three firms, after
allocating x
1
to Firm 1 what remains is S
2
:
Q R x
1
S
2
(4.53)
Allocating x
2
to Firm 2, leaves S
3
.
S
2
x
2
S
3
(4.54)
Finally, allocating x
3
to Firm 3 leaves S
4
.
S
3
x
3
S
4
(4.55)
Figure 4.6 shows the different values of each of these
states, S
j
, and decision-variables x
j
beginning with a
quantity Q R 10. Our task is to find the best path
through the network, beginning at the left-most node
having a state value of 10. To do this we need to know the
net benefits we will get associated with all the links (rep-
resenting the allocation decisions we could make) at each
node (state) for each firm (stage).
Figure 4.7 shows the same network as in Figure 4.6;
however the numbers on the links represent the net ben-
efits obtained from the associated water allocations. For
the three firms j 1, 2 and 3, the net benefits, NB
j
(x
j
),
associated with allocations x
j
are:
NB
1
(x
1
) maximum(12 p
1
)p
1
3(p
1
)
1.30
where p
1
0.4(x
1
)
0.9
(4.56)
NB
2
(x
2
) maximum(20 1.5p
2
)p
2
5(p
2
)
1.20
where p
2
0.5(x
2
)
0.8
(4.57)
NB
3
(x
3
) maximum(28 2.5p
3
)p
3
6(p
3
)
1.15
where p
3
0.6(x
3
)
0.7
(4.58)
respectively.
The discrete dynamic programming algorithm or pro-
cedure is a systematic way to find the best path through
this network, or any other suitable network. What makes
a network suitable for dynamic programming is the fact
E
0
1
1
2
1
7
x
firm 1 firm 2 firm 3
3 3
x NB ( )
2 2 x
NB ( )
1 1 x
NB ( )
0.0
0 3.7
6.3
8.6
8.6
8.6
3
3
3
5.7
5.7
5.7
27.9
27.9
27.9
27.9
2
2
33.7
33.7
0
9
8
4
5
6
3 7
2
0
Figure 4.7. Network representing integer value allocations of
water to three water-consuming firms. The circles or nodes
represent the discrete quantities of water available, and the
links represent feasible allocation decisions. The numbers on
the links indicate the net benefits obtained from these
particular allocation decisions.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 91
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
that all the nodes can be lined up in a sequence of
columns and each link connects a node in one column to
another node in the next column of nodes. No link passes
over or through any other column(s) of nodes. Links also
do not connect nodes in the same column. In addition,
the contribution to the overall objective value (in this
case, the total net benefits) associated with each discrete
decision (link) in any stage or for any firm is strictly a
function of the allocation of water to the firm. It is
not dependent on the allocation decisions associated with
other stages (firms) in the network.
The main challenge in using discrete dynamic
programming to solve an optimization problem is to struc-
ture the problem so that it fits this dynamic programming
network format. Perhaps surprisingly, many water resources
planning and management problems do. But it takes
practice to become good at converting optimization
problems to networks of states, stages and decisions suitable
for solution by discrete dynamic programming algorithms.
In this problem the overall objective is to:
Maximize
j

1
3
NB
j
(x
j
) (4.59)
where NB
j
(x
j
) is the net benefit associated with an
allocation of x
j
to firm j
.
. Equations 4.56, 4.57 and 4.58
define these net benefit functions. As before, the index j
represents the particular firm, and each firm is a stage for
this problem. Note that the index or subscript used in the
objective function often represents an object (like a water-
using firm) at a place in space or in a time period. These
places or time periods are called the stages of a dynamic
programming problem. Our task is to find the best path
from one stage to the next: in other words, the best
allocation decisions for all three firms.
Dynamic programming is called a multi-stage
decision-making process. Instead of deciding all three
allocations in one single optimization procedure, like
Lagrange multipliers, the dynamic programming proce-
dure divides the problem up into many optimization
problems, one for each possible discrete state (e.g.,
amount of water available) in each stage (e.g., for each
firm). Given a particular state S
j
and stage j that is, a
particular node in the network what decision (link) x
j
will result in the maximum total net benefits, designated
as F
j
(S
j
), given this state S
j
for this and all remaining stages
or firms j, j 1, j 2 ? This question must be
answered for each node in the network before one can
92 Water Resources Systems Planning and Management
find the overall best set of decisions for each stage: in
other words, the best allocations to each firm (represented
by the best path through the network) in this example.
Dynamic programming networks can be solved in
two ways beginning at the most right column of nodes
or states and moving from right to left, called the
backward-moving (but forward-looking) algorithm, or
beginning at the left most node and moving from left to
right, called the forward-moving (but backward-looking)
algorithm. Both methods will find the best path through
the network. In some problems, however, only the
backward-moving algorithm produces a useful solution.
This is especially relevant when the stages are time
periods. We often want to know what we should do next
given a particular state we are in, not what we should
have just done to get to the particular state we are in. We
cannot alter past decisions, but we can, and indeed must,
make future decisions. We will revisit this issue when
we get to reservoir operation where the stages are time
periods.
4.2. Backward-Moving Solution Procedure
Consider the network in Figure 4.7. Again, the nodes
represent the discrete states the water available to allocate
to all remaining users. The links represent particular
discrete allocation decisions. The numbers on the links are
the net benefits obtained from those allocations. We want
to proceed through the node-link network from the state of
10 at the beginning of the first stage to the end of the
network in such a way as to maximize total net benefits.
But without looking at all combinations of successive
allocations we cannot do this beginning at a state of 10.
However, we can find the best solution if we assume we
have already made the first two allocations and are at any
of the nodes or states at the beginning of the final, third,
stage with only one allocation decision remaining. Clearly
at each node representing the water available to allocate to
the third firm, the best decision is to pick the allocation
(link) having the largest net benefits.
Denoting F
3
(S
3
) as the maximum net benefits we can
achieve from the remaining amount of water S
3
, then for
each discrete value of S
3
we can find the x
3
that maximizes
F
3
(S
3
). Those shown in Figure 4.7 include:
F
3
(7) Maximum{NB
3
(x
3
)}
x
3
7, the total flow available.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 92
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 93
4 x
3
6, the allowable range of allocation
decisions.
Maximum{27.9, 31.1, 33.7} 33.7
when x
3
6 (4.60)
F
3
(6) Maximum{NB
3
(x
3
)}
x
3
6
4 x
3
6
Maximum{27.9, 31.1, 33.7} 33.7
when x
3
6 (4.61)
F
3
(5) Maximum{NB
3
(x
3
)}
x
3
5
4 x
3
6
Maximum{27.9, 31.1} 31.1
when x
3
5 (4.62)
F
3
(4) Maximum{NB
3
(x
3
)}
x
3
4
4 x
3
6
Maximum{27.9} 27.9 when x
3
4 (4.63)
These computations are shown on the network in
Figure 4.8. Note that there are no benefits to be obtained
after the third allocation, so the decision to be made for
each node or state prior to allocating water to Firm 3 is
simply that which maximizes the net benefits derived
from that last (third) allocation. In Figure 4.8 the links
representing the decisions or allocations that result in the
largest net benefits are shown with arrows.
Having computed the maximum net benefits, F
3
(S
3
),
associated with each initial state S
3
for Stage 3, we can now
move backward (to the left) to the discrete states S
2
at the
beginning of the second stage. Again, these states represent
the quantity of water available to allocate to Firms 2 and 3.
Denote F
2
(S
2
) as the maximum total net benefits obtained
from the two remaining allocations x
2
and x
3
given the
quantity S
2
water available. The best x
2
depends not only
on the net benefits obtained from the allocation x
2
but also
on the maximum net benefits obtainable after that, namely
the just-calculated F
3
(S
3
) associated with the state S
3
that
results from the initial state S
2
and a decision x
2
. As
defined in Equation 4.54, this final state S
3
in Stage 2 obvi-
ously equals S
2
x
2
. Hence for those nodes at the begin-
ning of Stage 2 shown in Figure 4.8:
F
2
(10) Maximum{NB
2
(x
2
) F
3
(S
3
10 x
2
)} (4.64)
x
2
10
3 x
2
5
Maximum{15.7 33.7, 18.6 33.7,
21.1 31.1} 52.3 when x
2
4
F
2
(9) Maximum{NB
2
(x
2
) F
3
(S
3
9 x
2
)} (4.65)
x
2
9
3 x
2
5
Maximum{15.7 33.7, 18.6 31.1,
21.1 27.9} 49.7 when x
2
4
F
2
(8) Maximum{NB
2
(x
2
) F
3
(S
3
8 x
2
)} (4.66)
x
2
8
3 x
2
5
Maximum{15.7 31.1, 18.6 27.9} 46.8
when x
2
3
These maximum net benefit functions, F
2
(S
2
), could be
calculated for the remaining discrete states from 7 to 0.
Having computed the maximum net benefits obtain-
able for each discrete state at the beginning of Stage 2,
that is, all the F
2
(S
2
) values, we can move backward or left
to the beginning of Stage 1. For this problem there is only
E
0
1
1
2
1
7
y
firm 1 firm 2 firm 3
0.0
0 3.7
6.3
8.6
8.6
8.6
3
3
3
5.7
5.7
5.7
27.9
27.9
27.9
27.9
2
2
33.7
33.7
0
9
8
4
5
6
3 7
2
0
F
49.7
( 0) 1 =
=
3.7
53.4
+
1
F
31.1
(9) =
=
18.6
49.7
+
2
F
33.7
( 0) 1 =
=
18.6
52.3
+
2
F
F
F
F
F
31.1
(8)
(4)
(5)
(6)
(7)
= =
=
=
=
=
15.7
27.9
31.1
33.7
33.7
46.8
+
2
3
3
3
3
Figure 4.8. Using the backward-moving dynamic program-
ming method for finding the maximum remaining net benefits,
F
j
(S
j
), and optimal allocations (denoted by the arrows on the
links) for each state in Stage 3, then for each state in Stage 2
and finally for the initial state in Stage 1 to obtain the optimum
allocation policy and maximum total net benefits, F
1
(10). The
minimum flow to remain in the river, R, is in addition to the ten
units available for allocation and is not shown in this network.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 93
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
one state, the state of 10 we are actually in before making
any allocations to any of the firms. In this case, the maxi-
mum net benefits, F
1
(10), we can obtain from all three
allocations given 10 units of water available, is
F
1
(10) Maximum{NB
1
(x
1
) F
2
(S
2
10 x
1
)} (4.67)
x
1
10
0 x
1
2
Maximum{0 52.3, 3.7 49.7,
6.3 46.8} 53.4 when x
1
1
Equation 4.67 is another way of expressing Equation
4.59. Their values are the same. It is the maximum net
benefits obtainable from allocating the available 10 units
of water. From Equation 4.67 we know that we will get a
maximum of 53.4 net benefits if we allocate 1 unit of
water to Firm 1. This leaves 9 units of water to allocate to
the two remaining firms. This is our optimal state at the
beginning of Stage 2. Given a state of 9 at the beginning
of Stage 2, we see from Equation 4.65 that we should allo-
cate 4 units of water to Firm 2. This leaves 5 units of
water for Firm 3. Given a state of 5 at the beginning
of Stage 3, Equation 4.62 tells us we should allocate all
5 units to Firm 3. All this is illustrated in Figure 4.8.
Compare this discrete solution with the continuous
one defined by Lagrange multipliers as shown in
Table 4.2. The exact solution, to the nearest tenth, is 1.2,
3.7, and 5.1 for x
1
, x
2
and x
3
respectively. The solution
just derived from discrete dynamic programming that
assumed only integer values is 1, 4, and 5 respectively.
To summarize, a dynamic programming model was
developed for the following problem:
Maximize Net_benefit (4.68)
Subject to:
Net_benefit Total_return Total_cost (4.69)
Total_return (12 p
1
)p
1
(20 1.5p
2
)p
2
(28 2.5p
3
)p
3
(4.70)
Total_cost 3(p
1
)
1.30
5(p
2
)
1.20
6(p
3
)
1.15
(4.71)
p
1
0.4(x
1
)
0.9
(4.72)
p
2
0.5(x
2
)
0.8
(4.73)
p
3
0.6(x
3
)
0.7
(4.74)
x
1
x
2
x
3
10 (4.75)
94 Water Resources Systems Planning and Management
The discrete dynamic programming version of this
problem required discrete states S
j
representing the
amount of water available to allocate to firms j, j 1, , .
It required discrete allocations x
j
. Next it required the
calculation of the maximum net benefits, F
j
(S
j
), that could
be obtained from all firms j, beginning with Firm 3, and
proceeding backwards as indicated in Equations 4.76
to 4.78.
F
3
(S
3
) maximum{NB
3
(x
3
)} over all x
3
S
3
, for all
discrete S
3
values between 0 and 10 (4.76)
F
2
(S
2
) maximum{NB
2
(x
2
) F
3
(S
3
)} over all x
2
S
2
and S
3
S
2
x
2
, 0 S
2
10 (4.77)
F
1
(S
1
) maximum{NB
1
(x
1
) F
2
(S
2
)} over all x
1
S
1
and S
2
S
1
x
1
and S
1
10 (4.78)
To solve for F
1
(S
1
) and each optimal allocation x
j
we must
first solve for all values of F
3
(S
3
). Once these are known
we can solve for all values of F
2
(S
2
). Given these F
2
(S
2
)
values, we can solve for F
1
(S
1
). Equations 4.76 need to
be solved before Equations 4.77 can be solved, and
Equations 4.77 need to be solved before Equations 4.78
can be solved. They need not be solved simultaneously,
and they cannot be solved in reverse order. These three
equations are called recursive equations. They are defined
for the backward-moving dynamic programming solution
procedure.
There is a correspondence between the non-linear
optimization model defined by Equations 4.68 to 4.75
and the dynamic programming model defined by the
recursive Equations 4.76, 4.77 and 4.78. Note that F
3
(S
3
)
in Equation 4.76 is the same as:
F
3
(S
3
) Maximum NB
3
(x
3
) (4.79)
Subject to:
x
3
S
3
(4.80)
where NB
3
(x
3
) is defined in Equation 4.58.
Similarly, F
2
(S
2
) in Equation 4.62 is the same as:
F
2
(S
2
) Maximum NB
2
(x
2
) NB
3
(x
3
) (4.81)
Subject to:
x
2
x
3
S
2
(4.82)
where NB
2
(x
2
) and NB
3
(x
3
) are defined in Equations 4.57
and 4.58.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 94
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 95
Finally, F
1
(S
1
) in Equation 4.63 is the same as:
F
1
(S
1
) Maximum NB
1
(x
1
) NB
2
(x
2
) NB
3
(x
3
) (4.83)
Subject to:
x
1
x
2
x
3
S
1
10 (4.84)
where NB
1
(x
1
), NB
2
(x
2
) and NB
3
(x
3
) are defined in
Equations 4.56, 4.57 and 4.58.
Alternatively, F
3
(S
3
) in Equation 4.76 is the same as:
F
3
(S
3
) Maximum(28 2.5p
3
)p
3
6(p
3
)
1.15
(4.85)
Subject to:
p
3
0.6(x
3
)
0.7
(4.86)
x
3
S
3
(4.87)
Similarly, F
2
(S
2
) in Equation 4.77 is the same as:
F
2
(S
2
) Maximum(20 1.5p
2
)p
2
(28 2.5p
3
)
p
3
5(p
2
)
1.20
6(p
3
)
1.15
(4.88)
Subject to:
p
2
0.5(x
2
)
0.8
(4.89)
p
3
0.6(x
3
)
0.7
(4.90)
x
2
x
3
S
2
(4.91)
Finally, F
1
(S
1
) in Equation 4.78 is the same as:
F
1
(S
1
) Maximum(12 p
1
)p
1
(20 1.5p
2
)p
2
(28 2.5p
3
)p
3
[3(p
1
)
1.30
5(p
2
)
1.20
6(p
3
)
1.15
] (4.92)
Subject to:
p
1
0.4(x
1
)
0.9
(4.93)
p
2
0.5(x
2
)
0.8
(4.94)
p
3
0.6(x
3
)
0.7
(4.95)
x
1
x
2
x
3
S
1
10 (4.96)
The transition function of dynamic programming defines
the relationship between two successive states S
j
and S
j1
and the decision x
j
. In the above example these transition
functions are defined by Equations 4.53, 4.54 and 4.55,
or, in general terms for all firms j, by:
S
j1
S
j
x
j
(4.97)
4.3. Forward-Moving Solution Procedure
We have just described the backward-moving dynamic
programming algorithm. In that approach at each node
(state) in each stage we calculated the best value of the
objective function that can be obtained from all further or
remaining decisions. Alternatively one can proceed forward,
that is, from left to right, through a dynamic programming
network. For the forward-moving algorithm at each node
we need to calculate the best value of the objective function
that could be obtained from all past decisions leading to that
node or state. In other words, we need to find how best to
get to each state S
j1
at the end of each stage j.
Returning to the allocation example, define f
j
(S
j1
) as
the maximum net benefits from the allocation of water to
firms 1, 2, , j, given a state S
j1
after having made those
allocations. For this example we begin the forward-
moving, but backward-looking, process by selecting each
of the ending states in the first stage j 1 and finding the
best way to have arrived at (or to have achieved) those
ending states. Since in this example there is only one way
to get to each of those states, as shown in Figure 4.6 or
4.7, the allocation decisions are obvious.
f
1
(S
2
) maximum{NB
1
(x
1
)}
x
1
10 S
2
(4.98)
Hence, f
1
(S
2
) is simply NB
1
(10 S
2
). Once the values
for all f
1
(S
2
) are known for all discrete S
2
between 0 and 10,
move forward (to the right) to the end of Stage 2 and find
the best allocations to have made given each final state S
3
.
f
2
(S
3
) maximum{NB
2
(x
2
) f
1
(S
2
)}
0 x
2
10 S
3
S
2
S
3
x
2
(4.99)
Once the values of all f
2
(S
3
) are known for all discrete states
S
3
between 0 and 10, move forward to Stage 3 and find the
best allocations to have made given each final state S
4
.
f
3
(S
4
) maximum{NB
3
(x
3
) f
2
(S
3
)}
for all discrete S
4
between 0 and 10.
0 x
3
10 S
4
S
3
S
4
x
3
(4.100)
Figure 4.9 illustrates a portion of the network represented
by Equations 4.98 through 4.100, and the f
j
(S
j1
) values.
From Figure 4.9, note the highest total net benefits are
obtained by ending with 0 remaining water at the end of
Stage 3. The arrow tells us that if we are to get to that state
optimally, we should allocate 5 units of water to Firm 3.
Thus we must begin Stage 3, or end Stage 2, with
10 5 5 units of water. To get to this state at the end
of Stage 2 we should allocate 4 units of water to Firm 2.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 95
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The arrow also tells us we should have had 9 units of
water available at the end of Stage 1. Given this state of 9
at the end of Stage 1, the arrow tells us we should allocate
1 unit of water to Firm 1. This is the same allocation
policy as obtained using the backward-moving algorithm.
4.4. Numerical Solutions
The application of discrete dynamic programming to
most practical problems will require writing some software.
There are no general dynamic programming computer
96 Water Resources Systems Planning and Management
programs available that will solve all dynamic program-
ming problems. Thus any user of dynamic programming
will need to write a computer program to solve a particu-
lar problem. Most computer programs written for solving
specific dynamic programming problems create and store
the solutions of the recursive equations in tables. Each
stage is a separate table, as shown in Tables 4.3, 4.4 and
4.5 for this example water allocation problem. These
tables apply to only a part of the entire problem, namely
that part of the network shown in Figures 4.8 and 4.9. The
backward solution procedure is used.
Table 4.3 contains the solutions of Equations 4.60
to 4.63 for the third stage. Table 4.4 contains the solu-
tions of Equations 4.64 to 4.66 for the second stage.
Table 4.5 contains the solution of Equation 4.67 for the
first stage.
From Table 4.5 we see that, given 10 units of water
available, we will obtain 53.4 net benefits and to get this
we should allocate 1 unit to Firm 1. This leaves 9 units of
water for the remaining two allocations. From Table 4.4
we see that for a state of 9 units of water available we
should allocate 4 units to Firm 2. This leaves 5 units.
From Table 4.3 for a state of 5 units of water available we
see we should allocate all 5 of them to Firm 3.
Performing these calculations for various discrete total
amounts of water available, say from 0 to 38 in this
example, will define an allocation policy (such as the
one shown in Figure 4.6 for a different allocation
problem) for situations when the total amount of water is
less than that desired by all the firms. This policy can then
be simulated using alternative time series of available
amounts of water, such as streamflows, to obtain
estimates of the time series (or statistical measures of
those time series) of net benefits obtained by each firm,
assuming the allocation policy is followed over time.
0.0
0 3.7
6.3
8.6
8.6
8.6
3
3
3
5.7
5.7
5.7
27.9
27.9
27.9
27.9
2
2
33.7
33.7
0
9
8
4
5
6
3 7
2
0
F
F
F
F
F
F
F
F
F
F
F
(4)
(0)
(8)
(5)
( ) 1
(9)
(6)
(2)
( 0) 1
(7)
(3)
=
=
=
=
=
=
=
=
=
=
=
24.9
53.4
6.3
22.3
50.5
3.7
19.4
47.3
0
15.7
43.6
2
3
1
2
3
1
2
3
1
2
3
E
0
1
1
2
1
7
z
firm 1 firm 2 firm 3
Figure 4.9. Using the forward-moving dynamic programming
method for finding the maximum accumulated net benefits,
F
j
(S
j+1
), and optimal allocations (denoted by the arrows on the
links) that should have been made to reach each ending state,
beginning with the ending states in Stage 1, then for each ending
state in Stage 2 and finally for the ending states in Stage 3.
E
0
2
0
8
2
0
j
4
7 33.7 6
6 33.7 6
5 31.1 5
4 27.9 4
state S 5 6
3
27.9
27.9
27.9
27.9 33.7
---
---
31.1
31.1
31.1
33.7
F
3
S
3
( )
x
3
decisions: x
3
remaining net benefits
3
NB
3
S
( )
---
Table 4.3. Computing the values of
F
3
(S
3
) and optimal allocations x
3
for all
states S
3
in Stage 3.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 96
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 97
4.5. Dimensionality
One of the limitations of dynamic programming is handling
multiple state variables. In our water allocation example we
had only one state variable: the total amount of water
available. We could have enlarged this problem to include
other types of resources the firms require to make their
products. Each of these state variables would need to be
discretized. If, for example, only m discrete values of each
state variable are considered, for n different state variables
(e.g., types of resources) there are m
n
different combinations
of state variable values to consider at each stage. As the
number of state variables increases, the number of discrete
combinations of state variable values increases exponen-
tially. This is called dynamic programmings curse of dimen-
sionality. It has motivated many researchers to search for
ways of reducing the number of possible discrete states
required to find an optimal solution to large multi-state
problems.
4.6. Principle of Optimality
The solution of dynamic programming models or
networks is based on a principal of optimality (Bellman,
1957). The backward-moving solution algorithm is based
on the principal that no matter what the state and stage
(i.e., the particular node you are at), an optimal policy is
one that proceeds forward from that node or state and
stage optimally. The forward-moving solution algorithm
is based on the principal that no matter what the state and
stage (i.e., the particular node you are at), an optimal
policy is one that has arrived at that node or state and
stage in an optimal manner.
This principle of optimality is a very simple concept
but requires the formulation of a set of recursive equa-
tions at each stage. It also requires that either in the
last stage (j J) for a backward-moving algorithm, or
in the first stage (j 1) for a forward-moving algorithm,
the future value functions, F
J1
(S
J1
), associated with
the ending state variable values, or past value functions,
f
0
(S
1
), associated with the beginning state variable values,
respectively, all equal some known value. Usually that
value is 0 but not always. This condition is required in
order to begin the process of solving each successive
recursive equation.
4.7. Additional Applications
Among the common dynamic programming applications
in water resources planning are water allocations to
multiple uses, infrastructure capacity expansion, and
reservoir operation. The previous three-user water alloca-
tion problem (Figure 4.1) illustrates the first type of appli-
cation. The other two applications are presented below.
E
0
2
0
8
2
0
k
3
10 52.3 4
9 49.7 4
8 46.8 3
state S 4 5
2
15.7 + 33.7
15.7 + 31.1
15.7 + 33.7
18.6 + 33.7
18.6 + 27.9
18.6 + 31.1
21.1 + 31.1
21.1 + 27.9
F
2
S
2
( )
x
2
decisions: x
2
remaining net benefits
2
NB
2
S
( )
3
F
3 2
S S
( )
+ = -
2
x
---
Table 4.4. Computing the values of
F
2
(S
2
) and optimal allocations x
2
for all
states S
2
in Stage 2.
E
0
2
0
8
2
0
m
0
10 53.4 1
state S 1 2
2
0 + 52.3 3.7 + 49.7 6.3 + 46.8
F
2
S
2
( )
x
2
decisions: x
2
remaining net benefits
1
NB S
( )
2
F S S
( )
+ = -x
1 2 1 1
Table 4.5. Computing the values of
F
1
(S
1
) and optimal allocations x
1
for all
states S
1
in Stage 1.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 97
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
4.7.1. Capacity Expansion
How much infrastructure should be built, when and why?
Consider a municipality that must plan for the future
expansion of its water supply system or some component
of that system, such as a reservoir, aqueduct, or treatment
plant. The capacity needed at the end of each future
period t has been estimated to be D
t
. The cost, C
t
(s
t
, x
t
), of
adding capacity x
t
in each period t is a function of that
added capacity as well as of the existing capacity s
t
at the
beginning of the period. The planning problem is to find
that time sequence of capacity expansions that minimizes
the present value of total future costs while meeting the
predicted capacity demand requirements. This is the usual
capacity-expansion problem.
This problem can be written as an optimization model:
The objective is to minimize the present value of the total
cost of capacity expansion.
Minimize
t
C
t
(s
t
, x
t
) (4.101)
where C
t
(s
t
, x
t
) is the present value of the cost of capacity
expansion x
t
in period t given an initial capacity of s
t
.
The constraints of this model define the minimum
required final capacity in each period t, or equivalently
the next periods initial capacity, s
t1
, as a function of the
known existing capacity s
1
and each expansion x
t
up
through period t.
s
t1
s
t

t
1
x

for t 1, 2, , T (4.102)
Alternatively these equations may be expressed by a series
of continuity relationships:
s
t1
s
t
x
t
for t 1, 2, , T (4.103)
In this problem, the constraints must also ensure that the
actual capacity s
t1
at the end of each future period t is no
less than the capacity required D
t
at the end of that period.
s
t1
D
t
for t 1, 2, , T (4.104)
There may also be constraints on the possible expansions
in each period defined by a set
t
of feasible capacity
additions in each period t:
x
t

t
(4.105)
Figure 4.10 illustrates this type of capacity-expansion
problem. The question is how much capacity to add and
when. It is a significant problem for several reasons.
One is that the cost functions C
t
(s
t
, x
t
) typically exhibit
98 Water Resources Systems Planning and Management
fixed costs and economies of scale, as illustrated in
Figure 4.11. Each time any capacity is added there are
fixed as well as variable costs incurred. Fixed and variable
costs that show economies of scale (decreasing average
costs associated with increasing capacity additions) moti-
vate the addition of excess capacity, capacity not needed
immediately but expected to be needed eventually to
meet an increased demand for additional capacity in the
future.
The problem is also important because any estimates
made today of future demands, costs and interest rates are
demand & possible capacity additions
demand
for capacity
capacity
time
E
0
2
0
1
0
3
a
Figure 4.10. A demand projection (solid line) and a possible
capacity-expansion schedule for meeting that projected
demand over time.
C
o
c
o
s
t
added capacity
E
0
2
0
1
0
3
b
Figure 4.11. Typical cost function for additional capacity given
an existing capacity. The cost function shows the fixed costs,
C
0
, required if additional capacity is to be added, and the
economies of scale associated with the concave portion of the
cost function.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 98
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 99
likely to be wrong. The future is uncertain. Its uncertain-
ties increase the further we look into the future. Capacity-
expansion planners need to consider the future if their
plans are to be cost-effective. Just how far into the future
do they need to look? And what about the uncertainty in
all future estimates? These questions will be addressed
after showing how the problem can be solved for any
fixed-planning horizon and estimates of future demands,
interest rates and costs.
The constrained optimization model defined by
Equations 4.101 to 4.105 can be restructured as a multi-
stage decision-making process and solved using either a
forward or backward-moving discrete dynamic program-
ming solution procedure. The stages of the model will be
the time periods t. The states will be either the capacity
s
t1
at the end of a stage or period t if a forward-moving
solution procedure is adopted, or the capacity s
t
, at the
beginning of a stage or period t if a backward-moving
solution procedure is used.
A network of possible discrete capacity states and deci-
sions can be superimposed onto the demand projection of
Figure 4.10, as shown in Figure 4.12.
The solid blue circles in Figure 4.12 represent possible
discrete states, S
t
, of the system, the amounts of additional
capacity existing at the end of each period t 1 or equiv-
alently at the beginning of period t.
Consider first a forward-moving dynamic program-
ming algorithm. To implement this, define f
t
(s
t1
) as the
minimum cost of achieving a capacity s
t1
, at the end of
period t. Since at the beginning of the first period t 1,
the accumulated least cost is 0, f
0
(s
1
) 0.
Hence, for each final discrete state s
2
in stage
t 1 ranging from D
1
to the maximum demand D
T
,
define
f
1
(s
2
) min{C
1
(s
1
, x
1
)} in which the discrete
x
1
s
2
and s
1
0 (4.106)
Moving to stage t 2, for the final discrete states s
3
rang-
ing from D
2
to D
T
,
f
2
(s
3
) min{C
2
(s
2
, x
2
) f
1
(s
2
)} over all discrete x
2
between 0 and s
3

D
1
where s
2
s
3

x
2
(4.107)
Moving to stage t 3, for the final discrete states s
4
ranging from D
3
to D
T
,
f
3
(s
4
) min{C
3
(s
3
, x
3
) f
2
(s
3
)} over all discrete x
3
between 0 and s
4

D
2
where s
3
s
4

x
3
(4.108)
In general for all stages t between the first and last:
f
t
(s
t1
) min{C
t
(s
t
, x
t
) f
t1
(s
t
)} over all discrete x
t
between 0 and s
t1

D
t1
where s
t
s
t1

x
t
(4.109)
For the last stage t T and for the final discrete
state s
T1

D
T
,
f
T
(s
T1
) min{C
T
(s
T
, x
T
) f
T1
(s
T
)} over all discrete x
T
between 0 and D
T

D
T1
where s
T
s
T1

x
T
(4.110)
The value of f
T
(s
T1
) is the minimum present value of the
total cost of meeting the demand for T time periods.
To identify the sequence of capacity-expansion decisions
that results in this minimum present value of the total
cost requires backtracking to collect the set of best
decisions x
t
for all stages t. A numerical example will
illustrate this.
A numerical example
Consider the five-period capacity-expansion problem
shown in Figure 4.12. Figure 4.13 is the same network
d
e
m
a
n
d

&

p
o
s
s
i
b
l
e
c
a
p
a
c
i
t
y

a
d
d
i
t
i
o
n
s
t = 1
time
2 3 4 5
E
0
2
0
1
0
3
c
Figure 4.12. Network of discrete capacity-expansion
decisions (links) that meet the projected demand.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 99
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
with the present value of the expansion costs on each
link. The values of the states, the existing capacities,
represented by the nodes, are shown on the left vertical
axis. The capacity-expansion problem is solved on Figure
4.14 using the forward-moving algorithm.
From the forward-moving solution to the dynamic
programming problem shown in Figure 4.14, the present
value of the cost of the optimal capacity-expansion sched-
ule is 23 units of money. Backtracking (moving left
against the arrows) from the farthest right node, this
schedule adds 10 units of capacity in period t 1, and
15 units of capacity in period t 3.
Next consider the backward-moving algorithm
applied to this capacity-expansion problem. The general
recursive equation for a backward-moving solution is
F
t
(s
t
) minimum{C
t
(s
t
, x
t
) F
t1
(s
t1
)} over all discrete
x
t
from D
t
s
t
to D
T
s
t
for all discrete states s
t
from D
t1
to D
T
(4.111)
where F
T1
(D
T
) 0 and as before each cost function is
the discounted cost.
Once again, as shown in Figure 4.15, the minimum
total present value cost is 23 if 10 units of additional
100 Water Resources Systems Planning and Management
capacity are added in period t 1 and 15 in period
t 3.
Now we look to the question of the uncertainty of
future demands, D
t
, discounted costs, C
t
(s
t
, x
t
), as well as
to the fact that the planning horizon T is only 5 time
periods. Of importance is just how these uncertainties
and finite planning horizon affect our decisions. While
the model gives us a time series of future capacity-
expansion decisions for the next 5 time periods, what
is important to decision-makers is what additional
capacity to add now, not what capacity to add in future
periods. Does the uncertainty of future demands and
costs and the 5-period planning horizon affect this first
decision, x
1
? This is the question to address. If the answer
is no, then one can place some confidence in the value
of x
1
. If yes, then more study may be warranted to deter-
mine which demand and cost scenario to assume, or, if
applicable, how far into the future to extend the planning
horizon.
Future capacity-expansion decisions in Periods 2, 3
and so on can be based on updated information and
analyses carried out closer to the time those decisions
are to be made. At those times, the forecast demands and
time
a
d
d
e
d

c
a
p
a
c
i
t
y
t = 1 2 3 4 5
25
E
0
2
0
1
0
3
d
0 0 0 0
0
9
7
0
0
8 11
14 15 21 27 20
8 10 13
9
0
10
0
7 6 5 4
20
15
10
5
12
9
6
9
Figure 4.13. A discrete capacity-
expansion network showing the
present value of the expansion
costs associated with each
feasible expansion decision.
Finding the best path through
the network can be done using
forward or backward-moving
discrete dynamic programming.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 100
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 101
time
a
d
d
e
d

c
a
p
a
c
i
t
y
t = 1 2 3 4 5
25
E
0
2
0
1
0
3
f
0
12 12
17
23
9
4 4 4
0 0 0 0
9
4
0
0 0 0
0
9
7
0
0
8 11
14 15 21 27 20
12
8 10 13
9
0
9
6 10
9
0
7 6 5 4
20
15
10
5
Figure 4.15. A capacity-
expansion example, showing the
results of a backward-moving
dynamic programming
algorithm. The numbers next to
the nodes are the minimum
remaining cost to have the
particular capacity required at
the end of the planning horizon
given the existing capacity of the
state.
time
a
d
d
e
d

c
a
p
a
c
i
t
y
t = 1 2 3 4 5
25
E
0
2
0
7
2
3
o
0 0 0 0
0
9
7
0
0
8 11
14 15 21 27 20
12
8 10 13
9
0
9
6 10
9
0
7 6 5 4
20
15
10
5
11 11
9
15
21 20 20
24 23 23 23
15
21
27
0
Figure 4.14. A capacity-
expansion example, showing the
results of a forward-moving
dynamic programming
algorithm. The numbers next to
the nodes are the minimum cost
to have reached that particular
state at the end of the particular
time period t.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 101
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
economic cost estimates can be updated and the planning
horizon extended, as necessary, to a period that again
does not affect the immediate decision. Note that in
the example problem shown in Figures 4.14 and 4.15, the
use of 4 periods instead of 5 would have resulted in the
same first-period decision. There is no need to extend
the analysis to 6 or more periods.
To summarize: What is important to decision-makers
now is what additional capacity to add now. While the
current periods capacity addition should be based on the
best estimates of future costs, interest rates and demands,
once a solution is obtained for the capacity expansion
required for this and all future periods up to some distant
time horizon, one can then ignore all but that first
decision, x
1
: that is, what to add now. Then just before
the beginning of the second period, the forecasting and
analysis can be redone with updated data to obtain an
updated solution for what capacity to add in Period 2,
and so on into the future. Thus, these sequential decision-
making dynamic programming models can be designed to
be used in a sequential decision-making process.
102 Water Resources Systems Planning and Management
4.7.2. Reservoir Operation
Reservoir operators need to know how much water to
release and when. Reservoirs designed to meet demands for
water supplies, recreation, hydropower, the environment
and/or flood control need to be operated in ways that meet
those demands in the most reliable and effective manner.
Since future inflows or storage volumes are uncertain,
the challenge, of course, is to determine the best reservoir
release or discharge for a variety of possible inflows and
storage conditions.
Reservoir release policies are often defined in the form
of what are called rule curves. Figure 4.16 illustrates a
rule curve for a single reservoir on the Columbia River in
the northwestern United States. It combines components
of two basic types of release rules. In both of these, the
year is divided into various within-year time periods.
There is a specified release for each value of storage in
each within-year time period. Usually higher storage
zones are associated with higher reservoir releases. If
the actual storage is relatively low, then less water is
1300 CFS max for Stl. Hd. fishing
6000
5000
4000
3000
2000
1000
0
1600
full pool if possible full pool if possible
Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul
1550
1500
1450
max pool 1600'
min pool 1445'
assured refill curve
operating curve for power
minimum pool for boat access and campsite use during recreation season
required for flood control
additional FC requirement depends on forecast
E
0
2
0
1
0
3
g
control for
winter Stl. Hd. fishing
and goose nesting
10 acre feet of storage
3
feet elevation
Figure 4.16. An example
reservoir rule curve specifying
the storage targets and some of
the release constraints, given
the particular current storage
volume and time of year. The
release constraints also include
the minimum and maximum
release rates and the maximum
downstream channel rate of
flow and depth changes that can
occur in each month.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 102
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 103
usually released so as to hedge against a continuing water
shortage or drought.
Release rules may also specify the desired storage level
for the time of year. The operator is to release water as
necessary to achieve these target storage levels. Maximum
and minimum release constraints might also be specified
that may affect how quickly the target storage levels can
be met. Some rule curves define multiple target storage
levels depending on hydrological (e.g., snow pack) con-
ditions in the upstream watershed, or on the forecast
climate conditions as affected by ENSO cycles, solar geo-
magnetic activity, ocean currents and the like. (There is
further discussion of this topic in Appendix C).
Reservoir release rule curves for a year, such as that
shown in Figure 4.16, define a policy that does not vary
from one year to the next. The actual releases will vary,
however, depending on the inflows and storage volumes
that actually occur. The releases are often specified inde-
pendently of future inflow forecasts. They are typically
based only on existing storage volumes and within-year
periods.
Release rules are typically derived from trial and error
simulations. To begin these simulations it is useful to have
at least an approximate idea of the expected impact of
different alternative policies on various system perform-
ance measures or objectives. Policy objectives could be
the maximization of expected annual net benefits from
downstream releases, reservoir storage volumes, hydro-
electric energy and flood control, or the minimization of
deviations from particular release, storage volume, hydro-
electric energy or flood flow targets or target ranges.
Discrete dynamic programming can be used to obtain
initial estimates of reservoir operating policies that meet
these and other objectives. The results of discrete
dynamic programming can be expressed in the form
shown in Figure 4.16.
A numerical example
Consider, as a simple example, a reservoir having an
active storage capacity of 20 million cubic metres, or for
that matter any specified volume units. The active storage
volume in the reservoir can vary between 0 and 20. To
use discrete dynamic programming, this range of possible
storage volumes must be divided into a set of discrete
values. These will be the discrete state variable values. In
this example let the range of storage volumes be divided
into intervals of 5 storage volume units. Hence, the initial
storage volume, S
t
, can assume values of 0, 5, 10, 15 and
20 for all periods t.
For each period t, let Q
t
be the mean inflow, L
t
(S
t
, S
t1
)
the evaporation and seepage losses that depend on the
storage volume in the reservoir, and R
t
the release or
discharge from the reservoir. Each variable is expressed as
volume units for the period t.
Storage volume continuity requires that in each period
t the initial active storage volume, S
t
, plus the inflow, Q
t
,
less the losses, L
t
(S
t
, S
t1
), and release, R
t
, equals the final
storage, or equivalently the initial storage, S
t1
, in the
following period t 1. Hence:
S
t
Q
t
R
t
L
t
(S
t
, S
t1
) S
t1
for each period t.
(4.112)
To satisfy the requirement (imposed for convenience in
this example) that each storage volume variable be a dis-
crete value ranging from 0 to 20 in units of 5, the releases,
R
t
, must be such that when Q
t
R
t
L
t
(S
t
, S
t1
) is added
to S
t
the resulting value of S
t1
is one of the 5 discrete
numbers between 0 and 20.
Assume four within-year periods t in each year
(kept small for this illustrative example). In these four
seasons assume the mean inflows, Q
t
, are 24, 12, 6 and
18 respectively. Table 4.6 defines the evaporation and
seepage losses based on different discrete combinations of
initial and final storage volumes for each within-year
period t.
Rounding these losses to the nearest integer value,
Table 4.7 shows the net releases associated with initial
and final storage volumes. They are computed using
Equation 4.112.
The information in Table 4.7 allows us to draw a
network representing each of the discrete storage volume
states (the nodes), and each of the feasible releases (the
links). This network for the four seasons t in the year is
illustrated in Figure 4.17.
This reservoir-operating problem is a multistage
decision-making problem. As Figure 4.17 illustrates, at
the beginning of any season t, the storage volume can be
in any of the five discrete states. Given that state, a release
decision is to be made. This release will depend on the
state: the initial storage volume and the mean inflow, as
well as the losses that may be estimated based on the
wrm_ch04.qxd 8/31/2005 11:51 AM Page 103
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
initial and final storage volumes, as in Table 4.6. The
release will also depend on what is to be accomplished
that is, the objectives to be satisfied.
For this example, assume there are various targets that
water users would like to achieve. Downstream water
users want reservoir operators to meet their flow targets.
Individuals who use the lake for recreation want the reser-
voir operators to meet storage volume or storage level
targets. Finally, individuals living on the downstream
floodplain want the reservoir operators to provide storage
104 Water Resources Systems Planning and Management
capacity for flood protection. Table 4.8 identifies these
different targets that are to be met, if possible, for the
duration of each season t.
Clearly, it will not be possible to meet all these storage
volume and release targets in all four seasons, given
inflows of 24, 12, 6 and 18, respectively. Hence, the
objective in this example will be to do the best one can:
to minimize a weighted sum of squared deviations from
each of these targets. The weights reflect the relative
importance of meeting each target in each season t. Target
deviations are squared to reflect the fact that the marginal
l
o
s
s
e
s
0
10
15
20
5
p
e
r
i
o
d
t

=

3
l
o
s
s
e
s
0
10
15
20
5
p
e
r
i
o
d
t

=

4
0
10
15
20
5
0 0.1 0.3 0.4 0.5
0.3 0.4 0.6 0.8 1.0
0.4 0.6 0.8 1.0 1.2
0.6 0.8 1.0 1.2 1.4
0 0.5 0.7 0.8 1.0
0.5 0.7 0.8 1.0 1.2
0.7 0.8 1.0 1.2 1.4
0.8 1.0 1.2 1.4 1.6
1.0 1.2 1.4 1.6 1.8
0.1 0.2 0.6 0.8 0.4
0 0.7 0.9 1.0 1.2
0.7 0.9 1.0 1.2 1.4
0.9 1.0 1.2 1.4 1.6
1.0 1.2 1.4 1.6 1.8
1.2 1.4 1.6 1.8 2.0
0 0.1 0.2 0.3 0.4
0.1 0.2 0.3 0.4 0.5
0.2 0.3 0.4 0.5 0.6
0.3 0.4 0.5 0.6 0.7
0.4 0.5 0.6 0.7 0.8
0 5 10 15 20
final storage initial
storage
l
o
s
s
e
s
l
o
s
s
e
s
p
e
r
i
o
d
t
=
0
10
15
20
5
p
e
r
i
o
d
t

=

2
E
0
2
0
8
2
0
n
Table 4.6. Evaporation and seepage losses based on initial
and final storage volumes for example reservoir operating
problem.
0
10
15
20
5
p
e
r
i
o
d
t

=

3
0
10
15
20
5
p
e
r
i
o
d
t

=

4
0
10
15
20
5
0 5 10 15 20
final storage initial
storage
r
e
l
e
a
s
e
s
r
e
l
e
a
s
e
s
r
e
l
e
a
s
e
s
r
e
l
e
a
s
e
s
p
e
r
i
o
d
t

=
0
10
15
20
5
p
e
r
i
o
d
t

=

2
E
0
2
0
8
2
0
o
i
n
f
l
o
w
Q
1
=

2
4
i
n
f
l
o
w
Q
2
=

1
2
i
n
f
l
o
w
Q
3
=

6
i
n
f
l
o
w
Q
4
=

1
8
24 19 14 8 3
29 24 18 13 8
34 29 23 18 13
39 33 28 23 18
43 38 33 28 23
12 7 1 - -
17 11 6 1 -
21 1 6 11 6 1
26 21 16 11 5
31 26 21 15 10
6 0 - - -
10 5 0 - -
15 10 5 0 -
20 15 10 4 -
25 20 14 9 4
18 13 8 3 -
23 18 13 8 3
28 23 18 13 7
33 28 23 17 12
38 33 27 22 17
Table 4.7. Discrete releases associated with initial and final
storage volumes.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 104
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 105
period or
season t
storage targets
TS
R
t
, TS
F
t
release target
TR
t
15 flood control
2 20 15 recreation
3 20 20 recreation
4 15
>
>
>
>
E
0
2
0
8
2
0
p
10
Table 4.8. Storage volume and release
targets for the example reservoir operation
problem.
20, 24 20, 12 20, 60 20, 18 20, 24
15, 24 15, 12 15, 60 15, 18 15, 24
10, 24 10, 12 10, 60 10, 18 10, 24
5, 24 5, 12 5, 60 5, 18 05, 24
0, 24 0, 12 0, 60 0, 18 0, 24
release: R R R R
2 3 1 4
1 2 3 4 stage:
states:
t t t t = = = =
E
0
2
0
1
0
3
h
possible release decision
t
S
t
, Q
Figure 4.17. Network
representation of the four-
season reservoir release
problem. Given any initial
storage volume S
t
at the
beginning of a season t, and an
expected inflow of Q
t
during
season t, the links indicate the
possible release decisions
corresponding to those in
Table 4.7.
losses associated with deviations increase with increasing
deviations. Small deviations are not as serious as
larger deviations, and it is better to have numerous small
deviations rather than a single larger one.
During the recreation season (Periods 2 and 3),
deviations below or above the recreation storage lake
volume targets are damaging. During the flood season
(Period 1), any storage volume in excess of the flood
wrm_ch04.qxd 8/31/2005 11:51 AM Page 105
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
106 Water Resources Systems Planning and Management
control storage targets of 15 reduces the flood storage
capacity. Deviations below that flood control target are
not penalized. Flood control and recreation storage
targets during each season t apply throughout the season,
thus they apply to the initial storage S
t
as well as to the
final storage S
t1
.
The objective is to minimize the sum of total weighted
squared deviations, TSD
t
, over all seasons t from now on
into the future:
Minimize
t
TSD
t
(4.113)
where
TSD
t
ws
t
[(TS
R
t
S
t
)
2
(TS
R
t
S
t1
)
2
]
wfs
t
[(ES
t
)
2
(ES
t1
)
2
] wr
t
[DR
t
2
] (4.114)
In the above equation, when t 4, the last period of the
year, t 1 1, is the first period in the following year.
Each ES
t
is the storage volume in excess of the flood stor-
age target volume, TS
F
t
. Each DR
t
is the difference between
the actual release, R
t
, and the target release TR
t
, when the
release is less than the target. The excess storage, ES
t
, at
the beginning of each season t can be defined by the
constraint:
S
t
TS
F
t
ES
t
(4.115)
for periods t 1 and 2, and the deficit release, DR
t
,
during period t can be defined by the constraint:
R
t
TR
t
DR
t
(4.116)
This constraint applies for all periods t.
The first component of the right side of Equation
4.114 defines the weighted squared deviations from a
recreation storage target, TS
R
t
, at the beginning and end of
season t. In this example the recreation season is during
Periods 2 and 3. The weights ws
t
associated with the
recreation component of the objective are 1 in Periods 2
and 3. In Periods 1 and 4 the weights ws
t
are 0.
The second component of Equation 4.114 is for flood
control. It defines the weighted squared deviations
associated with storage volumes in excess of the flood
control target volume, TS
F
t
, at the beginning and end of
the flood season, period t 1. In this example, the
weights wfs
t
are 1 for Period 1 and 0 for Periods 2, 3 and
4. Note the conflict between flood control and recreation
at the end of Period 1 or equivalently at the beginning of
Period 2.
Finally, the last component of Equation 4.114 defines
the weighted squared deficit deviations from a release
target, TR
t
, In this example all release weights wr
t
equal 1.
Associated with each link in Figure 4.17 is the release,
R
t
, as defined in Table 4.7. Also associated with each link
is the sum of weighted squared deviations, TSD
t
, that
result from the particular initial and final storage volumes
and the storage volume and release targets identified in
Table 4.8. They are computed using Equation 4.114, with
the releases defined in Table 4.7 and targets defined in
Table 4.8, for each feasible combination of initial and final
storage volumes, S
t
and S
t1
, for each of the four seasons
in a year. These computed weighted squared deviations
for each link are shown in Table 4.9.
The goal in this example problem is to find the path
through a multi-year network each year of which is as
shown in Figure 4.17 that minimizes the sum of the
squared deviations associated with each of the paths
links. Again, each links weighted squared deviations are
given in Table 4.9. Of interest is the best path into the
future from any of the nodes or states (discrete storage
volumes) that the system could be in at the beginning of
any season t.
These paths can be found using the backward-moving
solution procedure of discrete dynamic programming.
This procedure begins at any arbitrarily selected time
period or season when the reservoir presumably produces
no further benefits to anyone (and it doesnt matter when
that time is just pick any time) and proceeds backward,
from right to left one stage (i.e., one time period) at a
time, towards the present. At each node (representing a
discrete storage volume S
t
and inflow Q
t
), we can calcu-
late the release or final storage volume in that period that
minimizes the remaining sum of weighted squared
deviations for all remaining seasons. Denote this
minimum sum of weighted squared deviations for all n
remaining seasons t as F
t
n
(S
t
, Q
t
). This value is dependent
on the state (S
t
, Q
t
), and stage, t, and the number n of
remaining seasons. It is not a function of the decision R
t
or S
t1
.
This minimum sum of weighted squared deviations for
all n remaining seasons t is equal to:
F
t
n
(S
t
, Q
t
) min
t
n
TSD
t
(S
t
, R
t
, S
t1
) over all feasible
values of R
t
(4.117)
wrm_ch04.qxd 8/31/2005 11:51 AM Page 106
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 107
where
S
t1
S
t
Q
t
R
t
L
t
(S
t
, S
t1
) (4.118)
and
S
t
K, the capacity of the reservoir (4.119)
The policy we want to derive is called a steady-state policy.
Such a policy assumes the reservoir will be operating for a
relatively long time with the same objectives. We can find
this steady-state policy by first assuming that at some time
all future benefits, losses or penalties, F
t
0
(S
t
, Q
t
), will be
0. We can begin in that season t and work backwards
towards the present, moving left through the network one
season t at a time. We can continue for multiple years until
the annual policy begins repeating itself each year. In other
words, when the optimal R
t
associated with a particular
state (S
t
, Q
t
) is the same in two or more successive years,
and this applies for all states (S
t
, Q
t
) in each season t, a
steady-state policy has probably been obtained (a more
definitive test of whether or not a steady-state policy has
been reached will be discussed later). A steady-state policy
will occur if the inflows, Q
t
, and objectives, TSD
t
(S
t
, R
t
,
S
t1
), remain the same from year to year. This steady-state
policy is independent of the assumption that the operation
will end at some point.
To find the steady-state operating policy for this exam-
ple problem, assume the operation ends in some distant
year at the end of Season 4 (the right-hand side nodes in
Figure 4.17). At the end of this season the number of
remaining seasons, n, equals 0. The values of the remain-
ing minimum sums of weighted squared deviations,
F
1
0
(S
1
, Q
1
) associated with each state (S
1
, Q
1
), i.e., each
node, equal 0. Now we can begin the process of finding
the best releases R
t
in each successive season, moving
backward to the beginning of stage t 4, then stage
t 3, then to t 2, and then to t 1, and then to t 4
of the preceding year, and so on, each move to the left
increasing the number n remaining seasons by one.
At each stage, or season, we can compute the release R
t
or equivalently the final storage volume S
t1
, that
minimizes
F
t
n
(S
t
, Q
t
) Minimum{TSD
t
(S
t
, R
t
, S
t1
)
F
t1
n1
(S
t1
, Q
t1
)} for all 0 S
t
20
(4.120)
The decision-variable can be either the release, R
t
, or the
final storage volume, S
t1
. If the decision-variable is the
release, then the constraints on that release R
t
are:
R
t
S
t
Q
t
L
t
(S
t
, S
t1
) (4.121)
R
t
S
t
Q
t
L
t
(S
t
, S
t1
) 20 (the capacity) (4.122)
and
S
t1
S
t
Q
t
R
t
L
t
(S
t
, S
t1
) (4.123)
0
10
15
20
5
p
e
r
i
o
d
t

=

3
0
10
15
20
5
p
e
r
i
o
d
t

=

4
0
10
15
20
5
0 5 10 15 20
final storage initial
storage
T
S
D
T
S
D
T
S
D
T
S
D
0
10
15
20
5
p
e
r
i
o
d
t

=

2
E
0
2
0
8
2
0
q
T
R
4
1
5
>
T
S
T
R
3
3
=

2
0
,
2
0
>
T
S
T
R
2
2
=

2
0
,
1
5
>
0 0 0 4 74
0 0 0 0 29
0 0 0 0 25
0 0 0 0 25
25 25 25 25 50
809 689 696
---
625 466 406 446
500 325 216 206 296
425 250 125 66 125
400 225 100 25 25
996 1025 ---
725 675 725
525 425 425 525
425 275 225 306
400 225 136 146 256
0 4 49 144
---
0 0 4 49 144
0 0 0 4 64
0 0 0 0 9
0 0 0 0 0
---
---
---
---
---
---
---
---
p
e
r
i
o
d
t

=
T
R
T
S
1
1
0
,
1
5
>
t
>
Table 4.9. Total sum of squared deviations, TSD
t
, associated
with initial and final storage volumes. These are calculated
using Equations 4.114 through 4.116.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 107
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
If the decision-variable is the final storage volume, the
constraints on that final storage volume S
t1
are:
0 S
t1
20 (4.124)
S
t1
S
t
Q
t
L
t
(S
t
, S
t1
) (4.125)
and
R
t
S
t
Q
t
S
t1
L
t
(S
t
, S
t1
) (4.126)
Note that if the decision-variable is S
t1
in season t, this
decision becomes the state variable in season t 1. In
both cases, the storage volumes in each season are limited
to discrete values 0, 5, 10, 15 and 20.
Tables 4.10 through 4.19 show the values obtained
from solving the recursive equations for 10 successive
seasons or stages (2.5 years). Each table represents a stage
or season t, beginning with Table 4.10 at t 4 and the
108 Water Resources Systems Planning and Management
number of remaining seasons n 1. The data in each
table are obtained from Tables 4.7 and 4.9. The last two
columns of each table represent the best release and
final storage volume decision(s) associated with the state
(initial storage volume and inflow).
Note that the policy defining the release or final storage
for each discrete initial storage volume in season
t 3 in Table 4.12 is the same as in Table 4.16, and
similarly for season t 4 in Tables 4.13 and 4.17, and for
season t 1 in Tables 4.14 and 4.18, and finally for season
t 2 in Tables 4.15 and 4.19. The policy differs over each
state, and over each different season, but not from year to
year for any specified state and season. We have reached a
steady-state policy. If we kept on computing the release
and final storage policies for preceding seasons, we would
E
0
2
0
8
2
7
a
0
10
15
20
5
0 5 10 15 20
final storage S
1
F
4
1
R
4
S
1
TSD
4
period t = 4 n = Q = 18
4
F
(
4
1
S
4
, Q
4
)
= min
4
TSD
initial
storage
S
4
0 4 49 144 - 0 18 0
0 0 4 49 144 0 18-23 0-5
0 0 0 4 64 0 18-28 0-10
0 0 0 0 9 0 17-33 0-15
0 0 0 0 0 0 17-38 0-20
Table 4.10. Calculation of
minimum squared deviations
associated with various discrete
storage states in season t = 4
with only n = 1 season remaining
for reservoir operation.
E
0
2
0
8
2
7
b
0
10
15
20
5
0 5 10 15 20
final storage S
4
F
3
2
R
3
S
4
period t = 3 n = 2 Q = 6
3
F
(
3
2
S
3
, Q
3
)
= min
3
{ } TSD
(
S
3
, R
3
)
, S
4
F
(
4
1
S
4
, Q
4
)
+
initial
storage
S
3
1
TSD
3
F
(
4
S , Q
) +
4 4
996 1025 - - - 996 6 0
725 675 725 - - 675 5 5
525 425 425 525 - 425 5-10 5-10
425 275 225 306 - 225 10 10
400 225 136 146 256 136 14 10
Table 4.11. Calculation of
minimum squared deviations
associated with various discrete
storage states in season t = 3
with n = 2 seasons remaining for
reservoir operation.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 108
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 109
E
0
2
0
8
2
7
c
0
10
15
20
5
0 5 10 15 20
final storage S
3
F
2
R S
period t = 2 n = 3 Q = 12
2
F
(
2
3
S
2
, Q
)
= min { } TSD
(
S , R
)
, S F
(
2
S , Q
)
+
initial
storage
S
2
TSD
2
F
(
S , Q
) +
2 2 2 2 3 3 3 3
3 3 3
2
3
2 3
809+996 689+675 696+425 - - 1121 1 10
625+996 466+675 406+425 446+225 - 671 1 15
500+996 325+675 216+425 206+225 296+136 431 6 15
425+996 250+675 125+425 66+225 125+136 261 5 20
400+996 225+675 100+425 25+225 25+136 161 10 20
Table 4.12. Calculation of minimum
squared deviations associated with
various discrete storage states in
season t = 2 with n = 3 seasons
remaining for reservoir operation.
E
0
2
0
8
2
7
d
0
10
15
20
5
0 5 10 15 20
final storage S
2
F
1
R S
period t = n = 4 Q = 24
1
F
(
1
4
S , Q
)
= min { } TSD
(
S , R
)
, S F
(
3
S , Q
)
+
initial
storage
S
1
TSD F
(
S , Q
) +
2
2
3
4
2
1 1 1 1 1 2 2 2
1 2 2
1
0+1121 0+671 0+431 4+261 74+161 235 3 20
0+1121 0+671 0+431 0+261 29+161 190 8 20
0+1121 0+671 0+431 0+261 25+161 186 13 20
0+1121 0+671 0+431 0+261 25+161 186 18 20
25+1121 25+671 25+431 25+261 50+161 211 23 20
Table 4.13. Calculation of minimum
squared deviations associated with
various discrete storage states in
season t = 1 with n = 4 seasons
remaining for reservoir operation.
E
0
2
0
8
2
7
e
0
10
15
20
5
0 5 10 15 20
final storage S
F R S
period t = 4 n = 5 Q = 18
4
F
(
5
S , Q
)
= min { } TSD
(
S , R
)
, S F
(
S , Q
)
+
initial
storage
S
4
TSD F
(
S , Q
) +
4
1
1
4 4 4 4 4 4
4
1 1 1
4 1
4
1 1
1 5
4
0+235 4+190 49+186 144+186 - 194 13 5
0+235 0+190 4+186 49+186 144+211 190 13-18 5-10
0+235 0+190 0+186 4+186 64+211 186 18 10
0+235 0+190 0+186 0+186 9+211 186 17-23 10-15
0+235 0+190 0+186 0+186 0+211 186 22-27 10-15
Table 4.14. Calculation of minimum
squared deviations associated with
various discrete storage states in
season t = 4 with n = 5 seasons
remaining for reservoir operation.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 109
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
110 Water Resources Systems Planning and Management
E
0
2
0
8
2
7
f
0
10
15
20
5
0 5 10 15 20
final storage S
F R S
period t = 3 n = 6 Q = 6
3
F
(
6
S , Q
)
= min { } TSD
(
S , R
)
, S F
(
S , Q
)
+
initial
storage
S
3
TSD F
(
S , Q
) +
4
3 4
5
6
3
3 3 3 3 3 4 4 4
3
5
4 4 4
4
3
996+194 1025+190 - - - 1190 6 0
725+194 675+190 725+186 - - 865 5 5
525+194 425+190 425+186 525+186 - 611 5 10
425+194 275+190 225+186 306+186 - 411 10 10
400+194 225+190 136+186 146+186 256+186 322 14 10
Table 4.15. Calculation of
minimum squared deviations
associated with various discrete
storage states in season t = 3
with n = 6 seasons remaining for
reservoir operation.
E
0
2
0
8
2
7
g
0
10
15
20
5
0 5 10 15 20
final storage S
F R S
period t = 2 n = 7 Q = 12
2
F
(
7
S , Q
)
= min { } TSD
(
S , R
)
, S F
(
S , Q
)
+
initial
storage
S
2
TSD F
(
S , Q
) +
2
6
7
2
3
3
2 2 2 2 2 3 3 3
2
6
3 3 3
3
2
809+1190 689+865 696+611 - - 1307 1 10
625+1190 466+865 406+611 446+411 - 857 1 15
500+1190 325+865 216+611 206+411 296+322 617 6 15
425+1190 250+865 125+611 66+411 125+322 447 5 20
400+1190 225+865 100+611 25+411 25+322 347 10 20
Table 4.16. Calculation of
minimum squared deviations
associated with various discrete
storage states in season t = 2
with n = 7 seasons remaining for
reservoir operation.
E
0
2
0
8
2
7
h
0
10
15
20
5
0 5 10 15 20
final storage S
2
F
1
8
R
2
S
period t = n = 8 Q = 24
2
F
(
1
8
S , Q
)
= min { } TSD
(
S , R
)
, S
2
F
(
2
7
S , Q
)
+
initial
storage
S
TSD F
(
S , Q
) +
1
1 1 1 1 1 2 2
1 2
7
2 2
1
0+1307 0+857 0+617 4+447 74+347 421 3 20
0+1307 0+857 0+617 0+447 29+347 376 8 20
0+1307 0+857 0+617 0+447 25+347 372 13 20
0+1307 0+857 0+617 0+447 25+347 372 18 20
25+1307 25+857 25+617 25+447 50+347 397 23 20
Table 4.17. Calculation of
minimum squared deviations
associated with various discrete
storage states in season t = 1
with n = 8 seasons remaining for
reservoir operation.
wrm_ch04.qxd 8/31/2005 11:51 AM Page 110
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 111
get the same policy as that found for the same season in the
following year. The policy is dependent on the state the
initial storage volume in this case and on the season t,
but not on the year. This policy as defined in Tables
4.164.19 is summarized in Table 4.20.
This policy can be defined as a rule curve, as shown in
Figure 4.18. It provides a first approximation of a reser-
voir release rule curve that one can improve upon using
simulation.
Table 4.20 and Figure 4.18 define a policy that can
be implemented for any initial storage volume condition
at the beginning of any season t. This can be simulated
under different flow patterns to determine just how well
it satisfies the overall objective of minimizing the
weighted sum of squared deviations from desired, but
conflicting, storage and release targets. There are other
performance criteria that may also be evaluated using
simulation, such as measures of reliability, resilience, and
vulnerability (Chapter 10).
Assuming the inflows that were used to derive this
policy actually occurred each year, we can simulate the
derived sequential steady-state policy to find the storage
volumes and releases that would occur in each period,
year after year, once a repetitive steady-state condition
were reached. This is done in Table 4.21 for an arbitrary
initial storage volume of 20 in season t 1. You can try
other initial conditions to verify that it requires only two
years at most to reach a repetitive steady-state policy.
As shown in Table 4.21, if the inflows were repetitive
and the optimal policy was followed, the initial storage
E
0
2
0
8
2
7
k
0
10
15
20
5
0 5 10 15 20
final storage S
1
F
4
R
4
S
1
period t = 4 n = 9 Q = 18
4
F
(
4
9
S
4
, Q
4
)
= min
4
{ } TSD
(
S
4
, R
4
)
, S
1
F
(
8
S , Q
)
+
initial
storage
S
4
TSD F
(
S , Q
) +
1 1 1
4
8
1 1 1
9
0+421 4+376 49+372 144+372 - 380 13 5
0+421 0+376 4+372 49+372 144+397 376 13-18 5-10
0+421 0+376 0+372 4+372 64+397 372 18 10
0+421 0+376 0+372 0+372 9+397 372 17-23 10-15
0+421 0+376 0+372 0+372 0+397 372 22-27 10-15
Table 4.18. Calculation of minimum
squared deviations associated with
various discrete storage states in
season t = 4 with n = 9 seasons
remaining for reservoir operation.
E
0
2
0
8
2
7
m
0
10
15
20
5
0 5 10 15 20
final storage S
4
F
3
10
R
4
S
period t = 3 n = 10 Q = 6
3
F
(
3
10
S , Q
)
= min { } TSD
(
S , R
)
, S
4
F
(
4
9
S
4
, Q
4
)
+
initial
storage
S
3
TSD
3
F
(
4
S , Q
) +
4 4
3 3 3 3 3
9
3
996+380 1025+376 - - - 1376 6 0
725+380 675+376 725+372 - - 1051 5 5
525+380 425+376 425+372 525+372 - 797 5 10
425+380 275+376 225+372 306+372 - 597 10 10
400+380 225+376 136+372 146+372 256+372 508 14 10
Table 4.19. Calculation of minimum
squared deviations associated with
various discrete storage states in
season t = 3 with n = 10 seasons
remaining for reservoir operation.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 111
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
volumes and releases would begin to repeat themselves
once a steady-state condition has been reached. Once
reached, the storage volumes and releases will be the
same each year (since the inflows are the same). These
storage volumes are denoted as a blue line on the rule
curve shown in Figure 4.18. The annual total squared
deviations will also be the same each year. As seen in
Table 4.21, this annual minimum weighted sum of
squared deviations for this example equals 186. This is
what would be observed if the inflows assumed for this
analysis repeated themselves.
112 Water Resources Systems Planning and Management
E
0
2
0
8
2
7
n
0
10
15
20
5
2 3 4
season t
initial
storage
S
release
2 3 4
season t
final storage volume
3 1 6 13 20 10 0 5
8 1 5 13-18 20 15 5 5-10
18 5 10 17-23 20 20 10 10-15
13 6 5 15 10 10 18 10
23 10 14 22-27 20 20 10 10-15
Table 4.20. The discrete steady-
state reservoir operating policy
as computed for this example
problem in Tables 4.16 to 4.19.
20
15
10
5
0
expected storage volume
full reservoir storage target for recreation
( t = 1 )
( t = 2, 3 )
s
t
o
r
a
g
e

v
o
l
u
m
e
4
season of year t storage zone release volume
3
27
2 1
22
14 10 23
18
13
10
5
18
5 6 13
5 1 8
6 1 3
E
0
2
0
1
0
3
j max storage target for flood control
Figure 4.18. Reservoir rule curve based on policy defined in
Table 4.20. Each season is divided into storage volume zones.
The releases associated with each storage volume zone are
specified. Also shown are the storage volumes that would
result if in each year the actual inflows equalled the inflows
used to derive this rule curve.
Note from Tables 4.12 4.15 and 4.16 4.19 that
once the steady-state sequential policy has been reached
for any specified storage volume, S
t
, and season t, the
annual difference of the accumulated minimum sum of
squared deviations, F
t
n
(S
t
, Q
t
), equals a constant, namely
the annual value of the objective function. In this case
that constant is 186.
F
t
n4
(S
t
, Q
t
) F
t
n
(S
t
, Q
t
) 186 for all S
t
, Q
t
and t.
(4.127)
This condition indicates a steady-state policy has been
achieved.
This policy in Table 4.21 applies only for the assumed
inflows in each season. It does not define what to do if
the initial storage volumes or inflows differ from those for
which the policy is defined. Initial storage volumes and
inflows can and will vary from those specified in the
solution of any deterministic model. One fact is certain:
no matter what inflows are assumed in any model, the
actual inflows will always differ. Hence, a policy as
defined in Table 4.20 and Figure 4.18 is much more
useful than that in Table 4.21. But as for any output from
a relatively simple optimization model, this policy should
be simulated, evaluated and further refined in an effort to
identify the policy that best meets the operating policy
objectives.
4.8. General Comments on Dynamic
Programming
Before ending this discussion of dynamic programming
methods and its applicability for analysing water
resources planning problems, we should examine a major
wrm_ch04.qxd 8/31/2005 11:52 AM Page 112
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 113
assumption that has been made in each of the applica-
tions presented. The first is that the net benefits or costs
or other objective values resulting from each decision at
each stage of the problem are dependent only on the state
variable values in each stage. They are independent of
decisions made at other stages. If the returns at any stage
are dependent on the decisions made at other stages in a
way not captured by the state variables. For example, if
the release and storage targets or the reservoir capacity
were unknown in the previous reservoir operating policy
problem, then dynamic programming, with some excep-
tions, becomes much more difficult to apply. Dynamic
programming models can be applied to design problems,
such as the capacity-expansion problem or a reservoir
storage capacityyield relationship as will be discussed
later, or to operating problems, such as the water
allocation and reservoir operation problems, but rarely to
problems having both unknown design and operating
policy decision-variables. While there are some tricks that
may allow dynamic programming to be used to find the
best solutions to both design and operating problems
encountered in water resources planning and manage-
ment studies, other optimization methods, perhaps
combined with dynamic programming where appropri-
ate, are often more useful.
5. Linear Programming
If the objective function and constraints of an optimiza-
tion model are all linear, there are many readily available
computer programs one can use to find their solutions.
These programs are very powerful, and unlike many other
optimization methods, they can be applied successfully to
very large optimization problems. Many water resources
problems contain many variables and constraints, too
many to be easily solved using non-linear or dynamic
programming methods. Linear programming procedures
or algorithms for solving linear optimization models are
often the most efficient ways to find solutions to such
problems.
Because of the availability of computer programs that
can solve linear programming problems, linear program-
ming is arguably the most popular and commonly applied
optimization algorithm. It is used to identify and evaluate
alternative plans, designs and management policies in
agriculture, business, commerce, education, engineering,
finance, the civil and military branches of government,
and many other fields.
Many models of complex water resources systems are,
or can be made, linear. Many are also very large. The
number of variables and constraints simply defining mass
balances and capacity limitations alone at many river
basin sites and for numerous time periods can become
so big as to preclude the practical use of most other
optimization methods. Because of the power and
availability of computer programs that can solve large
linear programming problems, a variety of methods have
been developed to approximate non-linear (especially
separable) functions with linear ones just so linear
programming can be used to solve various otherwise non-
linear problems. Some of these methods will be described
shortly.
year season
t
2
3
4
TSD
t
S
t
+ Q R L S
t t t t+1
- = -
2
2
2
2
3
year
year
season
t
season
t
2
3
4
TSD
TSD
t
t
S
S
t
t
+
+
Q
Q
R
R
L
L
S
S
t
t
t
t
t
t
t+1
t+1
-
-
=
=
-
-
20 24 23 1 20
20 12 10 2 20
20 6 14 2 10
10 18 18 0 10
10
10 etc ... repeating ...
24 14 0 20 25
20 12 10 2 20 25
20 6 14 2 10 136
10 18 18 0 10 0
E
0
2
0
8
2
7
p
186 total sum of squared deviations, , = TSD
Table 4.21. A simulation of the derived operating policy in
Table 4.20. The storage volumes and releases in each period
t will repeat themselves each year, after the first year. The
annual total squared deviations, TSD
t
for the specific initial
and final storage volumes and release conditions are obtained
from Table 4.9.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 113
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
In spite of its power and popularity, for most real-
world water resources planning and management prob-
lems, linear programming, like the other optimization
methods already discussed in this chapter, is best viewed
as a preliminary screening tool. Its value is more for
reducing the number of alternatives for further more
detailed simulations than for finding the best decision.
This is not just because approximation methods may have
been used to convert non-linear functions to linear ones,
but more likely because it is difficult to incorporate all the
complexity of the system and all the objectives considered
important to all stakeholders into the linear model.
Nevertheless, linear programming, like other optimiza-
tion methods, can provide initial designs and operating
policy information that simulation models require before
they can simulate those designs and operating policies.
Equations 4.45 and 4.46 define the general structure
of any constrained optimization problem. If the objective
function F(X) of the vector X of decision-variables x
j
is linear
and if all the constraints g
i
(X) in Equation 4.46 are linear,
then the model becomes a linear programming model. The
general structure of a linear programming model is:
Maximize or minimize
j
n
P
j
x
j
(4.128)
Subject to:
j
n
a
ij
x
j
b
i
for i 1, 2, 3, , m (4.129)
x
j
0 for all j 1, 2, 3, , n. (4.130)
If any model fits this general form, where the constraints
can be any combination of equalities () and inequalities
( or ), then a large variety of linear programming
computer programs can be used to find the optimal
values of all the unknown decision-variables x
j
. With
some exceptions, variable non-negativity is enforced
within the solution algorithms of most commercial linear
programming programs, eliminating the need to have to
specify these conditions in any particular application.
Users of linear programming need to know how to
construct linear models and how to use the computer
programs that are available for solving them. They do not
have to understand all the mathematical details of the
solution procedure incorporated in the linear program-
ming codes. But users of linear programming computer
programs should understand what the solution procedure
does and what the computer program output means.
To begin this discussion of these topics, consider some
simple examples of linear programming models.
114 Water Resources Systems Planning and Management
5.1. Reservoir Storage CapacityYield Models
Linear programming can be used to define storage
capacityyield functions for a single or multiple reser-
voirs. A storage capacityyield function defines the
maximum constant dependable reservoir release or yield
that will be available, at a given level of reliability, during
each period of operation, as a function of the active
storage volume capacity. The yield from any reservoir or
group of reservoirs will depend on the active storage
capacity of each reservoir and the water that flows into
each reservoir, its inflow. Figure 4.19 illustrates two
typical storageyield functions for a single reservoir.
To describe what a yield is and how it can be
increased, consider a sequence of 5 annual flows, say 2, 4,
1, 5 and 3, at a site in an unregulated stream. Based on
this admittedly very limited record of flows, the minimum
(historically) dependable annual flow yield of the stream
at that site is 1, the minimum observed flow. Assuming
the flow record is representative of what future flows
might be, a discharge of 1 can be guaranteed in each of
non-zero the five time-periods of record. (In reality, that
or any non-zero yield will have a reliability less than 1, as
will be considered in Chapter 11).
If a reservoir having an active storage capacity of 1 is
built, it could store 1 volume unit of flow when the flow
is equal to or greater than 2, and then release it along with
the natural flow when the natural flow is 1, increasing the
E
0
2
0
1
0
3
m
r
e
s
e
r
v
o
i
r

y
i
e
l
d
active storage capacity of a reservoir
lower
higher
reliability
Figure 4.19. Two storageyield functions for a single
reservoir defining the maximum minimum dependable
release. These functions can be defined for varying levels of
yield reliability.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 114
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 115
minimum dependable flow to 2 units in each year.
Storing 2 units when the flow is 5, releasing 1 and the
natural flow when that natural flow is 2, and storing 1
when the flow is 4, and then releasing the stored 2 units
along with the natural flow when the natural flow is 1,
will permit a yield of 3 in each time period with 2 units
of active capacity. This is the maximum annual yield that
is possible at this site, again based on these five annual
inflows and their sequence. The maximum annual yield
cannot exceed the mean annual flow, which in this
example is 3. Hence, the storage capacityyield function
equals 1 when the active capacity is 0, 2 when the active
capacity is 1, and 3 when the active capacity is 2. The
annual yield remains at 3 for any active storage capacity
in excess of 2.
This storageyield function is dependent not only on
the natural unregulated annual flows but also on their
sequence. For example if the sequence of the same 5
annual flows were 5, 2, 1, 3, 4, the needed active storage
capacity is 3 instead of 2 volume units as before to obtain
a dependable flow or yield of 3 volume units. In spite
of these limitations of storage capacityyield functions,
historical records are still typically used to derive them.
(Ways of augmenting the historical flow record are
discussed in Chapter 7.)
There are many methods available for deriving
storageyield functions. One very versatile method,
especially for multiple reservoir systems, uses linear
programming. Others are discussed in Chapter 11.
To illustrate a storage capacityyield model, consider a
single reservoir that must provide a minimum release
or yield Y in each period t. Assume a record of known
(historical or synthetic) streamflows at the reservoir site
is available. The problem is to find the maximum uniform
yield Y obtainable from a given active storage capacity.
The objective is to
maximize Y (4.131)
This maximum yield is constrained by the water available
in each period, and by the reservoir capacity. Two sets of
constraints are needed to define the relationships
among the inflows, the reservoir storage volumes, the
yields, any excess release, and the reservoir capacity. The
first set of continuity equations equate the unknown final
reservoir storage volume S
t1
in period t to the unknown
initial reservoir storage volume S
t
plus the known inflow Q
t
,
minus the unknown yield Y and excess release, R
t
, if any, in
period t. (Losses are being ignored in this example.)
S
t
Q
t
Y R
t
S
t1
for each period t 1, 2, 3, , T. T1 1 (4.132)
If, as indicated in Equation 4.132, one assumes that
Period 1 follows the last Period T, it is not necessary to
specify the value of the initial storage volume S
1
and/or
final storage volume S
T1
. They are set equal to each other
and that variable value remains unknown. The resulting
steady-state solution is based on the inflow sequence that
is assumed to repeat itself.
The second set of required constraints ensures that the
reservoir storage volumes S
t
at the beginning of each period
t are no greater than the active reservoir capacity K.
S
t
K t l, 2, 3, , T (4.133)
To derive a storageyield function, the model defined by
Equations 4.131, 4.132 and 4.133 must be solved for
various assumed values of capacity K. Only the inflow
values Q
t
and reservoir active storage capacity K are
known. All other storage, release and yield variables are
unknown. Linear programming will be able to find their
optimal values. Clearly, the upper bound on the yield
regardless of reservoir capacity will equal the mean inflow
(less any losses if they were included).
Alternatively, one can solve a number of linear
programming models that minimize an unknown storage
capacity K needed to achieve various specified yields Y.
The resulting storageyield function will be same; the
minimum capacity needed to achieve a specified yield will
be the same as the maximum yield obtainable from the
corresponding specified capacity K. However, the speci-
fied yield Y cannot exceed the mean inflow. If it does,
there will be no feasible solution to the linear program-
ming model.
Box 4.1 illustrates an example storageyield model
and its solution to find the storageyield function. For
this problem, and others in this chapter, the program
LINGO(www.lindo.com) is used.
Before moving to another application of linear
programming, consider how this storageyield problem,
Equations 4.1314.133, can be formulated as a discrete
dynamic programming model. The use of discrete
dynamic programming is clearly not the most efficient
way to define a storageyield function but the problem of
wrm_ch04.qxd 8/31/2005 11:52 AM Page 115
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
finding a storageyield function provides a good exercise
in dynamic programming. The dynamic programming
network has the same form as shown in Figure 4.19,
where each node is a discrete storage and inflow state, and
the links represent releases. Let F
t
n
(S
t
) be the maximum
yield obtained given a storage volume of S
t
in period t of
116 Water Resources Systems Planning and Management
a year with n periods remaining of reservoir operation.
For initial conditions, assume all values of F
t
0
(S
t
) for some
final period t with no more periods n remaining equal a
large number that exceeds the mean annual inflow. Then
for the set of feasible discrete total releases R
t
:
F
t
n
(S
t
) max{min[R
t
, F
t1
n1
(S
t1
)]} (4.134)
Box 4.1. Example storage capacityyield model
and its solution from LINGO
E
0
2
0
9
0
3
a
K Y S1 S2 S3 S4 S5 R1 R2 R3 R4 R5
0 5 0 0 0 0 0 5 0 25 15 10
5 10 5 5 0 5 5 0 0 15 10 5
10 12.5 10 7.5 0 2.5 10 0 0 15 0 2.5
15 15 10 10 0 15 15 0 0 0 5 0
18 16 17 11 0 14 18 0 0 0 0 0
model solutions for specified values of K in model 1 or values of Y in model 2 :
! Reservoir Storage-Yield Model:
Define St as the initial active res. storage, period t,
Y as the reliable yield in each period t,
Rt as the excess release from the res., period t,
Qt as the known inflow volume to the res., period t
K as the reservoir active storage volume capacity.
;
Max = Y ; !Applies to Model 1. Must be omitted for Model 2;
Min = K ; !Applies to Model 2. Must be omitted for Model 1;
!
Subject to:
Mass balance constraints for each of 5 periods t.
;
S1 + Q1 - Y - R1 = S2;
S2 + Q2 - Y - R2 = S3;
S3 + Q3 - Y - R3 = S4;
S4 + Q4 - Y - R4 = S5;
S5 + Q5 - Y - R5 = S1; ! assumes a steady -state condition;
!
Capacity constraints on storage volumes.
;
S1 < K; S2 < K; S3 < K; S4 < K; S5 < K;
Data:
Q1 = 10; Q2 = 5; Q3 = 30; Q4 = 20; Q5 = 15;
!Note mean = 16;
K = ? ; ! Use for Model 1 only. Allows user to enter
any value of K during model run.;
Y = ? ; ! Use for Model 2 onl y. Allows user to enter
any value of Y during model run.
Enddata
wrm_ch04.qxd 8/31/2005 11:52 AM Page 116
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 117
This applies for all discrete storage volumes S
t
and for
all within-year periods t and remaining periods n. The
constraints on the decision-variables R
t
are:
R
t
S
t
Q
t
,
R
t
S
t
Q
t
K, and
S
t1
S
t
Q
t
R
t
(4.135)
These recursive Equations 4.134 together with constraint
Equations 4.135 can be solved, beginning with n 1
and then for successive values of seasons t and remaining
periods n, until a steady-state solution is obtained, that
is, until
F
t
n
(S
t
) F
t
nT
(S
t
) 0 for all values of S
t
and periods t.
(4.136)
The steady-state yields F
t
(S
t
) will depend on the storage
volumes S
t
. High initial storage volumes will result in
higher yields than will lower ones. The highest yield will
be that associated with the highest storage volumes and it
will equal the same value obtained from either of the two
linear programming models.
5.2. A Water Quality Management Problem
Some linear programming modelling and solution
techniques can be demonstrated using the simple water
quality management example shown in Figure 4.20. In
addition, this example can serve to illustrate how models
can help identify just what data are needed and how
accurate they must be for the decisions that are being
considered.
The stream shown in Figure 4.20 receives wastewater
effluent from two point sources located at Sites 1 and 2.
Without some wastewater treatment at these sites, the
concentration of some pollutant, P
j
mg/l, at sites j 2
and 3, will continue to exceed the maximum desired
concentration P
j
max
. The problem is to find the level of
wastewater treatment (waste removed) at sites i 1 and
2 that will achieve the desired concentrations at sites
j 2 and 3 at a minimum total cost.
This is the classic water quality management problem
that is frequently found in the literature, although least-
cost objectives have not been applied in practice. There
are valid reasons for this that we will review later.
Nevertheless, this particular problem can serve to
illustrate the development of some linear models for
determining data needs, estimating the values of model
parameters, and for finding, in this case, cost-effective
treatment efficiencies. This problem can also serve to
illustrate graphically the general mathematical procedures
used for solving linear programming problems.
The first step is to develop a model that predicts the
pollutant concentrations in the stream as a function of
the pollutants discharged into it. To do this we need some
notation. Define P
j
to be the pollutant concentration in
the stream at site j. The total mass per unit time of the
pollutant (M/T) in the stream at site j will be its
concentration P
j
(M/L
3
) times the streamflow Q
j
(L
3
/T).
For example if the concentration is in units of mg/l and
Q
stre
a
m
flo
w
firm
producing W
1 recreation
park
1
x W
site
( - 1 )
1 1
site 3
x W
site 2
( - 1 )
2 2
firm 2
producing W
2
E
0
2
0
1
0
3
n
Figure 4.20. A stream pollution
problem that requires finding
the waste removal efficiencies
(x
1
, x
2
) of wastewater treatment
at Sites 1 and 2 that meet the
stream quality standards at
Sites 2 and 3 at minimum total
cost. W
1
and W
2
are the amounts
of pollutant prior to treatment at
Sites 1 and 2.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 117
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the flow is in terms of m
3
/s, and mass is to be expressed
as kg/day:
Mass at site j (kg/day) P
j
(mg/l) Q
j
(m
3
/s) (10
3
litres/m
3
)
(kg/10
6
mg) (86400 sec./day) 86.4 P
j
Q
j
(4.137)
Each unit of a degradable pollutant mass in the stream
at Site 1 in this example will decrease as it travels
downstream to Site 2. Similarly each unit of the pollutant
mass in the stream at Site 2 will decrease as it travels
downstream to Site 3. The fraction a
12
of the mass at
Site 1 that reaches Site 2 is often assumed to be:
a
12
exp(kt
12
) (4.138)
where k is a rate constant (1/time unit) that depends on
the pollutant and the temperature, and t
12
is the time
(number of time units) it takes a particle of pollutant to
flow from Site 1 to Site 2. A similar expression, a
23
,
applies for the fraction of pollutant mass at Site 2 that
reaches Site 3. The actual concentration at the down-
stream end of a reach will depend on the streamflow at
that site as well as on the initial pollutant mass, the time
of travel and rate constant k.
In this example problem, the fraction of pollutant
mass at Site 1 that reaches Site 3 is the product of the
transfer coefficients a
12
and a
23
:
a
13
a
12
a
23
(4.139)
In general, for any site k between sites i and j:
a
ij
a
ik
a
kj
(4.140)
Knowing the a
ij
values for any pollutant and the time
of flow t
ij
permits the determination of the rate constant
k for that pollutant and reach, or contiguous series
of reaches, from sites i to j, using Equation 4.138. If
the value of k is 0, the pollutant is called a conservative
pollutant; salt is an example of this. Only increased dilution
by less polluted water will reduce its concentration.
For the purposes of determining wastewater treatment
efficiencies or other capital investments in infrastructure
designed to control the pollutant concentrations in the
stream, some design streamflow conditions have to be
established. Usually the design streamflow condition is
set at some very low flow value (e.g., the lowest monthly
average flow expected once in twenty years, or the
minimum seven-day average flow expected once in ten
years). Low design flows are based on the assumption that
118 Water Resources Systems Planning and Management
pollutant concentrations will be higher in low-flow
conditions than in higher-flow conditions because of
less dilution. While low-flow conditions may not provide
as much dilution, they do result in longer travel times,
and hence greater reductions in pollutant masses between
water quality monitoring sites. Hence the pollutant
concentrations may well be greater at some downstream
site when the flow conditions are higher than those of
the low-flow design value.
In any event, given particular design streamflow and
temperature conditions, our first job is to determine the
values of these dimensionless transfer coefficients a
ij
.
They will be independent of the amount of waste
discharged into the stream. To determine both a
12
and a
23
in this example problem (Figure 4.20) requires a number
of pollutant concentration measurements at Sites 1, 2 and
3 during design streamflow conditions. These measure-
ments of pollutant concentrations must be made just
downstream of the wastewater effluent discharge at Site 1,
just upstream and downstream of the wastewater effluent
discharge at Site 2, and at Site 3.
Assuming no change in streamflow and no extra
pollutant entering the reach that begins just downstream
of Site 1 and ends just upstream of Site 2, the pollutant
concentration P
2
just upstream of Site 2 will equal the
concentration just downstream of Site 1, P
1
, times the
transfer coefficient a
12
:
P
2
P
1
a
12
(4.141)
Under similar equal flow conditions in the following
reach beginning just downstream from Site 2 and extend-
ing to Site 3, the pollutant concentration P
3
will equal
the concentration just downstream of Site 2, P
2

, times
the transfer coefficient a
23
.
If the streamflows Q
i
and Q
j
at sites i and j differ, then
the downstream pollutant concentration P
j
resulting from
an upstream concentration of P
i
will be:
P
j
P
i
a
ij
(Q
i
/Q
j
) (4.142)
5.2.1. Model Calibration
Sample measurements are needed to estimate the
values of each reachs pollutant transport coefficients a
ij
.
Assume five pairs of sample pollutant concentration
measurements have been taken in the two stream reaches
wrm_ch04.qxd 8/31/2005 11:52 AM Page 118
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 119
during design flow conditions. For this example, also
assume that the design streamflow just downstream of
Site 1 and just upstream of Site 2 are the same and equal
to 12 m
3
/s. The concentration samples taken just down-
stream from Site 1 and just upstream of Site 2 during this
design flow condition can be used to solve Equation
4.142 after adding error terms. More than one sample is
needed to allow for measurement errors and other ran-
dom effects such as those from wind, incomplete mixing
or varying wasteload discharges within a day.
Denote the concentrations of each pair of sample
measurements s in the first reach (just downstream of Site
1 and just upstream of Site 2) as SP
1s
and SP
2s
and their
combined error as E
s
. Equation 4.142 becomes
SP
2s
E
s
SP
1s
a
12
(Q
1
/Q
2
) (4.143)
The problem is to find the best estimates of the unknown
a
12
. One way to do this is to define best as those values
of a
12
and all E
s
that minimize the sum of the absolute
values of all the error terms E
s
. This objective could be
written
Minimize
s
| E
s
| (4.144)
The set of Equations 4.143 and 4.144 is an optimization
model. If the absolute value signs can be removed, these
equations together with constraints that require all
unknown variables to be non-negative would form a
linear programming model.
The absolute value signs in Equation 4.144 can be
removed by writing each error term as the difference
between two non-negative variables, PE
s
NE
s
. Thus for
each sample pair s:
E
s
PE
s
NE
s
(4.145)
If any E
s
is negative, PE
s
will be 0 and NE
s
will equal E
s
.
The actual value of NE
s
is non-negative. If E
s
is positive, it
will equal PE
s
, and NE
s
will be 0. The objective function,
Equation 4.154, that minimizes the sum of absolute value
of error terms, can now be written as one that minimizes
the sum of the positive and negative components of E
s
:
Minimize
s
(PE
s
NE
s
) (4.146)
Equations 4.143 and 4.145, together with objective
function 4.146 and a set of measurements, SP
1s
and SP
2s
,
upstream and downstream of the reach between Sites 1
and 2 define a linear programming model that can be
solved to find the transfer coefficient a
12
. An example
illustrating the solution of this model for the stream reach
between Site 1 and just upstream of Site 2 is presented in
Box 4.2. Again, the program LINGO (www.lindo.com) is
used to solve the models.
Box 4.3 contains the model and solution for the
reach beginning just downstream of Site 2 to Site 3.
In this reach the design streamflow is 12.5 m
3
/s due to
the addition of wastewater flow at Site 2.
In this example, based on the solutions for a
ij
and
flows given in Boxes 4.2 and 4.3, a
12
0.25, a
23
0.60,
and thus from Equation 4.140, a
12
a
23
a
13
0.15.
5.2.2. Management Model
Now that these parameter values a
ij
are known, a water
quality management model can be developed. The water
quality management problem, illustrated in Figure 4.20,
involves finding the fractions x
i
of waste removal at
sites i 1 and 2 that meet the stream quality standards
at the downstream Sites 2 and 3 at a minimum total cost.
The pollutant concentration, P
2
, just upstream of Site 2
that results from the pollutant concentration at Site 1
equals the total mass of pollutant at Site 1 times the frac-
tion of that mass remaining at Site 2, a
12
, divided by the
streamflow just upstream of Site 2, Q
2
. The total mass of
pollutant at Site 1 at the wastewater discharge point is the
sum of the mass just upstream of the discharge site, P
1
Q
1
,
plus the mass discharged into the stream, W
1
(1 x
1
), at
Site 1. The parameter W
1
is the total amount of pollutant
entering the treatment plant at Site 1. Similarly for Site 2.
Hence the concentration of pollutant just upstream of
Site 2 is:
P
2
[P
1
Q
1
W
1
(1x
1
)]a
12
/Q
2
(4.147)
The terms P
1
and Q
1
are the pollutant concentration
(M/L
3
) and streamflow (L
3
/T) just upstream of the waste-
water discharge outfall at Site 1. Their product is the mass
of pollutant at that site per unit time period (M/T).
The pollutant concentration, P
3
, at Site 3 that results
from the pollutant concentration at Site 2 equals the total
mass of pollutant at Site 2 times the fraction a
23
. The total
mass of pollutant at Site 2 at the wastewater discharge point
is the sum of what is just upstream of the discharge site,
P
2
Q
2
, plus what is discharged into the stream, W
2
(1x
2
).
Hence the concentration of pollutant at Site 3 is:
P
3
[P
2
Q
2
W
2
(1x
2
)]a
23
/Q
3
(4.148)
wrm_ch04.qxd 8/31/2005 11:52 AM Page 119
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Equations 4.147 and 4.148 will become the predictive
portion of the water quality management model. The
remaining parts of the model include the restrictions on
the pollutant concentrations at Sites 2 and 3, and limits
on the range of values that each waste removal efficiency
x
i
can assume.
P
j
P
j
max
for j 2 and 3 (4.149)
0 x
i
1.0 for i 1 and 2. (4.150)
Finally, the objective is to minimize the total cost of meeting
the stream quality standards P
2
max
and P
3
max
specified in
Equations 4.149. Letting C
i
(x
i
) represent the wastewater
treatment cost function for site i, the objective can be written:
Minimize C
1
(x
1
) C
2
(x
2
) (4.151)
The complete optimization model consists of Equations
4.147 through 4.151. There are four unknown decision
120 Water Resources Systems Planning and Management
variables, x
1
, x
2
, P
2
, and P
3
. All variables are assumed to
be non-negative.
Some of the constraints of this optimization model can
be combined to remove the two unknown concentration
values, P
2
and P
3
. Combining Equations 4.147 and 4.149,
the concentration just upstream of Site 2 must be no
greater than P
2
max
:
[P
1
Q
1
W
1
(1x
1
)]a
12
/Q
2
P
2
max
(4.152)
Combining Equations 4.148 and 4.149, and using
the fraction a
13
(see Equation 4.139) to predict the
contribution of the pollutant concentration at Site 1 on
the pollutant concentration at Site 3:
{[P
1
Q
1
W
1
(1x
1
)]a
13
[W
2
(1x
2
)]a
23
}/Q
3
P
3
max
(4.153)
Box 4.2. Calibration of water quality model transfer
coefficient parameter a
12
E
0
2
0
9
0
3
b
! Calibration of Water Quality Model parameter a .
Define variables:
SP1(k) = sample pollutant concentration just downstream of site 1 (mg/l).
SP2(k) = sample pollutant concentration just upstream of site 2 (mg/l).
PE(k) = positive error in pollutant conc. sample just upstream of site 2 (mg.l).
NE(k) = negative error in pollutant conc. sample just upstream of site 2 (mg.l).
Qi = streamflow at site i (i=1, 2), (m3/s).
a12 = pollutant transfer coefficient for stream reach between sites 1 and 2. ;
Sets:
Sample / 1..5 / : PE, NE, SP1, SP2 ;
Endsets
!
Objective: Minimize total sum of positive and negative errors.
;
Min = @sum( Sample: PE + NE )
;
! Subject to constraint for each sample k:
;
@For (Sample: a12 * SP1 = ( SP2 + PE - NE )* (Q2/Q1));
Data:
SP1 = 232, 256, 220, 192, 204;
SP2 = 55, 67, 53, 50, 51;
Q1 = 12; !Flow downstream of site 1; Q2 = 12; !Flow upstream of site 2;
Enddata
Solution: a = 0.25; Total sum of absolute values of deviations = 10.0
12
12
wrm_ch04.qxd 8/31/2005 11:52 AM Page 120
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 121
Equation 4.153 assumes that each pollutant discharged
into the stream can be tracked downstream, independent
of the other pollutants in the stream. Alternatively,
Equation 4.148 computes the sum of all the pollutants
found at Site 2 and then uses that total mass to compute
the concentration at Site 3. Both modelling approaches
give the same results if the parameter values and cost
functions are the same.
To illustrate the solution of either of these models,
assume the values of the parameters are as listed in
Table 4.22. Rewriting the water quality management
model defined by Equations 4.150 to 4.153 and substi-
tuting the parameter values in place of the parameters,
and recalling that kg/day 86.4 (mg/l)(m
3
/s):
Minimize C
1
(x
1
) C
2
(x
2
) (4.154)
Subject to:
Water quality constraint at Site 2:
[P
1
Q
1
W
1
(1x
1
)]a
12
/Q
2
P
2
max
(4.155)
[(32)(10) 250000(1x
1
)/86.4]0.25/12 20
that when simplified is: x
1
0.78
Water quality constraint at Site 3:
[P
1
Q
1
W
1
(1x
1
)]a
13
[W
2
(1x
2
)]a
23
}/Q
3
P
3
max
(4.156)
{[(32)(10) 250000(1x
1
)/86.4]0.15
[80000(1x
2
)/86.4]0.60}/13 20
that when simplified is: x
1
1.28x
2
1.79
Restrictions on fractions of waste removal:
0 x
i
1.0 for sites i 1 and 2 (4.157)
Box 4.3. Calibration of water quality model transfer
coefficient parameter a
23
E
0
2
0
9
0
3
c
! Calibration of Water Quality Model parameter a .
Define variables:
SP2(k) = sample pollutant concentration just downstream of site 2 (mg/l).
SP3(k) = sample pollutant concentration at site 3. (mg/l).
PE(k) = positive error in pollutant conc. sample at site 3 (mg/l).
NE(k) = negative error in pollutant conc. sample at site 3 (mg/l).
Qi = streamflow at site i (i=2, 3), (m3/s).
a23 = pollutant transfer coefficient for stream reach between sites 2 and 3.
;
Sets:
Sample / 1..5 / : PE, NE, SP2, SP3 ;
Endsets
!
Objective: Minimize total sum of positive and negative errors.
;
Min = @sum( Sample: PE + NE )
;
! Subject to constraint for each sample k:
;
@For (Sample: a23 * SP2 = (SP3 + PE NE )* (Q3/Q2) );
Data:
SP2 = 158, 180, 140, 150, 135;
SP3 = 96, 107, 82, 92, 81;
Q2 = 13; !Flow just downstream of site 2; Q3 = 13; !Flow at site 3;
Enddata
Solution: a = 0.60; Total sum of absolute values of deviations = 6.2
23
23
wrm_ch04.qxd 8/31/2005 11:52 AM Page 121
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The feasible combinations of x
1
and x
2
can be shown
on a graph, as in Figure 4.21. This graph is a plot of
each constraint, showing the boundaries of the region
of combinations of x
1
and x
2
that satisfy all the
constraints. This red shaded region is called the feasible
region.
To find the least-cost solution we need the cost func-
tions C
1
(x
1
) and C
2
(x
2
) in Equations 4.151 or 4.154.
Suppose these functions are not known. Can we
122 Water Resources Systems Planning and Management
determine the least-cost solution without knowing these
costs? Models like the one just developed can be used to
determine just how accurate these cost functions (or any
model parameters) need to be for the decisions being
considered.
While the actual cost functions are not known, their
general form can be assumed, as shown in Figure 4.22.
Since the wasteloads produced at Site 1 are substantially
greater than those produced at Site 2, and given similar
site, transport, labour, and material cost conditions, it
seems reasonable to assume that the cost of providing a
specified level of treatment at Site 1 would exceed (or
certainly be no less than) the cost of providing the same
specified level of treatment at Site 2. It would also seem
the marginal cost at Site 1 would be greater than, or at
least no less than, the marginal cost at Site 2 for the same
amount of treatment. The relative positions of the cost
functions shown in Figure 4.22 are based on these
assumptions.
Rewriting the cost function, Equation 4.154, as a
linear function converts the model defined by Equations
4.150 through 4.151 into a linear programming model.
For this example problem, the linear programming model
can be written as:
Minimize c
1
x
1
c
2
x
2
(4.158)
E
0
2
0
8
2
7
r
parameter
f
l
o
w
Q m 10 flow just upstream of site 1
1
Q
2
Q
3
w
a
s
t
e
W 250,000 pollutant mass produced at site 1 kg/day
1
P 32 concentration just upstream of site 1 mg/l
1
P
2
P
3
W
2
a
12
a
13
a
23
p
o
l
l
u
t
a
n
t
c
o
n
c
.
d
e
c
a
y
f
r
a
c
t
i
o
n
3
/s
unit
--
12 flow just upstream of site 2
80,000 pollutant mass produced at site 2
13 flow at park
20 maximum allowable concentration upstream of 2
20 maximum allowable concentration at site 3
0.25 fraction of site 1 pollutant mass at site 2
0.15 fraction of site 1 pollutant mass at site 3
0.60 fraction of site 2 pollutant mass at site 2
value remark
m
3
/s
m
3
/s
kg/day
mg/l
mg/l
--
--
Table 4.22. Parameter values selected for
the water quality management problem
illustrated in Figure 4.20.
0.0
1.5
1.0
0.5
0.0
X
2
X
1
0.5 1.0 1.5 2.0
feasible region
equation 160
equation 160
equation 162
equation 163
E
0
2
0
1
0
3
o
Figure 4.21. Plot of the constraints of water quality
management model identifying those values of the unknown
(decision) variables x
1
and x
2
that satisfy all the constraints.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 122
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 123
Subject to:
x
1
0.78 (4.159)
x
1
1.28 x
2
1.79 (4.160)
0 x
i
1.0 for i 1 and 2 (4.161)
where the values of c
1
and c
2
depend on the values of x
1
and x
2
and both pairs are unknown. Even if we knew the
values of x
1
and x
2
before solving the problem, in this
example the cost functions themselves (Figure 4.22) are
unknown. Hence, we cannot determine the values of the
marginal costs c
1
and c
2
. However, we might be able to
judge which will likely be greater than the other for any
particular values of the decision-variables x
1
and x
2
.
First, assume c
1
equals c
2
. Let c
1
x
1
c
2
x
2
equal c and
assume c/c
1
1. Thus x
1
x
2
1.0. This line can be plot-
ted onto the graph in Figure 4.23, as shown by line a in
Figure 4.23.
Line a in Figure 4.23 represents equal values for c
1
and c
2
, and the total cost, c
1
x
1
c
2
x
2
, equal to 1. Keeping
the slope of this line constant and moving it upward, rep-
resenting increasing total costs, to line b, where it covers
the nearest point in the feasible region, will identify the
least-cost combination of x
1
and x
2
, again assuming the
marginal costs are equal. In this case the solution is
approximately 80% treatment at both sites.
Note this particular least-cost solution applies for
any value of c
1
greater than c
2
(for example line c in
Figure 4.23). If the marginal cost of 80% treatment at Site
1 is no less than the marginal cost of 80% treatment at
Site 2, then c
1
c
2
and indeed the 80% treatment effi-
ciencies will meet the stream standards for the design
streamflow and wasteload conditions at a total minimum
cost. In fact, from Figure 4.23 and Equation 4.160, it is
clear that c
2
has to exceed c
1
by a multiple of 1.28 before
the least-cost solution changes to another solution. For
any other assumption regarding c
1
and c
2
, 80% treatment
at both sites will result in a least-cost solution to meeting
the water quality standards for those design wasteload
and streamflow conditions.
If c
2
exceeds 1.28c
1
, as illustrated by line d, then the
least-cost solution would be x
1
100% and x
2
62%.
Clearly, in this example the marginal cost, c
1
, of provid-
ing 100% wasteload removal at Site 1 will exceed the
marginal cost, c
2
, of 60% removal at Site 2, and hence,
that combination of efficiencies would not be a least-cost
one. Thus we can be confident that the least-cost solution
is to remove 80% of the waste produced at both waste-
generating sites.
Note the least-cost wasteload removal efficiencies have
been determined without knowing the cost functions.
Why spend money defining these functions more pre-
cisely? The answer: costs need to be known for financial
0.0
t
o
t
a
l

c
o
s
t
s

(
?
)
treatment efficiency, x , at site i
i
0.3 0.6
E
0
2
0
1
0
3
p
0.1 0.2 0.4 0.5 0.8 0.7 0.9 1.0
c
c
2
1
s
it
e
2
s
ite
Figure 4.22. General form of total cost functions for
wastewater treatment efficiencies at Sites 1 and 2 in Figure
4.20. The dashed straight-line slopes c
1
and c
2
are the average
cost per unit (%) removal for 80% treatment. The actual
average costs clearly depend on the values of the waste
removal efficiencies x
1
and x
2
respectively.
0.0
1.5
1.0
0.5
0.0
X
2
X
1
0.5 1.0 1.5 2.0
feasible region
E
0
2
0
1
0
3
q
b
d
a
c
X
X 0.79
=
=
1
2
X
X
1.00
0.62
=
=
1
2
0.78
Figure 4.23. Plots of various objective functions (dashed
lines) together with the constraints of the water quality
management model.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 123
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
planning, if not for economic analyses. No doubt the
actual costs of installing the least-cost treatment efficien-
cies of 80% will have to be determined for issuing bonds
or making other arrangements for paying the costs.
However, knowing the least-cost removal efficiencies
means we do not have to spend money defining the entire
cost functions C
i
(x
i
). Estimating the construction and oper-
ating costs of achieving just one wastewater removal effi-
ciency at each site, namely 80%, should be less expensive
than defining the total costs for a range of practical treat-
ment plant efficiencies that would be required to define
the total cost functions, such as shown in Figure 4.22.
Admittedly this example is relatively simple. It will not
always be possible to determine the optimal solutions
to linear programming problems, or other optimization
problems, without knowing more about the objective
function than was assumed for this example. However,
this exercise illustrates the use of modelling for purposes
other than finding good or optimal solutions. Models can
help define the necessary precision of the data needed to
find those good solutions.
Modelling and data collection and analysis should take
place simultaneously. All too often planning exercises are
divided into two stages: data collection and then analysis.
Until one knows what data one will need, and how accu-
rate those data must be, one should not spend money and
time collecting them. Conversely, model development in
the absence of any knowledge of the availability and cost
of obtaining data can lead to data requirements that are
costly, or even impossible, to obtain, at least in the time
available for decision-making. Data collection and model
development are activities that should be performed
simultaneously.
Because software is widely available to solve linear
programming programs, because these software programs
can solve very large problems containing thousands of
variables and constraints, and finally because there is less
chance of obtaining a local non-optimal solution when
the problem is linear (at least in theory), there is an
incentive to use linear programming to solve large
optimization problems. Especially for large optimization
problems, linear programming is often the only practical
alternative. Yet models representing particular water
resources systems may not be linear. This motivates the
use of methods that can approximate non-linear functions
with linear ones.
124 Water Resources Systems Planning and Management
The following simple groundwater supply problem
illustrates the application of some linearization methods
commonly applied to non-linear separable functions
functions of only one unknown variable.
These methods typically increase the number of vari-
ables and constraints in a model. Some of these methods
require integer variables, or variables that can have values
of only 0 or 1. There is a practical limit on the number of
integer variables any linear programming software
program can handle. Hence, for large models there may
be a need to perform some preliminary screening
designed to reduce the number of alternatives that should
be considered in more detail. This example can be used to
illustrate an approach to preliminary screening.
5.3. A Groundwater Supply Example
Consider a water-using industry that plans to obtain
water from a groundwater aquifer. Two wellfield sites
have been identified. The first question is how much
the water will cost, and the second, given any specified
amount of water delivered to the user, is how much
should come from each wellfield. This situation is illus-
trated in Figure 4.24.
Wells and pumps must be installed and operated to
obtain water from these two wellfields. The annual cost of
wellfield development will depend on the pumping
capacity of the wellfield. Assume that the annual costs
associated with various capacities Q
A
and Q
B
for Wellfields
industry
w
ellfields
A B
E
0
2
0
1
0
3
r
Q
Q
A
Q
B
Figure 4.24. Schematic of a potential groundwater supply
system that can serve a water using industry. The unknown
variables are the flows, Q
A
and Q
B
, from each wellfield.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 124
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 125
A and B respectively are as shown in Figure 4.25. These
are non-linear functions that contain both fixed and vari-
able costs and hence are discontinuous. The fixed costs
result from the fact that some of the components required
for wellfield development come in discrete sizes. As indi-
cated in the figure, the maximum flow capacity of
Wellfields A and B are 17 and 13, respectively.
In Figure 4.25, the non-linear functions on the left
have been approximated by piecewise linear functions on
the right. This is a first step in linearizing non-linear
separable functions. Increasing the number of linear
segments can reduce the difference between the piecewise
linear approximation of the actual non-linear function
and the function itself. At the same time it will increase
the number of variables and possibly constraints.
When approximating a non-linear function by a series
of straight lines, model developers should consider two
factors. The first is just how accurate need be the approx-
imation of the actual function for the decisions that will
be made, and second is just how accurate is the actual (in
this case non-linear) function in the first place. There is
little value in trying to eliminate relatively small errors
caused by the linearization of a function when the
function itself is highly uncertain. Most cost and benefit
functions, especially those associated with future activi-
ties, are indeed uncertain.
5.3.1. A Simplified Model
Two sets of approximations are shown in Figure 4.25.
Consider first the approximations represented by the light
blue dotdash lines. These are very crude approximations
a single straight line for each function. In this example these
straight-line cost functions are lower bounds of the actual
non-linear costs. Hence, the actual costs may be somewhat
higher than those identified in the solution of a model.
Using the blue dotdash linear approximations in
Figure 4.25, the linear programming model can be
written as follows:
Minimize CostA CostB (4.162)
Subject to:
CostA 8I
A
[(408)/17]Q
A
{linear approximation of C(Q
A
)} (4.163)
CostB 15I
B
[(2615)/13]Q
B
{linear approximation of C(Q
B
)} (4.164)
I
A
, I
B
are 0, 1 integer variables (4.165)
Q
A
17I
A
{limits Q
A
to 17 and forces I
A
1
if Q
A
0} (4.166)
Q
B
13I
B
{limits Q
B
to 13 and forces I
B
1
if Q
B
0} (4.167)
5
3
40
26
35
10
10
17
13
30
26
20
8
20
18
15
0
0
0
0
flow Q well field A
A
flow Q
B
a
n
n
u
a
l

c
o
s
t
a
n
n
u
a
l

c
o
s
t
(
C
Q
)
A
a
n
n
u
a
l

c
o
s
t
(
C
Q
)
B
a
n
n
u
a
l

c
o
s
t
E
0
2
0
1
0
3
s
well field B
Figure 4.25. Annual cost
functions associated with the
Wellfields A and B as shown in
Figure 4.24. The actual functions
are shown on the left, and two
sets of piecewise linear
approximations are shown on
the right.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 125
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Q Q
A
Q
B
{mass balance} (4.168)
Q, Q
A
, Q
B
0 {non-negativity of all decision variables}
(4.169)
Q some specified amount from 0 to 30. (4.170)
The expressions within the square brackets, [ ], above rep-
resent the slopes of the dotdash linear approximations of
the cost functions. The integer 0, 1 variables are required
to include the fixed costs in the model.
Solving this model for various values of the water
demand Q provides some interesting results. Again, they
are based on the dotdash linear cost functions in Figure
4.25. As Q increases from 0 to just under 6.8, all the water
will come from the less expensive Wellfield A. For any Q
from 6.8 to 13, Wellfield B becomes less expensive and all
the water will come from it. For any Q greater than the
capacity of Wellfield B of 13 but no greater than the
capacity of Wellfield A, 17, all of it will come from
Wellfield A. Because of the fixed costs, it is cheaper to use
one rather than both wellfields. Beyond Q 17, the max-
imum capacity of A, water needs to come from both well-
fields. Wellfield B will pump at its capacity, 13, and the
additional water will come from Wellfield A.
Figure 4.26 illustrates these solutions. One can under-
stand why in situations of increasing demands for Q over
time, capacity-expansion modelling might be useful. One
would not close down a wellfield once developed, just to
achieve what would have been a least-cost solution if the
existing wellfield had not been developed.
126 Water Resources Systems Planning and Management
5.3.2. A More Detailed Model
A more accurate representation of these cost functions
may change these solutions for various values of Q,
although not significantly. However consider the more
accurate cost minimization model that includes the red
solid-line piecewise linearizations shown in Figure 4.25.
Minimize CostA CostB (4.171)
Subject to:
CostA {8I
A1
[(208)/5]Q
A1
} {26I
A2
[(3026)/
(105)]Q
A2
} {35I
A3
[(4035)/
(1710)]Q
A3
} {linear approximation of C(Q
A
)}
(4.172)
CostB {15I
B1
[(1815)/3]Q
B1
}
{18I
B2
[(2018)/(103)]Q
B2
[(2620)/(1310)]Q
B3
}
{linear approximation of C(Q
B
)} (4.173)
Q
A
Q
A1
(5I
A2
Q
A2
) (10I
A3
Q
A3
)
{Q
A
defined} (4.174)
Q
B
Q
B1
(3I
B2
Q
B2
Q
B3
) {Q
B
defined} (4.175)
I
Ai
, I
Bi
are 0, 1 integer variables for all segments i (4.176)
Q
A1
5I
A1
, Q
A2
(10 5)I
A2
, Q
A3
(17 10)I
A3
{limits Q
Ai
to width of segment i and forces
I
Ai
1 if Q
Ai
0} (4.177)
I
A1
I
A2
I
A3
1 (4.178)
{limits solution to at most only one cost function
segment i}
Q
B1
3I
B1
, Q
B2
(103)I
B2
, Q
B3
(1310)I
B2
(4.179)
{limits Q
Bi
to width of segment i and forces I
Bi
1
if Q
Bi
0}
I
B1
I
B2
1 (4.180)
{limits solution to at most only the first segment or to
the second and third segments of the cost function.
Note that a 0, 1 integer variable for the fixed cost of
the third segment of this function is not needed since
its slope exceeds that of the second segment. However
the flow, Q
B3
, in that segment must be bounded
using the integer 0, 1 variable, I
B2
, associated with the
second segment, as shown in the third of Equations
4.179}
0 5 10 15 20 25 30
total demand Q
w
e
l
l
f
i
e
l
d

w
i
t
h
d
r
a
w
a
l
20
15
10
5
0
A
B
E
0
2
0
7
2
3
p
Figure 4.26. Least-cost wellfield use given total demand Q
based on model defined by Equations 4.162 to 4.170.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 126
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 127
Q Q
A
Q
B
{mass balance} (4.181)
Q, Q
A
, Q
B
0 {non-negativity of all decision variables}
(4.182)
Q some specified amount from 0 to 30. (4.183)
The solution to this model, shown in Figure 4.27, differs
from the solution of the simpler model, but only in the
details. Wellfield A supplies all the water for Q 4.3. For
values of Q between 4.4 and 13 all the water comes from
Wellfield B. For values of Q in excess of 13 to 14.8, the
capacity of Wellfield B remains at 13 and Wellfield A
provides the additional amount of needed capacity over
13. For Q 14.9 to 17, the capacity of Wellfield B drops
to 0 and the capacity of Wellfield A increases from 14.9
to 17. For values of Q between 17 and 18 Wellfield B
provides 13, its capacity, and the capacity of A increases
from 4 to 5. For values of Q from 18.1 to 27, Wellfield B
provides 10, and Wellfield A increases from 8.1 to 17.
For values of Q in excess of 27, Wellfield A remains at 17,
its capacity, and Wellfield B increases from 10 to 13.
As in the previous example, this shows the effect
on the least-cost solution when one cost function has
relatively lower fixed and higher variable costs compared
with another cost function having relatively higher fixed
and lower variable costs.
5.3.3. An Extended Model
In this example, the simpler model (Equations 4.162 to
4.170) and the more accurate model (Equations 4.171
to 4.183) provided essentially the same allocations of
wellfield capacities associated with a specified total
capacity Q. If the problem contained a larger number of
wellfields, the simpler (and smaller) model might have
been able to eliminate some of these wellfields from
further consideration. This would reduce the size of any
new model that approximates the cost functions of the
remaining wellfields more accurately.
The model just described, like the capacity-expansion
model and water quality management model, is another
example of a cost-effective model. The objective was to
find the least-cost way of providing a specified amount of
water to a water user. Next, consider a costbenefit analy-
sis in which the question is just how much water
the user should use. To address this question we assume
the user has identified the annual benefits associated with
various amounts of water. The annual benefit function,
B(Q), and its piecewise linear approximations, are shown
in Figure 4.28.
The straight, blue, dotdashdash linear approxima-
tion of the benefit function shown in Figure 4.28 is an
0 5 10 15 20 25 30
total demand Q
w
e
l
l
f
i
e
l
d

w
i
t
h
d
r
a
w
a
l
20
15
10
5
0
A
B
E
0
2
0
7
2
3
q
Figure 4.27. Least-cost wellfield use given total demand Q
based on Equations 4.171 to 4.183.
9
50
21 30
45
25
10
-10
0
flow Q
a
n
n
u
a
l

b
e
n
e
f
i
t
B

(
Q
)
a
n
n
u
a
l

b
e
n
e
f
i
t
B

(
Q
)
E
0
2
0
1
0
3
t
33
Figure 4.28. Benefit function of
the amount of water provided to
the water user. Piecewise linear
approximations of that function
of flow are shown on the right.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 127
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
upper bound of the benefits. Incorporating it into a
model that uses the dotdash linear lower bound
approximations of each cost function, as shown in Figure
4.25, will produce an optimistic solution. It is unlikely
that the value of Q that is based on more accurate and
thus less optimistic benefit and cost functions will be any
greater than the one identified by this simple optimistic
model. Furthermore, if any wellfield is not in the solution
of this optimistic model, with some care we might be able
to eliminate that wellfield from further consideration
when developing a more accurate model.
Any component of a water resources system that does
not appear in the solution of a model that includes
optimistic approximations of performance measures that
are to be maximized, such as benefits, or that are to be
minimized, such as costs, are candidates for omission in
any more detailed model. This is an example of the
process of preliminary screening.
The model defined by Equations 4.162 through 4.170
can now be modified. Equation 4.170 is eliminated and the
cost-minimization objective Equation 4.162 is replaced with:
Maximize Benefits (CostA CostB) (4.184)
where
Benefits 10 [(45 25)/(21 9)]Q
{linear approximation of B(Q)} (4.185)
The solution of this model, Equations 4.163 to 4.169,
4.184, and 4.185 (plus the condition that the fixed
benefit of 10 only applies if Q 0, added because it is
clear the benefits would be 0 with a Q of 0) indicates that
only Wellfield B needs to be developed, and at a capacity
of 10. This would suggest that Wellfield A can be omitted
in any more detailed modelling exercise. To see if this
assumption, in this example, is valid, consider the more
detailed model that incorporates the red, solid-line linear
approximations of the cost and benefit functions shown
in Figures 4.25 and 4.28.
Note that the approximation of the generally concave
benefit function in Figure 4.28 will result in negative
values of the benefits for small values of Q. For example,
when the flow Q, is 0 the approximated benefits are 10.
Yet the actual benefits are 0 as shown in the left part of
Figure 4.28. Modelling these initial fixed benefits the
same way as the fixed costs have been modelled, using
another 0, 1 integer variable, would allow a more accurate
representation of the actual benefits for small values of Q.
128 Water Resources Systems Planning and Management
Alternatively, to save having to add another integer
variable and constraint to the model, one can allow the
benefits to be negative. If the model solution shows nega-
tive benefits for some small value of Q, then obviously
the more preferred value of Q, and benefits, would be 0.
This more approximate trial-and-error approach is often
preferred in practice, especially when a model contains
a large number of variables and constraints. This is
the approach taken here.
5.3.4. Piecewise Linearization Methods
There are a number of ways of modelling the piecewise
linear concave benefit function shown on the right side of
Figure 4.28. Several are defined in the next several sets
of equations. Each method will result in the same model
solution.
One approach to modelling the concave benefit
function is to define a new unrestricted (possibly negative-
valued) variable. Let this variable be Benefits. When being
maximized this variable cannot exceed any of the linear
functions that bound the concave benefit function:
Benefits 10 [(25 (10))/9]Q (4.186)
Benefits 10 [(45 25)/(219)]Q (4.187)
Benefits 33 [(50 45)/(3021)]Q (4.188)
Since most linear programming algorithms assume all
unknown variables are non-negative (unless otherwise
specified), unrestricted variables, such as Benefits, can be
replaced by the difference between two non-negative
variables, such as Pben Nben. Pben will equal Benefits if
its value is greater than 0. Otherwise Nben will equal
Benefits. Thus in place of Benefits in the Equations 4.186
to 4.188, and those below, one can substitute Pben
Nben.
Another modelling approach is to divide the variable
Q into parts, q
i
, one for each segment i of the function.
These parts sum to Q. Each q
i
, ranges from 0 to the
width of the user-defined segment i. Thus for the
piecewise linear benefit function shown on the right of
Figure 4.28:
q
1
9 (4.189)
q
2
21 9 (4.190)
q
3
30 21 (4.191)
wrm_ch04.qxd 8/31/2005 11:52 AM Page 128
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 129
and
Q q
1
q
2
q
3
(4.192)
The linearized benefit function can now be written as the
sum over all three segments of each segment slope times
the variable q
i
:
Benefits 10 [(25 10)/9]q
1
[(45 25)/(21 9)]q
2
[(50 45)/(30 21)]q
3
(4.193)
Since the function being maximized is concave (decreas-
ing slopes as Q increases), we are assured that each q
i1
will be greater than 0 only if q
i
is at its upper limit, as
defined by constraint Equations 4.189 to 4.191.
A third method is to define unknown weights w
i
associated with the breakpoints of the linearized function.
The value of Q can be expressed as the sum of a weighted
combination of segment endpoint values. Similarly, the
benefits associated with Q can be expressed as a weighted
combination of the benefits evaluated at the segment
endpoint values. The weights must also sum to 1.
Hence, for this example:
Benefits (10)w
1
25w
2
45w
3
50w
4
(4.194)
Q 0w
1
9w
2
21w
3
30w
4
(4.195)
1 w
1
w
2
w
3
w
4
(4.196)
For this method to provide the closest approximation
of the original non-linear function, the solution must
include no more than two non-zero weights and those
non-zero weights must be adjacent to each other. Since
a concave function is to be maximized, this condition
will be met, since any other situation would yield less
benefits.
The solution to the more detailed model defined by
Equations 4.184, 4.172 to 4.182, and either 4.186 to
4.188, 4.189 to 4.193, or 4.194 to 4.196, indicates a
value of 10 for Q will result in the maximum net benefits.
This flow is to come from Wellfield B. This more precise
solution is identical to the solution of the simpler model.
Clearly the simpler model could have successfully served
to eliminate Wellfield A from further consideration.
5.4. A Review of Linearization Methods
This section presents a review of the piecewise lineariza-
tion methods just described and some other approaches
for incorporating special conditions into linear program-
ming models.
Ifthenelse conditions
There exist a number of ways ifthenelse and and and
or conditions (that is, decision trees) can be included in
linear programming models. To illustrate some of them,
assume X is an unknown decision-variable in a model
whose value may depend on the value of another
unknown decision-variable Y. Assume the maximum
value of Y would not exceed Ymax and the maximum
value of X would not exceed Xmax. These upper bounds
and all the linear constraints representing ifthenelse
conditions must not restrict the values of the original
decision-variable Y. Four ifthenelse (with and/or)
conditions are presented below using additional integer
0.1 variables, denoted by Z. All the X, Y and Z variables
in the constraints below are assumed to be unknown.
These constraints would be included in the any linear
programming model where the particular ifthenelse
conditions apply.
These illustrations are not unique. At the boundaries
of the if constraints in the examples below, either of the
then or else conditions can apply.
a) If Y 50 then X 10, else X 15.
Define constraints:
Y 50Z Ymax(1 Z) where Z is a 0, 1
integer variable.
Y 50(1 Z)
X 10Z Xmax(1 Z)
X 15(1 Z)
b) If Y 50 then X Y, else X Y.
Define constraints:
Y 50Z where Z is a 0, 1 integer variable.
Y 50(1 Z) YmaxZ
X Y XmaxZ
X Y Ymax(1Z)
c) If Y 20 or Y 80 then X 5, else X 10.
Define constraints:
Y 20Z
1
80Z
2
Ymax(1 Z
1
Z
2
)
Y 20Z
2
80(1 Z
1
Z
2
)
Z
1
Z
2
1 where each Z is a 0, 1 integer variable.
X 5(Z
1
(1 Z
1
Z
2
)) XmaxZ
2
X 5(Z
1
(1 Z
1
Z
2
))
X 10Z
2
wrm_ch04.qxd 8/31/2005 11:52 AM Page 129
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
d) If 20 Y 50 or 60 Y 80, then X 5,
else X 10.
Define constraints:
Y 20Z
1
50Z
2
60Z
3
80Z
4
Ymax(1 Z
1
Z
2
Z
3
Z
4
)
Y 20Z
2
50Z
3
60Z
4
80(1 Z
1
Z
2
Z
3
Z
4
)
Z
1
Z
2
Z
3
Z
4
1 where each Z is a 0, 1
integer variable.
X 5 (Z
2
Z
4
) Xmax*(1 Z
2
Z
4
)
X 10* ((Z
1
Z
3
) (1 Z
1
Z
2
Z
3
Z
4
))
Fixed costs in cost functions:
Cost C
0
CX if X 0,
0 otherwise.
To include these fixed costs in a LP model, define
Cost C
0
I CX and constrain X MI where M is
the maximum value of X, and I is an unknown 0, 1
variable.
Minimizing the maximum or maximizing the
minimum
Let the set of variables be {X
1
, X
2
, X
3
, , X
n
}
Minimize maximum {X
1
, X
2
, X
3
, , X
n
} is equivalent
to:
Minimize U subject to U X
j
, j 1, 2, 3, , n.
Maximize minimum {X
1
, X
2
, X
3
, , X
n
} is equivalent to:
Maximize L subject to L X
j
, j 1, 2, 3, , n.
130 Water Resources Systems Planning and Management
Minimizing the absolute value of the difference
between two unknown non-negative variables:
Minimize | X Y | is equivalent to
Minimize D subject to X Y D; Y X D;
X, Y, D 0.
or Minimize (PDND)
subject to: X Y PD ND; PD, ND, X, Y 0.
Minimizing convex functions or maximizing concave
functions
Maximize G(X) Maximize B
Subject to: I
1
S
1
X B
I
2
S
2
X B
I
3
S
3
X B
F(X) S
1
x
1
S
2
x
2
S
3
x
3
;
G(X) S
1
x
1
S
2
x
2
S
3
x
3
X x
1
x
2
x
3
; x
1
a; x
2
b a
E
0
2
0
1
0
3
u
c
o
s
t
X
C
0
=
fixed cost if X >0
C=
variable
cost / unit
G
(
X
)
X
0 a b
3
2
1
S slopes S
S
S
1
2
3
s
E
0
2
0
1
0
3
v
F
(
X
)
G

(
X
)
X X
E
0
2
0
1
0
3
w
0 b a
1
1
1 1
1
2 3 3 2 1
variables:
slopes:
x
S
x
S
2
2
x
S
3
3
x
S
x
S
2
2
x
S
3
3
segments:
a b 0
wrm_ch04.qxd 8/31/2005 11:52 AM Page 130
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 131
Minimize F(X) F(0)w
1
F(a)w
2
F(b)w
3
F(c)w
4
Maximize G(X) G(0)w
1
G(a)w
2
G(b)w
3
G(c)w
4
Subject to:
X 0w
1
aw
2
bw
3
cw
4
;
w
1
w
2
w
3
w
4
1
Minimizing concave functions or maximizing
convex functions
Minimizing or maximizing combined
concaveconvex functions
Maximize C(X) (5z
1
6x
1
3x
2
) (53z
3
5x
3
)
Subject to: (x
1
x
2
) (12z
3
x
3
) X;
x
1
4z
1
; x
2
8z
1
; x
3
99z
3
; z
1
z
3
1;
z
1
, z
3
0, 1.
Minimize C(X) (5z
1
6x
1
) (29z
2
3x
2
5x
3
)
Subject to: x
1
(4z
2
x
2
x
3
) X;
x
1
4z
1
; x
2
8z
2
; x
3
99 z
2
; z
1
z
2
1;
z
1
, z
2
0, 1.
F
(
X
)
G
(
X
)
X X
E
0
2
0
1
0
3
x
0 b a
2 3 3 2 1
weight:
w w
2 1 1
segments: 1
c c
w
3
w
4
w w
2
w
3
w
4
a 0 b
G
(
)
X
X
0
S
slopes S
S
S
1
2
3
s
E
0
2
0
1
0
3
y
x
1 2 3 segments:
x
2
x
3
=5
=3
=2
44
20
0
4 12
1
F
X
(
)
X
0
E
0
2
0
1
0
8
b
x x
2 1
x
3
35
32
22
5
4 12 17 x
4
6
3
-2
0
53
C

(
X
)
x
29
5
X
0 12
x
2 1
x
3
6
3
5
E
0
2
0
1
0
8
a
4
Minimize G(X) 5x
1
(20z
2
3x
2
) (44z
3
2x
3
)
Subject to:
x
1
(4z
2
x
2
) (12z
3
x
3
) X;
z
s
0 or 1 s;
x
1
4z
1
; x
2
8z
2
; x
3
99 z
3
; z
1
z
2
z
3
1.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 131
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Maximize or Minimize F(X)
F(X) (5z
1
6x
1
) (35z
2
3x
2
) (32z
3
2x
3
) 22z
4
Subject to:
x
1
(4z
2
x
2
) (12z
3
x
3
) (17z
4
x
4
) X;
x
1
4z
1
; x
2
8z
2
; x
3
5z
3
; x
4
99z
4
;
z
s
1; z
s
0, 1 s.
s

132 Water Resources Systems Planning and Management

decision-variables and multiple constraints. Also like their
real-world counterparts, there are multiple feasible solu-
tions to each of these problems. Hence, the task is to find
the best solution, or a number of near-best solutions.
Each solution must satisfy all the constraints.
Constraints can reflect physical conditions, environ-
mental regulations and/or social or economic targets.
Especially with respect to environmental or social condi-
tions and goals, it is often a matter of judgement to decide
what is considered an objective that is to be minimized or
maximized and what is considered a constraint that has to
be met.
Except for relatively simple problems, the use of these
optimization models and methods is primarily for reduc-
ing the number of alternatives that need to be further
analysed and evaluated using simulation methods.
Optimization is generally used for preliminary screening
eliminating inferior alternatives before more detailed
analyses are carried out. Presented were some approaches
to preliminary screening involving calculus-based
Lagrange multiplier and non-linear programming meth-
ods, discrete dynamic programming methods, and linear
programming methods. Each method has its strengths
and limitations.
The example problems used to illustrate these model-
ling and model solution methods have been relatively
simple. However, simple applications such as these can
form the foundation of models of more complex prob-
lems, as will be shown in later chapters.
7. Reference
BELLMAN, R.E. 1957. Dynamic programming. Princeton,
N.J., Princeton University Press.
Additional References (Further Reading)
BASSSON, M.S.; ALLEN, R.B.; PEGRAM, G.G.S. and
VAN ROOYEN, J.A. 1994. Probabilistic management of
water resource and hydropower systems. Highlands Ranch,
Colo., Water Resources Publications.
BEROGGI, G.E.G. 1999. Decision modelling in policy man-
agement: an introduction to analytic concepts. Boston, Mass.,
Kluwer Academic.
C

(
X
)
E
0
2
0
1
0
8
c
17
5
X
0
4
x x
2 1
x
3
-17
6
3
5
12
Maximize C(X) (5z
1
6x
1
3x
2
) (17z
3
5x
3
)
Subject to: (x
1
x
2
) x
3
X; z
1
, z
3
0, 1.
x
1
4z
1
; x
2
8z
1
; x
3
99z
3
; z
1
z
3
1;
Minimize C(X) (5z
1
6x
1
) (17z
2
3x
2
5x
3
)
Subject to: x
1
(4z
2
x
2
x
3
) X; z
1
, z
2
0, 1.
x
1
4z
1
; x
2
12z
2
; x
3
99z
2
; z
1
z
2
1
6. A Brief Review
Before proceeding to some other optimization and simu-
lation methods in the following chapters, it may be useful
to review the topics covered so far. The focus has been on
model development as well as model solution. Several
types of water resources planning and management
problems have been used to illustrate model development
and solution processes. Like their real-world counter-
parts, the example problems all had multiple unknown
wrm_ch04.qxd 8/31/2005 11:52 AM Page 132
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Optimization Methods 133
BLANCHARD, B.S. and FABRYCKY, W.J. 1998. Systems
engineering and analysis, 3rd edn. Upper Saddle River,
N.J., Prentice Hall.
BURAS, N. 1966. Dynamic programming and water
resources development, advances in hydroscience, Vol. 3.
New York, Academic Press.
DORFMAN, R.; JACOBY, H.D. and THOMAS, H.A. Jr.
1972. Models for managing regional water quality.
Cambridge, Mass., Harvard University Press.
ESOGBUE, A.O. (ed.). 1989 Dynamic programming for
optimal water resources systems analysis. Englewood Cliffs,
N.J., Prentice Hall.
HALL, W.A. and DRACUP, J.A. 1970. Water resources sys-
tems engineering. New York, McGraw-Hill.
HILLIER, F.S. and LIEBERMAN, G.J. 1990. Introduction to
operations research, 5th edn. New York, McGraw-Hill.
HILLIER, F.S. and LIEBERMAN, G.J. 1990. Introduction
to stochastic models in operations research. New York,
McGraw-Hill.
HUFSCHMIDT, M.M. and FIERING, M.B. 1966. Simulation
techniques for design of water-resource systems. Cambridge,
Mass., Harvard University Press.
KARAMOUZ, M.; ZINSSER, W.K. and SZIDAROVSZKY, F.
2003. Water resources systems analysis. Boca Raton, Fla.,
Lewis.
LOUCKS, D.P.; STEDINGER, J.S. and HAITH, D.A.
1981. Water resource systems planning and analysis.
Englewood Cliffs, N.J., Prentice Hall.
MAASS, A.; HUFSCHMIDT, M.M.; DORFMAN, R.;
THOMAS, H.A. Jr.; MARGLIN, S.A. and FAIR, G.M.
1962. Design of water-resource systems. Cambridge, Mass.,
Harvard University Press.
MAJOR, D.C. and LENTON, R.L. 1979. Applied water
resources systems planning. Englewood Cliffs, N.J., Prentice
Hall.
MAYS, L.W. and TUNG, Y.-K. 1992. Hydrosystems
engineering and management. New York, McGraw-Hill.
NEWNAN, D.G.; LAVELLE, J.P. and ESCHENBACH, T.G.
2002. Essentials of engineering economic analysis, 2nd edn.
New York, Oxford University Press.
RARDIN, R.L. 1998. Optimization in operations research.
Upper Saddle River, N.J., Prentice Hall.
REVELLE, C. 1999. Optimizing reservoir resources. New York,
John Wiley.
SOMLYODY, L. 1997. Use of optimization models in
river basin water quality planning: WATERMATEX97.
In: M.B. Beck and P. Lessard (eds.), Systems analysis and
computing in water quality management: towards a new
agenda. London, IWA.
TAYLOR, B.W. III. 1999. Introduction to management
science, 6th edn. Upper Saddle River, N.J., Prentice Hall.
WILLIS, R. and YEH, W.W.-G. 1987. Groundwater
systems planning and management. Englewood Cliffs, N.J.,
Prentice Hall.
WINSTON, W.L. 1994. Operations research: applications
and algorithms, 3rd edn. Boston, Mass., Duxbury.
WINSTON, W.L. and ALBRIGHT, S.C. 2001. Practical
management science, 2nd edn. Pacific Grove, Calif., Duxbury.
WRIGHT, J.R. and HOUCK, M.H. 1995. Water resources
planning and management. In: W.F. Chen (ed.), Civil
engineering handbook, Chapter 37. Boca Raton, CRC.
WURBS, R.A. 1996. Modelling and analysis of reservoir
system operations. Upper Saddle River, N.J., Prentice Hall.
wrm_ch04.qxd 8/31/2005 11:52 AM Page 133
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
5. Fuzzy Optimization
1. Fuzziness: An Introduction 135
1.1. Fuzzy Membership Functions 135
1.2. Membership Function Operations 136
2. Optimization in Fuzzy Environments 136
3. Fuzzy Sets for Water Allocation 138
4. Fuzzy Sets for Reservoir Storage and Release Targets 139
5. Fuzzy Sets for Water Quality Management 140
6. Summary 144
7. Additional References (Further Reading) 144
wrm_ch05.qxd 8/31/2005 11:52 AM Page 134
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Fuzziness: An Introduction
Large, small, pure, polluted, satisfactory, unsatisfactory,
sufficient, insufficient, excellent, good, fair, poor and so
on are words often used to describe various attributes or
performance measures of water resources systems. These
descriptors do not have crisp, well-defined boundaries
that separate them from others. A particular mix of eco-
nomic and environmental impacts may be more acceptable
to some and less acceptable to others. Plan A is better than
Plan B. The water quality and temperature is good for
swimming. These qualitative, or fuzzy, statements con-
vey information despite the imprecision of the italicized
adjectives.
This chapter illustrates how fuzzy descriptors can be
incorporated into optimization models of water resources
systems. Before this can be done some definitions are
needed.
1.1. Fuzzy Membership Functions
Consider a set A of real or integer numbers ranging from
say 18 to 25. Thus A[18, 25]. In classical (crisp) set
theory, any number x is either in or not in the set A. The
statement x belongs to A is either true or false depending
135
Fuzzy Optimization
The precise quantification of many system performance criteria and parameter and
decision variables is not always possible, nor is it always necessary. When the values
of variables cannot be precisely specified, they are said to be uncertain or fuzzy. If
the values are uncertain, probability distributions may be used to quantify them.
Alternatively, if they are best described by qualitative adjectives, such as dry or wet,
hot or cold, clean or dirty, and high or low, fuzzy membership functions can be used
to quantify them. Both probability distributions and fuzzy membership functions of
these uncertain or qualitative variables can be included in quantitative optimization
models. This chapter introduces fuzzy optimization modelling, again for the
preliminary screening of alternative water resources plans and management policies.
5
on the value of x. The set A is referred to as a crisp set.If
one is not able to say for certain whether or not any num-
ber x is in the set, then the set A could be referred to as
fuzzy. The degree of truth attached to that statement is
defined by a membership function. This function ranges
from 0 (completely false) to 1 (completely true).
Consider the statement, The water temperature
should be suitable for swimming. Just what temperatures
are suitable will depend on the person asked. It would
be difficult for anyone to define precisely those temper-
atures that are suitable if it is understood that temper-
atures outside that range are absolutely not suitable.
A membership function defining the interval or range
of water temperatures suitable for swimming is shown in
Figure 5.1. Such functions may be defined on the basis of
the responses of many potential swimmers. There is a
zone of imprecision or disagreement at both ends of the
range.
The form or shape of a membership function depends
on the individual subjective feelings of the members or
individuals who are asked their opinions. To define this
particular membership function, each individual i could
be asked to define his or her comfortable water temper-
ature interval (T
1i
, T
2i
). The membership value associated
with any temperature value T equals the number of
wrm_ch05.qxd 8/31/2005 11:52 AM Page 135
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
individuals who place that T within their range (T
1i
, T
2i
),
divided by the number of individual opinions obtained.
The assignment of membership values is based on
subjective judgements, but such judgements seem to be
sufficient for much of human communication.
1.2. Membership Function Operations
Denote the membership function associated with a fuzzy
set A as m
A
(x). It defines the degree or extent to which any
value of x belongs to the set A. Now consider two fuzzy
sets, A and B. Set A could be the range of temperatures
that are considered too cold, and set B could be the range
of temperatures that are considered too hot. Assume these
two sets are as shown in Figure 5.2.
The degree or extent that a value of x belongs to either
of two sets A or B is the maximum of the two individual
membership function values. This union membership
function is defined as:
m
AB
(x) maximum(m
A
(x), m
B
(x)) (5.1)
136 Water Resources Systems Planning and Management
This union set would represent the ranges of tempera-
tures that are either too cold or too hot, as illustrated in
Figure 5.3.
The degree or extent that a value of a variable x is
simultaneously in both sets A and B is the minimum of
the two individual membership function values. This
intersection membership function is defined as:
m
AB
(x) minimum (m
A
(x), m
B
(x)) (5.2)
This intersection set would define the range of tempera-
tures that are considered both too cold and too hot.
Of course it could be an empty set, as indeed it is in this
case, based on the two membership functions shown in
Figure 5.2. The minimum of either function for any value
of x is 0.
The complement of the membership function for
fuzzy set A is the membership function, m
A
c
(x), of A
c
.
m
A
c
(x) 1 m
A
(x) (5.3)
The complement of set A (defined in Figure 5.2) would
represent the range of temperatures considered not
too cold for swimming. The complement of set B (also
defined in Figure 5.2) would represent the range of
temperatures considered not too hot for swimming. The
complement of the union set as shown in Figure 5.3
would be the range of temperatures considered just right.
This complement set is the same as shown in Figure 5.1.
2. Optimization in Fuzzy
Environments
Consider the problem of finding the maximum value of
x given that x cannot exceed 11. This is written as:
Maximize U x (5.4)
E
0
2
0
1
0
8
d
0 5 10 15 20 25 30 35 40 45 50
0
1
C
o
E
0
2
0
1
0
8
e
0 5 10 15 20 25 30 35 40 45 50
0
1
C
o
set too cold A set too hot B
Figure 5.1. A fuzzy membership function for suitability of
water temperature for swimming.
Figure 5.2. Two membership functions relating to swimming
water temperature. Set A is the set defining the fraction of all
individuals who think the water temperature is too cold, and
Set B defines the fraction of all individuals who think the
water temperature is too hot.
E
0
2
0
1
0
8
0 5 10 15 20 25 30 35 40 45 50
0
1
C
o
f
m
A B
(x)
Figure 5.3. Membership function for water temperatures that
are considered too cold or too hot.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 136
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Fuzzy Optimization 137
subject to:
x 11 (5.5)
The obvious optimal solution, x 11, is shown in
Figure 5.4.
Now suppose the objective is to obtain a value of x sub-
stantially larger than 10 while making sure that the maxi-
mum value of x should be in the vicinity of 11. This is no
longer a crisp optimization problem; rather, it is a fuzzy one.
What is perceived to be substantially larger than 10
could be defined by a membership function, again repre-
senting the results of an opinion poll of what individuals
think is substantially larger than 10. Suppose the mem-
bership function for this goal, m
G
(x), reflecting the results
of such a poll, can be defined as:
m
G
(x) 1/{1 [1/(x 10)
2
]} if x 10
m
G
(x) 0 otherwise (5.6)
This function is shown in Figure 5.5.
The constraint on x is that it should be in the vicinity of
11. Suppose the results of a poll asking individuals
to state what they consider to be in the vicinity of 11 results
in the following constraint membership function, m
C
(x):
m
C
(x) 1/[1 (x 11)
4
] (5.7)
This membership function is shown in Figure 5.6.
Recall the objective is to obtain a value of x substan-
tially larger than 10 while making sure that the maximum
value of x should be in the vicinity of 11. In this fuzzy
environment the objective is to maximize the extent to
which x exceeds 10 while keeping x in the vicinity of 11.
The solution can be viewed as finding the value of x that
maximizes the minimum values of both membership
functions. Thus, we can define the intersection of both
membership functions and find the value of x that maxi-
mizes that intersection membership function.
The intersection membership function is:
m
D
(x) minimum{m
G
(x), m
C
(x)}
minimum{1/(1 [1/(x 10)
2
]),
1/(1 (x 11)
4
)} if x 10
0 otherwise (5.8)
This intersection set, and the value of x that maximizes its
value, is shown in Figure 5.7.
This fuzzy decision is the value of x that maximizes the
intersection membership function m
D
(x), or equivalently:
Maximize m
D
(x) max min{m
G
(x), m
C
(x)} (5.9)
Using LINGO

, the optimal solution is x 11.75 and

m
D
(x) 0.755.
U
E
0
2
0
1
0
8
g
20
15
10
5
0
X
0 5 10 15 20
x =
<
11
U x =
maximum feasible value
of U , and x
Figure 5.4. A plot of the crisp optimization problem defined by
Equations 5.4 and 5.5.
1.0
0.5
0.0
19 18 17 16 15 14 13 12 11 10 9 8 7 6 5
m
G
x
E
0
2
0
1
0
8
h
Figure 5.5. Membership function defining the fraction of
individuals who think a particular value of x is substantially
greater than 10.
1.0
0.5
0.0
19 18 17 16 15 14 13 12 11 10 9 8 7 6 5
m
C
x
E
0
2
0
1
0
8
j
Figure 5.6. Membership function representing the vicinity of 11.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 137
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
3. Fuzzy Sets for Water Allocation
Next consider the application of fuzzy modelling to the
water allocation problem illustrated in Figure 5.8.
Assume, as in the previous uses of this example prob-
lem, the problem is to find the allocations of water to each
firm that maximize the total benefits TB(X):
Maximize TB(X) 6x
1
x
1
2
7x
2
1.5x
2
2

8x
3
0.5x
3
2
(5.10)
These allocations cannot exceed the amount of water
available, Q, less any that must remain in the river, R.
Assuming the available flow for allocations, Q R, is 6,
the crisp optimization problem is to maximize Equation
5.10 subject to the resource constraint:
x
1
x
2
x
3
6 (5.11)
The optimal solution is x
1
1, x
2
1, and x
3
4 as
previously obtained in Chapter 4 using several different
138 Water Resources Systems Planning and Management
optimization methods. The maximum total benefits,
TB(X), from Equation 5.10, equal 34.5.
To create a fuzzy equivalent of this crisp model, the
objective can be expressed as a membership function
of the set of all possible objective values. The higher the
objective value the greater the membership function
value. Since membership functions range from 0 to 1, the
objective needs to be scaled so that it also ranges from
0 to 1.
The highest value of the objective occurs when there is
sufficient water to maximize each firms benefits. This
unconstrained solution would result in a total benefit of
49.17 and this happens when x
1
3, x
2
2.33, and
x
3
8. Thus, the objective membership function can be
expressed by:
m(X) 6x
1
x
1
2
7x
2
1.5x
2
2

8x
3
0.5x
3
2
49.17 (5.12)
It is obvious that the two functions (Equations 5.10 and
5.12) are equivalent. However, the goal of maximizing
objective function 5.10 is changed to that of maximizing
the degree of reaching the objective target. The optimiza-
tion problem becomes:
maximize m(X) 6x
1
x
1
2
7x
2
1.5x
2
2

8x
3
0.5x
3
2
49.17
subject to:
x
1
x
2
x
3
6 (5.13)
The optimal solution of (5.13) is the same as (5.10 and
5.11). The optimal degree of satisfaction is m(X) 0.70.
1.0
0.5
0.0
19 18 17 16 15 14 13 12 11 10 9 8 7 6 5
m
D
x
constraint objective
E
0
2
0
1
0
8
k
Q
x
x
x
1
2
3
E
0
2
0
1
0
8
m
firm 2
B
2
=7x
2
1.5x
2
2
3
firm 3
B =8x 0.5x
2
3 3
firm 1
B
1
=6x x
2
1 1
Figure 5.7. The intersection membership function and the
value of x that represents a fuzzy optimal decision.
Figure 5.8. Three water-
consuming firms i obtain
benefits B
i
from their allocations
x
i
of water from a river whose
flow is Q.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 138
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Fuzzy Optimization 139
Next, assume the amount of resources available to be
allocated is limited to about 6 units more or less, which
is a fuzzy constraint. Assume the membership function
describing this constraint is defined by Equation 5.14 and
is shown in Figure 5.9.
m
c
(X) 1 if x
1
x
2
x
3
5
m
c
(X) [7(x
1
x
2
x
3
)]/2 if 5 x
1
x
2
x
3
7
m
c
(X) 0 if x
1
x
2
x
3
7 (5.14)
The fuzzy optimization problem becomes:
Maximize minimum (m
G
(X), m
C
(X))
subject to:
m
G
(X) 6x
1
x
1
2
7x
2
1.5x
2
2

8x
3
0.5x
3
2
49.17
m
C
(X) [7 (x
1
x
2
x
3
)]/2 (5.15)
Solving (5.15) using LINGO

to find the maximum of a

lower bound on each of the two objectives, the optimal
fuzzy decisions are x
1
0.91, x
2
0.94, x
3
3.81,
m(X) 0.67, and the total net benefit, Equation 5.10, is
TB(X) 33.1. Compare this with the crisp solution of
x
1
1, x
2
1, x
3
4, and the total net benefit of 34.5.
4. Fuzzy Sets for Reservoir Storage
and Release Targets
Consider the problem of trying to identify a reservoir
storage volume target, T
S
, for the planning of recreation
facilities given a known minimum release target, T
R
, and
reservoir capacity K. Assume, in this simple example,
these known release and unknown storage targets must
apply in each of the three seasons in a year. The objective
will be to find the highest value of the storage target, T
S
,
that minimizes the sum of squared deviations from actual
storage volumes and releases less than the minimum
release target.
Given a sequence of inflows, Q
t
, the optimization
model is:
Minimize D
t
3
(T
S
S
t
)
2
DR
t
2
0.001T
S
(5.16)
subject to:
S
t
Q
t
R
t
S
t1
t 1, 2, 3; if t 3, t 1 1
(5.17)
S
t
K t 1, 2, 3 (5.18)
R
t
T
R
DR
t
t 1, 2, 3 (5.19)
Assume K 20, T
R
25 and the inflows Q
t
are 5, 50 and
20 for periods t 1, 2 and 3. The optimal solution, yield-
ing an objective value of 184.4, obtained by LINGO

is
listed in Table 5.1.
Now consider changing the objective function into
maximizing the weighted degrees of satisfying the reser-
voir storage volume and release targets.
Maximize
t
(w
S
m
St
w
R
m
Rt
) (5.20)
where w
S
and w
R
are weights indicating the relative impor-
tance of storage volume targets and release targets respec-
tively. The variables m
St
are the degrees of satisfying storage
volume target in the three periods t, expressed by Equation
5.21. The variables m
Rt
are the degrees of satisfying release
target in periods t, expressed by Equation 5.22.
m
S
=
{
S
t
/ T
S
for S
t
T
S
(5.21)
(KS
t
)/(KT
S
) for T
S
S
t
m
R
= R
t
/T
R
for R
t
T
R
(5.22)
{
1 for R
t
T
R
1.0
0.5
0.0
10 9 8 7 6 5 4 3 2 1 0
m
C
Q- R
E
0
2
0
1
0
8
n
Figure 5.9. Membership function for about 6 units more or less.
E
0
2
0
8
2
7
q
T
S
S
R
S
target storage for each period
reservoir storage volume at beginning of period 1
reservoir storage volume at beginning of period 2
reservoir storage volume at beginning of period 3
reservoir release during period 1
reservoir release during period 2
reservoir release during period 3
variable
R
R
s
1
1
2
2
3
3
value
15.6
19.4
7.5
20.0
14.4
27.5
18.1
remarks
Table 5.1. The LINGO

solution to the reservoir optimization

problem.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 139
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Equations 5.21 and 5.22 are shown in Figures 5.10 and
5.11, respectively.
This optimization problem written for solution using
LINGO

is as shown in Box 5.1.

Given weights w
S
0.4 and w
R
0.6, the optimal
solution obtained from solving the model shown in Box
5.1 using LINGO

is listed in Table 5.2.

140 Water Resources Systems Planning and Management
If the objective Equation 5.20 is changed to one of
maximizing the minimum membership function value,
the objective becomes:
Maximize m
min
maximize minimum {m
St
, m
Rt
} (5.23)
A common lower bound is set on each membership
function, m
St
and m
Rt
, and this variable is maximized. The
optimal solution changes somewhat and is as shown in
Table 5.3.
This solution differs from that shown in Table 5.2
primarily in the values of the membership functions. The
target storage volume operating variable value, T
S
, stays
the same in this example.
5. Fuzzy Sets for Water Quality
Management
Consider the stream pollution problem illustrated in
Figure 5.12. The stream receives waste from sources
1.0
0.5
0.0
K 0
m
S
E
0
2
0
1
0
8
o
storage S
t
T
s
Figure 5.10. Membership function for storage volumes.
1.0
0.5
0.0
35 30 25 20 15 10 5 0
m
R
E
0
2
0
1
0
8
p
release R
t
T
R
Figure 5.11. Membership function for releases.
E
0
2
0
8
2
7
s
degree
target storage volume
total weighted sum membership function values
variable
T
s
S
S
S
1
2
3
M
M
M
1
2
3
R
R
R
1
2
3
value remarks
M
M
M
M
M
M
1
1
2
2
3
3
s
R
s
s
R
R
20.00
1.00
25.00
0.60
30.00
0.88
20.00
0.00
1.00
1.00
0.80
1.00
1.00
20.00
20.00
0.00
2.48
storage volume at beginning of period 1
storage volume at beginning of period 2
storage volume at beginning of period 3
reservoir release in period 1
reservoir release in period 2
reservoir release in period 3
sum weighted membership values period 1
sum weighted membership values period 2
sum weighted membership values period 3
storage volume membership value period 1
storage volume membership value period 2
storage volume membership value period 3
reservoir release membership value period 1
reservoir release membership value period 2
reservoir release membership value period 3
Box 5.1. Reservoir model written for
solution by LINGO
SETS:
PERIODS /1..3/: I, R, m, ms, mr , s1, s2, ms1, ms2;
NUMBERS /1..4/: S;
ENDSETS
!*** OBJECTIVE ***; max = degree + 0.001*TS;
!Initial conditions; s(1) = s(TN + 1);
!Total degree of satisfaction; degree = @SUM(PERIODS(t): m(t));
!Weighted degree in period t; @FOR (PERIODS(t):
m(t) = ws*ms(t) + wr*mr(t);
S(t) = s1(t) + s2(t);
s1(t) < TS ; s2(t) < K TS ;
!ms(t) = (s1(t) / TS) (s2(t)/(K TS)) = rewritten in case dividing by 0;
ms1(t)*TS = s1(t); ms2(t)*(K TS) = s2(t); ms(t) = ms1(t) ms2(t);
mr(t) < R(t) / TR ; mr(t) < 1; S(t+1) = S(t) + I(t) R(t););
DATA:
TN = 3; K = 20; ws = ?; wr = ?; I = 5, 50, 20; TR = 25;
ENDDATA
E
0
2
0
9
0
3
d
Table 5.2. Solution of fuzzy model for reservoir storage
volumes and releases based on objective 5.20.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 140
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Fuzzy Optimization 141
located at Sites 1 and 2. Without some waste treatment at
these sites, the pollutant concentrations at Sites 2 and 3
will exceed the maximum desired concentration. The
problem is to find the level, x
i
, of wastewater treatment
(fraction of waste removed) at Sites i 1 and 2 required
to meet the quality standards at Sites 2 and 3 at a
minimum total cost. The data used for the problem
shown in Figure 5.12 are listed in Table 5.4.
The crisp model for this problem, as discussed in the
previous chapter, is:
Minimize C
1
(x
1
) C
2
(x
2
) (5.24)
subject to:
Water quality constraint at site 2:
[P
1
Q
1
W
1
(1x
1
)]a
12
/Q
2
P
2
max
(5.25)
[(32)(10) 250000(1x
1
)/86.4] 0.25/12 20
which, when simplified, is: x
1
0.78
Water quality constraint at site 3:
{[P
1
Q
1
W
1
(1x
1
)]a
13
[W
2
(1x
2
)]a
23
}/Q
3
P
3
max
(5.26)
{[(32)(10) 250000(1x
1
)/86.4] 0.15
[80000(1x
2
)/86.4] 0.60}/13 20
which, when simplified, is: x
1
1.28x
2
1.79
Restrictions on fractions of waste removal:
0 x
i
1.0 for sites i 1 and 2 (5.27)
For a wide range of reasonable costs, the optimal solution
found using linear programming was 0.78 and 0.79, or
essentially 80% removal efficiencies at Sites 1 and 2.
Compare this solution with that of the following fuzzy
model.
To develop a fuzzy version of this problem, suppose
the maximum allowable pollutant concentrations in the
stream at Sites 2 and 3 were expressed as about 20 mg/l
or less. Obtaining opinions of individuals of what
E
0
2
0
8
2
7
t
MMF
target storage volume
minimum membership function value
variable
T
s
S
S
S
1
2
3
R
R
R
1
2
3
value remarks
20.00
13.88
41.11
20.00
20.00
20.00
11.11
0.556
storage volume at beginning of period 1
storage volume at beginning of period 2
storage volume at beginning of period 3
reservoir release in period 1
reservoir release in period 2
reservoir release in period 3
M
M
M
M
M
M
1
1
2
2
3
3
s
R
s
s
R
R
0.556
1.000
0.800
0.556
0.556
0.556
storage volume membership value period 1
storage volume membership value period 2
storage volume membership value period 3
reservoir release membership value period 1
reservoir release membership value period 2
reservoir release membership value period 3
Table 5.3. Optimal solution of reservoir operation model
based on objective 5.23.
Q
stre
a
m
flo
w
firm
producing W
1 recreation
park
1
x W
site 1
( 1 )
1 1
site 3
x W
site 2
( 1 )
2 2
firm 2
producing W
2
E
0
2
0
1
0
8
q
Figure 5.12. A stream pollution
problem of finding the waste
removal efficiencies (x
1
, x
2
) that
meet the stream quality
standards at least cost.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 141
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
they consider to be 20 mg/l or less, a membership
function can be defined. Assume it is as shown in
Figure 5.13.
Next, assume that the government environmental
agency expects each polluter to install best available
technology (BAT) or to carry out best management
practices (BMP) regardless of whether or not this is
required to meet stream-quality standards. Asking experts
just what BAT or BMP means with respect to treatment
efficiencies could result in a variety of answers. These
responses can be used to define membership functions
for each of the two firms in this example. Assume these
membership functions for both firms are as shown in
Figure 5.14.
142 Water Resources Systems Planning and Management
Finally, assume there is a third concern that has to do
with equity. It is expected that no polluter should be
required to treat at a much higher efficiency than any
other polluter. A membership function defining just what
differences are acceptable or equitable could quantify this
concern. Assume such a membership function is as
shown in Figure 5.15.
Considering each of these membership functions as
objectives, a number of fuzzy optimization models can be
defined. One is to find the treatment efficiencies that
maximize the minimum value of each of these member-
ship functions.
Maximize m max min{m
P
, m
T
, m
E
} (5.28)
E
0
2
0
8
2
7
u
parameter
f
l
o
w
Q m 10 flow just upstream of site 1
1
Q
2
Q
3
w
a
s
t
e
W 250,000 pollutant mass produced at site 1 kg/day
1
P 32 concentration just upstream of site 1 mg/l
1
P
2
P
3
W
2
a
12
a
13
a
23
p
o
l
l
u
t
a
n
t
c
o
n
c
.
d
e
c
a
y
f
r
a
c
t
i
o
n
3
/s
unit
--
12 flow just upstream of site 2
80,000 pollutant mass produced at site 2
13 flow at park
20 maximum allowable concentration upstream of 2
20 maximum allowable concentration at site 3
0.25 fraction of site 1 pollutant mass at site 2
0.15 fraction of site 1 pollutant mass at site 3
0.60 fraction of site 2 pollutant mass at site 2
value remark
m
3
/s
m
3
/s
kg/day
mg/l
mg/l
--
--
Table 5.4. Parameter values selected for
the water quality management problem
illustrated in Figure 5.12.
1.0
0.5
0.0
24 22 20 18 16 14 12 10
m
concentration
E
0
2
0
1
0
8
r
Figure 5.13. Membership function for about 20 mg/l or less.
1.0
0.5
0.0
95 90 85 80 75 70 65 60
m

E
0
2
0
1
0
8
s
% treatment efficiency x
i
Figure 5.14. Membership function for best available
treatment technology.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 142
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Fuzzy Optimization 143
If we assume that the pollutant concentrations at sites
j 2 and 3 will not exceed 23 mg/l, the pollutant
concentration membership functions m
Pj
are:
m
Pj
1 p
2j
/5 (5.29)
The pollutant concentration at each site j is the sum of
two components:
p
j
p
1j
p
2j
(5.30)
where
p
1j
18 (5.31)
p
2j
(2318) (5.32)
If we assume the treatment plant efficiencies will be
between 70 and 90% at both Sites i 1 and 2, the treat-
ment technology membership functions m
Ti
are:
m
Ti
(x
2i
/0.05) (x
4i
/0.10) (5.33)
and the treatment efficiencies are:
x
i
0.70 x
2i
x
3i
x
4i
(5.34)
where
x
2i
0.05 (5.35)
x
3i
0.05 (5.36)
x
4i
0.10 (5.37)
Finally, assuming the difference between treatment
efficiencies will be no greater than 14, the equity mem-
bership function, m
E
, is:
m
E
Z (0.5/0.05) D
1
0.5(1 Z)
(0.5/(0.14 0.05)) D
2
(5.38)
where
D
1
0.05Z (5.39)
D
2
(0.14 0.05) (1Z) (5.40)
x
1
x
2
DP DM (5.41)
DP DM D
1
0.05(1Z) D
2
(5.42)
Z is a binary 0, 1 variable. (5.43)
The remainder of the water quality model remains the same:
Water quality constraint at site 2:
[P
1
Q
1
W
1
(1x
1
)] a
12
/Q
2
P
2
(5.44)
[(32)(10) 250000(1x
1
)/86.4] 0.25/12 P
2
Water quality constraint at site 3:
{[P
1
Q
1
W
1
(1x
1
)] a
13
[W
2
(1x
2
)] a
23
}/Q
3
P
3
(5.45)
{[(32)(10) 250000(1x
1
)/86.4] 0.15
[80000(1x
2
)/86.4] 0.60}/13 P
3
Restrictions on fractions of waste removal:
0 x
i
1.0 for sites i 1 and 2. (5.46)
Solving this fuzzy model using LINGO

yields the results

shown in Table 5.5.
This solution confirms the assumptions made
when constructing the representations of the member-
ship functions in the model. It is also very similar to
the least-cost solution found from solving the crisp LP
model.
1.0
0.5
0.0
14 12 10 8 6 4 2 0
m
E
E
0
2
0
1
0
8
t
x
1
- x
2
Figure 5.15. Equity membership function in terms of the
absolute difference between the two treatment efficiencies.
E
0
2
0
8
2
7
v
M minimum membership value
variable
X
P
X
P
1
2
3
value remarks
M
M
2
3
p
p
0.93
0.94
0.81
18.28
0.81
18.36
0.93
treatment efficiency at site 1
pollutant concentration just upstream of site 2
treatment efficiency at site 2
pollutant concentration just upstream of site 3
membership value for pollutant concentration site 2
membership value for pollutant concentration site 3
2
M
M
1
T
T
0.93
0.93 membership value for treatment level site 1
membership value for treatment level site 2
2
M membership value for difference in treatment 1.00
E
Table 5.5. Solution to fuzzy water quality management model
Equations 5.28 to 5.46.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 143
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
6. Summary
Optimization models incorporating fuzzy membership
functions are sometimes appropriate when only qualita-
tive statements are made when stating objectives and/or
constraints of a particular water management problem or
issue. This chapter has shown how fuzzy optimization
can be applied to some simple example problems associ-
ated with water allocations, reservoir operation, and
pollution control. This has been only an introduction.
Those interested in more detailed explanations and
applications may refer to any of the additional references
listed in the next section.
7. Additional References (Further
Reading)
BARDOSSY, A. and DUCKSTEIN, L. 1995. Fuzzy rule-
based modeling with applications to geophysical, biological,
and engineering systems. Boca Raton, Fla., CRC Press.
CHEN, S.Y. 1994. Theory and application of fuzzy system
decision-making. Dalian, China, Dalian University of
Technology Press. (In Chinese.).
144 Water Resources Systems Planning and Management
KINDLER, J. 1992. Rationalizing water requirements with
aid of fuzzy allocation model. Journal of Water Resources
Planning and Management, ASCE, Vol. 118, No. 3,
pp. 30823.
KUNDZEWICZ, W. (ed.). 1995. New uncertainty concepts
in hydrology and water resources. Cambridge, UK,
Cambridge University Press.
LOOTSMA, F.A. 1997. Fuzzy logic for planning and
decision-making. Boston, Mass., Kluwer Academic.
TERANO, T.; ASAI, K. and SUGENO, M. 1992. Fuzzy
systems theory and its application. San Diego, Calif., Academic
Press.
TILMANT, A.; VANCLOSSTER, M.; DUCKSTEIN, L. and
PERSOONNS, E. 2002. Comparison of fuzzy and
nonfuzzy optimal reservoir operating policies, Journal of
Water Resources Planning and Management, ASCE,
Vol. 128, No. 6, pp. 3908.
ZHOU, H.-C. 2000. Notes on fuzzy optimization and
applications. Dalian, China, Dalian University of Technology
Press.
ZIMMERMANN, H.-J. 1987. Fuzzy sets, decision-making,
and expert systems. Boston, Mass., Kluwer Academic.
wrm_ch05.qxd 8/31/2005 11:52 AM Page 144
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
6. Data-Based Models
1. Introduction 147
2. Artificial Neural Networks 148
2.1. The Approach 148
2.2. An Example 151
2.3. Recurrent Neural Networks for the Modelling of Dynamic Hydrological Systems 153
2.4. Some Applications 153
2.4.1. RNN Emulation of a Sewerage System in the Netherlands 154
2.4.2. Water Balance in Lake IJsselmeer 155
3. Genetic Algorithms 156
3.1. The Approach 156
3.2. Example Iterations 158
4. Genetic Programming 159
5. Data Mining 163
5.1. Data Mining Methods 163
6. Conclusions 164
7. References 165
wrm_ch06.qxd 8/31/2005 11:53 AM Page 146
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Most optimization and simulation models used for water
resources planning and management describe, in mathe-
matical terms, the interactions and processes that take
place among the various components of the system. These
mechanistically or process-based models usually contain
parameters whose values are determined from observed
data during model calibration. These types of models are
contrasted to what are typically called black-box models,
or statistical models. Statistical models do not describe
physical processes. They attempt to convert inputs (e.g.,
rainfall, inflows to a reservoir, pollutants entering a waste-
water treatment plant or discharged to a river) to outputs
(e.g., runoff, reservoir releases, pollutant concentrations
in the effluent of a treatment plant or downstream from a
point of discharge to a river) using any mathematical
equation or expression that does the job.
Regression equations, such as of the forms
output a b (input) (6.1)
output a b (input)
c
(6.2)
or
output a b
1
(input
1
)
c1
b
2
(input
2
)
c2
(6.3)
are examples of such data-based or statistical models.
They depend on data consisting of both observed inputs
147
Data-Based Models
Data sets contain information, often much more than can be learned from just
looking at plots of those data. Models based on observed input and output data
help us abstract and gain new information and understanding from these data
sets. They can also serve as substitutes for more process-based models in
applications where computational speed is critical or where the underlying
relationships are poorly understood. Models based on data range from the
commonly used regression models to models based on, for example, evolutionary
or biological processes. This chapter provides a brief introduction to these types of
models. They are among a wide range of data-mining tools designed to identify and
abstract information and/or relationships from large sets of data.
6
and observed outputs for the estimation of the values of
their parameters (a, b, c etc.) and for further refinement of
their structure. They lack an explicit, well-defined repre-
sentation of the processes involved in the transformation
of inputs to outputs.
While these statistical models are better at interpolating
within the range of data used to calibrate them, rather than
extrapolating outside that range (as illustrated in Figure 6.1),
many have proven quite successful in representing com-
plex physical systems within those ranges.
Regression equations are examples of data-based
modelling methods. Other examples of such methods
are based on evolutionary principles and concepts. These
are a class of probabilistic search procedures known as
evolutionary algorithms (EAs). Such algorithms include
genetic algorithms (GAs), genetic or evolutionary pro-
gramming (GP or EP) and evolutionary strategy (ES).
Each of these methods has many varieties but all use com-
putational methods based on natural evolutionary
processes and learning. Perhaps the most robust and
hence the most common of these methods are genetic
algorithms and all their varieties, and genetic program-
ming. Alternatively, an extension of regression is artificial
neural networks (ANN). The development and applica-
tion of black-box statistical models like GP and ANNs
emulate larger, deterministic, process-oriented models.
Their use may be advantageous when for some reason a
wrm_ch06.qxd 8/31/2005 11:53 AM Page 147
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
large number of model solutions must be obtained in a
short period of time.
Examples of such situations where multiple solutions
of a model must be obtained include sensitivity or uncer-
tainty analysis, scenario evaluations, risk assessment,
optimization, inverse modelling to obtain parameter
values, and/or when model runs must be extremely fast,
as for rapid assessment and decision support systems,
real-time predictions/management/control, and so on.
Examples of the use of data-driven models for model
emulation are given in the following sections.
Genetic algorithms and genetic programming are auto-
mated, domain-independent methods for evolving solu-
tions to existing models or for producing new models that
emulate actual systems, such as rainfallrunoff relation-
ships in a watershed, wastewater removal processes in a
treatment plant or discharges of water from a system of
natural lakes, each subject to random inputs.
Search methods such as genetic algorithms and genetic
programming are inspired by our understanding of
biology and natural evolution. They start initially with
a number of randomly created values of the unknown
variables or a number of black-box models, respectively.
The variable values or structure of each of these models
is progressively improved over a series of generations.
The evolutionary search uses the Darwinian principal of
survival of the fittest and is patterned after biological
operations including crossover (sexual recombination),
mutation, gene duplication and gene deletion.
148 Water Resources Systems Planning and Management
Artificial neural networks are distributed, adaptive,
generally nonlinear networks built from many different
processing elements (PEs) (Principe et al., 2000). Each
processing element receives inputs from other processing
elements and/or from itself. The inputs are scaled by
adjustable parameters called weights. The processing
elements sum all of these weighted inputs to produce an
output that is a nonlinear (static) function of the sum.
Learning (calibration) is accomplished by adjusting the
weights. The weights are adjusted directly from the
training data without any assumptions about the datas
statistical distribution or other characteristics (Hagan
et al., 1996; Hertz et al., 1991).
The following sections are intended to provide some
background helpful to those who may be selecting one
among all the available computer codes for implementing
a genetic algorithm, genetic program or artificial neural
network. These sections are largely based on Babovic and
Bojkov (2001) and from material provided by Mynett and
Van den Boogaard of Delft Hydraulics.
2. Artificial Neural Networks
2.1. The Approach
Before the development of digital computers, any infor-
mation processing necessary for thinking and reasoning
was carried out in our brains. Much of it still is. Brain-
based information processing continues today and will
unsupported area x
true relation E(y/x)
model y(x)
density
data
y
E
0
2
0
7
2
3
a
Figure 6.1. Data-driven models
are able to estimate relatively
accurately within their
calibrated ranges, but not
outside those ranges. The
bottom curve represents the
relative density of data used in
model calibration. The arrows
point to where the model does
not predict well.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 148
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 149
continue in the future even given our continually improv-
ing digital processing technology capabilities. While
recent developments in information technology (IT) have
mastered and outperformed much of the information pro-
cessing one can do just using brain power, IT has not mas-
tered the reasoning power of our brains. Perhaps because
of this, some computer scientists have been working on
creating information processing devices that mimic the
human brain. This has been termed neurocomputing. It
operates with networks (ANNs) representing simplified
models of the brain. In reality, it is just a more complex
type of regression or statistical (black-box) model.
The basic structure of an ANN is shown in Figure 6.2.
There are a number of input layer nodes on the left side
of the figure and a number of output layer nodes on the
right. The middle columns of nodes between these input
and output nodes are called hidden layers. The number of
hidden layers and the number of nodes in each layer are
two of the design parameters of any ANN.
Most applications require networks that contain at
least three types of layers:
The input layer consists of nodes that receive an input
from the external environment. These nodes do not
perform any transformations upon the inputs but
just send their weighted values to the nodes in the
immediately adjacent, usually hidden, layer.
The hidden layer(s) consists of nodes that typically receive
the transferred weighted inputs from the input layer or
previous hidden layer, perform their transformations on
it, and pass the output to the next adjacent layer, which
can be another hidden layer or the output layer.
The output layer consists of nodes that receive the
hidden-layer output and send it to the user.
The ANN shown in Figure 6.2 has links only between
nodes in immediately adjacent layers or columns and is
often referred to as a multi-layer perceptron (MLP) net-
work, or a feed-forward (FF) network. Other architec-
tures of ANNs, which include recurrent neural networks
(RNN), self-organizing feature maps (SOFMs), Hopfield
networks, radial basis function (RBF) networks, support
vector machines (SVMs) and the like, are described in
more detail in other publications (for example, Haykin,
1999; Hertz et al., 1991).
Essentially, the strength (or weight) of the connection
between adjacent nodes is a design parameter of the ANN.
The output values O
j
that leave a node j on each of its out-
going links are multiplied by a weight, w
j
. The input I
k
to
each node k in each middle and output layer is the sum of
each of its weighted inputs, w
j
O
j
from all nodes j provid-
ing inputs (linked) to node k.
Input value to node k: I
k

j
w
j
O
j (6.4)
input layer middle layer output layer
E
0
2
0
7
2
3
b
Figure 6.2. A typical multi-layer
artificial neural network
showing the input layer for ten
different inputs, the hidden
layer(s), and the output layer
having three outputs.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 149
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Again, the sum in Equation 6.4 is over all nodes j provid-
ing inputs to node k.
At each node k of hidden and output layers, the input
I
k
is an argument to a linear or non-linear function
f
k
(I
k

k
), which converts the input I
k
to output O
k
. The
variable
k
represents a bias or threshold term that
influences the horizontal offset of the function. This trans-
formation can take on a variety of forms. A commonly
used transformation is a sigmoid or logistic function as
defined in Equation 6.5 and graphed in Figure 6.3.
O
k
1/[1 exp{(I
k

k
)}] (6.5)
The process of converting inputs to outputs at each
hidden layer node is illustrated in Figure 6.4. The same
process also happens at each output layer node.
150 Water Resources Systems Planning and Management
The design issues in artificial neural networks are
complex and are major concerns of ANN developers. The
number of nodes in the input as well as in the output
layer is usually predetermined from the problem to be
solved. The number of nodes in each hidden layer and the
number of hidden layers are calibration parameters that
can be varied in experiments focused on getting the best
fit of observed and predicted output data-based on the
same input data. These design decisions, and most impor-
tantly the determination of the values of the weights and
thresholds of each connection, are learned during the
training of the ANN using predefined (or measured) sets
of input and output data.
Some of the present-day ANN packages provide
options for building networks. Most provide fixed
network layers and nodes. The design of an ANN can
have a significant impact on its data-processing capability.
There are two major connection topologies that define
how data flows between the input, hidden and output
nodes. These main categories are:
Feed-forward networks in which the data flow through
the network in one direction from the input layer to
the output layer through the hidden layer(s). Each
output value is based solely on the current set of
inputs. In most networks, the nodes of one layer are
fully connected to the nodes in the next layer (as
shown in Figure 6.2); however, this is not a require-
ment of feed-forward networks.
Recurrent or feedback networks in which, as their name
suggests, the data flow not only in one direction but in
the opposite direction as well for either a limited or a
complete part of the network. In recurrent networks,
information about past inputs is fed back into and
mixed with inputs through recurrent (feedback) con-
nections. The recurrent types of artificial neural net-
works are used when the answer is based on current
data as well as on prior inputs.
Determining the best values of all the weights is called
training the ANN. In a so-called supervised learning
mode, the actual output of a neural network is compared
to the desired output. Weights, which are usually ran-
domly set to begin with, are then adjusted so that the next
iteration will produce a closer match between the desired
and the actual output. Various learning methods for
weight adjustments try to minimize the differences or
E
0
2
0
7
2
3
c
1.0
0.5
0.0

k
Figure 6.3. The sigmoid or logistic threshold function with
threshold
k
.
nodes j node k
O w
4 4
O
O
O
w
w
w
4
4
4
4
4
4
input:
I
k
=
j
w
j
O
j
output:
I
k
= f O (
k

k
+ )
E
0
2
0
7
2
3
d
Figure 6.4. A hidden-layer node k converting input values to
an output value using a non-linear function f (such as defined
by Equation 6.5) in a multi-layer ANN.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 150
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 151
errors between observed and computed output data.
Training consists of presenting input and output data to
the network. These data are often referred to as training
data. For each input provided to the network, the corre-
sponding desired output set is provided as well.
The training phase can consume a lot of time. It is
considered complete when the artificial neural network
reaches a user-defined performance level. At this level the
network has achieved the desired statistical accuracy as it
produces the required outputs for a given sequence of
inputs. When no further learning is judged necessary, the
resulting weights are typically fixed for the application.
Once a supervised network performs well on the train-
ing data, it is important to see what it can do with data it
has not seen before. If a system does not give a reasonable
output for this test set, this means that the training period
should continue. Indeed, this testing is critical to ensure
that the network has learned the general patterns involved
within an application and has not simply memorized a
given set of data.
Smith (1993) suggests the following procedure for
preparing and training an ANN:
1. Design a network.
2. Divide the data set into training, validation and testing
subsets.
3. Train the network on the training data set.
4. Periodically stop the training and measure the error on
the validation data set.
5. Save the weights of the network.
6. Repeat Steps 2, 3 and 4 until the error on the valida-
tion data set starts increasing. This is the moment
where the overfitting has started.
7. Go back to the weights that produced the lowest error
on the validation data set, and use these weights for
the trained ANN.
8. Test the trained ANN using the testing data set. If it
shows good performance use it. If not, redesign the
network and repeat entire procedure from Step 3.
There is a wide selection of available neural network
models. The most popular is probably the multi-layer
feed-forward network, which is typically trained with
static back propagation. They are easy to use, but they
train slowly, and require considerable training data. In
fact, the best generalization performance is produced if
there are at least thirty times more training samples than
network weights (Haykin, 1999). Adding local recurrent
connections can reduce the required network size,
making it less sensitive to noise, but it may get stuck on a
solution that is inferior to what can be achieved.
2.2. An Example
To illustrate how an ANN might be developed, consider
the simple problem of predicting a downstream pollutant
concentration based on an upstream concentration and
the streamflow. Twelve measurements of the streamflow
quantity, velocity and pollutant concentrations at two
sites (an upstream and a downstream site) are available.
The travel times between the two measurement sites have
been computed and these, plus the pollutant concentra-
tions, are shown in Table 6.1.
Assume at first that the ANN structure consists of two
input nodes, one for the upstream concentration and the
travel
time
(days)
2.0
2.0
1.5
1.0
0.5
1.0
0.5
1.5
1.5
2.0
1.0
0.5
concentration
downstream upstream
20.0
15.0
30.0
20.0
20.0
15.0
30.0
25.0
15.0
30.0
30.0
25.0
6.0
4.5
12.2
11.0
14.8
8.2
22.2
10.2
6.1
9.0
16.5
18.5
E
0
2
0
7
2
3
e
Table 6.1. Streamflow velocities and pollutant concentrations.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 151
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
other for the travel time, a single hidden layer of only
one node, and a single output node, the downstream
concentration expressed as a fraction of the upstream
concentration. This is shown in Figure 6.5.
The model output is the fraction of the upstream
concentration that reaches the downstream site. That
fraction can be any value from 0 to 1. Hence the sigmoid
function (Equation 6.5) is applied at the middle node
and at the output node. Using two or more data sets
to train or calibrate this ANN (Figure 6.5) results in a
poor fit as measured by the minimum sum of absolute
deviations between calculated and measured con-
centration data. The more data samples used, the worse
the fit. This structure is simply too simple. Hence, another
152 Water Resources Systems Planning and Management
node was added to the middle layer. This ANN is shown
in Figure 6.6.
Using only half the data (six data sets) for training or
calibration, the weights obtained provided a near perfect
fit. The weights obtained are shown in Table 6.2.
Next the remaining six data sets were applied to
the network with weights set to those values shown in
Table 6.2. Again the sum of absolute deviations was
essentially 0. Similar results were obtained with increas-
ing numbers of data sets.
The weights in Table 6.2 point out something water
quality modellers typically assume, and that is that the
fraction of the upstream pollutant concentration that
reaches the downstream site is independent of the actual
upstream concentration
input nodes
w1
w2
E
0
2
0
7
2
3
g
w6
weights: weights:
travel time
middle layer nodes output node
w3
w4
w5
weights
w
w
w
w
E
0
2
0
7
2
3
h
value value weights
1
2
3
4
w
w
5
6
0.0
0.0
-0.6
3.9
8.1
-2.8
Table 6.2. Weights for each link of the ANN shown in
Figure 6.6 based on six data sets from Table 6.1.
All bias variables (
k
in Equation 6.5) were 0.
Figure 6.6. Modified ANN for example
problem.
upstream concentration
input nodes
w1
w2
E
0
2
0
7
2
3
f
w3
weights: weights:
travel time
middle layer node output node
Figure 6.5. Initial ANN for example problem.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 152
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 153
upstream concentration (see Chapter 12). This ANN
could have had only one input node, namely that for
travel time. This conforms to the typical first order
decay function:
Fraction of pollutant concentration downstream per unit
concentration upstream exp{k*(travel time)} (6.6)
where the parameter k is the decay rate constant (travel
time units
1
).
2.3. Recurrent Neural Networks for the
Modelling of Dynamic Hydrological Systems
The flexibility, efficiency, speed and emulation capacity of
data-based models are particularly relevant within the
fields of hydrology, meteorology and hydraulics, where
one often has to model dynamic systems and processes
that evolve in time. The previous illustration of an ANN
was typical of many static ANNs. In training and/or appli-
cations the inputoutput patterns could be processed
independently in the sense that an input pattern is not
affected by the ANNs response to previous input
patterns. The serial order of the inputoutput patterns
does not really matter.
This situation is the same as with the more com-
monly applied linear or non-linear regression and inter-
polation techniques where curves or surfaces are fitted
to a given set of data points. In fact, standard ANNs are
universal function approximators (Cybenko, 1989;
Hornik et al., 1989) and, as such, are static non-linear
regression models without explicit consideration of
time, memory, and time evolution and/or interaction of
inputoutput patterns.
For dynamic systems, the dependencies and/or interac-
tion of inputoutput patterns cannot be ignored. This is per-
haps most recognizable from the statespace property of
dynamic models. In a dynamic statespace model, the sys-
tems state in time t is based upon the systems state at time
t1 (and/or more previous time steps) and external forcings
on the system. (See, for example, the applications of
dynamic programming to reservoir operation in Chapter 4.)
This statespace property of dynamic models can be
included within an ANN to create data-based models
better suited for non-linear modelling and the analysis of
dynamic systems and time series. This extension leads to
the use of recurrent neural networks. These models are
equipped with a time propagation mechanism that
involves feedback of computed outputs rather than feed-
back of observed outputs, as shown in Figure 6.7. More
detail concerning RNNs can be found in Van den
Boogaard et al. (2000).
2.4. Some Applications
Here two hydrological RNN applications are presented.
The first case applies a RNN for the emulation of a
dynamic conceptual model of a sewerage system. The
t-2
X
(1)
t-2
X
(2)
t-2
t-1
t
X
X
(1)
(1)
t-1
t
X
(2)
t
U
t-2
U
t-1
U
t
E
0
2
0
7
2
3
j
Figure 6.7. Illustration of the architecture of a recurrent
neural network (RNN) where the inputs include external
data U as well as computed data X from previous periods t.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 153
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
second case applies a RNN to the modelling of the water
balance in Lake IJsselmeer in the Netherlands.
2.4.1. RNN Emulation of a Sewerage System in
the Netherlands
A major problem in sewerage systems is overflow during
severe rainstorm events. An overflow occurs when the
capacity of the sewerage system or treatment plants is
exceeded during a rainfall event. Overflow devices are
often no more than a storage chamber with a weir as an
element for flow control. An overflow results when the
upstream water depth exceeds the weirs crest.
Approximately 90% of all sewerage systems in the
Netherlands are combined systems, spilling diluted
sewage into open water systems during extreme storm
events. For most towns, these systems have mild slopes,
if any, which implies that the system forms an inter-
connected network with many loops. When all street
sewer lines are included, the resultant networks may have
thousands or even tens of thousands of pipes.
For the simulation of such complex sewerage systems,
models based on the numerical solution of the Saint Venant
equations can be used. One such deterministic model is
the SOBEK-URBAN model jointly developed by three
organizations in the Netherlands: WL | Delft Hydraulics,
RIZA and DHV. As one might expect, this is the principal
tool used in the Netherlands for the simulation of sewerage
systems (Stelling, 2000). This modelling platform provides
an integrated approach for a 1-D simulation of processes
in rivers, sewers and drainage systems. Flows in the pipe
network and the receiving waters can easily be combined.
Another well-known deterministic model applicable
for sewerage systems is the INFOWORKS model of
Wallingford Software in the UK.
Over the past years, legislation regarding sewage
spilling to open water systems has placed more and more
constraints on the amounts and frequency of overflows
allowed during storm events. To assess whether or not a
particular sewerage system meets the current legislation,
simulations over a ten-year period are required. In the
near future, legislation may require that this period be
extended to about twenty-five years.
Simulations of sewerage systems are performed on the
basis of recorded historical series of rainstorm events. For a
given layout, or for proposed rehabilitation designs (with,
154 Water Resources Systems Planning and Management
say, additional storage capacities), this historical set of
storm events must be simulated. For a time period of ten or
twenty-five years this will typically involve at least several
hundred events where sewer overflows result from storms.
For large systems containing several thousands of
pipes, such simulations are, and will continue to be,
computationally burdensome despite increased computer
speed and memory capacity. Even when restricted to
subsets of events with potential overflows, the computa-
tional effort will still be large, especially when alternative
designs must be compared. For this reason, RNNs as fast
model emulators and/or model reduction devices have
been developed (Price et al., 1998; Proao et al., 1998).
The external input consists of a rainfall time series dur-
ing a storm event. The output of the RNN is a time series
of overflow discharges at one or more overflow structures.
This output must be zero when no overflow occurs.
The RNN is calibrated on the basis of an ensemble of
rainfalloverflow combinations with the overflow time
series generated by a numerical model. The calibration
(and verification) set includes rainstorm events with
overflow as well as events without overflows. In this way,
the output time series of a few nodes of the numerical
model are emulated, rather than the complete state of the
sewerage system.
An important question is whether or not the emulation
can be based on a limited subset of the large original
ensemble of events. Otherwise the inputoutput of virtu-
ally all events must be pre-computed with the numerical
model, and the desired reduction of computational cost
will not be achieved. Apart from that, emulation would
no longer make sense as the frequency and quantity of
overflows could then be established directly from the
numerical models predictions.
This model emulation/reduction was applied to the
sewerage system of Maartensdijk, a town with about
10,000 inhabitants in the central Netherlands. Based on
the system emptying time and the storage/pump capaci-
ties, 200 potential overflow events were selected from a
rainfall series of ten years. For all forty-four events of
the calibration and verification set, a simulation with a
SOBEK-URBAN model of the Maartensdijk sewerage sys-
tem was carried out, providing time series of water depths
and discharges at three overflow structures in the system.
With these depth and discharge series, a RNN-model was
calibrated and verified. The weights of the RNN were
wrm_ch06.qxd 8/31/2005 11:53 AM Page 154
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 155
optimized for all training events simultaneously and not
for each event separately.
This study showed that the water depth rather than
the discharge determines the state of the system. In fact,
the discharges are a function of the water depths (rating
curve), but this relation is not one to one (for example:
the discharge is zero for any water depth below the
weirs crest). Therefore, discharges cannot be used for
the representation of the system state. This aspect nicely
illustrates that within black-box procedures too, physical
principles or system knowledge must often be used to
guarantee a meaningful model set-up.
After several training experiments, a proper architec-
ture of the RNN was identified and accurate emulations of
the SOBEK-URBAN model were obtained. Validation tests
on events not contained in the training and verification
set demonstrated the RNNs capability to emulate the
deterministic model.
2.4.2. Water Balance in Lake IJsselmeer
The most important factors that affect the water balance
of Lake IJsselmeer are the inflow of the River IJssel and
outflows at the sluices at Den Oever and Korn-
werderzand. Figure 6.8 is a map of the area with the
positions of the River IJssel and the sluices denoted.
The sluices form an interface of the lake to the tidal Dutch
Wadden Sea and are used to discharge excess water. This
spilling is only possible during low tides when the outside
water levels are lower than the water levels within the
lake. The spilled volumes are determined by the differ-
ence of these water levels. The inner water level can be
significantly affected by wind.
An RNN was prepared for the dynamic modelling of
the water levels of Lake Ijsselmeer (Van den Boogaard
et al., 1998). The model included five external system
forcings. These were the discharge of the River IJssel,
the northsouth and eastwest components of the wind,
and the outside tidal water levels at the sluices of
Kornwerderzand and Den Oever. The (recurrent) output
consists of the water level of Lake IJsselmeer. The time
step in the model is one day, adopted from the sampling
rate of the observed discharges of the River IJssel and the
lakes water levels. For synchronization, the time series of
the wind and the outer tidal water levels are also repre-
sented by daily samples.
The discharges of the River IJssel are a major source of
uncertainty. They were not directly observed but on the
basis of empirical relations using water-level registrations
of the River Rhine at Lobith near the DutchGerman
border. These relations may involve errors, especially for
winter seasons when water levels and/or discharges are
large or even extreme. In all the other external forcings,
such as the wind, large uncertainties may also be present.
For applications under operational or real-time condi-
tions (e.g., short to medium-term water-level forecasting)
these uncertainties can be taken into account more explic-
itly by modelling the uncertainties in a statistical sense
and applying online data assimilation techniques to
improve the models ability to forecast. Van den Boogaard
et al., (2000) present applications of online data assimila-
tion to this situation. The results from this RNN model
are probably as good as, if not better than, could be
expected from a much more detailed hydraulic process-
based model.
For other applications of artificial neural networks
applied to problems in meteorology, oceanography,
hydraulics, hydrology, and ecology, see Wst (1995),
Minns (1996), Minns and Hall (1996), Scardi (1996),
Abrahart et al., (2004), Recknagel et al., (1997), Clair and
Ehrman (1998), Hsieh and Tang (1998), Lange (1998),
Sanchez et al., (1998), Shen et al., (1998), Van Gent and
Amsterdam
Lake IJssel
Lake Marker
Wadden Sea
North Sea
E
0
2
0
7
2
3
k
Haarlem
Zaandam
Almere
Lelystad
25 20 15 10 5 0 km
Figure 6.8. Plan view of Lake Ijsselmeer.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 155
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Van den Boogaard (1998), Wen and Lee (1998) and See
and Openshaw (1999). For the identification of other
inputoutput relations see for example Haykin (1999)
and Beale and Jackson (1990).
3. Genetic Algorithms
3.1. The Approach
Genetic algorithms are randomized general-purpose
search techniques used for finding the best values of
the parameters or decision-variables of existing models. It
is not a model-building tool like genetic programming
or artificial neural networks. Genetic algorithms and their
variations are based on the mechanisms of natural selec-
tion (Goldberg, 1989). Unlike conventional optimization
search approaches based on gradients, genetic algorithms
work on a population of possible solutions, attempting to
find a solution set that either maximizes or minimizes the
value of a function of those solution values. This function
is called the objective function. Some populations of solu-
tions may improve the value of the objective function,
others may not. The ones that improve its value play a
greater role in the generation of new populations of
solutions than those that do not.
Each individual solution set contains the values of
all the parameters or variables whose best values are
being sought. These solutions are expressed as strings
of values. For example, if the values of three variables
x, y and z are to be obtained, these variables are
arranged into a string, xyz. If each variable is expressed
156 Water Resources Systems Planning and Management
using three digits, then the string 056004876 would
represent x 56, y 4, and z 876. These strings
are called chromosomes. A chromosome is an array of
numbers. The numbers on the chromosome are called
genes. Pairs of chromosomes from two parents join
together and produce offspring, who in turn inherit
some of the genes of the parents. Altered genes may
result in improved values of the objective function.
These genes will tend to survive from generation to
generation, while those that are inferior will tend to die.
Chromosomes are usually represented by strings of
binary numbers. While much of the literature on genetic
algorithms focuses on the use of binary numbers,
numbers of any base may be used.
To illustrate the main features of genetic algorithms,
consider the problem of finding the best allocations of
water to the three water-consuming firms shown in
Figure 6.9. Only integer solutions are to be considered.
The maximum allocation to any single user cannot exceed
5, and the sum of all allocations cannot exceed the value
of Q, say 6.
0 x
i
5 for i 1, 2, and 3. (6.7)
x
1
x
2
x
3
6 (6.8)
The objective is to find the values of each allocation that
maximizes the total benefits, B(X).
Maximize B(X)
(
6x
1
x
1
2
)

(
7x
2
1.5x
2
2
)

(
8x
3
0.5x
3
2
)
(6.9)
A population of possible feasible solutions is generated
randomly. A GA parameter is the size of the sample
Q
x
x
x
1
2
3
E
0
2
0
1
0
8
u
firm 2
B
2
=7x
2
-1.5x
2
2
3
firm 3
B =8x -0.5x
2
3 3
firm 1
B
1
=6x -x
2
1 1
Figure 6.9. Water allocation to
three users from a stream
having a flow of Q.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 156
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 157
solution population the number of solutions being
considered. The best values of genetic algorithm para-
meters are usually determined by trial and error.
Using numbers to the base 10, a sample individual
solution (chromosome) could be 312, representing the
allocations x
1
3, x
2
1, and x
3
2. Another individ-
ual solution, picked at random, might be 101. These two
individuals or chromosomes, each containing three genes,
can pair up and have two children.
The genes of the children are determined by crossover
and mutation operations. These pairing, crossover and
mutation operations are random. The crossover and
mutation probabilities are among the parameters of the
genetic algorithm.
Suppose a crossover is to be performed on the pair of
strings, 312 and 101. Crossover involves splitting the two
solution strings into two parts, each string at the same
place. Assume the location of the split was randomly
determined to be after the first digit,
3 | 1 2
1 | 0 1
Crossover usually involves switching one part of one string
with the corresponding part of the other string. After a
crossover, the two new individuals are 301 and 112.
Another crossover approach is to determine for each
corresponding pair of genes whether or not they will be
exchanged. This would be based on some pre-set proba-
bility. For example, suppose the probability of a crossover
were set at 0.30. Thus, an exchange of each correspon-
ding pair of genes in a string or chromosome has a 30%
chance of being exchanged. The result of this uniform
crossover involving, say, only the middle gene in the pair
of strings 312 and 101 could be, say, 302 and 111. The
literature on genetic algorithms describes many crossover
methods for both binary as well as base-10 numbers. The
interesting aspect of GA approaches is that they can be,
and are, modified in many ways to suit the analyst in the
search for the best solution set.
Random mutation operations apply to each gene in
each string. Mutation involves changing the value of the
gene being mutated. If these strings contained binary
numbers, a 1 would be changed to 0, and a 0 would be
changed to 1. If numbers to the base 10 are used as they
are here, mutation has to be defined. Any reasonable
mutation scheme can be defined. For example, suppose
the mutation of a base-10 number reduces it by 1, unless
the resulting number is infeasible. Hence in this example,
a mutation could be defined such that if the current value
of the gene being mutated (reduced) is 0, then the new
number is 5. Suppose the middle digit 1 of the second
new individual, 112, is randomly selected for mutation.
Thus, its value changes from 1 to 0. The new string is
102. Mutation could just as well increase any number by
1 or by any other integer value. The probability of a muta-
tion is usually much smaller than that of a crossover.
Suppose these paring, crossover and mutation opera-
tions have been carried out on numerous parent strings
representing possible feasible solutions. The result is a
new population of individuals (children). Each childs
fitness, or objective value, can be determined. Assuming
the objective function (or fitness function) is to be
maximized, the higher the value the better. Adding up all
the objective values associated with each child in the
population, and then dividing each childs objective value
by this total sum yields a fraction for each child. That
fraction is the probability of that child being selected for
the new population of possible solutions. The higher the
objective value, the higher the probability of its becoming
a parent in a new population.
In this example the objective is to maximize the total
benefit derived from the allocation of water, Equation 6.9.
Referring to Equation 6.9, the string 301 has a total ben-
efit of 16.5. The string 102 has a total benefit of 19.0. The
sum of these two individual benefits is 35.5. Thus the
string 301 has a probability of 16.5/35.5 0.47 of being
selected for the new population, and the string 102 has a
probability of 19/35.5 0.53 of being selected. Drawing
from a uniform distribution of numbers ranging from
0 to 1, if a random number is in the range 0 to 0.47, then
the string 301 would be selected. If the random number
exceeds 0.47, then the string 102 would be selected.
Clearly in a more realistic example the new population
size should be much greater than two, and indeed it typ-
ically involves hundreds of strings.
This selection or reproduction mechanism tends to
transfer to the next generation the better individuals
of the current generation. The higher the fitness (i.e.
the objective value) of an individual in other words,
the larger the relative contribution to the sum of objective
function values of the entire population of individual
solutions the greater will be the chances of that
wrm_ch06.qxd 8/31/2005 11:53 AM Page 157
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
individual string of solution values being selected for the
next generation.
Genetic algorithms involve numerous iterations of
the operations just described. Each iteration (or genera-
tion) produces populations that tend to contain better
solutions. The best solution of all populations of solu-
tions should be saved. The genetic algorithm process
can end when there is no significant change in the
values of the best solution that has been found. In this
search process, there is no guarantee this best solution
will be the best that could be found, that is, a global
optimum.
This general genetic algorithm process just described
is illustrated in the flow chart in Figure 6.10.
158 Water Resources Systems Planning and Management
3.2. Example Iterations
A few iterations with a small population of ten individual
solutions for this example water-allocation problem can
illustrate the basic processes of genetic algorithms. In prac-
tice, the population typically includes hundreds of indi-
viduals and the process involves hundreds of iterations. It
would also likely include some procedures the mod-
eller/programmer may think would help identify the best
solution. Here we will keep the process relatively simple.
The genetic algorithm process begins with the random
generation of an initial population of feasible solutions,
proceeds with the paring of these solution strings,
performs random crossover and mutation operations,
computes the probability that each resulting child will be
stop continue
yes
no
define population size, and probabilities for crossover
and mutation
randomly generate population of feasible solution
values called strings
define pairs of strings ( ) parents
perform crossover operations on each pair of strings,
if applicable
perform mutation operations on each gene in each
string, if applicable
evaluate 'fitness' of each new string ( ) and its
probability of being selected for the new population
child
save best solution found thus far, and compare
best fitness value with that of all previous populations
E
0
2
0
1
0
8
v
Figure 6.10. Flow chart
of genetic algorithm
procedure.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 158
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 159
selected for the next population, and then randomly
generates the new population. This process repeats itself
with the new population and continues until there is no
significant improvement in the best solution found from
all past iterations.
For this example, we will
1. Randomly generate an initial population of strings of
allocation variable values, ensuring that each alloca-
tion value (gene) is no less than 0 and no greater than
5. In addition, any set of allocations x
1
, x
2
and x
3
that
sums to more than 6 will be considered infeasible and
discarded.
2. Pair individuals and determine if a crossover is to be
performed on each pair, assuming the probability of a
crossover is 50%. If a crossover is to occur, we will
determine where in the string of numbers it will take
place, assuming an equal probability of a crossover
between any two numbers.
3. Determine if any number in the resulting individual
strings is to be mutated, assuming the probability of
mutation of any particular number (gene) in any string
(chromosome) of numbers is 0.10. For this example, a
mutation reduces the value of the number by 1, or if the
original number is 0, mutation changes it to 5. After
mutation, all strings of allocation values (the genes in the
chromosome) that sum to more than 6 are discarded.
4. Using Equation 6.9, evaluate the fitness (total bene-
fits) associated with the allocations represented by
each individual string in the population. Record the
best individual string of allocation values from this and
previous populations.
5. Return to Step 1 above if the change in the best solu-
tion and its objective function value is significant;
Otherwise terminate the process.
These steps are performed in Table 6.3 for three iterations
using a population of 10.
The best solution found so far is 222: that is, x
1
2,
x
2
2, x
3
2. This process can and should continue.
Once the process has converged on the best solution it
can find, it may be prudent to repeat the process, but this
time, change the probabilities of crossover or mutation or
let mutation be an increase in the value of a number
rather than a decrease. It is easy to modify the procedures
used by genetic algorithms in an attempt to derive the
best solution in an efficient manner.
4. Genetic Programming
One of the challenges in computer science is to learn how
to program computers to perform a task without telling
them how to do it. In other words, how can we enable
computers to learn to program themselves for solving par-
ticular problems? Since the 1950s, computer scientists
have tried, with varying degrees of success, to give com-
puters the ability to learn. The name for this field of study
is machine learning (ML), a phrase used in 1959 by the
first person to make a computer perform a serious learn-
ing task, Arthur Samuel. Originally, machine learning
meant the ability of computers to program themselves.
That goal has, for many years, proven very difficult. As a
consequence, computer scientists have pursued more
modest goals. A good present-day definition of machine
learning is given by Mitchell (1997), who identifies
machine learning as the study of computer algorithms
that improve automatically through experience.
Genetic programming (GP) aspires to do just that: to
induce a population of computer programs or models
(objects that turn inputs to outputs) that improve
automatically as they experience the data on which they
are trained (Banzhaf et al., 1998). Genetic programming
is one of many machine-learning methods. Within the
machine learning community, it is common to use genetic
programming as shorthand for any machine learning
system that evolves tree structures (Koza et al., 1992).
While there is no GP today that will automatically gen-
erate a model to solve any problem, there are some exam-
ples where GP has evolved programs that are better than the
best programs written by people to solve a number of diffi-
cult engineering problems. Some examples of these human-
competitive GP achievements can be seen in Koza (1999),
as well as in a longer list on the Internet (www.genetic-
programming.com/humancompetitive.html). Since Babovic
(1996) introduced the GP paradigm in the field of water
engineering, a number of researchers have used the tech-
nique to analyse a variety of water management problems.
The main distinctive feature of GP is that it conducts
its search for a solution to a given problem by changing
model structure rather than by finding better values of
model parameters or variables. There is no guarantee,
however, that the resulting structure (which could be as
simple as regression Equations 6.1, 6.2 or 6.3) will give us
any insight into the actual workings of the system.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 159
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The task of genetic programming is to find at the same
time both a suitable functional form of a model and the
numerical values of its parameters. To implement GP, the
user must define some basic building blocks (mathemati-
cal operations and variables to be used); the algorithm
160 Water Resources Systems Planning and Management
then tries to build the model using the specified building
blocks.
One of the successful applications of GP in automatic
model-building is that of symbolic regression. Here GP
searches for a mathem